196 The Assessment of Hazard and Risk

Figure 2.5.22 Height of the deposits foreseen in the high hazard and very high hazard areas of the alluvial fan
of the Cant Torrent in case of ordinary (a), serious (b), and catastrophic (c) events.

geomorphological information and complete an inventory of the control

measures while measuring critical sections.
At the first level of analysis, for a number of cases typical of the
larger Alpine valleys (100 alluvial fans), the evaluations were supported
by an estimate of the magnitude of flow events through the application of
regional empirical formulas. The application of these formulas requires
physiographical data, an estimate of the GI, and slope stability maps.
A higher level of detail was reached on representative basins for which
a recognisable geological and geomorphological pattern needed to be
extended in more detail throughout the basin in order to identify potential
sources of debris, the effective availability of the debris in the main chan-
nel, the position of eventual intermediary deposition zones, the efficiency
of the defence works and the contribution to the production of sediment
from the landslides and zones susceptible to high erosion rates.
 Torrential Processes 197

It is important to notice that a certain margin of subjectivity in param-

eterization of the factors that influence hazard is intrinsic to this method,
but does not affect individual parameters (geomorphology, acclivity, veg-
etation, flow conditions), objectively quantifiable in a system of relative
values; in case of attribution of relative weight for the various parameters
some themes should be considered as having more influence on hazard
then others: for this reason it was decided not to attribute relative weights
other than 1 to individual themes.
Even though somewhat inadequate from a rheological point of view,
the method in its detailed section provided a semi-quantitative estimate
of the basin fragility in terms of its propensity to generate torrential mass
transport, while allowing the forecasting of scenarios with different hypo-
thetical intensities through the comparison of the potential debris available
in the basin and the volume of sediment potentially stored in the higher
hazard level areas on the alluvial fan.

BIBLIOGRAPHY

The works marked by an asterisk have not been quoted explicitly in the text.
AA.VV. (2005a), Final Report. INTERREG IIIB Spazio Alpino Programme, Pro-
getto CatchRisk: Mitigation of Hydrogeological risk in alpine catchments.
Regione Lombardia. (Region of Lombardia).
AA.VV. (2005b), Final Report. INTERREG IIIB Spazio Alpino Programme, Pro-
getto CatchRisk: Mitigation of Hydrogeological risk in alpine catchments.
Regione Lombardia. (Region of Lombardia) Arpa Piemonte.
Anselmo V. (1985), Massime portate osservate o indirettamente valutate nei corsi
d’acqua subalpini. Atti e rassegna tecnica della Società degli Ingegneri e degli
Architetti in Torino (Papers and technical reports of the Society of Engineers
and Architects of Turin), New series, A. 39, n. 10–12, October–December 1985,
pp. 245–275 (in Italian).
*Arpa Piemonte, Geographic Information System – Portal for information
systems, (http://www.web-gis.csi.it/arpagis/).
*Arpa Piemonte – Sistema Informativo Geologico (Geological Information Sys-
tem) – Sottosistema Processi-Effetti – managed by the Arpa Piemonte Regional
Centre for Territorial and Geological Research.
*Arpa Piemonte (2002), Pericolosità geologica in conoide: ricerca bibliografica
e analisi critica. Quaderno n. 18 della Collana informativa Tecnico-scientifica
edita dal Centro Regionale per le Ricerche Territoriali e Geologiche di Arpa
Piemonte (in Italian).
*Aulitzky H. (1973), Vorläufige Wildbach-Gefährlichkeits-Klassifikation für
Schwemmkegel. 100 Jahre Hochschule für Bodenkultur, Band IV, Teil 2,
pp. 81–113 (in German).
Aulitzky H. (1980), Preliminary two-fold Classification of Torrent. Proc. Int.
Symp. Interpraevent, Bad Ischl, vol. 4, pp. 285–309.
Bianco G. & Franzi L. (2000), Estimation of debris-flow volumes from
storm events Debris-flow Hazards Mitigation: Mechanics, Prediction
198 The Assessment of Hazard and Risk

and Assessment.Wieczorek & Naeser (eds), 2000 Balkema, Rotterdam,

pp. 441–448.
*Bull W.B. (1977), The alluvial fan environment. Progress in Physical Geography,
1: pp. 222–270.
*Ceriani M., Crosta G., Frattini P. & Quattrini S. (2000), Evaluation of hydroge-
ological hazard on alluvial fans. Proc. Int. Symp. Interpraevent 2000, Villach,
Austria.
*Ceriani M., Fossati D. & Quattrini S. (1998), Valutazione della peri-
colosità idrogeologica sulle conoide alpine; esempio della metodologia di
Aulitzky applicata al conoide del torrente Re di Gianico – Valcamonica
(BS), Alpi Centrali. Convegno di Idraulica e costruzioni idrauliche, 3, pp.
15–26 (in Italian).
D’agostino V., Cerato M. & Coali R. (1996), Il trasporto solido di eventi estremi
nei torrenti del Trentino Orientale. Proc. Int. Symp. Interprevent 1996, 1,
pp. 377–386 (in Italian).
*ESRI (1998), ArcView, ArcGis, Spatial Analyst, Environmental System
Research Institute inc., Redlands, California, (http://www.esri.com).
*FEMA (1996), Alluvial fan flooding. Federal Emergency Management Agency.
National Academy Press, Washington D.C.
Ghilardi P., Natale L. & Savi F. (1999), Il rischio idraulico nelle aree di conoide.
Quaderni Regionali di Ricerca n. 34. Regione Lombardia. (in Italian).
Hampel R. (1977), Geschiebewirtschaft in Widbächen. Wildbach und Lawinen-
verbau, vol. 41, pp. 3–34 (in German).
*HYDRODATA S.p.A. (2004), Rapporto tecnico sulle attività svolte per conto di
Arpa Piemonte nell’ambito del Progetto CatchRisk, Work Package n◦ 3 Alluvial
fan flooding: Fornitura di consulenza tecnico-scientifica comprendente rilievi
geomorfologici e successive analisi di sintesi finalizzata alla delimitazione delle
aree di conoide a diverso grado di pericolosità in ambiente alpino. Rapporto
interno (in Italian).
Marchi L. &Tecca P.R. (1996), Magnitudo delle colate detriche nelleAlpi Orientali
Italiane. Geoingegneria Ambientale e Mineraria, 33(2/3), pp. 79–86 (in Italian).
Melton M.A. (1965), The geomorphic and paleoclimatic significance of alluvial
deposits in Southern Arizona. J. of Geology, 73, pp. 1–38.
Mizuyama T., Ikeya H. & Imamura R. (1984), Methods for zoning debris flow
vulnerable areas. Int. Symp. on Effects of Forest Land Use on Erosion and
Slope Stability, pp. 281–287.
Perla R., Cheng T. & Mc Clung T. (1980), A two-parameter model of snow
avalanche motion. J. of Glaciology, 26(4), pp. 197–207.
*PROGETTO IFFI (2004), (http://www.webgis.csi.it/arpagis/index.htm; http://
www.apat.gov.it/site/it-IT/Progetti/IFFI/).
*Regione Piemonte (1997), Esame di alcuni dati storici relativi ad eventi allu-
vionali e fenomeni di instabilità naturale nelle valli dei Torrenti Orco e Soana,
Alto Canavese, Provincia di Torino. Quaderno n˚ 7 della Collana informativa
Tecnico-scientifica edita dal Settore Studi e Ricerche Banca Dati Geologica
della Regione Piemonte (in Italian).
*Regione Piemonte (2000), Eventi alluvionali in Piemonte 10–14 Giugno 2000.
Direzione Servizi Tecnici di Prevenzione, p. 120 (in Italian).
Regione Piemonte (2001a), Raccolta e organizzazione di dati territoriali; val-
utazione di pericolosità e rischio dei fenomeni naturali e predisposizione di
 Torrential Processes 199

Piani Comunali di Protezione Civile, Programma Operativo UE Interreg IIC

(in Italian).
Regione Piemonte (2001b), Catasto Opere di Difesa , Direzione Difesa del Suolo,
(http://www.regione.piemonte.it/sit/argomenti/difesasuolo /catasto/home. htm).
Regione Piemonte (2002), Piani Forestali Territoriali (PFT) in scala 1:10,000,
Direzione Economia Montana e Foreste, (http://www.regione.piemonte.it/
montagna/foreste/pianifor/home.htm).
Rickenmann D. (1990), Debris flows 1987 in Switzerland: modelling and sedi-
ment transport. International Association of Hydrological Sciences Publ.,
pp. 371–378.
Rickenmann D. & Koch T. (1997), Comparison of debris flow modelling
approaches. Proc. of First Int. Conf. Debris-Flow Hazard Mitigation: Mechan-
ics, Prediction, and Assessment. ASCE, Cheng Iung Chen (ed), pp. 576–585.
Rickenmann D. & Zimmermann M. (1993), The 1987 debris flows in Switzerland:
documentation and analysis. Geomorphology, vol. 8, pp. 175–189.
Takahashi T., Ashida K. & Sawai K. (1991), Debris Flow. IAHR Monograph,
Balkema, Rotterdam.
Takei A. (1984), Interdependence of sediment budget between individual torrents
and a river system. Int. Symp., Villach, Austria, 2, pp. 35–48.
Voellmy A. (1955), Über die Zerstörungskraft von Lawinen. Schweizerische
Bauzeitung, 73(12), pp. 159–162; (15) pp. 212–217; (17) pp. 246–249; (19)
pp. 280–285 (in German).
Whipple K.X. & Dunne T. (1992), The influence of debris flow rheology on fan
morphology, Owens Valley, California. Geological Society of America Bulletin,
104, pp. 887–900.
 2.6 Avalanches

2.6.1 INTRODUCTION

The historical analysis of avalanches and their effects over the past 150
years, carried out through the consultation of archive data, monographs,
and period newspaper articles, has led to the estimation of serious events
in the Piedmontese Alps recurring over every 20–30 years approximately.

2.6.2 AVALANCHES IN THE TERRITORIAL CONTEXT OF

THE PIEDMONTESE ALPS

Historic events with decidedly catastrophic characteristics were those that

occurred in Piemonte during the 19th century, especially in January 1845,
January 1885, and February 1888. During the latter event the quantity of
snow that accumulated over several consecutive days reached the height
of 325 cm of snow on the ground in Oropa (BI), 350 cm in Prali (TO),
432 cm in Entracque (CN), 144 cm in Cuneo (Denza, 1889).
In 1888 avalanches of enormous dimensions resulted in serious dam-
age throughout the Italian Alpine Arc, causing a total of 248 victims. The
most serious damage occurred in mountain zones of an intermediate quota
in that exceptional avalanche phenomena involved many hamlets located
on the slopes and valley bottom, even at altitudes below 1000 meters, that
were densely inhabited at the time.
At the beginning of the 1900’s, a historically significant event tragically
struck the community of Pragelato (TO): an avalanche of huge dimensions

Figure 2.6.1 The Deveis

(Exilles – TO) avalanche
caused the destruction of 14
houses and 39 victims on
January 18th, 1885. Drawing
by Oreste Silvestri (Serra,
1997).
 202 The Assessment of Hazard and Risk

Figure 2.6.2 The avalanches during the month of April 1986 caused heavy damage to road and traffic conditions in
the Sesia and Ossola Valleys.

caused the death of 81 workers in the Beth mines in Val Troncea on April
19, 1904 (Avondo et al., 2003). This event was counted among the most
serious avalanche incidents in modern history because of the number of
victims involved. The occurrence of such exceptionally intense snows
over the regional area today would surely cause fewer victims than it did
in the past. This is due to both the higher level of technological and scien-
tific development over recent decades in the field of short and mid-term
forecasting snow meteorological events and the improvements in commu-
nications and warnings for civil protection purposes. Furthermore, even
the environmental and socio-economic contexts have changed consider-
ably: a more widespread wooded coverage of the slopes in comparison to
the past reduced the amount of land susceptible to the release of avalanches
and the concentration of the population in larger towns of mountain ter-
ritory, which are generally located in low-risk zones, further reduces the
number of people exposed to danger in comparison to the past.
However, the knowledge of the effects of past catastrophic events and
of statistic probability, even if low, that historically significant emergency
situations may in any case repeat themselves in the future, establishes the
 Avalanches 203

Figure 2.6.3 The nivometric

stations of the Arpa Piemonte
automatic meteorological
network constitute an
important source of
information on the snow
conditions for the forecasting
of avalanche risk.

need to take the risk of avalanches into consideration when planning for
the use of mountain territory.
Even in relatively recent times, highly intense snowy precipitation
has caused situations of high criticality for the Piedmontese territory. For
example, a significant event that involved mostly the northern sector of
the Piedmontese Alpine Arc in April 1986 caused numerous interruptions
in the road and traffic conditions along main and secondary roads of
the Sesia Valley and the Formazza Valley, luckily without causing any
victims.
On this occasion there was damage to several road and traffic infra-
structures, some of which could be traced to an underestimation of
avalanche risk during the planning phases. One example was the destruc-
tion of the viaduct of the State Highway of Alagna over the Sesia River
in the village of Isolello di Riva Valdobbia (VC) by an enormous ground
avalanche of wet snow.
The sustainable development of human activities in mountain terri-
tories must therefore include the risk assessment of avalanches that may
pose a threat to the structures and infrastructures of the mountain and
tourism economy.
204 The Assessment of Hazard and Risk

The prevention of this type of risk is possible above all through correct
planning of mountain land use based on two essential requisites:
• the availability of knowledge concerning past avalanches and their
effects, facilitated by the use of a Geographic Information System that
allows quick and efficient access;
• the application of scientific criteria in outlining the dangerous zones
through the use of models for calculating avalanche dynamics or
historical-statistical analysis tools.
The Region of Piemonte boasts advanced regulations in the planning
of land use in mountain territories, as outlined in Regional Law 56/1977
Land preservation and use (Tutela ed uso del suolo), and regarding the
aspects of hydrogeological, hydraulic, and avalanche risk prevention, by
the P.G.R. (President of the Regional Administration) Circular n. 7/LAP of
May 8th, 1996, Technical guidelines for the elaboration of geological stud-
ies to support urban planning tools (Specifiche tecniche per l’elaborazione
degli studi geologici a supporto degli strumenti urbanistici). The principle
behind these regulations is that of a balanced and sustainable development
of the territory through the development of construction and infra-
structures following an adequate assessment of the natural risks and the
application of prevention and protection provisions for their minimization.
Over the past twenty years the Regions and the Autonomous Provinces
of the Italian Alpine Arc have been coordinating their efforts to improve
avalanche forecasting services and the public information services for
the prevention of avalanche incidents following the 1983 institution of the
AINEVA (Associazione Interregionale Neve eValanghe), or Inter-regional
Snow and Avalanche Association, in which Arpa Piemonte represents the
Region of Piemonte in the Association’s Directive Technical Committee.
The results of this collaboration between the technical structures of the
individual Regional and Provincial Administrations specialised in activ-
ities related to the prevention of avalanche risk, are above all the adoption
of shared methodologies for actuating provisions and verification of snow
cover analyses, of standards for their elaboration, and the diffusion of
snow-weather information to the public. Moreover, the association plays
a leading role in the diffusion of knowledge in the field of snow science
through the publication of the four-monthly journal “Neve e Valanghe”
(Snow and Avalanches) and in the field of professional training in snow
science on a national level.
The forecasting of avalanche hazard constitutes a determining element
in the prevention of avalanche risk and is actuated on two territorial levels:
regional and local.
On a regional level, Arpa Piemonte contributes with the issuing of a
three-weekly snow bulletin recognised as a forecasting instrument within
the Piedmontese Civil Protection warning system, managed by the Cen-
tro Funzionale according to the operative guidelines provided by the
D.P.C.M. (President of the Ministerial Council Decree) of February 27th,
2004.
 Avalanches 205

Figure 2.6.4 The avalanche

hazard bulletin for the Region
of Piemonte.

The regional snow bulletin is edited according to short and mid-term

meteorological forecasts and the snow data acquired from the auto-
matic regional network, from the manual observatory network, from
survey results, stratigraphic analyses of the snow cover, and snow cover
stability tests.
The bulletin contains information on the snow cover (average depth
of the cover at 2000 metres, distribution of the cover, quantity of new
snow), on snow cover conditions (stratigraphy, consolidation, identifi-
cation of elements that may lead to the release of an avalanche), on the
degree of avalanche hazard (type and dimensions of the avalanches that
206 The Assessment of Hazard and Risk

have fallen or are expected, identification of hazard zones), and on the

evolution during the successive 24–48 hours. The degree of danger is
established according to the definition of the European Avalanche Danger
Scale, which is unified and differentiated by 7 geographic areas in the
Piedmontese Alpine Arc.
On a local level, the technical support to the Mayor in carrying out
his institutional tasks in order to guarantee public safety in avalanche
hazard situations is the responsibility of the Local Avalanche Com-
mission, instituted by the Mountain Communities as per Article 40
Difesa dalle valanghe (Defence from Avalanches) of Regional Law
02/07/1999, n. 16, Testo unico delle leggi sulla montagna (Unified
Code of Laws concerning Mountains), “for the actuation of snow phe-
nomena controls and for the purpose of indicating avalanche hazard
on towns or inhabited nuclei, public works and facilities, or public
infrastructures.”
The corresponding regulation, enacted by P.G.R. Decree n. 4R of
June 7th, 2002, Modalità costitutive e di funzionamento delle Commis-
sioni Locali Valanghe (Constitutive and Functional Modalities of Local
Avalanche Commissions), assigns Arpa Piemonte, by virtue of the transfer
of function as provided for by Regional Law 28/2002, the responsibility
of providing technical-scientific support to the Local Avalanche Com-
mission for the adoption of standard operative methodologies on regional
territories for the availability of snow and meteorological data.

2.6.3 THEMATIC AVALANCHE CARTOGRAPHY FOR LAND

USE PLANNING

2.6.3.1 Maps of probable avalanche locations

The knowledge of the difficulties related to the verification of avalanches
in mountain territories is of fundamental importance for technical bodies
of the Public Administration occupied on the various levels of planning
and management of the Alpine territory.
The Carte di Localizzazione Probabile delle Valanghe (CLPV) (Maps
of ProbableAvalanche Locations) constitute a synthetic cartographic prod-
uct of an avalanche knowledge base as effectuated by the Region and the
Autonomous Provinces belonging to AINEVA, according to a method-
ology that was shared and elaborated at the beginning of the 1970’s by the
French National Geographic Institute. This product was adopted as a refer-
ence document for the realization of follow-up in-depth studies finalized
at planning the use of mountain territories.
CLPVs are the integrated product of desk study through photo-
interpretation and field surveys on the territory in question, including
interviews and local testimony, the researching of archive data (parish
 Avalanches 207

Figure 2.6.5 Excerpt from a

Map of Probable Avalanche
Locations.

archives, forestry archives, town and editorial archives). The work was
carried out in three consequential phases:
• identification of avalanche sites by means of photo-interpretation and
summer aerial photograms;
• carrying out field surveys on the territory being studied and the
collection of oral testimony;
• verification of land data with historical information from archives and
bibliographies.
The CLPV reports the borders of avalanche sites, or the areas sus-
ceptible to simultaneous and sudden movement of the snow pack in the
maximum extension known at the time of drawing the map, even if this
border refers to events that occurred in past epochs and with high return
times of even more than 100 years.
Analytical files are compiled that report the topographical and mor-
phological data of the three areas that characterise each avalanche site: the
release zone, the run out zone, and the stopping zone. The description of
damages ascertained with the relative data of occurrence, as well as the
defence works actuated (represented also in the thematic map with specific
symbols), complete the characterisation of the site, together with eventual
photographic, journalistic, or technical documentation. Up until about
twenty years ago, the only example of an organic collection of archived
data on avalanches in Piemonte was the Archivio Storico-Topografico delle
Valanghe in Italia (Historical-Topographical Archive of Avalanches in
Italy) (Capello, 1977, 1980) for the territory of the Provinces of Cuneo
and Torino.
More or less at the same time as the birth of the Regional Nivometric
Service (Servizio Nivometrico Regionale) in 1983, on the occasion of
major precipitations, a permanent survey was initiated on avalanches in
Piemonte. These studies were structured around the use of special forms
standardised for the whole Alpine Arc as part of the inter-regional coor-
dination carried out by AINEVA (Model 7) and recorded by collaborators
of the regional nivometric network.
 208 The Assessment of Hazard and Risk

Figure 2.6.6 The Avalanche Information System shared by Arpa Piemonte and the Province of Torino.

In 1993, the Meteo-hydrographic Sector of the Region of Piemonte

began the realisation of the CLPV on a scale of 1:25,000 in collaboration
with the Soil Conservation Service of the Province of Torino and the
Cartographic Service of the Province of Cuneo.
In order to further detail the knowledge base of avalanches from a
distant past, a survey of the archives of local newspapers from the province
of Torino for the period 1885–1951 (Tomasuolo, 2001) was carried out;
the study resulted in the census of 203 historic avalanches associated with
162 different sites. The information extracted from the newspaper articles
of that time allowed the validation of the border of known avalanche sites
and, in many cases, increased the data base relative to events and damage
produced by the avalanches, even if in many cases it was not possible to
deduce the areal extension of the avalanches with precision from reading
the articles. The information obtained from thematic cartography and
documents are available in the Avalanche Information System, produced
with the collaboration of the Soil Conservation Service of the Province of
Torino and can be consulted on the Arpa Piemonte web site.
 Avalanches 209

2.6.3.2 Hazard maps

The adoption of risk mitigation measures in urbanized areas calls for
progressively in-depth studies, beginning with the representation of as
much historical data as possible concerning avalanches.
The availability of a registry of past avalanches and its synthesis in a
CLPV is considered to be the first step in a progressive process for refin-
ing prevention tools in the context of territorial planning. This level of
detail is appropriate on the planning scale of Territorial Plans and con-
stitutes the methodological base, for example, for the drafting of the
Hydraulic and Hydrogeological Risk Atlas attached to the Master Plan
for the Hydrogeological Safeguard (PAI, or Piano Stralcio per l’Assetto
Idrogeologico) adopted by the Po River Basin Authority with Resolution
of the Institutional Committee n. 18 on April 26th, 2001.
A correct and detailed evaluation of the interaction between avalanches
and forecasts to be used in local urban planning must be based on the
elaboration of specific studies aimed at outlining the areas with different
degrees of exposure to avalanche hazard to be effectuated during the pre-
liminary phases of drafting the General Municipal Regulatory Plan. The
identification of the perimeters is generally done with reference to the
frequency and intensity of the phenomena considered.
Over the past four years, AINEVA has, through an agreement stipu-
lated with the Department of Hydraulic Engineering of the University of
Padua and in collaboration with Arpa Piemonte, promoted research ori-
ented towards the definition of the scientific criteria for identifying the
perimeter of the zones at risk of avalanches as a support for the legislation
pertinent to regional territorial planning.
The research resulted in the publication of two specific documents:
Linee di indirizzo per la gestione del pericolo di valanghe nella piani-
ficazione territoriale (Guidelines for avalanche risk management in
territorial planning) and Criteri per la perimetrazione e l’utilizzo delle
aree soggette al pericolo di valanghe (Criteria for identifying the perim-
eters and the use of areas at risk of avalanches), approved by the AINEVA
Assembly respectively on February 23rd, 2001 and June 19th, 2002.
These two specific documents, which defined the framework of the
various phases of urban and territorial planning, were later supplemented
by the Linee guida metodologiche per la perimetrazione delle aree esposte
al pericolo di valanghe (Methodological guidelines for identifying the
perimeters of the areas exposed to the risk of avalanches), which developed
the technical-engineering aspects of the process while providing criteria
for the elaboration of detailed studies aimed at the drafting of Piani delle
zone esposte a valanga (Plans for zones exposed to avalanches) (Barbolini
et al., 2004a).
The document entitled Criteri per la perimetrazione e l’utilizzo delle
aree soggette al pericolo di valanghe defines the criteria adopted for out-
lining the perimeters of the areas exposed and suggests the most suitable
 210 The Assessment of Hazard and Risk

categories of potential use with a view to urban planning. The defining of

the different degrees of exposure to hazard, expressed essentially through
the parameters of return time of the avalanche and its impact pressure,
results in the outlining (even if temporary and subject to potential con-
tinual updating) of the areas subjected to avalanches, through the use of
mathematical models supported by historical and land data. The outlining
of areas subjected to different hazards therefore becomes a fundamental
instrument for defining the correct land usage system as part of the draft-
ing of urban planning tools for mountain communities, and consequently,
of Municipal Civil Protection Plans. The classification proposed calls for
the defining of three zones of decreasing hazard.

Figure 2.6.7 Example of

boundaries of different
avalanche hazard areas
obtained by the analysis of
historic and land data and
with the application of an
avalanche dynamics
simulation model.

Red Zone (high risk)

Red zones (zones of high risk) are those portions of the territory that can
be impacted by avalanches with a certain frequency, even if with modest
destructive potential, or more rarely by highly destructive avalanches. In
particular, a portion of territory is attributed to the red zone when there is
a possibility of:
• frequent avalanches (for which a conventional return time equal to 30
years is used as a reference) that cause pressure equal or greater than
3 kPa;
• rare avalanches (for which a conventional return time equal to 100 years
is used as a reference) that cause pressure equal or greater than 15 kPa.

Only one of these two conditions is in itself sufficient to attribute

the portion of territory in question to the red zone; for such areas, the
 Avalanches 211

30
29
28
27
26
25
24
23
22
21
20
19
18
17
P (kPa)

16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260 270 280 290 300

T (anni)

Figure 2.6.8 Representation of the different hazard zones according to frequency and impact pressure of a design
avalanche (Barbolini, personal communication).

application of urban planning restrictions for areas subjected to building

bans is advised.

Blue Zone (moderate risk)

Blue zones (zones of moderate risk) are those portions of the territory
that can be impacted by the residual effects of avalanches with a certain
frequency, or more rarely by moderately destructive avalanches. In par-
ticular, a portion of territory is attributed to the blue zone when there is a
possibility of:
• frequent avalanches (for which a conventional return time equal to 30
years is used as reference) that cause pressure less than 3 kPa;
• rare avalanches (for which a conventional return time equal to 100 years
is used as a reference) that cause pressure between 3 and 15 kPa.
Only one of these two conditions is in itself sufficient to attribute
the portion of territory in question to the blue zone; for such areas, the
application of restrictions for urban planning and building is advised.

Yellow zone (low risk)

Yellow zones (zones of low risk) are those portions of the territory that
can rarely be impacted by the residual effects of avalanches. In particular,
a portion of territory is attributed to the yellow zone when there is the
possibility that it produces rare avalanches (for which a conventional return
212 The Assessment of Hazard and Risk

time equal to 100 years is used as a reference) that cause pressure of less
than 3 kPa.
A yellow classification will also be assigned to portions of the territory
involved in the stopping of avalanches of exceptional occurrence (those
for which an indicative return time of 300 years used as a reference). For
such areas, building may be allowed with some reservations.

2.6.4 MITIGATION MEASURES FOR AVALANCHE RISK

IN THE PLANNING AND USE OF MOUNTAIN
TERRITORIES

In dealing with problems related to avalanche risk and the outlining of

relative mitigation measures, it is better to introduce some terminological
and conceptual definitions that identify some structural and non-structural
prevention measures, the role of short and mid-term forecasting, the sub-
jects involved in the various scales of the warning system (from a regional
level to that of a single town or a portion of it – for example, ski facilities),
the necessary education for acting when confronted with risk, and finally,
some aspects concerning regulations related to the specific problem.
Prevention includes all the measures adopted before the risk arises, in
order to diminish, avoid, or mitigate the damages expected as a conse-
quence of the occurrence in question. These are long term measures and
generally imply adequate territorial planning and management according
to the results obtained from a propaedeutic project (which was extensively
discussed in the previous section) that essentially amounts to the elabor-
ation of a zoning of risk for areas exposed to avalanches. These measures
back up and follow structural actions (works of classical engineering,
soil bio-engineering, improvements in forest management, etc.) or non-
structural actions (legislation and regulations on land use in risk zones,
civil protection plans, risk management plans in ski facilities, etc.). Until
recently, structural measures in inhabited centres and along major traffic
arteries was the most common solution sustained by a culture unwilling
to coexist with, and therefore to manage, such risks (for example through
temporary suspension of ski-lift facility operations or temporary closing
of roadways, etc.) and that aim instead at radical solutions, even if much
more costly.
In recent years, especially abroad, a more ecological line of inter-
vention based on risk management through a balance of structural and
non-structural measures has become more common in dealing with ski
facilities and road infrastructures, with the possibility of actively interven-
ing through the programmed release of snow masses during the preventive
closure of areas at risk following the development of critical situations.
Prevention therefore implies actions related to the identification and
in-depth characterisation of the phenomena and the outlining of areas
 Avalanches 213

that may potentially be involved (CLPV, Plan for Zones Exposed to

Avalanches, Local and Territorial Urban Planning, etc.), to national and
regional legislation, to specific sector regulations (limitations for land use
in areas at risk, Civil Protection Plans, regulations regarding the ski-lift
facility safety, etc.), and to parallel structural measures (active and passive
defence works) that can be actuated to mitigate risk.
With regards to the role of Arpa Piemonte avalanche structures, fore-
casting instead refers to a series of operations that, according to the
recording of the phenomena that involve the territorial region and the
short and mid-term meteorological forecasts, results in the publication of
regional avalanche risk forecast bulletins.

2.6.4.1 Ski slopes and lift facilities

In the case of ski facilities, which represent a fundamental element from
a socio-economic and tourism point of view for various Piedmontese
Alpine valleys, below is an overview of the principal national and regional
regulations that define the avalanche safety requirements for cableway
installations:
• Ministerial Decree 04/08/1998, n. 400 of August 4th, 1998. Regola-
mento generale per le funicolari aeree e terrestri in servizio pubblico
destinate al trasporto di persone (General regulation for air and land
cableway installations for the transport of persons for public use);
• Ministerial Decree n. 392 of December 5th, 2003, Regolamento con-
cernente modifica dell’art.7 del D.M. 4 agosto 1998, n. 400 (Regulation
concerning the modification of Article 7 of Ministerial Decree n. 400
of August 4th, 1998);
• Regional Law dated December 14th, 1989. Disciplina degli impianti
funiviari in servizio pubblico per il trasporto di persone (Regula-
tions for public service cableway installations for the transport of
persons);
• Decree n. 13/R of the P.G.R., dated November 29th, 2004, Regola-
mento regionale relativo alle procedure per l’approvazione dei progetti
per la realizzazione di nuovi impianti a fune e per l’ammodernamento
di impianti esistenti da parte delle comunità montane (Regional reg-
ulation concerning the approval procedure for cableway installation
construction projects and for the modernisation of existing installations
by mountain communities).
With regards to components relative to the stability of the slopes and
the environmental compatibility of the interventions, new regional regu-
lations, including actuation provisions that define conformity standards
for ski lift and ski slope installations are applied. These include: Regional
Law 09/08/1989, n. 45 concerning interventions to be carried out in lands
subject to hydrogeological restrictions, Regional Law 14/12/1998, n. 40
with provisions concerning the environmental compatibility and assess-
ment procedures, Regional Law 05/12/1977, n. 56, and P.G.R. Circular
 214 The Assessment of Hazard and Risk

Figure 2.6.9 A steel snow-fence to defend a lift facility from avalanches in the Rio Nero Valley in the Town of Cesana
Torinese (Via Lattea ski district).

n. 7/LAP 08/05/1996 regarding technical specifications for the elabora-

tion of geological studies to support the drafting of Municipal General
Zoning Plans.
The various regulations define a somewhat detailed and complex legi-
slative framework with a view to different objectives, primarily that of
guaranteeing the safety of persons, property, and infrastructures while
never losing sight of the need to preserve slope stability and optimize the
compatibility of the works within the Alpine environmental context.
In the preliminary procedure for obtaining authorization for the real-
ization and operation of new cableway installations, the recent Ministerial
Decree n. 392 of December 5th, 2003 states that conditions “for immunity
from avalanche risk, or the efficiency of the proposed interventions” must
be verified by competent authorities.
For this reason, those who make such proposals must program defence
works to avoid the possibility of avalanches impacting the fixed struc-
tural elements of the installation during the planning phase; this refers
 Avalanches 215

Figure 2.6.10 Active

avalanche defence works
and gas blasters for
programmed release of
avalanches (indicated with
red circles) on the NE slope
of Monte Fraiteve (Via Lattea
ski district).

to all types of active structural interventions (stabilization of the snow

cover, snow nets, snow fences, etc.) or passive types (deviation dams or
wedges, etc.) that impede the mass from reaching man-made structures.
The innovation introduced by Ministerial Decree 392/03 is the alterna-
tive of actuating active measures of risk management, consisting of the
controlled release of snowy masses that in any case must not reach the
fixed structural elements of the installation. If ever the risk of avalanches
should involve only the borderline of the facility and not the actual struc-
tural element, the temporary closure (prevention measure) of the facility
is allowed until the high risk situation has passed.
If conditions call for the actuation of risk management measures
through the preventive closure of the facilities and programmed release, a
Safety Management Plan that identifies the operative procedures and the
safeguards to be taken and an eventual Intervention Plan for the Artifi-
cial Release of Avalanches (PIDAV – Piano di Intervento per il Distacco
Artificiale di Valanghe) must be written and approved.
The Safety Management Plan must be actuated by personnel certified
in the issues related to the role they are called upon to play within the safety
plan: this qualification must be issued byAINEVA or by specialised Italian
or foreign institutions.
The Region of Piemonte does not yet have a law that regulates the safety
of ski slopes, including that concerning avalanche risk, even if the relative
Council Commission is discussing a bill that, among other things, should
absorb and enact the recent law L. 24/12/2003, n. 363 Norme in materia di
sicurezza nella pratica degli sport invernali da discesa e da fondo (Safety
regulations for practicing downhill and cross-country winter sports).
 216 The Assessment of Hazard and Risk

Figure 2.6.11
Snow-meteorological surveys
for local forecasting of
avalanche hazard.

Preliminary procedures concerning the verification phase, accord-

ing to Article 10 of Regional Law 40/1998, Disposizioni concernenti la
compatibilità ambientale e le procedure di valutazione (Regulations con-
cerning environmental compatibility and assessment procedures), for ski
slope projects of a total length of 1,500 metres also analyse the safety
aspects of the work, especially in relation to avalanche risk.
In particular situations and with sufficient technical motivations, for
example where the adoption of fixed structural works would be too costly
or have an excessive impact on the landscape, avalanche safety on ski
slopes may also be guaranteed through management measures, as have
been implemented for several years in France and in the Autonomous
 Avalanches 217

Province of Trento, and which are provided for in the implemental regu-
lation of Provincial Law 21/04/1987 Disciplina delle linee funiviarie in
servizio pubblico e delle piste da sci (Regulation of cableway installations
for public service and for ski slopes).
The Safety Management Plan is based on the continual monitoring
of meteorological and snow cover conditions in the ski district for local
forecasting of avalanche risk and the application of suitable procedures
for the preventative closure of the slopes under critical conditions.
An operative job description defines the procedures for the acquisition
of daily snow science-meteorological data to be carried out by specialised
personnel or acquired by automatic nivometrical stations; in these cases
nivometric poles may be installed in the sectors susceptible to avalanche
release for a remote reading of the height of the snow cover.
The need for the preventative closure of ski slopes at risk and their
successive reopening is established by the safety manager through the
adoption of scientific reference material (critical precipitation levels, wind
activity, temperature, etc.) and other information (stability tests, effects
of eventual artificial release with explosives).
In many cases the management of an avalanche risk safety plan is also
backed by the operative support of programmed avalanche release.
The programmed release of avalanches can be carried out with diverse
systems, from the more traditional ones with classic explosives to more
innovative ones that use gaseous substances (hydrogen or propane, mixed
with oxygen) that are easier to use in that they are not subjected to safety
regulations for transport and application.
This latter system consists of the high-quota installation of remote-
controlled release mechanisms; the plan for using these devices requires
a rigorous safety procedure to avoid the risk of accidents and a model
study of avalanche dynamics in order to evaluate the operative modes and
timing that allow the triggering of the release under optimal conditions
according to the height of the snow, which in turn allows the analyses of
avalanche stopping distances and eventual effects. An in-depth review of
the programmed avalanche release theory and the various systems being
used is given in Cresta (2002).
An example of the application of avalanche safety management pro-
cedures is represented by the case of the ski district of Belvedere in
Macugnaga (VB). The typically Alpine morphological characteristics of
the eastern slope of the Monte Rosa Massif considerably condition tourism
during the winter, when even very large avalanches may strike the area.
In 1991, the Region of Piemonte commissioned the Snow Science
Division of the CEMAGREF (Centre National du Machinisme Agricole,
du Génie Rural, des Eaux et des Forets) of Grenoble (France) to carry out
a specific study entitled Study of the avalanche risks on the ski district
west of Macugnaga (Buisson & Charlier, 1993), which resulted in the
development of avalanche risk scenarios for lift facilities and ski slopes
located in the area of Pecetto-Burki-Belvedere.
 218 The Assessment of Hazard and Risk

Figure 2.6.12 Direct snow

cover stability test
(Rutschblock).

This study included the analysis of potentially feasible solutions to

guarantee the operational safety of ski district facilities by specifically
evaluating the possible alternatives between structural interventions and
risk management procedures through programmed avalanche release
similar to systems implemented in comparable contexts in France. The
realization of active defence structures (snow nets or bridges) resulted as
not being feasible from technical and economical points of view for the
extension of the release area to be developed and for the presence of areas
with glacial coverings on which it is not possible to build.
According to these restrictions and analogous experiences conducted
for some time in France, CEMAGREF technicians prospect the actuation
of risk management procedures through the preventive closing of the ski
district, carried out according to scientific criteria based on the acquisition
and analysis of snow science and meteorological data, snow cover stability
tests, and observations by a team of qualified experts to be nominated by
the managing Company.
Since 1993 until today, an Avalanche Safety Plan that constitutes an
integral part of the ski lift operations regulations has been actuated in
the western ski district of Macugnaga. This plan is based on specific
provisions as indicated by the results of the above mentioned study and
defined by Arpa Piemonte. In order to back up the Plan Management,
another actuation of a PIDAV is being planned, thanks to the installation
of four gas blasters in quota and the use of explosives transported by
helicopter for interventions on slopes with critical snow conditions.
 Avalanches 219

Figure 2.6.13 Effect of a powder snow avalanche fallen from Cima Jazzi on a larch wood at Burki,
Macuganaga (VB) Photo by R. Cresta.

2.6.5 A CASE STUDY: THE RISK OF AVALANCHES ALONG

THE ROADWAYS OF THE UPPER SUSA AND
CHISONE VALLEYS

2.6.5.1 Scope
In order to support the activities of the Safety Planning Group for the
2006 XX Winter Olympic Games in Torino, instituted by the Prefecture
of Torino, within which the Civil Protection Service of the Province of
Torino was responsible for defining the risk scenario following intense
and prolonged snowfalls, the Arpa Piemonte Regional Forecasting and
Environmental Monitoring Area carried out an avalanche risk study on
the roadways and transportation system in the Olympic area.
The study constituted a detailed consolidation of the knowledge base
for avalanche phenomena represented in the CLPV of the Upper Susa
Valley and the Upper Chisone Valley and contained in the Avalanche
Information System shared by Arpa Piemonte and the Province of Torino.
It represented a contribution from a viewpoint of managing a situation
of critical snowfall – meteorological conditions that may have caused
avalanches on access roads to the Olympic Alpine area. In consideration of
 220 The Assessment of Hazard and Risk

the strategic importance that mobility played in the context of the Olympics
(held February 10th–26th, 2006) and the Paralympics (held March
10th–19th, 2006), the results of the study were aimed at identifying
stretches of transport routes along the bottom of the valley exposed
to avalanches if there should be critical snowfall conditions. This was
done through opportune elaborations of the available knowledge of past
avalanches and the morphological conditions of the territory with a view
to avalanche release.
The scenarios that were derived can be used to assist the agencies and
authorities responsible for traffic safety in optimising the preventive trans-
portation route closing procedures and the safeguarding measures to be
implemented in order to guarantee public safety with a secondary view of
rapidly recovering use of the road network once the emergency situation
has passed.
This study was carried out with another secondary objective of evalu-
ating the site-by-site applicability of programmed avalanche release
procedures with either conventional explosives or helicopter transport
innovations. The adoption of programmed release systems, where feasible,
would facilitate emergency management.

2.6.5.2 The study area

The study area was within the Alpine sector of the 2006 Torino Win-
ter Olympic Games and focused especially on the mountain territories
susceptible to avalanches (Figure 2.6.14).
More specifically, the valleys studied were:
• Upper Chisone Valley, from the municipal territory of Roure as far as
that of Sestriere;
• Upper Susa Valley, territory upstream of Salbertrand, in the Cesana and
the Bardonecchia Dora River Basins.

Figure 2.6.14 Identification of

the mountain territory of the
2006 Torino Winter Olympic
Games, object of the study.
 Avalanches 221

Only avalanche sites that interfere with the transport network that were
strategic during the Olympic period and within the portions of these valleys
indicated were studied.

2.6.5.3 The methodology

The work was organised according to a classical scientific methodology.
The first phase consisted of the bibliographic collection of documentary
and cartographic data from the archives of the Arpa Piemonte Regional
Forecasting and Environmental Monitoring Area.
The historic data were collected mostly from the Historic-Topographic
Archive of Italian Avalanches (Capello, 1980) and a specific study carried
out on local newspaper articles from the period (Tomasuolo, 2001). The
1:25,000 scale CLPV of the communities involved were referred to as
thematic cartography.
The second phase consisted of a land survey campaign, site by site, to
verify and eventually correct past data and to collect data on avalanche sites
not yet known or not sufficiently characterised by the bibliography. This
phase was also supported by the collection of oral testimony, in particular
from personnel involved in the removal of snow from the roadway network,
in order to optimise the outlining of the perimeter and characterise the
frequency of phenomena on each individual avalanche site. Together with
the land survey, a stereoscopic photo interpretative analysis of colour
aerial photographs from 1979 and 2000 was carried out.
The third phase consists of the graphic and conceptual elaboration
of the data collected during the first two phases. The attribution of
each individual site to a class of probability (high, moderate, or low)
of avalanche release under specific critical threshold snowfall conditions
was defined according to the geomorphological, topographical, and plant
covering features of the site. The evaluations carried out were verified
through a comparison with the available historical documentation and
with the effects of a moderately critical snowfall, which, during early
March 1993, caused avalanches of a certain importance on the road-
ways of the Chisone Valley and the effects of which were recorded in
detail.
The objective assessment of the land inclination values in order to
identify the potential release areas was obtained through the elaboration
of 10 metre grids with the Digital Elevation Model realized by Arpa
Piemonte Regional Centre for Territorial and Geological Research. The
use of ESRI© ArcGis software allowed the subdivision of the study area
into classes of significant slope (Figure 2.6.16). The areas susceptible to
release, or with slopes between 28◦ and 60◦ , were later examined with
orthophoto maps from the Regional Technical Map to verify the presence
of tree coverings, allowing the identification of zones where the tree cov-
ering guarantees the stability of the snow covering from those areas where
the retention of the snow cover is negligible or inexistent.
 222 The Assessment of Hazard and Risk

METHODOLOGY

1 PRELIMINARY STEP
- COLLECTION OF:
• HISTORICAL DATA
• BIBLIOGRAPHY
• MAPS
• SNOW-METEOROLOGICAL DATA
- STEREOSCOPIC PHOTO INTERPRETATIVE ANALYSIS

2 LAND SURVEY
- VALIDATION OF PAST DATA
- MORPHOLOGICAL AND VEGETATIONAL SURVEY
- CATALOGUE OF PROTECTION WORKS
- CATALOGUE OF AVALANCHES DAMAGES
- COLLECTION OF ORAL TESTIMONY

3 DATA ELABORATION
- STATISTICAL ANALYSIS OF SNOW-METEOROLOGICAL DATA
- MODELLING AVALANCHES
- AVALANCHE SITES MAPPING
- ANALYSIS OF SNOWFALL CONDITIONS VS AVALANCHES PAST
EFFECTS RELATION
- ASSESSMENT OF RISK SCENARIOS

Figure 2.6.15 Scheme of the ATTRIBUTION OF EACH INDIVIDUAL SITE TO A CLASS OF PROBABILITY
methodology implemented.

Figure 2.6.16 Excerpt of the

slope map for the study area,
discriminated with an
algorithm applied in ESRI©
ArcGis.
 Avalanches 223

Maximal Pressure of Simulation

120
1600
100

Max Pressure (kPa)

Height (m asl)

1500 80

1400 60

40
1300
20
1200
0
0 200 400 600 800 1000 1200 1400
Track width (m)

250
200
150
100
50
0
0 200 400 600 800 1000 1200 1400
Projected Track Length (m)

Figure 2.6.17 Output of a simulated model obtained with AVAL-1D software for calculating the run out distance of
an avalanche.

The morphological tendency of individual release areas to function as

sites where snow drifts accumulate in relation to the vicinity to the main
crests or secondary watersheds that may influence the dynamics of the
transport and the drifts was considered as an aggravating element in the
site risk rating. Risk assessment is intended as the expected recurrence
of avalanches and not their magnitude, which was not considered as a
differentiating element in identifying stretches of roadways at risk of being
interrupted. Other parameters considered in the assessment of potential
avalanche sites and the long distance run out of the snow mass are the
roughness of the land and the transversal geometry of the slope, or its
degree of channelling and the tortuosity of the track channels.
In case of doubtful interpretation in areas where the moderate slope and
the linear development of the deposition area of the avalanche impeded,
due to scarce historical data or the semi-quantitative assessment, a reliable
definition of involvement of the transportation system in an avalanche and
the snow cover conditions, recourse was made to a modelling of avalanche
dynamics through the use of a one-dimensional mathematical model to
identify the stopping distances of an avalanche in order to calculate the
height of the snow at the release point. The AVAL-1D dynamic numerical
calculation program used was developed by Davos by the Federal Institute
for Snow and Avalanches (Christen et al., 2002), allows the calculation
 224 The Assessment of Hazard and Risk

Avalanche n.
Topographic map
Pictures
General description
Province
Municipality
Place
Site denomination
Basin
Morphometric features
Max height of release (m asl)
Min height of stopping (m asl)
Difference in level (m)
Real length (m)
Real area (m2)
Release zone
Position
Morphology
Bedrock
Max length (m)
Max width (m)
Mean inclination (°)
Aspect (°)
Real area (m2)
Run out zone
Position
Morphology
Bedrock
Mean inclination (°)
Profile
Planimetric profile
Stopping zone
Position
Morphology
Mean inclination (°)
Buildings
Presence of evidences
Defense works
Typology
Possible defense works
Typology
Damages
Figure 2.6.18 Avalanche site Sure
information form. Potential

of the run out distance, speed, and pressure of the both dense and pow-
dery type avalanches. According to the results obtained, it is possible to
distinguish situations that present greater uncertainty through a critical
and careful assessment of the modelling results.

2.6.5.4 Cartographic documentation

Each individual avalanche studied was graphically represented in a Geo-
graphic Information System with ESRI© ArcView software to allow the
management of all the existing georeferenced documentation for a single
site in one single information system. The study led to the identifica-
tion and characterisation of about 60 avalanche sites that interfere with
mountain roadway and traffic systems of Olympic interest.
 Avalanches 225

The cartographic elaboration was done on a scale of 1:10,000 (with an

overall framework on a scale of 1:25,000): each avalanche site mapped
was defined with a limit and characterised with a specific colour.
A descriptive form that includes the geographic, morphological, and
dimensional characteristics was completed for each site; furthermore,
each description form also includes, when possible, the data relative to
the damage caused by the avalanches (date of the event, detail of dam-
ages, eventual injuries and/or victims). Each form is subdivided into fields
completed with a series of codified definitions.
The photographic documentation consists of some photographs with
captions that describe the different zones of each avalanche site and the
documented damage in detail (when possible).

2.6.5.5 The map of probability of exposure of the transport

system to avalanches
The Map of Probability of Exposure of the Roads to Avalanches (Carta
di Probabilità di Esposizione della Viabilità alle Valanghe) is a thematic
cartographic document drawn on a scale of 1:25,000 (with details on a
scale of 1:5,000 and/or 1:10,000) on a topographic base of the Regional
Technical Map.
Each avalanche site mapped is characterised with a specific colour –
red, orange, or yellow – according to the legend found in Figure 2.6.19.
The three classes of risk, intended in this case as a degree of probability that
the avalanche will involve major roads and transportation systems under
specific critical snow conditions, are compared to an event with daily
snowfall equal to 60–80 cm. This snowfall can be characterised, according
to statistical analysis of climatic series referred to four nivometric stations
in the Olympic area, with a return time (Tr) of decades. In other words,
each class is compared to an event that has a 10% probability of occurring
annually in this area.
In the elaboration of the study, such values were considered as thresh-
olds that identify critical snow conditions for the triggering of spontaneous
avalanche activity; the three classes of risk defined in this way allow
the identification of roadway and traffic sectors that are more exposed
to the risk of avalanches. During critical threshold snowfall, these sec-
tors should be immediately subjected to a preventive closing to traffic
while monitoring and assessment of the stability of the snow cover is
carried out.
The objective of the study was therefore to outline a scenario of moder-
ate criticality for road and traffic conditions in the Olympic area in relation
to avalanche risk, to define the snow conditions comparable to those that
occurred during an intense snowfall early in March 1993, for which there
exists detailed information concerning avalanches that occurred along the
ex-State Highway n. 23 in the Chisone Valley.
 226 The Assessment of Hazard and Risk

Avalanche site with high probability of involving major

roads under specific critical snow conditions (daily
snowfall equal to 60-80 cm)

Avalanche site with moderate probability of involving

major roads under specific critical snow conditions
(daily snowfall equal to 60-80 cm)

Avalanche site with low probability of involving major

roads under specific critical snow conditions (daily
snowfall equal to 60-80 cm)

Avalanche site not involving major roads under

specific critical snow conditions (daily snowfall equal to
60-80 cm), in Map of Probable Avalanche Locations

Figure 2.6.19 Excerpt of thematic Map of Probability of Exposure of the Roads to Avalanches on orthophoto
paper.
 Avalanches 227

Figure 2.6.20 Avalanche sites on the ex-State Highway n. 23 in the Val Chisone.

2.6.6 CONCLUSIONS

The functions and competencies of the Arpa Piemonte-Regional Fore-

casting and Environmental Monitoring Area cover the various sectors
concerning the prevention of avalanche risk.
The management of the regional nivometric network, consisting of
more than 70 automatic teletransmission stations and 40 station managed
manually, allows continual monitoring of the snow science-meteorological
conditions of the entire Piedmontese Alpine Arc.
The forecasting of avalanche risk on the regional territory constitutes
a public information service that frequents the mountain territory during
the winter and constitutes an important element of the regional warning
system managed by the regional Centro Funzionale in compliance with
the D.P.C. (Decree of the Prime Minister) provision dated February 27,
2004.
In the same context of prevention, Arpa Piemonte provides technical-
scientific support for the prevention activities carried out by the Local
Avalanche Commissions instituted by the Mountain Communities in com-
pliance with Article 40 of the Regional Law n. 16/1999, in their function
of controlling snowfall phenomena in order to signal avalanche risk for
urban centres or hamlets, public works and facilities, or infrastructures.
The realisation and continual updating of the shared Avalanche Infor-
mation System makes all the information concerning known avalanche
sites on regional territory available, representing the knowledge synthe-
sized in the cartography or the archive documentation for each site on
different informative layers.
The knowledge of avalanches in the Information System offers an
updated overview of the problems that may interfere with the use of moun-
tain territory, either during the planning phase or during the adoption of
protection measures for the construction of buildings or infrastructures.
The interaction with analogous technical structures of the regions or
the autonomous provinces of the Italian Alpine Arc belonging to AINEVA
228 The Assessment of Hazard and Risk

allow reciprocal comparison and professional personnel updating on

avalanche risk prevention methodologies actuated in each individual
region from both technical and legislative points of view.
Research in territorial planning promoted by AINEVA and in which
Arpa Piemonte participates actively, allows the introduction of risk pre-
vention principles into preliminary procedures for the authorisation of lift
facilities, ski slopes, or artificial snow equipment. Even if not specif-
ically required by regional regulations, such intervention tends to instil
ski facility management with an increasing awareness for the need to
adopt preventive measures, be they structural and managerial, to avoid
avalanche-related accidents.
This perpetually updated knowledge base allows Arpa Piemonte to
carry out its own institutional task of providing technical-scientific
support to the Region, to the Provinces, and to Municipalities during the
preliminary phases of authorisation procedures for works and interven-
tions related to the tourist uses of the mountain territory, guaranteeing
the application of criteria and methods that are indispensable for the
sustainable development of the Piedmont Alpine areas.
The profound knowledge of the regional mountain territory and its cli-
matic conditions, together with the use of modelling applications and study
methodologies shared on an international level, allows the Regional Fore-
casting and Environmental Monitoring Area to play a role as consultant
in avalanche warning problems that may involve roadway infrastruc-
tures of strategic importance, as demonstrated by the study conducted
on avalanche safety for the Olympic road system.
The adoption of a regional sector regulation that calls for the safe
management of ski districts through the use of avalanche risk prevention
methods and systems, which has already been consolidated in practice
and recognised in sector regulations for some Alpine Arc regions, would
allow a more effective prevention of avalanche accidents.

BIBLIOGRAPHY

The works marked by an asterisk have not been quoted explicitly in the text.
Avondo G.V., Castellino D. & Rosselli D. (2003), Pragelato, il Beth e le sue miniere
ad un secolo dalla grande valanga. Pinerolo, Alzani editore (in Italian).
*Barbolini M. (1996), Sulla modellazione delle avalanches di neve densa.
Technical University of Milano. Thesis (in Italian).
*Barbolini M. & Cappabianca F. (2003), Calcolo della danger e analisi di rischi:
un nuovo metodo basato sull’utilizzo dei modelli di dinamica. Neve e Valanghe,
n. 50, pp. 62–71 (in Italian).
Barbolini M., Natale L., Cordola M. & Tecilla G. (2004a), Linee Guida
metodologiche per la perimetrazione delle aree esposte al pericolo di valanghe.
Neve e Valanghe, n. 53, pp. 6–13 (in Italian).
*Barbolini M., Cappabianca F. & Savi F. (2004b), Risk assessment in avalanche
prone areas. Annals of Glaciology, n. 38.
 Avalanches 229

*Bariffi A. (2002), GIS e modellizzazione del rischio valanga. Neve e Valanghe,

n. 45, pp. 32–41 (in Italian).
Buisson L. & Charlier C. (1993), Studio dei rischi valanghivi sul comprensorio
sciistico Ovest di Macugnaga. Grenoble (F), CEMAGREF. Consultancy report
for the Region of Piemonte (in Italian).
Capello C.F. (1977), Archivio storico-topografico delle valanghe italiane
(Provincia di Cuneo). Torino. Istituto di Geografia Alpina dell’Università di
Torino (in Italian).
Capello C.F. (1980), Archivio storico-topografico delle valanghe italiane
(Provincia di Torino). Torino. Istituto di Geografia Alpina dell’Università di
Torino (in Italian).
Christen M., Bartelt P. & Gruber U. (2002), Aval-1D Numerical calculation of
dense flow and powder snow avalanches. Davos (CH). Swiss Federal Institute
for Snow and Avalanche Research.
Cresta R. (2002), L’esplosivo e la neve. II ed. Milano, Club Alpino Italiano (in
Italian).
Denza F. (1889), Relazione sulle valanghe degli anni 1885 e 1888. Bollettino del
Club Alpino, vol. XXII, p. 181 (in Italian).
*Fontana E. (1991), Storie di antichi inverni. Varallo, ed. Club Alpino Italiano,
Sezione di Varallo (in Italian).
*Ghinoi A. (2004), STARTER – Un metodo per la valutazione della suscettibilità
valanghiva. Neve e Valanghe, n. 51, pp. 12–25 (in Italian).
*McClung, D.M. (2001), Characteristics of terrain, snow supply and forest
cover for avalanche initiation caused by logging. Annals of Glaciology, n. 32,
pp. 223–229.
Serra M. (1997), Le valanghe del 1885–1888. I terribili inverni dei nostri nonni.
Self-published (in Italian).
*Sobilla B., Bartelt P., Barbolini M., Christen M., Gruber U., Maggioni M., Tiefen-
bacher F., Turnbull B. & Amman W. (2001), La dinamica delle valanghe: lo
stato dell’arte e la ricerca. Professione Montagna, n. 64 (in Italian).
Tomasuolo E. (2001), Creazione di un archivio GIS degli eventi valanghivi in
Provincia di Torino nel periodo 1885–1951, Facoltà Scienze Matematiche,
Fisiche, Naturali dell’Università di Torino. Thesis (in Italian).
2.7 River Flooding

2.7.1 INTRODUCTION

The morphological analyses of waterways conducted by the Arpa

Piemonte – Regional Centre for Territorial and Geological Studies are
above all aimed at providing an overview of the evolutionary and cur-
rent dynamics of the main waterways, forecasting future tendencies to the
morphological variations that have already taken place. The activities con-
ducted are essentially morphological surveys (photo-interpretation and
field survey) and studies directed during floods for the production of
thematic maps aimed at the classification of flood probability and the
identification of the areas at risk.
The Arpa Piemonte Forecasting and Environmental Monitoring Area,
through its own Centro Funzionale, supplies a continuous forecasting
service for discharges of the main waterways of the Piedmontese hydro-
graphic network that allows the anticipated forecast of hydraulic risk
conditions, the issuing of forecast bulletins, and the alerting of the civil
protection structures involved. The warning requires a short-term forecast
that indicates when and where critical situations might occur in the most
reliable way possible.
The integration of knowledge and experiences within the two Struc-
tures of Arpa Piemonte is consolidated in interdisciplinary studies
finalised at the classification of the territory in function of the hydraulic
risk, with a view to completing the forecast of the temporal evolution of
the phenomena in real time, even with the indication of the areas involved,
especially those at higher risk, in function of the flooding dynamics and
the effects of the expected damage.

2.7.2 FLUVIAL PROCESSES

The genesis of waterways and their evolution depend on four principal

factors: morphodynamic processes, structural control, climatic variations,
and time. All of these factors operate on different scales, sometimes imper-
ceptible to man, conditioning the form of the landscape and its evolution.
The processes that influence the territory during floods are in any case
closely connected and conditioned by these factors; therefore studies can
increase our knowledge of them and about how they change, which is a
fundamental aspect of forecasting the effects of floods on the anthropic
fabric and implementing the instruments necessary for a correct territorial
management policy.
232 The Assessment of Hazard and Risk

The climatic variations that have involved the earth have had more or
less regular fluctuations with cycles of millenniums, like the succession
of the glacial and interglacial ages during the Quaternary and with shorter
century and decade-long cycles, like the more recent minor ice ages that
occurred between the years 500 and 700 A.D., and between 1500 and 1850
A.D. These latter cycles, even if brief, considerably influenced the life of
man and the morphology of the territory.
Recent climatic studies predict the beginning of a period marked by a
gradual increase in the temperatures, accentuated by the emission of gas
produced by fossil fuels into the atmosphere. Environmental modifica-
tions linked to the average increase of the temperature have been already
verified naturally, with 150–200-year cycles. The current period would
represent the climatic transition between the last Little Ice Age and the
next Greenhouse Effect. The latest hot periods noted are during Roman
times (150–350 A.D.) and Medieval times (1000–1300). Multidisciplinary
studies have allowed us to quantitatively evaluate the modifications of the
physical environment that have occurred at various latitudes over the past
thousands of years: during hot-arid periods, desertification took place
along the coastal areas of the Mediterranean with sudden 80% reductions
in rainfall, while in hilly and mountainous areas there was a reduction
of precipitation of more than 50% and a concentration of rainfall during
limited periods.
It has therefore been hypothesized that upcoming modifications,
according to the climatic-environmental history of previous periods and
other miscellaneous scientific data, together with other variations like the
desertification of coastal regions, a reduction in the pedogenetic processes,
a reduction in the vegetable covering, an increase of the soil erosion, etc.,
as well as the following changes:
• a strong reduction in total precipitation in mountainous and hilly areas
(more than 50% of the current level) but more violent occurrences with
the consequential increase of river discharges;
• a progressive shift of the neutral point towards the foothills area of the
waterways with a diminution of sediment transport;
• an increase in the erosion of sandy-gravelly coastlines;
• an increase in the freezing level with consequential progressive fusion
of permafrost and increase of the rainy precipitations in mountainous
Alpine areas (Ortolani & Pagliuca, 2001).

Geological processes have also influenced and continue to influence

the course of rivers and related processes. Below are three significant
examples of the geological conditioning of Piedmontese rivers.
During the Quaternary, continual geodynamic action was the main
cause of three important fluvial deviations during the Pleistocene that
involved the drainage system of the upper Po River Valley: the diversion
of the Po, that of the Dora Baltea River, and the capture of the Tanaro
River.
 River Flooding 233

The diversion of the Po was the consequence of the geodynamic evo-

lution of the Torino hills associated with erosive-depositional processes
of the glaciofluvial fan of the Rivoli-Avigliana Morainic Amphitheatre
during the lower Pleistocene. This combination of factors led to the north-
ern closure of the Southern Piedmontese Basin near Moncalieri. Since
then, the flow direction turned south of the Torino Hills and the reliefs of
Monferrato (Paleo Po).
The Quaternary evolution subsequently produced increased compres-
sion of theTorino Hills anticline together with an increase in the southwest-
wardly inclination of its axis. The main consequence of these processes
was the subsidence of the north-eastern sector of the Southern Piedmon-
tese Plain, causing a deformation structure now known as the Marginal
Flexure of the Poirino Highlands, with a north-south axis and a tectonic
lowering of the reliefs in the area of Asti. This progressive phenomenon
caused the return of the drainage of the Southern Piedmontese Basin north-
ward, inducing the diversion of the Po River, which began flowing north
of the Torino Hills and the reliefs of Monferrato, even if with a different
conformation (Polino & Sacchi, 1995; Nigrelli, 2005) (Figure 2.7.1).

Figure 2.7.1 The current

course of the Po River (in
blue), north of the Torino
Hills. The old course of the
river can be observed,
parallel to the current one,
incised in an east-west
direction between hilly reliefs.

The Tanaro, descending from the Maritime Alps, ran along the plain
and flowed into the Po River in correspondence with the town of Carig-
nano and, towards the end of the Upper Pleistocene, underwent a sort of
capture near Bra. The capture took place due to the regressive erosion
of the watershed that separated the ancient riverbed from the valley of
one of the torrents that moulded the hills to the east. The river waters
abandoned the old riverbed to flow into the Po in the plain of Alessandria
(Figure 2.7.2).
The third and last modification is that of the Dora Baltea River at the
closure of its valley run. Currently the Dora Baltea passes the City of
Ivrea through a narrow gorge in the rock of the crystalline substratum.
Northeast of the city the little Rio Ribes today runs in an ancient yet
clearly-marked channel of the river. The terminal part of this channel is
 234 The Assessment of Hazard and Risk

Figure 2.7.2 Current

eastward course of the
Tanaro River (in blue). Notice
the ancient course of the river
that deviates westwards,
running along the current
eastern boundaries of the
Cuneo plains.

now occupied by the Chiusella Torrent, which in post-glacial times had

undergone a capture due to regressive erosion caused by tectonic lifting
(Carraro, 1992). It is probable that the ancient glacial stream was divided
into several channels at its entry onto the plain downstream of Ivrea: one
of these channels passed through Ivrea along its current course, while
the others headed towards Fiorano. Probably the combination of tectonic
movements and erosion at the expense of filling deposits of the plain north
of Ivrea caused the formation of the present terraced surface of Fiorano
and the consequential abandonment of the Ribes channel. This particular
morphological situation, due to the presence of a minor watershed repre-
sented by the lower terrace of Fiorano, causes spillover during flooding
with discharges higher than 2,000 m3 /s and the reactivation of the aban-
doned branch. This is what happened during the floods of September 1993
(Bellardone & Forlati, 1996a) and October 2000 (Bellardone & Giampani,
2003), and one other time during the XX century in October 1920. Other
older reactivations were also noted in 1834 and in 1775 (Figure 2.7.3).
As already seen, the magnitude of the morphological changes can
vary greatly over time, but the most evident changes (because they are
on a humanly perceivable temporal scale) are associated with anthropic
activities and occur more quickly than the surrounding environment is
capable of sustaining, causing an imbalance of the entire system.
A stable river is dynamically balanced when, on a basin scale, the pro-
cesses of sedimentation and removal compensate each other. The boundary
conditions include the slope of the valley, the geo-lithological characteris-
tics, the type of soil, the characteristics of the vegetation, as well as climatic
conditions. Modifications in the quantity of solid transport during floods
and in the boundary conditions may disturb the balance of the waterway,
which may respond with evident and even sudden morphological changes,
until the system rebalances itself.
 River Flooding 235

Figure 2.7.3 Current course

of the Dora Baltea River (dark
blue), with flooded areas
during the September 1993
flood (light blue grid, fuchsia
arrows) and the October 2000
flood (light and dark blue grid,
red and fuchsia arrows) that
indicate not only the old
westward along the current
course of the Rio Ribes and
Chiusella Torrent, but also
other fluvial courses between
Ivrea and Pavone Canavese
that were reactivated in 2000.

When the erosive processes exceed the sedimentation, a natural inci-

sion is made in the channel. A reduction in the amount of sediment may
be related to climatic factors, but is also due to the creation of artificial
retention structures, the building of protection works, and the removal of
sediment from the riverbed for anthropic use.
On one hand, such actions favour the planimetric stability of the chan-
nel and the containment of minor or moderate flooding with decade-long
return periods, but on the other, they cause problems for both the environ-
ment and anthropic settlements. One of the major problems caused is the
lowering of the groundwater surface and the consequential lowering of the
quota reached by the capillary rise, which causes the impoverishment of
the soil. Among the problems caused to the anthropic fabric is the contin-
ual damage to bridges, even by flooding that is not particularly exceptional
(return periods of 20–50 years). Furthermore, the river training with the
reduction of areas involved by flows causes a reduction in the probability
of flooding for protected areas on one hand, but on the other causes an
increase in the velocity and the downstream discharge with the possibil-
ity of more widespread flooding. This problem had already been dealt
with by Visentini (1952): “the sight of these vast territories abandoned by
the river ways disgusts those who think of the pressing social needs that
spur the cultivation of new lands, but in spite of what has been said, the
narrowing of the riverbeds with embankments or other protection would
increase the value of the flood peaks in embanked rivers and therefore, at
least under current conditions, must be absolutely excluded”.
 236 The Assessment of Hazard and Risk

Figure 2.7.4 The Chisone

Stream near Pinerolo before
the flood of October 2000:
the waterway presented a
single active channel, with
the exception of a brief
secondary channel which is
now being abandoned.

Figure 2.7.5 The Chisone

Stream near Pinerolo,
following the flood of
October 2000: the flows are
distributed between two
channels on the right bank,
while the channel that was
previously active has been
abandoned.

The braided channel Piedmontese waterways have undergone modifi-

cations due to anthropic factors resulting in the almost total disappearance
of braided models and their replacement by single channel forms. Such
changes, which are also due to variation in hydrological patterns (Maraga,
1989) were already taking place as far back as the 17th century, but
accelerated considerably during the last decades of the past century. This
accelerated metamorphosis, which was induced essentially by anthropic
 River Flooding 237

Figure 2.7.6 The river

channel modelled by the flood
as photographed a few days
later (photo by V. Giraud).

Figure 2.7.7 Orco Stream

during the mid-19th century.
The width of the stream
channel varied from a
minimum of 60 metres at the
bridge of Cuorgnè, at the
mouth in the plains, up to a
maximum of more than 900
metres between the towns of
San Benigno Canavese and
Foglizzo.

factors, may be impeded by the occurrence of exceptional floods and a

return to the original morphologies, as exemplified by the following cases.
In 1858, the Chisone Torrent, in the zone between its mouth in the
plain and the State Highway 559 bridge, consisted of a riverbed with two
active channels that covered approximately 300 metres at its widest point
(Sheet LII, Pinerolo of Gran Carta degli Stati Sardi in Terraferma on a
scale of 1:50,000). Beginning in the first decades of the past century, a
progressive reduction of the width of the riverbed was noted, with the
concentration of the flow in a single channel that, at the beginning of the
1990s, had a width of approximately 70 m in the area near the bridge and
considerably narrower further upstream.
 238 The Assessment of Hazard and Risk

Figure 2.7.8 Channel of the Orco Stream following the 2000 flood (upstream of Rivarolo Canavese).

During the flood of October 2000 (estimated discharge at San Martino

of 980 m3 /s, return period approximately 130 years) (Barbero et al., 2003)
the torrent regained a braided channel in this zone, with a width calculated
at approximately 250 m.
Until the 1950s, the Orco Torrent had a wide riverbed with several
channels. A comparison of the torrent in different periods demonstrates
how the riverbed tended to diminish in width before the 1950s (Maraga,
1989). Its mutation into a single channel continued and was promoted by
the excavation of riverbed material and, above all, by the realisation of
structures aimed at containing the flow in increasingly narrow spaces. At
the end of the 1970s, the flow had been directed into a single channel,
but still with brief stretches of riverbed with two or more branches, in any
case characterised by main channels that were deeper than the lateral ones,
which were progressively being abandoned. This tendency persisted over
time, as can be observed by 1991 aerial photographs where the riverbed
is almost totally characterised by only one active channel (Bellardone &
Forlati, 1996b).
 River Flooding 239

Following the flood of September 24th–25th, 1993, with a peak dis-

charge in Pont Canavese of 1,500 m3 /s and return period of 50 years
(Ravera et al., 1996), the original river channel pattern prevailed with
the occurrence of the characteristic processes associated with flooding in
braided rivers: an increase of the bed section due to bank erosion, newly
formed channels, the reactivation of secondary channels not ordinarily
involved, flows, and bed load transport. Following the October 2000 flood
in Cuorgnè, with an estimated discharge of 1,650 m3 /s and return period
of 80 years (Barbero et al., 2003), the morphology of many stretches of the
bed was once again comparable to that of the middle of the past century,
Moreover, many stretches that were characterised by a single channel for
more than a century, like that near the bridge of Cuorgnè, had regained a
braided morphology.
Significant changes, also associated with the 2000 flood, towards the
original braided bed were observed along the stretch of the Dora Baltea,
downstream from the glacial deposits of the Morainic Amphitheatre of
Ivrea (Bellardone & Giampani, 2003).

2.7.3 THE GEOMORPHOLOGICAL DIAGNOSIS: AN

APPROACH FOR THE STUDY OF FLOOD
PROBABILITY

With the Interreg CatchRisk: Mitigation of Hydro-Geological Risk in

Alpine Catchments, the Arpa Piemonte Regional Centre for Territorial and
Geological Research conducted morphological studies aimed at the iden-
tification and the classification of natural flooding probability for the
purpose of supplying guidelines to be applied in analogous studies in the
valley bottoms and plains (AA.VV., 2005a). The morphological diagnosis
method is oriented towards the identification of areas potentially suscep-
tible to flooding and their cataloguing into classes of relative probability,
also by evaluating the existing and past morphological features of the river
(Maraga & Turitto, 1998).
The geomorphological approach presumes that the flow propagation
can occupy the entire system, and that for this reason it is first of all neces-
sary to identify this category defined as natural flood probability, and then
classify it by priority of submersion and by the energy of the potential
flows. The flows in areas naturally susceptible to flooding are often con-
trolled by the macro-morphology as well as by the micro-morphology as
represented by relic fluvial forms, with a low or null degree of incision.
The intersection of these features with the anthropic infrastructures, which
in turn condition routing, sometimes even in conflict with the former, may
determine an increase in the flood probability and its effects in the zone
of influence of these conditionings and on the infrastructure (especially
240 The Assessment of Hazard and Risk

roadways and artificial channels). Spatially locating such features is

therefore indispensable and can happen only through a careful morpholog-
ical analysis aimed at recognizing the past beds they were associated with.
The project calls for the analysis of two sample areas by selecting
of two stretches of river with very different morphological-environmental
characteristics and a corresponding availability of thematic cartographies.
The first area corresponds to the stretch of the Po River between the
confluences of the Stura in Lanzo and the Orco Torrent (for a stretch of
about 18 km). The second sample area is the Alpine stretch of the Sesia
River between the towns of Alagna Valsesia and Borgosesia (about 50 km).
For the Po River, the morphological reconstruction of the river channels
is founded essentially according to a topographic base (1858, 1923) and
existing thematic cartography (Dutto, 1993), while the reconstruction of
the river channel and the significant morphological elements modelled
from the October 2000 flood and the current terraced surfaces was carried
out through the interpretation of aerial photographs. In order to classify
the flood prone areas, existing maps of the flooded areas from the 1992,
1994, and 2000 floods were also used (Dutto, 1993; Mensio, 1996; Paro,
2003; Susella et al., 2003).
For the second sample area (Sesia River) the recognition of the forms
connected to the river in the 19th century is based on the analysis of topo-
graphic maps (1884–1886). The mapping of the river channel and other
significant morphological features from the second half of the 20th cen-
tury up to the year 2000 was carried out by means of photointerpretation
with optical stereoscope of aerial photographs (4 series) taken between
1954 and 2000, selected also because they were taken after significant
floods (1948–1951, 1976–1977, 1993–1994, 2000).
For the Sesia River, in addition to the planimetric reconstruction of the
channels, the morphometric parameters like the height of streambanks in
2000, the width of the riverbed and active channels, the riverbed length,
the length according to the valley axis, the length of the thalweg line, the
ramification index, the sinuosity index, the total length of bank erosion,
and the entrenchment index were all evaluated for homogeneous stretches.
The altimetric evolution of the water profile was estimated mostly
through indirect planimetric evidence, like the modification of the channel
pattern, the tendency of the waterway to channel, or on the contrary, a
tendency towards a ramified and unstable situation in the flood plain,
and secondly through direct observations on the state of undermining or
filling of channel work foundations (piles, defences), and finally, but
very limitedly, from a comparison of the cross-sections measured during
different times.
The significant parameters, for each homogeneous behaviour of bed
reach, were finally summarised in the descriptive notes and the tables.
The fluvial morphological features were then used for the identi-
fication of areas susceptible to flooding, flood prone areas seeking
to assign each individual area a degree of propensity to reactivation
 River Flooding 241

according to the height of the streambanks and the current fluvial struc-
ture, also in consideration of the typologies of associated vegetation
(sown, arboreal, shrubby). The areas susceptible to flooding were therefore
classified as:
1) area at very high possibility of flooding (Class 1). It corresponds to
the area containing the current bank full channel (2000), widened to
include the adjacent areas for the Po River, within the belt of recent his-
toric mobility (from 1923). For the Sesia River, this area was extended
to the belt of historic mobility (including the forms of the riverbed that
have been active since 1884–1886 in the presence of banks between 0
and 5 metres in height;
2) area susceptible to high flood possibility (Class II). For the Sesia River,
this corresponds mostly to first and second order terraces associated
with banks less than 5 m high. For the Po River, it corresponds to
the envelope of the areas involved in the 1994 and 2000 flooding and
to areas containing bed forms recognised on the oldest cartographies
(1858), in absence of embankments or reliefs in general;
3) area susceptible to average-low flooding (Class III). Corresponds to
areas usually outside of the bands of historic mobility, bordered by
older terraces located at quotas higher than the waterway, or areas
opposite stretches with banks of such a height as to be considered
at risk of limited flooding in case of particularly intense and violent
floods;
4) areas potentially susceptible to flooding only for catastrophic events
with a very low natural propensity of flooding (Class IV). It corres-
ponds to the areas of clear fluvial origins (envelope of the incised
or non-incised ancient fluvial forms), that for distance, or for average
quota in comparison with the waterway, presents a very limited genetic
propensity.
For the classification of natural propensity of flooding, the most dan-
gerous morphological situation for each individual area was taken into
account: for example, areas classifiable for their general characteristics,
like mid-low flood probability were placed in class II (high) due to the pres-
ence of critical points due to low banks upstream or for the branching off
of diversion channels, potential overflow flooding zones, or preferential
flow lines.

2.7.4 RESULTS OF THE APPLICATION OF THE

GEOMORPHOLOGICAL DIAGNOSIS

Here the results of the morphological analyses applied in the two different
contexts of the Po River in the Torino plain and the Alpine stretch of Sesia
River will be summarised.
 242 The Assessment of Hazard and Risk

Figure 2.7.9 The stretch of

the Po River analysed.

TABLE 2.7.1 MORPHOMETRIC FEATURES OF

ANALYSED STRETCH OF PO RIVER
Features Value
Max height 209.0 m asl
Min height 171.0 m asl
Low water channel length 21.0 km
Flood channel length 18.4 km
Channel slope 0.18%

In the stretch analysed (from Torino to Chivasso), the Po River can be

classified as a single channel, sinuous, meandering, type C as described
by Rosgen (1994), with an average sinuosity index of Is = 1.14.
Analysing the maps of the morphological aspects of the Po River dur-
ing different periods, it can be observed how the maximum planimetric
extension of the area occupied by the active channels was recorded at the
beginning of the 20th century and how, until the early 1990’s, was sub-
ject to a progressive reduction, which was successively interrupted by the
floods of November 1994 with a 20-year return period (Regione Piemonte,
1998), and of October 2000 with a return period of 80 years (Barbero
et al., 2003), reflecting the analogous behaviour of other Piedmontese
waterways.
The classification of flooded areas was based not only on the analysis
of their forms, but also on flood maps of the 1994 and 2000 events, which
 River Flooding 243

TABLE 2.7.2 MAXIMUM PLANIMETRIC EXTENSION

OF PO RIVER IN XX CENTURY
Period Channel Area [km2] Reduction [%]
1923 6.8 –
1958–1988 5.2 24
1992 3.9 43
1994 4.9 28
2000 5.8 15

TABLE 2.7.3 FLOOD PLAIN OF

PO RIVER STRETCH
CLASS FLOOD PLAIN [km2]

Class I 9.74
Class II 11.92
Class III 9.19
Class IV 9.93
Total 40.78

were comparable to reference floods of Warning Code 2 (less than or equal

to 20 years for the 1994 flood) and for Warning Code 3 (more than 20
years for the 2000 flood) of the Warning Bulletin for hydrogeological risk
of the Arpa Piemonte Natural Risk Situations Hall (Sala Situazioni Rischi
Naturali). All the areas flooded during the above-mentioned events, in
absence of morphological conditioning aimed at containing the flooding
(like terraced embankments), were included in the Class II for flood prob-
ability. The exceptions were those zones where flooding, caused in 1994
by the breakage of the flood embankments, no longer occurred in 2000 due
to the proper functioning of the reconstructed embankments; these areas
were assigned to a lower flood probability class (Class III), Table 2.7.3
includes the extensions of the flooded areas according to flood probability
class. The second sample area chosen for the floodability analyses is the
Alpine stretch of the Sesia River.
According to studies conducted, the River was subdivided into 12
reaches of homogeneous behaviour. Analyses of the dimensions of the bed
between 1884 and 1978 indicated that the braided stretches were subjected
to reductions in the average and maximum values nearly everywhere.
Between 1978 and 1994, both the average and maximum widths of the
bed increased and then, in 2000, maintained an extension comparable to
that of 1994.
Instead, by analysing the areas occupied by the active channels, it can
be observed that the maximum extension corresponding to 1884 under-
went a 20% reduction in 1978 and an increase in 1994 to once again
meet the 1884 measurement. In 2000, a minimal reduction over 1994 was
 244 The Assessment of Hazard and Risk

Figure 2.7.10 Second

sample area: Alpine stretch of
the Sesia River between the
towns of Alagna Valsesia and
Borgosesia.

Figure 2.7.11 Channel of the

Sesia in 1884 (IGM base
1:25,000). At that time all the
valley bottom was occupied
by the Sesia. For the legend,
please see the Figure 2.7.14.

recorded. The 1954 bed was not considered because the entire stretch
being studied was never mapped.
According to the analyses carried out, the Sesia River demonstrates a
limited streambank erosion along most of its single channel development
 River Flooding 245

Figure 2.7.12 Channel of the

Sesia in 1978 (CTR base
1:10,000). Note how the
flows are concentrated in a
single channel. The legend
for the bed forms is the same
as for Figure 2.7.14.

(where it is often incised into rock), while the wider stretches in the valley
bottom, where it acquires more channels, the lateral activity streambank
erosion is significant in keeping with the classic tendencies of braided
reaches.
Finally, from the value of the area of the riverbeds in 1994 and 2000,
it can be observed how also the Sesia River, like most of the braided
Piedmontese waterways, together with floods with not particularly signifi-
cant return periods, like those of 1993, 1994, and 2000, totally or almost
totally reacquired the width it had at the end of the 19th century, with a
reaffirmation of its multi-channel nature.
Arriving at the analysis of natural flood probability, of the 17 km2 of
classified valley bottom, 30% is in Class I (elevated propensity), 20% in
Class II (high propensity), 15% in Class III (mid-low), 11% in Class IV
(very low) and only 16% of the valley bottom would not be susceptible to
flooding.
Associating areas belonging to different classes of flood probability
with the information obtained from the morphological analyses resulted
in a further classification according to the expected impact, placing par-
ticular attention on the identification of areas that could be involved in
concentrated flows due to the presence of more or less remodelled chan-
nels associated with one or more active stream channels during the period
considered. The flood probability classes I, II, and III were therefore fur-
ther subdivided into categories A and B in function of the presence of such
elements.
246 The Assessment of Hazard and Risk

TABLE 2.7.4 EXAMPLE OF REPORT ABOUT MAIN MORPHOLOGIC

FEATURES OF RIVER STRETCH

PLANIMETRIC EVOLUTION OF THE RIVER (1885–2000)

PAST NOW
1954 – braided Braided irregular
CHANNEL TYPOLOGY 1978 – braided Locally braided
1994 – braided
1954 – 49.1
MEAN WIDTH [m] 1978 – 46.5 65.3
1994 – 66.1
1954 – 2
BRAIDING INDEX 1978 – 2 2
1994 – 2
1994 – left bank – 567.6 left bank – 310.64
(15.10%) (16.50%)
EROSIONE DI SPONDA [m]
1994 – right bank – 772.8 right bank – 1028.1
(20.28%) (28.54%)

ALTIMETRIC EVOLUTION OF CHANNEL BOTTOM

Present state
Evidences of deepening
REACTIVATION OF OXBOWS
Late abandoned meander 1 on left bank and 1 on the right one
Old abandoned meander –
GRAIN SIZE DISTRIBUTION
Boulders and cobbles
VEGETATION ALONG RIVER SPACE
Trees and shrub
Man-made measures

2.7.5 INTERDISCIPLINARY ANALYSES FOR THE PURPOSE

OF CLASSIFYING FLOOD RISK AREAS

The morphological approach to the flood probability analysis allows the

identification of areas of a fluvial nature (active channels, and flood-
prone areas), but allow only their classification according to their relative
probability of being flooded, while the morphological approach does not
allow the correlation of areas to probable flood events. Therefore, for the
recognition of areas at risk of flooding, it is necessary to associate the
above-mentioned morphological analyses with other studies for the pur-
pose of classifying the flooding of each individual waterway in terms of
discharge, frequency, and impact on the territory. Only through the inte-
gration of geomorphological analysis with hydraulic studies and with the
 River Flooding 247

Figure 2.7.13 Channel of

the Sesia River in 1994.
Observe the reactivation of
the channel making up the
bend in the valley (flow from
left to right). The legend for
the bed forms is the same
as for Figure 2.7.14.

Figure 2.7.14 The Sesia

River in 2000. The form of
the stream channel can be
compared to that of 1994,
but observe the lateral
channel upstream that has
been abandoned and the
formation of a small lateral
channel in the middle area
of the stretch.

recognition of the effects and damage caused by past floods, it is possible

to establish the identification of the boundaries of floodable areas with
known discharges (as reference for warning interventions, for example)
and the recognition of the other elements at risk. This information, asso-
ciated with a meteorological forecasting system and a real time flood
advisory system, allow the definition of scenarios thanks to which it is
248 The Assessment of Hazard and Risk

possible to actuate protection interventions as well as controls on works

and infrastructures in case the set warning thresholds should be surpassed.
Arpa Piemonte has experimented different methodologies for studying
Piedmontese waterways, allowing the development of a procedure used
in studies aimed at recognising risk areas, founded on an interdisciplinary
approach that combines historical information with geomorphological and
hydraulic analyses.
The next chapters will briefly summarise the main stages of these
analyses.

2.7.5.1 Analyses oriented at the morphological characterisation

of the water system and the elements at hydraulic risk
As already affirmed, the effects related to the passage of floodwaters
are the consequence of long-lasting physical and climatic factors and
anthropic conditioning that may aggravate or, on the contrary, reduce the
damages it causes; therefore, information on the impact of past floods,
if analysed with a view to the current state of the territory, in terms of
land occupation and the presence of defence/retention works, allow the
prediction of what may be the effects and damages caused by future floods.
In general, the study of past floods calls for a photointerpretive analysis
to recognise the morphological elements modelled by the floodwaters and
of the areas flooded. The directed analyses must be integrated with infor-
mation from historical or technical-scientific documentation (available in
various town, provincial, and regional archives, etc.) needed mostly for
the identification of the entities involved and for the identification, even if
only qualitative, of the damage (built-up zones, roadway infrastructures,
defence works, etc.). The GIS allows, through the informatization and
layering of data, a spatial and multi-temporal overview of the water system,
to visualise the historical evolution, and to forecast future developments
through suitable and targeted syntheses of the various informative layers.
The use of these techniques also allows researchers to spatially highlight
all those elements that have denoted signs of vulnerability to flooding over
time, frame them in their environmental context by individual episodes,
and classify them into classes of risk.
The analyses of the altimetry modifications of the river-bed complete
the data base by contributing to the construction of the picture of the cur-
rent and past morphological features of the waterway and its evolutionary
tendencies, according to bed reaches with homogeneous behaviour.
The morphological information on historic flooding of a waterway,
set into the current situation and integrated with land use information,
will lead to the forecasting of expected effects on the anthropic fabric,
with regards to risk scenarios classified and described according to grow-
ing levels of danger and specifically associated with suitable reference
discharges.
 River Flooding 249

Figure 2.7.15 Classes of

flood possibility divided into
categories A and B according
to the presence (A) or
absence (B) of associated
channels with at least one of
the stream channels
recognised during the period
analyses (1882/86–2000).

2.7.5.2 Hydraulic analyses

The reference discharges that are used to describe the scenarios are gen-
erally obtained from statistical elaborations of the historic series of the
waterways being studied. When discharge data is not available for a suffi-
cient period of time, the reference discharges may be gathered from basin
scale regionalization studies like those proposed by many Basin Author-
ities and by Operational Units of the National Group of Hydrogeological
Catastrophes of the Italian National Research Council.
For each bed reach with homogeneous behaviour, at least reference
discharges associated with moderately or elevated risk scenarios are iden-
tified; hydraulic verifications aimed at identifying the maximum flood
level for these discharges are conducted. These verifications are used to
examine the hydraulic freeboard of the longitudinal works (embankments,
defences, etc.) and transversal works (bridges, viaducts, etc.); these dis-
charges are to be used in the analyses for the defining the boundaries of
flood areas.
The modelling approach applied depends on the complexity of the
phenomenon to be represented. In some cases a continuous hydraulic
verification with one-dimension flow is sufficient, while in more com-
plex situations it may be opportune to use 2-D variable flow models
that take into account transversal components of the flow and mitigation
phenomena.
Particular attention should be paid to points that correspond to critical
stretches susceptible to overflowing, especially:
• in correspondence to transversal structures;
• in correspondence to diversion structures;
• in correspondence to embanked stretches know to be at risk of
overflowing;
• in correspondence to structures that impede water flow;
• upstream of zones that, according to the morphological analyses
described above, are indicated as susceptible to flooding with a high
relative probability.
It is always advisable to verify the degree of reliability of the model
applied whenever significant divergences should be found. The veri-
fication consists of a comparison of the modelled simulations with
250 The Assessment of Hazard and Risk

morphological effects and damages caused by past floods, for which it

is possible to reconstruct the flood areas and review the schematization
adopted.

2.7.5.3 Risk scenarios

The mapping of flood risk is a useful tool for warning activities if asso-
ciated with an adequate flood forecasting, activities that are the responsi-
bility of national and regional warning systems for hydrogeological and
hydraulic risk for civil protection purposes.
Operational guidelines for the organizational and functional manage-
ment of the system require the defining of an articulated scale on at least
two levels of criticality – moderate and high – for each type of risk, as
well as an ordinary situation base level in which possible criticalities
are believed to be generally accepted by the population. Excluding the
base level, the other levels can be classified according to the following
definitions:
• Level 2: moderate criticality, which corresponds to the surpassing of
ordinary flood levels defined as the discharge that transits and entirely
occupies the waterway bed with levels approaching those of ground or
dike levels. The erosion of the most vulnerable banks begins to become
significant and areas adjacent to the waterway begin to flood.
• Level 3: elevated criticality, which corresponds to the surpassing of
the extraordinary flood level defined as the discharge that cannot be
contained in the bed, therefore causing serious flood conditions.
Once the levels of criticality and the corresponding reference floods
have been defined, the corresponding risk scenarios are designated for
each bed reach with homogeneous behaviour. For each warning level
adopted, opportune maps with adequate topographic scales (at least
1:10,000) are drawn to indicate the flood fields forecasted according to
the morphological analyses and hydraulic models, as well as flood-related
works. It is of fundamental importance that the risk scenarios highlight
those structures that may influence the flow distribution, especially if they
determine preferential high hydraulic heads.

2.7.6 RESULTS OF THE INTERDISCIPLINARY APPROACH

TO THE FLOOD PROBABILITY ANALYSIS: THE
MAPPING OF THE HYDRAULIC RISK OF THE
SESIA RIVER

The methodology relative to the mapping of flood risk areas, briefly

described above, was developed and applied to a stretch of the Sesia River
(from Borgosesia to the confluence with the Cervo Torrent), as part of the
Floodaware Research Project – Flood Prevention and Forecasting and the
 River Flooding 251

Figure 2.7.16 The Sesia

River, Alpine tributary of the
Po, begins at Monte Rosa
and has a drainage basin
of 3,100 km2 . The main
contributions are taken from
the Sessera Torrent,
downstream from
Borgosesia, and Cervo
Torrent (drainage basin of
1,088 km2 ), just upstream
from Vercelli. Many of the
waterways from the Vercelli
plains converge in the Sesia.

European Union research programs activated following the extreme cli-

matic events that struck Europe during the early 1990s, dictating the need
to mitigate and manage flood events, given their high social-economic
impact (Figure 2.7.16).
The objective of the project was to supply emergency management bod-
ies with a decision-making support tool that was easily read, clear, and that,
in relation to the danger of forecasted floods, allowed the identification of
risk areas and the actuation of suitable civil protection measures. Within
the Floodaware Project, the scenarios were subdivided into two classes
that corresponded to the criticality levels described above. The reference
floods associated to the two levels were floods with return periods of 20
and 200 years respectively.
According to the analysis of the main flood events for the stretch in
question over the past century (Table 2.7.5), the floods of September 1948
and November 1968 were analysed. These events, according to the doc-
umentation analysed, were characterised by return periods much longer
than 20 years in the stretch in question (Autorità di Bacino del Fiume Po,
1997). Of the two of these, the 1968 flood was documented better and
was therefore considered as the reference flood for the elevated criticality
scenario.
252 The Assessment of Hazard and Risk

TABLE 2.7.5 HISTORICAL FLOODS AND RELATIVE WATER LEVELS OF SESIA RIVER

Hydrometric station August 1934 September 1948 November 1968 September 1993 November 1994
Q [m3/s] h [m] Q [m3/s] h [m] Q [m3/s] h [m] Q [m3/s] h [m] Q [m3/s] h [m]
Borgosesia 2,990 7.5 3,070 7.9 2,150 6.8 2,400 5.05 1,800 3.37
Vercelli 2,970 6.4 – – 3,900 6.9 3,400* 5.68* 3,200* 5.65*

* Hydrometric station at Palestro (downstream to Vercelli)

TABLE 2.7.6 DISCHARGE WITH GIVEN RETURN PERIOD FOR SESIA RIVER
(AUTORITÀ DI BACINO DEL FIUME PO, 1997)

Hydrometric station TR = 20 years TR = 200 years

Q [m3/s] q [m3/s km2] Q [m3/s] q [m3/s km2]
Borgosesia 2,730 3.93 4,220 6.07
Vercelli 3,370 1.48 4,900 2.15

With regards to the 1968 flood at the station of Vercelli “whose

hydrometrograph – (which was) seriously damaged – stopped working
before the peak, the flood diagram was partially reconstructed according
to direct readings. During the peak, which occurred at 10 PM on Novem-
ber 2nd, the fluvial level reached a height of 6.90 m. […]. The height of the
peak would certainly have been greater if, just before and in conjunction
with the same (N.d.R. time?), there had not been overflowing in the tribu-
taries and in the Sesia itself” (Ministero dei Lavori Pubblici, 1969). Other
sources indicate very different discharge values (Carraro et al., 1970): “the
flood wave was evaluated by technicians of the Consorzio Ovest Sesia at
about 6,000 m3 /s (other sources indicated lower values)”.
The September 1993 flood, with an estimated return period of about
20 years, was used as a reference for moderate criticality scenarios.
The definition of the past scenarios was therefore based on the study
of the processes and the effects related to historic floods as reconstructed
through the interpretation of aerial photographs of the major flood events
(1948 photographs for 1954 flood; photographs made in 1968, 1977, 1993
after each event). The photointerpretive analysis led to the recognition of
the flooded areas and the induced effects of floods (damage to works,
erosion, overflowing, and dike failure). Punctual information from the
Arpa Piemonte Geological Information System (SiGeo) is added to this
areal data. The information contained in the SiGeo, wherever it was not
sufficiently complete, was integrated with updates from various State
archives, town archives, and on-site interviews.
Even for the Sesia River, as for other Piedmontese waterways, the
flood analyses, especially that of 1968, demonstrated how the most signifi-
cant floods reactivate the relict fluvial morphology and how this strongly
influences the distribution of the flows. The level of energy reached by
the current in the marginal strip of the active channel are greater in cor-
respondence to the relict fluvial forms that, in branching off from the
 River Flooding 253

active channel, develop for considerable distances and become preferen-

tial courses for high energy flows. In this way, the significant effects of
the flood impact areas that are even very distant from the waterway and
sometimes occupied by settlements and infrastructures.
The scenarios reconstructed through the processes and the effects of
flooding mentioned were then applied in the current daily environmen-
tal context to verify their validity in keeping with the current territorial
landscape and existing defence works.
The current landscape was defined through the steps below:
• the plano-altimetric evolution of the bed was analysed and the points
with the most significant variations were identified;
• the territory studied was subjected to an in-depth morphological analy-
sis for the purpose of distinguishing the most important forms for the
flood probability classification and the reconstruction of risk scenarios;
• the effectiveness and the efficiency of hydraulic defence interventions
were verified;
• the photointerpretive analysis and the historical cartographies allowed
the reconstruction of the modifications of the active bed of the Sesia
River over time and the recognition of the belt of historic mobility.

The comparison of the bed limits of the Sesia River of 1882–1884 and
that of 1994 indicated a general reduction in its cross-section over the
past century. In particular, the study conducted on the modifications of
the Sesia River from 1954 until 1994 indicated a general tendency of the
active channel to narrow, characterised by the triggering of channel scour
with the possibility of serious consequences on the stability of works like
bridges or other types of transversal structures.
In order to evaluate the plano-altimetric variations of the Sesia River
more precisely, the cross-sections measured during different years and
the corresponding hydrometric levels for reference floods were compared.
The cross-sections obtained from documentation available (property of the
Po Magistrate) cover up to 1992, while six sections were specially surveyed
in 1997 in order to highlight the subsequent altimetric modification.
The comparison of the cross-sections in the stretch between Romag-
nano and the Torino-Milano motorway called attention to a general
increase in the discharge cross-section of the active channel as a con-
sequence of the removal of large quantities of bed materials between 1971
and 1992.
This tendency was confirmed by the comparison of the hydrometric
levels for reference floods. For a flood with a return period of 20 years, the
level reached in the cross-section in 1992 was generally less than 0.5–1 m
compared to that of the section in 1971 (a significant example is found in
Figure 2.7.17).
For the hydraulic analyses, the hydrometric level was calculated by
means of a hydraulic model that simulates the hydrodynamic routing of
the flood with a one dimensional schematization with non-stationary flow,
 254 The Assessment of Hazard and Risk

178

177

176

175 h 1971

174
Height (m als)
h 1992
173

172

171
Figure 2.7.17 Comparative
170
cross-section of the Sesia section n. 8 in 1971
River 1971–1992, upstream 169
section n. 8 in 1992
from the Torino-Milano
motorway and the 168
hydrometric levels for a flood 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500
with a 20-year return period. m

or MIKE11© of the Danish Hydraulic Institute (DHI, 2001–2004). In

the same stretch, the comparison of the cross-sections of 1992 and 1997
demonstrate an inverse tendency toward sedimentation (Figure 2.7.18)
that matches with the results of the morphological analyses of the Alpine
stretch of the Sesia River, conducted as part of the CatchRisk Project
described above. It can therefore be hypothesized that if a phenomenon
with a discharge similar to that of the 1968 event should occur, the water-
way would reacquire the width occupied by the river channel in that date
and the flooding could involve the same areas but would cause even more
serious damage because in the meantime the anthropic occupation of the
area in proximity to the waterway has increased.

216

214

212

210
Height (m asl)

208

206

204 section n. 13 in 1971

section n. 86 in 1992
section n. 02 in 1997
202
Figure 2.7.18 Comparative
cross-sections of the Sesia 200
River 1971–1992–1997 600 650 700 750 800 850 900 950 1000 1050 1100
upstream of Carpignano. m
 River Flooding 255

From Borgosesia to Romagnano and from the Torino-Milano motor-

way to the tributary of the Cervo Torrent, downstream of the head reach
described, it is not possible to make any considerations on the altimetric
variation of the channel bottom due to a lack of comparison cross-sections.
A second focal point for the definition of hydraulic risk scenarios was the
recognition of stretches of levees with a high risk of failure or collapse. The
analysis was carried out according to historical information on levee fail-
ure and the current state of adequacy and maintenance of the dikes. The
same criterion was applied to identify the defence works that are most
vulnerable to erosion. Beginning with topographic surveys of the bed
cross-sections (1992 survey of the Po River Authority) integrated with
cross sections of bridges, a one-dimensional non-stationary flow model,
as mentioned above, was used to reconstruct the flood discharge profiles
with return periods of 20 and 200 years.
In this way the stretches at risk and the degree of bridge safety were
verified. The stretches of dikes at risk of overflowing were identified
by comparing the height of the embankment with the hydraulic profile to
floods being used as reference. The degree of bridge safety, referred to the
dynamic action of the current on the floor system, was evaluated through
the comparison of the flood profile with the level of the bridge soffit.
The static safety of the bridge, in reference to the undermining of the
foundations, was classified according to the efficiency of the protection
works and the entity of the ongoing erosion. The overall lowering of the
channel bottom is particularly evident in proximity to the bridges; often
the footings present signs of undermining that, during flood conditions,
may compromise the static safety of the work itself. When analysing flood
effects, some important considerations must be made on the role played
by the presence of the levees along the Sesia River. In 1968, from Romag-
nano to Vercelli, there was a continuous right/left embankment system
that was proven to be inadequate during the flood of November of the
same year. The waters, which overflowed and broke down the embank-
ments in several points, were distributed over the outside areas that had
been considered protected; the embankments also aggravated the effects
of the flooding by impeding the return of the water to the riverbed. Due
to the fact that this embankment system has not improved in its func-
tionality over the next thirty years, the problem may recur in analogous
situations.
The integration of the information resulting from the analyses
described above led to the definition of risk scenarios associated with
warning levels.
For Level 2 (moderate criticality), the flood-prone area under ordin-
ary conditions was made to correspond to the bank-full channel of the
1993 flood (return period of about 20 years) and extended for as far
as the embankments continued; in this scenario the flood flow occurred
mainly inside the incised channel with the possibility of reactivating sec-
ondary channels near the waterway and flooding of lower adjacent areas.
 256 The Assessment of Hazard and Risk

Figure 2.7.19 Example of flood risk map.

 River Flooding 257

Figure 2.7.20 Table of scenarios associated with the risk map.

TABLE 2.7.7 BUILT-UP ZONES FOR THE TOWNS LOCATED IN AREAS NATURALLY PRONE TO FLOOD
OR FLOODED IN 1994 AND 2000 ALONG THE STRETCH OF THE RIVER PO STUDIED

Municipality Built-up zones subdivided in flood classes [m2] Flooded area [m2]

Name Buildings I II III IV Sum I-IV 1994 2000

[km2]
Brandizzo 1.29 0 17,565 0 177,059 194,624 17,565
Castiglione T.se 2.04 115 33,924 480,394 267,792 782,225 33,900 17,183

Chivasso 4.69 6,123 459,884 0 0 466,007 331,746 378,710

Gassino T.se 1.40 0 0 25,962 25,962 26,216
San Mauro T.se 3.91 47,143 720,008 673,434 414,022 1,854,607 85,666 788,112
San Raffaele Cimena 0.87 0 41 0 374,243 374,284 41
Settimo T.se 7.10 0 0 360,815 1,039,244 1,400,059

Torino 53.20 25,686 579,622 1,901,176 494 2,506,978 91,310 596,776

The effects on structures and infrastructures overlooking the bed were

forecasted only locally.
For Level 3 (elevated criticality), the area susceptible to extraordinary
flooding was obtained by the envelope of the flooded areas during the
flood that took place on November 2nd–4th, 1968 (return period between
20 and 200 years) with the 1882 active channels; in fact, the comparison
of the areas flooded in 1968 with the areas occupied by active chan-
nels of the Sesia in 1882 resulted in how the former were almost always
258 The Assessment of Hazard and Risk

contained within the second and how the now relict elements, associated
with the 1882 bed, were strongly influenced by the flow distribution. The
boundaries of the areas susceptible to flood were further extended where
indications of frequent damages had emerged from the historical analy-
sis of flood events. The resulting belt is very wide in some stretches in
that it takes into account the reactivation of relict morphologies or dis-
charge routing through artificial diversion channels. The recognition of
the 1882 riverbed brought to light how many pertinent areas were gradu-
ally occupied by civil and industrial settlements. Some of these areas were
flooded during the 1968 flood with the involvement of many structures,
while others that were subsequently settled and built-up, can currently be
considered at risk.
All of the information useful during the warning and monitoring phases
of the events, obtained from the hydraulic, morphological, and historical
analyses, were synthesized in the risk scenarios, classified according to
the warning level of the forecasted flood, and represented on a map with a
scale of 1:10,000 (Figure 2.7.19), with tables detailing the expected effects
and damage (Figure 2.7.20). The risk scenarios also include works that
must be kept under control and for which the effects of flooding for the
two warning levels have been summarily described.

2.7.7 CONCLUSIONS AND DEVELOPMENTS

The actions of Arpa Piemonte for forecasting floods and their potential
effects is aimed, on one side, at refining meteorological, hydrological, and
hydraulic forecasting and, on the other, at the realisation of a solid, at the
realisation geomorphological data base concerning waterways and their
fluvial environment, not only for civil protection actions but also with a
view to ordinary territorial management. The morphological analyses of
waterways conducted by theArpa Piemonte Regional Centre forTerritorial
and Geological Research had the objective of providing an overview of the
ongoing dynamics and forecasting of future tendencies in keeping with
the morphological variations that had already occurred. The results of the
studies were consolidated in geomorphological maps and in synthetic risk
maps and flood risk maps, to be used as support for decision-making with
regards to territorial planning and protection.
An example of application of the classification of areas genetically
susceptible to flooding in the Arpa Piemonte forecasting activities is to
identify the urbanized zones that may be potentially involved in floods and
classified according to their probability of inundation. This information
is important in warning activities because it allows an overview, even if
only indicative, of the distribution of subjects at risk; it is then possible
to focus on those municipalities where inhabited zones may be involved,
especially if located in flood probability classes I and II.
 River Flooding 259

The classification of the flood risk areas in function of the expected

effects also provides a priori knowledge of the areas that may be involved
in the passage of high energy currents capable of seriously damaging prop-
erty and making the eventual evacuation of persons present difficult. In
fact, transversal works in zones where there are known forms of aban-
doned fluvial forms that may be reactivated, like ancient channels, may
be struck and destroyed, causing serious danger for back areas in the case
of dikes and with serious risk of isolation for communities if roadways
should be involved. These classifications are also based on the recognition
of abandoned forms that are currently slightly incised or not at all incised,
and therefore recognisable only through the specialised analyses of aerial
photography or older maps.
For this reason, the morphological approach to flood probability also
allows the contextualization of hydraulic studies because it provides a
morphological view of the territory, highlighting even the minor features
that would be irrelevant from a topographic point of view, but that strongly
condition flow dynamics. In highly urbanized areas, the morphological
approach instead finds limited application in defining the boundaries of
flood risk areas. In fact, in the case in which the flood waters involve
urbanized areas, the extension of the inundated areas, the levels of energy
reach by the current, and the preferential flow paths are influenced more by
the distribution and density of structural works like buildings, roads, and
bridges than by the morphology of the land. From this point it becomes
necessary to implement other types of analyses like historical analyses
(Giampani & Loi, 2000) and hydraulic modelling. All the territorial analy-
ses described here were carried out according to the principles of the
Geographic Information Systems that allows the realisation of data bases
that can be shared and updated.
However, in order to better exploit the potential of the Geographic
Information Systems, homogeneous and semantically correct and stand-
ardised information must be available and based on common languages
and shared procedures to be applied immediately during the creation of the
data, or at the time of survey (whether field survey or photointerpretation).
For this reason, Arpa Piemonte has organised a series of instruments and
procedures dedicated to researchers so that they can develop numerical
cartographies according to an GIS perspective as soon as the data produc-
tion phase. The result of these efforts are IT tools and instruction manuals
aimed at integrating the cartographies of the individual experts involved
in fluvial analyses into a unicum of homogeneous and comprehensible
data to be shared by all.
The Floodaware Project represents a methodological example for
mapping flood risk areas where the hydraulic, morphological, and histor-
ical analyses of the water system are synthesized in an interdisciplinary
approach for the definition of real risk scenarios imposed onto the territory,
which indicate true actions to be taken during the warning and emergency
phases, according to the level of criticality established by the regional alert
260 The Assessment of Hazard and Risk

system. This work can be viewed as a basic risk study upon which ter-
ritorially competent administrations may base their civil protection plans
and define safeguards to be activated as a consequence of forecast and
surveillance evaluations of the regional Centro Funzionale assigned to
this task.

BIBLIOGRAPHY

The works marked by an asterisk have not been quoted explicitly in the text.
*APAT, Sistema Nazionale di Osservazione ed Informazione Ambientale (2005),
Prima definizione di modelli revisionali per qualificare e quantificare gli effetti
dei cambiamenti climatici sugli ecosistemi. Rapporto 03–19 (in Italian).
*ART,Ambiente Risorse eTerritorio (2004), Fiume Sesia, Relazione metodologica
e di sintesi. Programma Interreg IIIB Spazio alpino, Progetto CatchRisk: Mit-
igation of hydro-geological risk in alpine catchments – Definizione delle aree
geneticamente inondabili e della loro pericolosità relativa in ambiente fluviale,
pp. 158, Rapporto interno (in Italian).
*ART, Ambiente Risorse e Territorio (2004), Fiume Po, Relazione metodologica
e di sintesi. Programma Interreg IIIB Spazio alpino, Progetto Catch Risk: Miti-
gation of hydro-geological risk in alpine catchments – Definizione delle aree
geneticamente inondabili e della loro pericolosità relativa in ambiente fluviale,
pp. 11, Rapporto interno (in Italian).
*ART, Ambiente Risorse e Territorio (2004), Linee Guida. Programma Interreg
IIIB Spazio alpino, Progetto CatchRisk: Mitigation of hydro-geological risk in
alpine catchments, Definizione delle aree geneticamente inondabili e della loro
pericolosità relativa in ambiente fluviale, pp. 36, Rapporto interno (in Italian).
Autorità di Bacino del Fiume Po (1997), Piano Stralcio per la difesa idrogeologica
e della rete idrografica nel bacino del Po (in Italian).
AA.VV. (2005a), Guidelines. Programma Interreg IIIB Spazio alpino, Progetto
Catchrisk: Mitigation of hydro-geological risk in alpine catchments, Arpa
Piemonte.
*AA.VV. (2005b), Final Report. Programma Interreg IIIB Spazio alpino, Progetto
Catchrisk: Mitigation of hydro-geological risk in alpine catchments, Regione
Lombardia.
*Barbero S. & Giampani C. (1998), Mappatura delle aree a rischio di inondazione
finalizzata alla gestione dell’emergenza. Atti XXVI Convegno Idraulico e
Costruzioni Idrauliche, Catania, 9–12 settembre 1998, vol. III, pp. 3–14, Ed.
CUECM (in Italian).
Barbero S., Graziadei M. & Rabuffetti D. (2003), Analisi Idrometrica, in Regione
Piemonte, Eventi Alluvionali in Piemonte – Evento del 13–16 Ottobre 2000,
vol. I, pp. 45–57, ARES, Arti Grafiche Torino (in Italian).
Bellardone G. & Forlati F. (1996a), Torrente Orco tratto di pianura, in Regione
Piemonte, Gli eventi alluvionali del settembre – ottobre 1993 in Piemonte, pp.
71–80, L’artistica di Savigliano, Torino (in Italian).
Bellardone G. & Forlati F. (1996b), Bacino della Dora Baltea, in Regione
Piemonte, Gli eventi alluvionali del settembre – ottobre 1993 in Piemonte,
pp. 54–62, L’artistica di Savigliano, Torino (in Italian).
 River Flooding 261

Bellardone G. & Giampani C. (2003), Fiume Dora Baltea, nel tratto piemontese, in
Regione Piemonte, Eventi Alluvionali in Piemonte – Evento del 13–16 Ottobre
2000, vol. I, pp. 130–142, ARES, Arti Grafiche – Torino (in Italian).
∗Bortolami G.C., Campanino F., Carraro F., Clari P.A., Forno G., Ferrero E.,
Ghibaudo G., Maso V. & Ricci B. (1978), Dati preliminari sulla neotettonica
dei Fogli 56 (Torino), 68 (Carmagnola) e 80 (Cuneo). Estratto da: Contributi
preliminari alla realizzazione della Carta Neotettonica d’Italia. Pubb. N. 155
del Progetto Finalizzato Geodinamica, pp. 149, 180 (in Italian).
Carraro F., Dal Piaz G.V. & Sacchi R. (1970), Studio geologico di un settore a sud
di Vercelli, in Studi geologici nel Vercellese e nella Valle Strona, pp. 121–141,
Sirea Print, Torino (in Italian).
Carraro F. (1992), Evoluzione Plio Quaternaria, in Le Alpi, dal M. Bianco al
Lago Maggiore. Guide Geologiche d’Italia, 3/vol. I, pp 84–89, BE_Ma editrice,
Roma, 1992 (in Italian).
*Compagnoni R. & Forno M.G (1992), Significato geologico di depositi fluvi-
ali ghiaiosi Pleistocenici medi nella collina di Torino. Il Quaternario 5, (1),
pp. 105–122 (in Italian).
DHI Water & Environment (2001–2004), MIKE 11© A modelling system for river
channels. Reference Manual.
*Forno M.G. (1980), Evidenza di un drenaggio abbandonato nel settore settentri-
onale dell’altopiano di Poirino (prov. Di Torino). Geografia Fisica e Dinamica
del Quaternario, 3, pp. 61–65, 3 ff (in Italian).
Giampani C. & Loi B. (2000), L’analisi dei dati storici nella delimitazione delle
aree urbane a rischio di inondazione: il caso di Vercelli. Geam Geoingegneria
Ambientale e Mineraria, giugno-settembre 2000, pp. 135–142 (in Italian).
Maraga F. (1989), Ambiente fluviale in trasformazione: l’alveo-tipo pluricursale
verso un nuovo modellamento nell’alta pianura padana, Atti del Congresso
“Suolosottosuolo”, vol. 1, pp. 119–128, Associazione Mineraria Subalpina,
Torino 27–30 settembre 1989 (in Italian).
Maraga F. & Turitto O. (1998), Diagnosi geomorfologica d’inondabilità in casi
di studio sull’idrografia padana, in: Luino F (ed.), “La previsione delle catas-
trofi idrogeologiche: il contributo della ricerca scientifica.” Atti del Convegno
Internazionale “Alba 96”, vol. II, pp. 313–334, L’Artistica di Savigliano (in
Italian).
Ministero di Lavori Pubblici, Servizio Idrografico (1969), Alluvione dei giorni 1–3
Novembre 1968 sul territorio della Liguria e del Piemonte, suppl. al Bollettino
Idrografico del mese di Novembre 1968, pp. 14, Tipografia Edigraf, Roma (in
Italian).
Nigrelli G. (2005), Vezzolano: inquadramento naturalistico. Natura Web 2005 (in
Italian).
Ortolani F. & Pagliuca S. (2001), Le variazioni climatiche storiche e la prevedibil-
ità delle modificazioni relative all’effetto serra. Associazione Italiana Nucleare,
Roma (in Italian).
Polino R. & Sacchi R. (1995), Rapporti Alpi-Appennino, Atti II Incontro Inter-
nazionale dei Giovani Ricercatori in Geologia Applicata, Peveragno (CN) (in
Italian).
*Ramasco M., Giampani C. & Aliatta P. (2001), Analisi comparata dell’evoluzione
storica dell’alveo del fiume Sesia e delle piene del novembre 1968 e settem-
bre 1993 finalizzata alla prevenzione degli effetti indotti da eventi alluvionali.
Collana Informativa Tecnico Scientifica Arpa Piemonte Quaderno 8 (in Italian).
262 The Assessment of Hazard and Risk

Ravera S., Buffo M., Cavallero E. & Tresso F. (1996), Aspetti idrologici dell’evento
dei giorni 2224 settembre 1993, in Gli eventi alluvionali del settembre ottobre
1993 in Piemonte, pp. 27–31, L’artistica di Savigliano, Torino (in Italian).
Regione Piemonte (1998), Eventi alluvionali in Piemonte – 2–6 novembre 1994, 8
luglio 1996,7–10 ottobre 1996, vol. I page. 414, L’artistica di Savigliano, Torino
(in Italian).
Rosgen D. (1994), A classification of natural river. Catena, vol. 22, 169–199.
Visentini M. (1952), Igiene e terapeutica del Po, in La Rotta del Po, Ulisse, Anno,
VI, vol. III, Fascicolo XVII, pp. 552–565 (in Italian).

MAPS
*Bellardone G. & Tamberlani F., Carta numerica dei processi e degli effetti
associati al passaggio della piena del torrente Orco: evento alluvionale del
23–25 settembre 1993, tratto Cuorgné-confluenza Po. Scala 1:10.000, Regione
Piemonte Repertorio Cartografico.
*Bellardone G., Forlati F. & Tamberlani. F., Carta numerica dei processi e
degli effetti associati al passaggio della piena del fiume Dora Baltea: evento
alluvionale del 23–25 settembre 1993. tratto Carema-Ivrea. Scala 1:10.000,
Regione Piemonte Repertorio Cartografico.
*Bellardone G. & Giampani C. (2003), Carta numerica: del campo di inondazione
ed effetti indotti dalla piena del 13–16 ottobre 2000 della Dora Baltea nel tratto
piemontese, Scala 1:10.000. Arpa Piemonte.
Dutto F. (1993), Carta degli alvei attivi del F. Po tra il 1958 e il 1988, in Piano
territoriale del Fiume Po, 1993.
*Giampani C. & Ramasco M., Carta numerica del campo di inondazione ed effetti
indotti dalla piena del 2–3 novembre 1968 del fiume Sesia. tratto Borgosesia-
confluenza Cervo. Scala 1:10.000, Regione Piemonte Repertorio Cartografico.
*Giampani C. & Ramasco M., Carta numerica delle tendenze evolutive del fiume
Sesia dal 1954 al 1994, tratto Borgosesia-Vercelli. Scala 1:10.000, Regione
Piemonte Repertorio Cartografico.
*Giampani C. & Ramasco M., Carta numerica delle modificazioni dell’alveo
del fiume Sesia dal 1851 al 1994. tratto Romagnano-Vercelli. Scala 1:10.000,
Regione Piemonte Repertorio Cartografico.
Mensio L. (1996), Carta numerica del campo di inondazione ed effetti indotti
dalla piena del 04–06 novembre 1994 del fiume Po in provincia di Torino. Scala
1:10.000, Regione Piemonte Repertorio Cartografico.
*Giampani C. & Aliatta P. (1996), Carta numerica dei processi e degli effetti
associati al passaggio della piena del fiume Dora Baltea: evento alluvionale
del 23–25 settembre 1993, tratto Ivrea-confluenza Po. Scala 1:10.000, Regione
Piemonte Repertorio Cartografico.
*Giampani C., Carta numerica dei processi e degli effetti associati al passaggio
della piena del fiume Sesia: evento alluvionale del 23–25 settembre 1993, tratto
Borgosesia-confluenza Po, inedita.
Paro L. (2003), Carta numerica del campo di inondazione ed effetti indotti dalla
piena del 13–16 ottobre 2000 del fiume Po nel tratto Torino, confluenza Dora
Baltea. Scala 1:10.000. Arpa Piemonte.
Susella G., Bormioli D. & Gallo M. (2003), Carta numerica del campo di inon-
dazione ed effetti indotti dalla piena del 13–16 ottobre 2000 del torrente Orco
nel tratto Cuorgnè-Chivasso, Scala 1:10.000. Arpa Piemonte.
Forecasting and
Warning 3
 3.1 Meteorological Forecasting

3.1.1 PREDICTABILITY

The atmosphere is an intrinsically chaotic dynamic system with a high

number of degrees of freedom. At any given moment, its state is described
by the spatial distribution of wind, temperature, pressure, humidity and
other variables. The atmosphere, because it is a fluid, belongs to the branch
of physics called fluid dynamics and its evolution can be described in
mathematical terms like those for the conservation of mass, energy, and
momentum.

Figure 3.1.1 An example of

data coverage from WMO
(World Meteorological
Organization) soundings.

The solution of these differential equations requires the knowledge of

an initial condition from which integrations over time can be carried out.
This initial condition is the result of a complex procedure of assimilation
which evaluates the state of the atmosphere from available observations
(the so-called analyses). The fact that the supply of observed measure-
ments is limited and irregular introduces a significant factor of uncertainty
in the initial conditions that inevitably impact the forecast as a source of
error. For a forecast to be accurate, models must also be capable of real-
istically reproducing dominant atmospheric phenomena. As a matter of
fact, models can only simulate a limited number of processes according
 266 Forecasting and Warning

Figure 3.1.2 An example of

data coverage from WMO
surface observation stations.

to their space and time resolution scales. This is the second main source
of forecast errors.
Finally, in spite of the surprisingly fast development of computer
resources, including increasingly powerful and faster machines, compu-
tational resources often limit the complexity and resolution for numerical
models, making it necessary to find a compromise between model accu-
racy and computing times for operational purposes. These three main
sources of error cause the accuracy of the forecasts to deteriorate with an
increase forecast time. The uncertainty of the initial conditions can never
be completely removed due to both the intrinsic error of the measuring
instruments and the limits imposed by the irregular distribution coverage.
This means that minor uncertainties related to the observation system can-
not be eliminated from the initial conditions. As a consequence, even in
the hypothesis of a perfect equation system, two initial states that differ
even slightly tend to diverge more and more rapidly with an increase in
forecast time. Observation errors are amplified in time until they nega-
tively affect forecast reliability. Over the past 15 years, the accuracy of
forecasts for the first 5 days has improved substantially, but the reduction
of error for successive days is still very limited. The scientific community
is presently aiming to resolve the problem of predictability by tackling
forecasting in terms of probability through the so-called EPS (Ensemble
Prediction System). This approach takes into account the chaotic compo-
nent of atmospheric movements that are usually neglected by deterministic
numerical models. Lorenz (1963) was the first to find out that repeating
the same simulation with slightly different values as initial conditions
resulted in drastic differences in the evolution of the numerically elabo-
rated atmosphere. The results of his experiment led to the development of
the EPS forecast technique. This is obtained by carrying out a high number
of simulations which start from different initial conditions, which are in
 Meteorological Forecasting 267

turn obtained by slightly modifying the analysis. Every scenario obtained,

which may greatly differ from the others, is associated with a probability;
for example, if 25 out of 50 simulations forecast more than 20 mm of
precipitation, this event is assigned a 50% probability of occurrence.

Figure 3.1.3 Scheme of

possible atmospheric
evolution in time in a
deterministic model. Starting
from initial conditions that
differ even slightly from each
other, the trajectories that
describe the evolution in time
of the atmosphere diverge
greatly; the gap between the
arrival points evaluates the
forecast uncertainty.

While high resolution models have been proven to be very sensitive

to small variations in the initial conditions, the EPS approach derives an
advantage from this uncertainty. This technique is used by the Piemonte
meteorological service, which receives the daily results of a 51-member
ensemble produced by the ECMWF (European Centre for Medium-range
Weather Forecasts) for precipitation and temperature. These forecasts are
used for the so-called medium range forecast (from 4 to 7 days). An exam-
ple is given in Figure 3.1.4, which refers to the probability of precipitation
for October 15th and 16th, 2000. This flood was in fact forecasted with
sufficient notice thanks to this product. Currently in Piemonte, medium
term forecasts are considerably important in the Olympic system bulletin
for the description of meteorological trends.

3.1.2 MODELS

There are two main categories of numerical weather forecasting models:

• GCM (Global Circulation Model);
• LAM (Limited Area Model).
The GCMs are the starting point of the forecast process for atmospheric
phenomena. They simulate the evolution of the atmosphere over the entire
globe and a vertical range of 30 km from the earth’s surface (even if the
area involved in the evolution of weather is limited to the first 15 km from
the surface – Troposphere). What makes this atmospheric layer particu-
larly interesting for meteorology is the presence of water vapour in its
268 Forecasting and Warning

Figure 3.1.4 Precipitation probability forecasted by ECMWF for different thresholds (1, 5, 10, 20 mm), between
October 15th and 16th, 2000.

composition; consquentially, this is the place where all phase changes and
heat flows, which are at the basis of meteorological events, occur. Due to
this vast domain of integration (the entire globe), GCMs have a low spatial
resolution (about a few hundred kilometres). In order to understand the
concept of resolution, it is necessary to consider that the continuous func-
tions that represent the dynamic variables in models are represented as
matrixes. The latter assume their values on a domain, constituted by a set
of points in space, which in turn constitute the so-called grid of the model.
The resolution is given by the distance between two contiguous points:
the smaller the interval, the greater the resolution. The type of phenomena
that may be correctly simulated is defined according to the resolution. In
fact, a model is capable of realistically reproducing phenomena that occur
on a scale that is approximately twice its characteristic scale. On the other
hand, atmospheric phenomena present an enormous variety with regards
to type as well as in space-time scales. They include the so-called large
scale, of approximately 3,000 km, the mesoscale processes, or hundreds of
kilometres, breezes and squall line processes (tens of kilometres), convec-
tive phenomena (1–2 km) and even molecular motion. GCMs can usually
correctly simulate large scale and mesoscale phenomena.
 Meteorological Forecasting 269

LONG-WAVE SHORT-WAVE O3 CHEMISTRY

RADIATION RADIATION CH4 OXIDATION

CLOUD

CLOUD SUBGRID-SCALE
OROGRAPHIC DRAG
DEEP
CONVECTION

TURBULENT DIFFUSION
SHALLOW
CONVECTION

Latent Sensible
Long-wave Short-wave heat heat flux
flux flux CE
flux SURFA
WIND WAVES
OCEAN MODEL

Figure 3.1.5 Scheme of atmospheric processes.

LAMs, instead, are used on more limited areas. Reducing the domain
of integration reduces the load on the computer, making it possible to
increase the spatial resolution of the model (below 40 km), so that phe-
nomena that take place on smaller space and time scales can be correctly
simulated. LAM models are used to simulate mesoscale phenomena with
features that cannot be resolved by large scale models. According to
Orlanski’s definition (1975), such phenomena regard the following scales:
• microscale, less than 2 km (vortex turbulence, for example);
• meso γ, between 2 km and 20 km (for example, convection and flows
over complex orography);
• meso β, between 20 km and 200 km (sea and land breezes);
• meso α, between 200 km and 2,000 km (for example, frontal systems
and the development of low pressure areas).
GCM models typically have a longer forecast time and are gener-
ally used for medium term forecasts, which would be from the third day
onward. Instead, the first three days of forecasts are simulated more accu-
rately by limited area models. LAMs can in turn be divided into hydrostatic
and non-hydrostatic models, according to their horizontal resolution. The
former uses the so-called hydrostatic approximation, which is an equa-
tion that assumes a perfect balance between the gravity force and pressure
270 Forecasting and Warning

gradient. In this case, vertical acceleration is derived from the difference

of vertical velocities diagnosed by the model, imposing the conservation
of mass, instead of being explicitly computed. This approximation is valid
on a horizontal scale greater than 10 km.
Non-hydrostatic processes become non-negligible when the vertical
motions in the atmosphere change rapidly, or in other words, when ver-
tical accelerations reach the same scale as horizontal ones, as in the
case of intense convective phenomena. Given H as the vertical scale
and L as the horizontal scale of a process, if H/L  1, then the approx-
imation is valid and the vertical accelerations are negligible; otherwise,
if H/L ≥ 1, non-hydrostatic conditions prevail and other equations that
explicitly describe the vertical motions are needed. Usually, LAM models
include non-hydrostatic effects when their resolution is higher than 10 km.
If the resolution is increased, in fact, the hydrostatic equation is no
longer suitable for describing phenomena like some types of lee-waves or
cumulonimbus dynamics. In these cases non-hydrostatic models must be
used.
The advantages of LAMs in terms of resolution are, however, limited
by the fact that they evolve in a restricted portion of the atmosphere and
independently from the rest of the atmospheric system, they risk giving
unrealistic results. Obviously, it would be ideal to have a GCM with a
resolution as high as that of a LAM; this, however, is not achievable
with existing computing resources. In order to sum the advantages of
LAM resolution without losing those of the GCM, the so-called “nesting”
technique is applied. This technique uses the result of a low resolution
model as an initial and boundary condition to carry out the simulation
with a higher resolution model. Nesting is used for both passing from
a GCM to a LAM and for further refining the results by telescopically
increasing the resolution of the LAM itself.
As high as the resolution of the numerical model may be, it would not
be capable of solving the complexity of phenomena and the variety of
space and time scales involved in atmospheric processes.
The so-called dynamic variables, or the atmospheric variables explic-
itly computed by the model equations, are usually:
• horizontal components of wind velocity, U and V, and the vertical one,
W, if the model is non-hydrostatic;
• the temperature, T;
• relative (or specific) humidity, q;
• pressure, p.
These variables are derived directly from the Navier-Stokes equations:

dv
ρ = −∇p + ρg − 2 × (ρv) − ∇ · (ρv v )
dt



dp cpd cpd
=− p∇ · v + − 1 · Qh
dt cvd cvd
 Meteorological Forecasting 271

dT dp
ρcpd = + Qh
dt dt
dqv
ρ = −∇ · Fv − (I I + I f )
dt
dqI , f
ρ = −∇ · (P I , f + F I , f ) + I I , f
dt

 −1
Rv
ρ = p Rd (1 + − 1 qv − qI − qf )T (3.1.1)
Rd

where ρ is the air density, v is the vectorial wind speed, g is the accel-
eration of gravity,  is the angular velocity of the earth, and the terms
marked by superscripts represent turbulent terms not explicitly resolved
by the model; Cpd represents the specific heat at a constant pressure of dry
air, Cvd is the specific heat at constant volume, QH is the heat flow, q is the
specific humidity (the superscripts refer to its various states: vapour, liq-
uid, solid), I represents the microphysical developing processes of clouds
and precipitation, F is the evaporative flow, P precipitation flow, T is tem-
perature, and Rd is the gas constant for dry air while Rv is the one for
vapour. These prognostic equations constitute the dynamics of the model.
Other meteorological fields – like precipitation, snow cover, radiative
flow, and momentum in the planetary boundary layer – are not computed
directly, but rather by means of parameterization. This is intended as the
evolution of processes that occur on scales smaller than those resolved by
the model through empirical expressions. The latter are obtained through
direct observations, rather than through mathematical formulas that would
require heavier computations. Empirical formulas are derived from trends
observed during direct measurement campaigns and express a correlation
between the actual variable and one or more dynamic variables of the
model so that the variation of the first is a function of the second. An
example of parameterization is given in the following equation for the
computation of momentum flux τ (Wu, 1982):

τ = ρaria CD U10 2 (3.1.2)

where ρaria represents air density, U10 represents 10 m height model wind
speed, and CD is a constant (drag coefficient) given by the following
formula, once again as a function of U10 .

CD = 10−3 (0.8 + 0.065 · U10 ) (3.1.3)

Thus, the momentum flux is expressed as a function of dynamic variables

(in the specific case of wind speed at 10 m) instead of the explicit equation
for its evolution in time.
Another serious issue in atmospheric modeling is the initialization
problem. Both GCMs and LAMs, initiate their computations with a few
values observed within their domain of integration. These represent the
 272 Forecasting and Warning

initial condition of the actual model. Different data assimilation techniques

can be used to produce an analysis which represents the atmospheric con-
figuration from which the model starts. The more reliable and detailed this
analysis is, the more easily problems related to “spin up”, or the initial
divergence between the model and the real atmosphere, are avoided. In
a LAM configuration, an initial condition given by an analysis derived
from a global model is known as cold start. In this case the LAM slowly
reproduces the atmospheric circulation. During the first integration steps,
oscillations may develop, due to the model’s attempts to stabilise the
boundary conditions with its internal dynamics. On the other hand, when
the model analysis is obtained from a data assimilation system, the initial
condition is called warm start. Observed data are then assimilated homo-
geneously both in space and in time, through opportune weight functions
in the model, taking care to maintain its internal consistency.

Figure 3.1.6 LAMI domain

together with model
orography.

The Arpa Piemonte Centro Funzionale makes use of the global model
IFS, of ECMWF, which is distributed daily through the national meteoro-
logical centre of the Military Service (Aeronautica), and the Italian version
of the Lokal Modell, known as the Limited Area Model Italy LAMI. The
latter was developed under the auspices of the COSMO project (COnsor-
tium for Small-scale MOdelling, www.cosmo-model.org), in which the
national meteorological services of Italy, Germany, Switzerland, Greece,
and Poland joined forces with other regional centres, including Arpa
Piemonte and Arpa SIM (Emilia Romagna). These institutions stipulated a
convention that provides for the maintenance of the operational model and
its development. Since February 2001, the model produces daily forecasts
 Meteorological Forecasting 273

by means of numerical simulations carried out at the CINECA scientific

computation centre in Bologna. The current configuration is as follows:
• a domain covering the national territory;
• 7 km of horizontal resolution;
• 35 vertical levels;
• 72-hour forecast;
• output every 3 hours;
• 2 runs per day (00 UTC and 12 UTC);
• initial and boundary conditions given by the German global GME
model;
• assimilation of parameters aloft during the 12 hours prior to the run.
The model prognostic variables are: longitudinal and latitudinal and
vertical components of momentum, the temperature, pressure perturba-
tion from a reference condition, specific humidity, cloud water content.
The diagnostic variables are total density and fluxes of precipitation and
snow. The coordinate system is rotated and has uniform horizontal resolu-
tion and a generalised vertical coordinate system. The model uses a second
order finite differences space integration scheme and an Arakawa C-type
grid for horizontal variables and a Lorentz grid for vertical. The time inte-
gration is performed with a second order Leapfrog-type scheme (which
is explicit in a horizontal direction and implicit in a vertical direction).
The initial and boundary conditions are obtained from data interpolated
from the GME global model having higher horizontal resolution. Tiedtke’s
classic convection scheme is used.
The GME (German global circulation model) and the Swiss (called
aLMo) and the German versions (LM-DWD) of the Lokal Modell are
also used. The differences between the different LM versions are shown
in Table 3.1.1.

TABLE 3.1.1 DIFFERENCES AMONG DIFFERENT LM VERSIONS

LM CONFIGURATIONS ARPA DWD METEOSWISS
Domain Size (grid points) 234 x 272 325 x 325 385 x 325
Grid Spacing (horizontal) 0.0625° 0.0625° 0.0625°
(7 km) (7 km) (7 km)
Number of Layers 35 35 45
Time Step 40 sec 40 sec 40 sec
Forecast Range 72 hrs 48 hrs 72 hrs
Initial Time of Model Runs 00,12 UTC 00,12,18 UTC 00,12 UTC
Lateral Boundary Conditions GME GME IFS
LBC Update Frequency 1 hr 1 hr 3 hr
Initial State Nudging Scheme Nudging Scheme Nudging Scheme

External Analysis None SST, Snow Snow Depth

Depth, Soil from DWD
Moisture
 274 Forecasting and Warning

3.1.3 FORECASTS AND VALIDATION: THE IMPORTANCE

OF FEEDBACK

Forecast verification, especially the QPF (Quantitative Precipitation Fore-

cast) verification, is one of the most important activities of a Centro
Funzionale because it allows a better understanding of how the model
behaves in different meteorological configurations, highlights the sys-
tematic characteristics, and helps in evaluating reliability, whether for
average or extreme values, or over the long term or in the current situa-
tion. A good verification method represents different levels of feedback
concerning the warning system as a whole. It allows operators to under-
stand not only where interventions may bring improvements, but, above
all, affords the Centro Funzionale optimal exploitation of all the forecasted
fields in function of the forecast target.
In order to verify the quality of any forecasting model, there must be
the highest possible number of observed data so that a comparison is car-
ried out over a long period of time and over an area wide enough to be
considered a statistically valid sample. To reduce the problem of represen-
tativeness in verifying the mesoscale model, it is preferable to examine
detailed meteorological data like the ones provided by high-resolution
non-GTS stations, which may be integrated with those produced by radar.
For forecast validation, Arpa Piemonte uses the observed data of about
400 automatic ground stations managed by Arpa itself (Figure 3.1.7), the
data available from the COSMO project regarding northern Italy (Figure
3.1.8), and the data observed from the ground network of the Centri Fun-
zionali system that is now available at the Department of Civil Protection
(Figure 3.1.7).

Figure 3.1.7 Distribution of

the automatic stations and
warning areas in Piemonte
(left) and rain gauges
available at the Department
of Civil Protection (right).
 Meteorological Forecasting 275

Figure 3.1.8 Subdivision of warning areas (about 3,000 km2 ) (left); subdivision into boxes with sides of about
50 km (right); observation network available in the framework of the COSMO project.

The quality of the forecasted precipitation field is evaluated according

to statistical indices based on contingency tables. These are double entry
tables in which, for a specific threshold of precipitation accumulated over
a certain interval of time, the number of cases forecasted and observed
that surpass this threshold are evaluated as follows, where:
• A is the number of cases in which the event is neither forecasted nor
observed;
• B is the number of cases in which the event was forecasted but not
observed (false alarm);
• C is the number of cases in which the event observed was not forecasted
(misses);
• D is the number of cases in which the event observed was forecasted
(hits);
From the abovementioned values for the warning areas, the classical
quality indices are calculated (BIAS, which indicates the errors of overes-
timation or underestimation, ETS, which indicates the error of space-time
localization, FAR, which gives the number of false alarms, HRR, which
represents the number of hits), over a long period, to evaluate the average
reliability of the model, and seasonally, to evaluate the situation in func-
tion of the different climatic and phenomenological characteristics and the
modifications that the model undergoes periodically during its continual
evolution.
For further information concerning standard indices, please see
Stanski et al. (1989), Wilks (1995), and Ebert & McBride (1997, 2000).
Table 3.1.3 contains the definitions and ideal values of the classical
statistical indices.
In the following pages, the reader will find some of the results
of verifications of areal precipitation forecasts of the Lokal-Modell
276 Forecasting and Warning

TABLE 3.1.2 CONTINGENCY TABLE

EXPECTED

OBSERVED
NO YES
NO A B
YES C D

TABLE 3.1.3 DEFINITION AND IDEAL VALUES OF

CLASSICAL STATISTICAL INDEXES

ideal BIAS = 1

ideal FAR = 0

ideal HRR = 1

where

ideal ETS = 1

non-hydrostatic model developed within COSMO. The Italian version of

this model, called LAMI, is the limited area model used by the Depart-
ment of Civil Protection. Furthermore, the global model of the ECMWF
and the subjective numerical forecast of forecasters in Arpa Piemonte are
also subject to verification.
Oberto et al. (2003) and Milelli et al. (2003) examined the depen-
dence of LAMI QPF on the characteristics of the zone and, in particular,
on its dimension. They considered the different possible results accord-
ing to whether the model output was averaged on boxes of different areas
(0.25◦ ×0.25◦ , 0.50◦ ×0.50◦ , 0.75◦ ×0.75◦ ) or on warning areas. In these
studies, where the Piemonte network was considered, different perfor-
mances were obtained according to the different areas. Following up on
the work with warning areas for Northern Italy and using the observation
network available through the COSMO project (Figure 3.1.8), the above
results were confirmed (Figure 3.1.9).
The BIAS of LAMI QPF for thresholds greater than 35 mm/24 h indi-
cates a lower overestimate for the QPF on the warning area, while the ETS
 Meteorological Forecasting 277

1.8 0.45
1.6 0.40
1.4 0.35
1.2 0.30
BIAS

ETS
1.0 0.25
0.8 0.20
0.6 0.15
0.4 box 0.10 box
0.2 area 0.05 area
0.0 0.00
5 10 10 35 50 75 5 10 20 35 50 75
mm/24h mm/24h

Figure 3.1.9 BIAS (left) and ETS (right) for quantitative precipitation forecasts of the LAMI model (12 UTC
run) averaged over warning areas or boxes with sides of 50 km (Figure 3.1.8) during the period of June
2002-August 2003.

index shows better accuracy, considering each threshold, for the areas with
respect to the boxes. The QPF on the warning areas demonstrates better
results than the QPF calculated on boxes with dimensions comparable to
the zones. This result represents important positive feedback: the criteria
for defining warning areas are valid to obtain improved performance in
quantitative precipitation forecasts.
To evaluate improvements in the subjective analysis of the quantifica-
tion of precipitation information, the skills of subjective forecasting (PSA)
are compared with those of the LAMI model. Figure 3.1.10 provides the
results obtained varying the thresholds of the different statistical indices
relative to the areal average over the 24 hours of the second day of forecast
(+24/+48 hours of forecast) and to the period between January 2003 and
October 2004. In this analysis it is possible to carry out the hypothesis test
developed by Hamill (1999), in which a confidence interval is calculated
with the bootstrap method to establish the real difference between the skill
scores of the two models.

2.5 0.50
0.45
2.0 0.40
0.35
1.5 0.30
BIAS

ETS

0.25
1.0 0.20
0.15
psa psa
0.5 lami 0.10 lami
0.05
0.0 0.00
1 2 4 8 16 32 1 2 4 8 16 32
mm/24h mm/24h

Figure 3.1.10 BIAS and ETS indices as a function of the QPF threshold produced subjectively (PSA) and
by the LAMI model.
278 Forecasting and Warning

The error bar in Figure 3.1.10 indicates 2.5 and 97.5 percentiles of the
resampled distribution assigned to the reference model. It can be noted that
the test is symmetric: the reference model can be the competitive model
and vice-versa. The performance difference is statistically significant if
the skill score of the competitor model is outside a given confidence
interval (in this case 95%) of the resampled distribution.
Figure 3.1.10 indicates that both the forecasts overestimate the QPF,
especially the PSA for low thresholds (less than 4 mm/24 h): there are
no significant differences for thresholds greater than 8 mm/24 h in terms
of BIAS, while for thresholds greater than 16 mm/24 h, the PSA is more
accurate than the LAMI (ETS index).
In conclusion, we include a synthesis of more general results obtained
from validation studies (Turco et al., 2005):
• in spite of the overall improvement on a global scale and the fact that
the resolution of the limited area models has increased considerably
over recent years, the QPF produced by meteorological models has
not improved enough to allow it to be used directly. QPF subjective
forecasting continues to offer the best performance;
• in regions with complex orography (especially Alpine reliefs), the pat-
tern of precipitation forecasted is still not satisfactory; in particular,
there is overestimation in the maximum precipitation upwind and under-
estimation downwind (this effect is more pronounced during the winter
than the summer);
• generally, the 12 UTC runs demonstrate better skills than the 00 UTC
runs. In fact, the analysis of the 12 UTC can count on a longer cycle
of assimilation and on a greater number of available observations. Fur-
thermore, the initialization of the model from a more energetic state (12
UTC) allows a better representation of the triggering of precipitation
and a greater destabilisation of the atmosphere;
• the tendency of the model to perform worse during the summer;
• in general, the performance is less satisfactory when the threshold is
increased, or in other words, forecasting skills diminish for intense
rains. However, this result is also influenced by a small number of data
relative to the high thresholds that make the statistics less significant;
• there is a daily error cycle and, in particular, the worsening with the
forecast time is evident; furthermore, a skills periodicity of 24 with
greater BIAS values during the central hours of the day has been noted;
• considering more than 60% of the Italian territory, it can be stated
that LAMI overestimates most of the zones considered, especially the
Alpine ones, while the ECMWF model underestimates in most areas,
especially along the coasts;
• the new configurations introduced (for example, variational soil mois-
ture analysis, the scheme for prognostic precipitation, the correction
of some variables forecasted by means of assimilation of GTS obser-
vations through the Nudging technique) in operational versions of the
Lokal-Modell have brought significant improvements to the QPF.
 Meteorological Forecasting 279

In reference to the verification process of the quantitative use of mete-

orological forecasting in the Arpa Piemonte warning system, the main
conclusions are:
• the added value of the subjective analysis in providing QPF for warning
areas is significant: the performance of the warning system QPF is
generally good;
• the QPFs produced by meteorological models for the warning areas
demonstrate better results than the QPFs calculated on boxes with
dimensions comparable to the zones. This result represents an impor-
tant positive feedback: the criteria of definition of warning areas are
valid for obtaining improved QPF performance;
• it is difficult to determine long term systematic trends (in the Piemonte
warning system the quantitative precipitation forecasts on the warning
areas have been issued daily since March 2000) concerning the sub-
jective performance of the QPF, given the great annual variation in the
predictability and quantity of precipitation;
• performances are better for thresholds of 10 mm/24 h than for
1 mm/24 h; it is generally more difficult to obtain good performances
for weak and local rainfall forecasts that are difficult for models
to reproduce, while more consistent precipitation linked to synoptic
forcings are forecasted with greater precision;
• the quantitative precipitation forecasts in the warning areas present
decreasing overestimations considering the Alpine zones to the North,
then the southern areas, and finally the plains, where there is the best
space-time localization.

3.1.4 PROSPECTS AND TECHNIQUES FOR IMPROVING

METEOROLOGICAL FORECASTING

A meteorological model is not a static and immobile tool over time, but
a dynamic system that evolves according to the most recent innovations
of science and technology. The methods for improving a model may be
divided into two categories:
1) acting from inside by developing new parameterization schemes, new
numerical schemes, an increase in the space-time resolution, or by
inserting improved data assimilation techniques;
2) acting from outside with post-processing techniques applied to the
direct output of the model.
Concerning the development of new schemes, the model was tested
on some intense rainfall events in Piemonte as part of the Interreg IIIB
Hydroptimet project. In particular, the case of November 25th, 2002 was
taken into consideration. Different simulations were carried out differing
from the reference simulation (which was simply called s4) in the param-
eterization of some physical processes or in numerical schemes. The goal
280 Forecasting and Warning

was to highlight which of these schemes had the greatest impact on rainfall
forecasting. The results of the different simulations were mainly analysed
in terms of comparison between the values of precipitations forecasted
and observed over the warning areas in Piemonte.
More specifically, the following changes were made: the convection
scheme, the value of the Raymond filter applied to the orography, and the
number of vertical levels defined in the model; furthermore, other prog-
nostic variables like the ice phase of clouds (produced by the sublimation
of water vapour) and rainy and snowy precipitation were introduced.
The set of simulations produced is given in Table 3.1.4.
The results obtained with this set of simulations, limited to this case
study, indicate that the change that has the greatest effect on the localiza-
tion and quantification of rain forecasted is the introduction of the new
prognostic variables. This is especially true in a zone with a complex orog-
raphy, where the exact consideration of the advection of the precipitation
across the reliefs is fundamental. The results of this configuration are
further improved by increasing the number of vertical levels.
The meteorological model used at Arpa Piemonte has a resolution of
about 7 km, but may initially produce results at a higher resolution, pro-
vided some substantial changes are made in the equations describing the
physical phenomena. One possibility (from a purely numerical point of
view) of increasing the resolution of the model is to use a multiscale
approach, that is to say, multiple grids with space (and time) resolu-
tions that can also vary in dimensions and that transmit information to
each other. The advantage is a dynamic interaction between different
scales at a reduced computational cost, an improved representation of
the geometry of the domain and of the physical forcings, an improved
conservation of mass, and a more accurate and robust simulation of small
scale dynamic phenomena, as well as an improvement of the large scale
simulation. For this reason a testing phase has begun with numerical exper-
iments and applications to relevant cases for the Olympic simulations.
The purpose of these ongoing efforts is to obtain very high resolu-
tion forecasts (up to 1 km) by using innovative techniques for refining
the grids.
From the point of view of meteorological modelling and its appli-
cations, the prospect for the next few years is to reach an operational
resolution of approximately one km in areas that are large enough to
represent phenomena like the formation and propagation of storm clouds.
In addition to increasing the resolution in order to allow a more detailed
description of turbulent motion, it is necessary to refine the description of
those physical processes that occur in the atmosphere and on the earth’s
surface, causing exchanges of energy and the transformation of water into
its different states. For this purpose, other parameterization techniques
must be studied. The analysis of the interaction of surface geomorphologi-
cal parameters (both meteo-climatic factors and parameters inherent to the
vegetation and the ground) with the surface layer of the atmosphere, with
 Meteorological Forecasting 281

TABLE 3.1.4 LIST OF SIMULATIONS MADE

S4 Reference
S5 Without orographic filter
S6 With stronger orographic filter
S7 With ice and snow phase calculated
prognosticly
S8 With all water phases calculated
prognosticly
S9 Kain Fritsch convective scheme
SA Without convective scheme
SB 45 vertical levels without orographic filter
45 vertical levels without orographic filter
SC
and all water phases calculated prognosticly
45 vertical levels without orographic filter
SD
and ice and snow phase calculated prognosticly

particular reference to the processes of air-land and geosphere-biosphere

interaction from a hydrological point of view can be conducted with the
help of SVAT (Soil Vegetation Atmosphere Transfer) models. These mod-
els simulate the physical processes at the ground-atmosphere interface,
reproducing energetic and hydraulic balances.

1.3

1.2

1.1
BIAS

1.0 s4
Figure 3.1.11 Average BIAS s8
over Piemonte from 00 UTC, 0.9 sC
November 25th 2002 to 00
UTC, November 26th 2002 0.8
for different configurations of 5 10 20 35
the meteorological model. threshold (mm)

Arpa Piemonte uses a SVAT model for operative agricultural-

meteorological applications: the LSPM (Land Surface Process Model),
developed by the University of Torino (Loglisci et al., 2001). The struc-
ture of the LSPM can be outlined as follows: the spatial domain is limited
above by a reference level in the atmosphere above the vegetation and
below by the ground layer under the plants’ root system. In the atmo-
sphere, the boundary parameters are: temperature, humidity, wind, global
radiation, pressure, and precipitation at the height of the vegetation. The
vegetation is considered as a uniform layer (big-leaf approximation) in
both the atmosphere and the substratum, while the soil can be subdivided
282 Forecasting and Warning

into three or more layers (as the user prefers). Both are described by a
series of physical parameters.
The LSPM model assumes that the precipitation that reaches the
ground partially infiltrates into the first layer of the soil until it is saturated
(the excess water runs off), and that the roots extract humidity from the
layers of soil around them (transpiration). The sensitive and latent heat
fluxes are calculated according to the formulation of the analogy with
electrical circuits. The typical outputs of the model are: net radiation, heat
fluxes , temperature, and humidity in the substratum, in the air, and at the
vegetation level. The use of the model described above is important for the
forecasting of physical variables that cannot be easily measured, but that
are of fundamental importance from an agricultural point of view, like the
humidity content or the ground temperature, and find application in the
forecasting of late and early frosts or the establishment of agricultural-
meteorological conditions that are critical for the growth of particular
parasites.
The future of research in the field of mathematics as applied to
meteorology is, moreover, also related to the assimilation techniques of
meteorological data that aim at the optimal usage of all the atmosphere
observation data that arrives daily from fixed and mobile stations, weather
balloons, airplanes, ships, buoys, and satellites.
Making better use of these observations, especially satellite informa-
tion, which even though affected by notable errors of precision, are very
dense in space and time, requires mathematical instruments based on par-
ticular versions of the same forecasting models and so-called variational
algorithms. To solve these problems, a computational power greater than
the one used for meteorological models is needed.
Only through the use of such assimilation systems is it possible to
fully exploit certain observations like those from satellites. The satellite,
in fact, receives a great number of measurements of the radiance coming
from the surface below, from which it is possible to indirectly derive esti-
mates of the temperature and humidity of the atmosphere. For many years,
these remote measurements were of limited usefulness because they were
difficult to translate into information useful for atmospheric forecasting;
newer generations of meteorological satellites and new measuring instru-
ments that are much more articulated and sophisticated (radiometers) have
improved the situation.
Another field in which work is being done to improve model results is
the post-processing, or the treatment (usually based on statistics) of direct
meteorological model output. The post-processing methods used as part
of the services distributed by Arpa are:
• Kalman Filter;
• Multimodel Ensemble and SuperEnsemble.
These are currently applied to the surface meteorological parameters
that present variations on a space-time scale smaller than those resolved
 Meteorological Forecasting 283

by general circulation or limited area models. Therefore, the physical pro-

cesses that describe their evolution cannot be calculated explicitly by the
model, but must be parameterized off-line by using appropriate schemes.
In fact, the direct use of the output from a model does not yet provide sat-
isfactory forecasts; these techniques, in aiming to reduce the gap between
the model and the observations trajectories, tend to favour the latter.
In order to take into account extreme atmospheric variability, pre-
dictors and predictants could be used only during the final 2–3 weeks
and their statistical link should be evaluated every day. In this context,
recursive methods are the best option.
Attributing a greater weight to recent data results in a greater impact
on the estimates of the coefficients, which more faithfully reflect present
conditions. The Kalman filter technique (Kalman, 1960; Kalman & Bucy,

Figure 3.1.12 Example of soil

moisture calculated with the
LSPM model.

Figure 3.1.13 Example of soil

temperature calculated with
the LSPM model.
 284 Forecasting and Warning

Interpolator
Model input Notes
(forecasting the
(GRIB file) (text file)
measured data)

Covariance
matrix Kalman filter
(data/models)

Meteorological products

Regione Olympic Heat Wave

Piemonte Meteo Forecast ……… ………
Figure 3.1.14 Flow chart of Meteo Bulletin Bulletin Bulletin
the Kalman algorithm.

1961) allows the construction of an optimal recursive scheme. The Kalman

filter algorithm is presented below in more detail:
• calculation of the filter pay-off matrix;
• calculation of the innovation;
• calculation of filter pay-off;
• updating of the filter parameters;
• updating of the covariance matrix;
• updating of the filter pay-off matrix;
• calculation of the correct forecast.
The regional meteorological service currently performs several runs
of the Kalman filter on different models (ECMWF IFS, LAMI, in the
runs at 00 UTC and 12 UTC) and on different meteorological variables
(air temperature, humidity, dew point, wet bulb temperature, heat index,
humidex, discomfort index). The Kalman filter temperature forecast at
mountain stations are not sufficiently accurate. For this reason, work has
been done in two directions to improve the performance:
• modifications made to the internal parameters of the filter (x and p), in
order to give greater freedom to the filter’s response to input variations;
good results were not obtained because good and bad forecasts often
alternate in the performances of the models. Therefore the filter presents
a response to these variations that induces resonance and, therefore, the
subsequent worsening of the forecast;
• improvement of input (DMO, Direct Model Output): rather than using
the temperature at 2 m calculated by the model, the temperature fore-
casts at a pressure level higher and lower than the station altitude are
 Meteorological Forecasting 285

used and interpolated at the station altitude by means of the geopoten-

tial. In this way a reduction of the DMO error and, above all, a reduction
in the daily forecast variability were obtained, allowing an improved
functioning of the Kalman filter. Figures 3.1.15 and 3.1.16 respec-
tively demonstrate the reduction of the average error and of the average
quadratic errors obtained with the interpolation of the temperatures for
four stations of the Piemonte network.

ecm12 t2m DMO lkd00 t2m DMO alm00 int. t DMO ecm12 t2m KAL lkd00 t2m KAL alm00 int. t KAL
ecm12 int. t DMO lkd00 int. t DMO ecm12 int. t KAL lkd00 int. t KAL

2.0 2.0
MEAN ERROR (°C)

MEAN ERROR (°C)

0.0 0.0
-2.0 -2.0
-4.0 -4.0
-6.0 -6.0
-8.0 -8.0
-10.0 -10.0
-12.0 -12.0
Torino - 240 m Torino - 240 m
-14.0 -14.0
2.0 2.0
MEAN ERROR (°C)

MEAN ERROR (°C)

0.0 0.0
-2.0 -2.0
-4.0 -4.0
-6.0 -6.0
-8.0 -8.0
-10.0 -10.0
-12.0 -12.0
Bardonecchia Prerichard - 1353 m Bardonecchia Prerichard - 1353 m
-14.0 -14.0
2.0 2.0
MEAN ERROR (°C)

MEAN ERROR (°C)

0.0 0.0
-2.0 -2.0
-4.0 -4.0
-6.0 -6.0
-8.0 -8.0
-10.0 -10.0
-12.0 -12.0
Sestriere - 2020 m Sestriere - 2020 m
-14.0 -14.0
2.0 2.0
MEAN ERROR (°C)

MEAN ERROR (°C)

0.0 0.0
-2.0 -2.0
-4.0 -4.0
-6.0 -6.0
-8.0 -8.0
-10.0 -10.0
-12.0 -12.0
Sestriere Mt Banchetta - 2480 m Sestriere Mt Banchetta - 2480 m
-14.0 -14.0
12 18 24 30 36 42 48 12 18 24 30 36 42 48
forecast time forecast time

Figure 3.1.15 Mean error in the temperature forecast with respect to the observations of four regional network
stations at different altitudes (one in the plains, one in the mid mountains, and two in high mountains)
Left: direct output of models; right: Kalman filter output. ecm12 t2m: temperature at 2 m of the ECMWF IFS;
ecm12 int.t: interpolated temperature of the ECMWF IFS; lkd00 t2m: temperature at 2 m of the LM-DWD;
lkd00 int.t: Interpolated temperature of the LM-DWD; alm00 int.t: interpolated temperature of aLMo.

A second method of post-processing used at Arpa Piemonte is the

Multimodel approach that consists of the simultaneous use of more than
one model for estimating meteorological parameters. This method was
proposed by Krishnamurti et al. (1999) with the use of an Ensemble of
286 Forecasting and Warning

ecm12 t2m DMO lkd00 t2m DMO alm00 int. t DMO ecm12 t2m KAL lkd00 t2m KAL alm00 int. t KAL
ecm12 int. t DMO lkd00 int. t DMO ecm12 int. t KAL lkd00 int. t KAL

9.0 9.0
8.0 8.0
7.0 7.0
RMSE (°C)

RMSE (°C)
6.0 6.0
5.0 5.0
4.0 4.0
3.0 3.0
2.0 2.0
1.0 Torino - 240 m 1.0 Torino - 240 m
0.0 0.0
9.0 9.0
8.0 8.0
7.0 7.0

RMSE (°C)
RMSE (°C)

6.0 6.0
5.0 5.0
4.0 4.0
3.0 3.0
2.0 2.0
1.0 1.0
Bardonecchia Prerichard - 1353 m Bardonecchia Prerichard - 1353 m
0.0 0.0
9.0 9.0
8.0 8.0
7.0 7.0
RMSE (°C)

RMSE (°C)

6.0 6.0
5.0 5.0
4.0 4.0
3.0 3.0
2.0 2.0
1.0 1.0
Sestriere - 2020 m Sestriere - 2020 m
0.0 0.0
9.0 9.0
8.0 8.0
7.0 7.0
RMSE (°C)

RMSE (°C)

6.0 6.0
5.0 5.0
4.0 4.0
3.0 3.0
2.0 2.0
1.0 Sestriere Mt Banchetta - 2480 m
1.0 Sestriere Mt Banchetta - 2480 m
0.0 0.0
12 18 24 30 36 42 48 12 18 24 30 36 42 48
forecast time forecast time

Figure 3.1.16 Root mean square error in the temperature forecast with respect to the observations of four
regional network stations at different altitudes (one in the plains, one in the mid mountains, and two in
high mountains) Left: direct output of models; right: Kalman filter output. ecm12 t2m: temperature at 2 m
of the ECMWF IFS; ecm12 int.t: interpolated temperature of the ECMWF IFS; lkd00 t2m: temperature at 2 m
of the LM-DWD; lkd00 int.t: interpolated temperature of the LM-DWD; alm00 int.t: interpolated
temperature of aLMo.

suitably weighted models. Krishnamurti et al. (2000) also proposed the

Multimodel SuperEnsemble, which uses different weights for each model
considered. These weights are calculated for each model, in each point,
and at each time-step according to a certain previous period of time called
the training period. It is then hypothesized that the average systematic
error and the weight of each model are constant also for the following
forecast period.
In detail, the Multimodel Ensemble and SuperEnsemble algorithm
works in the following way:
1) the calculation of the forecast averages for each model and of the average
of the observations during the training period;
 Meteorological Forecasting 287

MODEL 1 MODEL 2 MODEL 3 …… MODEL N

(GRIB files) (GRIB files) (GRIB files) (GRIB files)

Interpolator Interpolator Interpolator Interpolator

…… Notes
(forecasting the (forecasting the (forecasting the (forecasting the
(text files)
measured data) measured data) measured data) measured data)

Calculation of weights during training period

Statistical values
Weights of
during training
each model
period

Multimodel Ensemble and SuperEnsemble application

Forecast Temperature
Wetness
on warning forecast in ………
forecast
areas alpine areas

Figure 3.1.17 Flow chart of the SuperEnsemble algorithm.

2) calculation of SuperEnsemble weights;

3) calculation of the forecast for the Ensemble and for the SuperEnsemble.
This technique was used for the temperature forecast at 2 m (Cane &
Milelli, 2005a) and its potential extension to other variables like relative
humidity and precipitation is currently being studied (Cane & Milelli,
2005b). It should be emphasized that the temperature forecasts of limited
area models contain serious errors in determining the daily cycle, espe-
cially in the mid mountains and in the high mountains, as noted in the
figure that follows.
Instead, notable improvements can be made at all altitudes with the
application of this technique (Figures 3.1.18, 3.1.19, and 3.1.20), with
results comparable to those obtained with the Kalman filter, but with the
advantage of being extendable to other variables. The Kalman filter, in
fact, is a complex technique and cannot be generalized to other parameters.
Furthermore, another disadvantage of the Kalman filter is that it cannot
take into account sudden variations of the meteorological conditions.
 288 Forecasting and Warning

Stations at altitude < 700 m

Figure 3.1.18 Comparison of January 2005 mean
6

Temperature (°C)
forecasted data from the
direct output of the model (in 4 Measured data
blue) and from the Multimodel 2
SuperEnsemble
SuperEnsemble (in red) with 0 model
observation data from the ⫺2 Model output
Arpa Piemonte network ⫺4
temperature for plains 12 18 24 30 36 42 48
stations (altitude < 700 m). Forecast time (h)

Figure 3.1.19 Comparison of Stations at altitude between 700 m and 1500 m

forecasted data from the January 2005 mean
Temperature (°C)

direct output of the model (in 5

blue) and from the Multimodel Measured data
SuperEnsemble (in red) with 0 SuperEnsemble
observation data from the model
⫺5
Arpa Piemonte network Model output
temperature for mid-mountain ⫺10
stations (700 m < 12 18 24 30 36 42 48
altitude < 1500 m). Forecast time (h)

Figure 3.1.20 Comparison of Stations at altitude >1500 m

forecasted data from the January 2005 mean
direct output of the model (in
Temperature (°C)

0
blue) and from the Multimodel Measured data
SuperEnsemble (in red) with
Super Ensemble
observation data from the ⫺5
model
Arpa Piemonte network
Model output
temperature for high ⫺10
mountain stations 12 18 24 30 36 42 48
(altitude > 1500 m). Forecast time (h)

3.1.5 THE WEATHER WATCH BULLETIN

The National System of Centri Funzionali promoted by the Department of

Civil Protection has proposed the realisation of a network of operational
centres for the national warning system with purposes of civil protection
that, through real time forecasting, monitoring, and surveillance of mete-
orological events and the consequential effects on the territory, would be a
support to the authorities in charge of warnings in making their decisions.
The Directive of the President of the Ministerial Council dated Febru-
ary 27th, 2004 (supp. G.U. n. 59 11/03/2004) modified by the Directive of
the President of the Ministerial Council (G.U. n. 55 8/03/2005) dictates the
operative guidelines for the management of this warning system. In par-
ticular, it defines the institutional subjects and territorial organs involved
and establishes the tools and modality of transmitting the information. The
national warning system therefore requires a forecasting phase with the
 Meteorological Forecasting 289

assessment of the expected meteorological situation as well as the effects

that such a situation may have on the ground, following by a monitor-
ing and surveillance phase. Based on this, the Department must issue a
national Weather Watch Bulletin to the public on a daily basis, as well
as a National Criticality Bulletin in a more reserved form. The individ-
ual meteorological services or the meteorological forecast areas of the
decentralised Centri Funzionali must instead issue, in a reserved form,
notification of adverse weather conditions and contextual criticality if nec-
essary in their own warning zone, indicated as significantly homogeneous
territorial environments for the expected occurrence of the typology and
severity of the intense precipitation and its relative impact. Indicators of the
probable occurrence of the prefigured scenarios as well as consequential
ground effects are identified for each zone and for each risk type.
Consequently, a set of values of the indicators that determines a thresh-
old system articulated on two levels of moderate and high criticality is
defined for each type of risk, as well as an ordinary situation base level
in which possible criticalities are believed to be acceptable by the popu-
lation. Following a Weather Advisory, the individual assessments of the
criticality level expected in the different warning areas of the regional
territory are given.
The Centro Funzionale Weather Watch Bulletin of the Piemonte
Region, which may be followed by a Regional Criticality Advisory, reports
the possibility of occurrence of the meteorological phenomena listed
below, over the following 60 hours (the afternoon and the two following
days), referred to the warning areas.

Thunderstorms, intense and localized phenomena

Highly localized, very intense precipitation over brief time periods (heavy
and very heavy showers or storms) is taken into consideration. These
phenomena may lead to the exceeding of critical hydrometric levels of

National Regional
Centro Funzionale Centro Funzionale
make

NATIONAL REGIONAL
METEOROLOGICAL METEOROLOGICAL
WATCH BULLETIN WATCH BULLETIN
issue

send

NATIONAL REGIONAL
CRITICALITY CRITICALITY
BULLETIN BULLETIN

Regional Civil
Protection

Territorial Government Regional Offices Provincial Offices

Figure 3.1.21 Flow chart of Offices
the information.
 290 Forecasting and Warning

Figure 3.1.22 Example of the

Weather Watch Bulletin.

the smaller drainage network (which have a torrential nature) with the
development of flooding in urban areas due to inadequate drainage, or of
mass transport and landslides localized in mountainous or hilly sectors.
Storms may also include strong wind gusts, especially in correspondence
to the most active cells, with consequential inconvenience for human
activities and, in particular, damage to vegetation, light poles, and roofs.
Furthermore, it is possible to have intense storms associated with hail,
which causes damages to crops, traffic systems, and people. In consider-
ation of the tremendous impact that storms have on the territory, specific
studies have been carried out in order to characterise the qualitative and
quantitative aspects of thunderstorms, with the aim of creating an oper-
ational instrument for warnings in case of significant convective events
capable of predicting the occurrence of a storm event in probabilistic or
deterministic terms (with a limited but known margin of error). In par-
ticular, a study of the most common thermodynamic indices (K, CAPE,
Lift, SWEAT) was carried out as predictor of storm events, beginning
with ECMWF analysis and observed lightning and precipitation data.
This analysis was carried out on several case studies and over several
time periods in order to verify the reliability of the indices and to identify
the thresholds that are representative on a local level. The verification of
the capability of LAMI in predicting storms through the more signifi-
cant indices obtained with the previously described analysis is currently
ongoing.

Rain
The precipitation is computed for each warning areas as the average value
of precipitation forecasted over a 12 or 24-hour interval. The quantities
of precipitation exceeding the reference thresholds are identified. The
thresholds are differentiated for each warning area and are represented by
a fraction of the moderate criticality threshold.
 Meteorological Forecasting 291

Exceeding these thresholds may lead to the exceeding of critical hydro-

metric levels along the main rivers with a consequential development of
flooding, mass transport along the rapid flow, and landslides in moun-
tainous and hilly sectors. For this reason, the most important field is
clearly the value of precipitation forecasted by the models. In particu-
lar, the global circulation models provide a better indication of average
values and are more reliable for long term forecasts, while more intense
quantities of precipitation are better forecasted by limited area models,
even if the maximum forecasted values are often is very different from
those observed, in terms of space-time localization and in quantity, as
demonstrated by the example in Figures 3.1.23 and 3.1.25, during the
convective event of September 15th, 2004.
For these reasons the forecaster must not directly use the field fore-
casted by the models, but must also elaborate (and filter) the information
according to a series of preliminary considerations such as:
• a comparison of the synoptic configuration forecasted by different
models paying attention to the flow over the region at all levels
(consistency);

Figure 3.1.23 Field of

observed precipitation.

Figure 3.1.24 Field of

precipitation forecasted by
the ECMWF (UK) global
model.
 292 Forecasting and Warning

Figure 3.1.25 Field of

precipitation forecasted by
the LAMI (Italy) model.

• assessment of the shift between the first forecast times (+12/+6 h) and
the measurements (satellite images and data observed on a synoptic
scale);
• comparison of the configuration forecasted by the same model with dif-
ferent initialization time and with forecasts of the previous day (internal
consistency);
• assessment of seasonal characteristics;
• assessment of the performance of models during previous days
(validation);
• comparison with previous reference situations (in case of doubts and/or
critics).

The analysis of the meteorological situation and of its probable evo-

lution is therefore defined in great detail. The regional area is taken
specifically into consideration and the attention is focused on each indi-
vidual expiration dates to which the forecast is extended (currently the
second day following the day of issue is reached) and on some specific
aspects:
• the presence of instability (convective and frontal);
• interaction of the flow with the orography (induction, modification,
intensification, control of precipitation);
• localisation of pressure lows (ground/altitude).
Only at the end of the evolution process is the meteorologist capable
of producing a reasonable quantitative forecast of precipitation over the
region.
Other important information associated with the intensity of precipi-
tation is the forecast of the level above which the precipitation becomes
snow. The forecasting of this quota is based firstly on the forecast of the
freezing level in the atmosphere: provided that that a falling snowflake
can remain in a solid state until about 2−3◦ C, the level of the snow alti-
tude is determined as a consequence, and is normally situated at about
400–500 m below the freezing level. However, the thermal stratification
 Meteorological Forecasting 293

of the atmosphere can be very different from the standard thermal pro-
file, which is equal to about 6◦ C/1,000 m. Recent studies revealed that
in almost isothermal conditions of temperature profile or with weakly
reducing temperature with altitude, or even in the presence of thermal
inversions, the difference between freezing and snow level can reach as
much as 1,500 m. Vice-versa, if the atmosphere presents a highly unsta-
ble stratification with a marked reduction in temperature with altitude,
the snow level can be only 100–200 m below freezing level. For the
forecast of the thermal stratification, thermodynamic soundings are pro-
duced at all levels by the meteorological limited area models in Torino,
Cuneo, Alessandria, Novara, Verbania, Susa, Cesana, Domodossola, and
Aosta. Another useful indication is provided by the examination of the
latest soundings observed at Milano Linate and Cuneo Levaldigi and
by their comparison with the soundings forecasted by the latest run of
the model.
For the various localities, and in particular for those in the plains, it
must be remembered that the temperature values at 2 m, which are also
fundamental in identifying the type of precipitation forecasted for the
locality, are subject to variations caused by radiative effects during the
day. Therefore, in order to obtain the most correct temperature forecast
possible for individual localities, it is necessary to correct systematic
errors and minimize casual errors by applying suitable statistical filters,
which have been previously elaborated by meteorological models, to the
direct output of temperature variable.

Snowfall
The forecasted snowfall intensity is examined in terms of classes. The
reference thresholds vary according to the altitude (lower or higher than
700 m) and according to the time interval considered (12 or 24 hours).
The exceeding of the established thresholds results in serious difficulties
in carrying out fundamental human activities like public and private trans-
portation. In particular, various problems may arise, including mobility
drawbacks due to slowdowns and snow-removal operations, the inter-
ruption in the supply of services (electricity, telephone, etc.) caused by
damage to distribution structures overloaded with snow, and the tempo-
rary isolation of hamlets and individual homes. Furthermore, persistent
intense snowfall may cause the structural collapse of the roofs of civil and
industrial buildings.

Frosts in the plains

Early and late frosts (before December and after February) will be sig-
nalled with particular reference to their impact on agriculture (with
possible crop damage) and traffic circulation (difficulty in travelling along
all main and secondary roadways). The Arpa Piemonte Regional Fore-
casting and Environmental Monitoring Area implemented and developed
294 Forecasting and Warning

a model that offers a 24-hour numerical forecast of temperature and road

surface conditions, beginning with the observed air temperature values,
dew point, relative humidity, cloudiness, wind speed, and precipitation.

Wind
Notification will be given regarding conditions in the mountains that may
cause inconveniences for human activities at high altitudes and in the
plains, especially in the case of foehn with possible local damages to
vegetation and buildings. Reference thresholds are defined according to
the altitude (1,500 or 3,000 m). The criticality is linked to potential damage
of temporary structures, problematic road and traffic conditions, potential
collapse of pavilions, inconveniences for human activities at high altitudes,
and flight safety. The highest criticality regards the forecast of katabatic
winds (foehn), which are strongly influenced by the complex orography
of the region. Studies are underway to empirically develop a foehn index
that takes into account the parameters of wind speed, surface pressure,
humidity, and temperature.
For this kind of forecast, an excellent support is the direct output of the
wind at the various altitudes supplied by the models, especially limited
area ones that have a high-resolution representation of the orography and
are therefore more realistic. Another indication is supplied by examining
the presence of strong baric gradients (higher than 4 hPa, responsible for
intense wind) and thermal gradients (4.5◦ C at 925 and 850 hPa and 6◦ C
at 700 hPa, which imply a stable atmosphere downwind) between the Po
River Valley and the Alps.

Fog
Low visibility will be signalled with particular reference to problems
related to road and traffic conditions (for both main and secondary roads)
and flight safety (airports). Currently this parameter is estimated mainly
according to the visibility and humidity data observed on the ground.
From a forecasting point of view, it is important to take into consid-
eration that fog situations normally occur under anticyclonic conditions
with the presence or advection of humidity in the lower levels of the
atmosphere. The examination of the situation, together with the analy-
sis of the forecasted soundings to determine the development of thermal
inversions, supplies useful indications. In the case of persistent fog,
the METEOSAT geostationary satellite images of the visible band and
the analysis of the observed soundings offer considerable support in
identifying the phenomena and in forecasting its short-term evolution. The
forecasting of postfrontal fogs is more difficult than the one for radiation
fogs since a precise forecast of clearing times and wind arrival is needed;
delays or anticipations of even a few hours may determine radically dif-
ferent situations. Even in this case, valid help is offered by the detailed
examination of the atmosphere in the lower layers through the planned
soundings.
 Meteorological Forecasting 295

A study is underway to develop an initial definition of a statistical index

for forecasting fogs based on the correlation between relative humidity,
wind speed, temperature, presence of a thermal inversion in lower layers as
in the LAMI model and corresponding visibility values. Moreover, initial
contacts have been made with the University of Bonn for the development
of a new model of cloud microphysics to be integrated into the operational
meteorological model, even if the direct model forecast of this variable
is very costly from a point of view of calculation time and therefore not
very useful.

Heat waves in the plains

This is where the assessments carried out in the Heat Wave Forecast Bul-
letin for provincial capitals are reported. This variable takes into account
the discomfort suffered by the population as a consequence of prolonged
heat with elevated humidity and no ventilation.
Meteorological conditions of extreme heat and their impact on health
constitute a very common problem in Italy, especially after the events of the
summer 2003, during which northern Italy was subjected to long periods
characterised by high temperatures and humidity with values well above
seasonal averages during both daytime and night-time. Epidemiologi-
cal studies indicate that these conditions cause physiological discomfort
among the population: the subjects mostly affected are persons suffer-
ing from cardiorespiratory diseases and elderly people who have reduced
thermoregulatory capacities.
The possibility of forecasting meteorological variations for brief peri-
ods with their potential impact on human health allows the planning and
management of prevention interventions and the reduction of damage. In
order to provide the city of Torino and the region of Piemonte with a spe-
cific warning system for preventing the effects of heat waves on health,
the Environmental Forecasting and Monitoring Area, in collaboration with
the Environmental Epidemiology Service, has activated a project to anal-
yse and study historical climatological and epidemiological data for the
purpose of realizing a forecasting model of adverse meteorological con-
ditions represented by the presence of heat waves and, in particular, for
the urban area of Torino aimed at estimating the impact of heat waves on
mortality.
The meteorological parameters used by the model are: maximum
temperature, apparent maximum and minimum temperature (Steadman,
1984), number of consecutive days with high apparent temperatures,
ozone levels, and other meteorological parameters necessary for the
calculation of some public health indices. The bio-meteorological thresh-
olds of these indices are chosen in function of the local climatic conditions
and criteria of specificity and sensitivity that do not derive directly from
literature values . The perception of the discomfort due to meteorological
conditions is peculiar to the region and to the local area. For this reason
it was decided to use a relative index, the Heat Stress Index (Kalkstein,
296 Forecasting and Warning

Calculation of meteorological indexes:

Issue of Metorological and - Heat Stress Index (HSI) Statistical estimate of the
Ozone levels forecast - Apparent Temperature (HI) number of health care
within 72 hours - Humidex Index (hum) events in excess (HCE)
- Discomfort Index (DI)

Yes Yes Yes

HSI ≥ 7 or HSI ≥ 9 or HCE ≥ 15
HCE ≥ 3 HCE ≥ 5

No No
No

No Yes
HI ≥ 27 and HCE ≥ 5
No hum ≥ 30 and Yes
DI ≥ 24

No HI ≥ 32 and HI ≥ 32 and
Yes No
hum ≥ 35 and hum ≥ 35 and
DI ≥ 27 DI ≥ 27

Yes
NO ALARM
ATTENTION
ALARM

No
Alarm for 2
days or more

EMERGENCY Yes

Figure 3.1.26 Decision tree of the risk levels.

2003). The Heat Stress Index is a bio-meteorological index that assesses

the physiological response of the population to the meteorological vari-
ables according to the localisation and the seasonal period. The assessment
is carried out through the distribution analysis of the probability of the
meteorological variable measured in the past.
This development resulted in a service capable of providing different
types of forecast information:
• meteorological: type of heat (according to the forecasted tempera-
tures and the expected relative humidity), maximum temperature, daily
average humidity, temperature perceived (maximum and minimum),
maximum daily ozone level, and synthetic index of stress caused by
heat;
 NO ALARM
0) No critical meteorological 5) Meteorological conditions
ALARM conditions and no health associated to HSI >= 9
care events in excess and concise bio-meteo
2
indexes in “high caution”
0 class

ATTENTION 1a) 15% < HCE <= 30%, ALARM 6a) 15% < HCE <= 100%,
meteorological conditions meteorological conditions
associated to HSI < 7 and associated to HSI >= 7 and
concise bio-meteorological concise bio-meteorological
1 2
indexes in “caution” class indexes in “high caution”
class
1b) 30% < HCE <= 100%,
meteorological conditions 6b) HCE > 100%,
associated to HSI < 7 and meteorological conditions
concise bio-meteorological associated to HSI >= 7 and
indexes not in “high concise bio-meteo indexes
caution” class in “caution” class

1c) 30% < HCE <= 100%,

meteorological conditions
associated to HSI >= 7 and
concise bio-meteorological
indexes in “wellbeing
class”

2) Meteorological conditions 7) Risk level equal to “Alarm”

ATTENTION associated to HSI >= 7 EMERGENCY for number of health care
and concise bio-meteo events in excess and
indexes in “caution” class persistence of risk level
1 3 for 2 days or more

3) 15% < HCE <= 100%, 8) Risk level equal to “Alarm”

ATTENTION meteorological conditions EMERGENCY for critical meteorological
associated to HSI >= 7 conditions and persistence
and concise bio-meteo of risk level for 2 days or
1 indexes in “caution” class 3 more

4a) HCE > 100%, 9) Risk level equal to “Alarm”

ALARM meteorological conditions EMERGENCY for number of health care
associated to HSI < 7 events in excess and
critical meteorological
4b) HCE > 100%, conditions and persistence
2 3 of risk level for 2 days or
meteorological conditions
associated to HSI >= 7 and more
concise bio-meteorological
indexes in “wellbeing
class”

4c) 30% < HCE <= 100%,

meteorological conditions
associated to HSI < 7 and
concise bio-meteorological
indexes in “high caution”
class

Figure 3.1.27 Legend of the risk levels.

298 Forecasting and Warning

• public health: the number of excess health care interventions;

• level of risk: no alarm, attention, alarm, or emergency.
Schemes for post-elaboration through statistical correlations among
the parameters forecasted by the LAMI limited area model and ground
parameters are used to provide a more accurate forecast of temperature,
humidity, and wind parameters, while predictors and predictants of the
previous 2–3 weeks prior to the forecast were used to take into account
for the extreme atmospheric variability. The apparent or perceived tem-
perature of Steadman (1984) is calculated with the forecasted values: an
index that allows the estimation of physiological discomfort due to expo-
sure to meteorological conditions characterised by high temperatures and
high hygroscopic levels, as well as reduced surface wind speed.
The forecast model of the heat waves used the daily perceived tem-
perature maximums and minimums as input variables. Epidemiological
studies in fact verified that the high temperature and humidity during
the night time do not allow the human body to physiologically recover
from the discomfort accumulated during the day time. Consecutive days
of extreme heat are observed with the maximum perceived temperature
exceeds the average climatological value for the period of 1990–2003 by
a standard deviation with no solution of continuity. Other input data con-
sist of the maximum daily concentration of ozone forecasted by a neural
network statistical model.
The synthetic index for heat stress used is Heat Stress Index (HSI),
a relative bio-meteorological index that is calculated following the pro-
cedure indicated by Kalkstein (2003) and using the data of a measuring
station located in the urban centre of the city of Torino.
The variable derived from these data and used in the HSI elaboration
are: the apparent temperature, the degree of daily cooling, the number of
consecutive days of extreme heat, and the cloud covering. From the anal-
ysis of climatological data measured over the past 14 years and relative
to the decade centred on the considered day, the statistical distribution
of each variable and the corresponding curve of cumulative probability
is determined. For each value assumed by the variable , the correspond-
ing percentile is then calculated. The percentile value of each value is
them summed, giving the SUM dimension. The HSI is nothing but the
percentile value associated with the position of the daily SUM value on
the relative cumulative probability curve. For its forecast, in addition to
the meteorological data previously described, the values of cloud cover
forecasted through the LAMI are used, after being verified subjectively.
The associated epidemiological model is a multivariate model that esti-
mates the number of health care events in excess (the health care events
are intended as the daily number of deaths of people above 64 years old.
The excess being defined as the difference between the number of events
expected and the number estimated from the theoretical model; the daily
events expected – an average of 15 – are calculated according to the histori-
cal series of May–September 1990–2002, with non-parametric regression
 Meteorological Forecasting 299

methods along the axes of the years, over about 5 days and, successively,
along summer periods of 153 days).
Multivariate regression methods led to the selection of cases observed
in relationship to expectations: HIS of the day, maximum apparent tem-
perature recorded or forecasted during the two previous days, consecutive
days of extreme heat, maximum daily ozone level of the day and the day
before. The multivariate model then suggested parameters to be inserted
into the estimation algorithm for the number of cases in excess, around
which four possible classes were defined: none, low, average, high. It is
important to observe how the model takes into account the input variable
relative to the day before the forecast date: a time interval between severe
meteorological conditions and the consequential excess in the mortality
rate was frequently observed. A complex decision tree synthesises the
expected level of risk (0-1-2-3) by joining the meteorological and health
care forecast information, which may derive from critical meteorologi-
cal conditions, from an excess of health care events, or from both. The
objective of this bulletin is to supply the local health care service with
the information needed to organize medical intervention and prevention
in case of forecasted heat waves.

BIBLIOGRAPHY

The works marked by an asterisk have not been quoted explicitly in the text.

*Bluestein H.B. (1992), Synoptic-Dynamic Meteorology in Midlatitudes. Oxford

University Press, Oxford.
Cane D. & Milelli M. (2005a), A new method for T2m forecast in complex
orography areas. COSMO Newsletter, submitted.
Cane D. & Milelli M., (2005b), Use of Multimodel SuperEnsemble Technique for
Mountain-area weather forecast in the Olympic Area of Torino 2006. Croatian
Meteorological Journal 40.
Ebert E.E. & McBride J.L. (1997), Methods for verifying quantitative precipi-
tation forecasts: Application to the BMRC LAPS model 24-hour precipitation
forecasts. BMRC Techniques Development Report n. 2, p. 87.
Ebert E.E. & McBride J.L. (2000), Verification of precipitation in weather
systems: Determination of systematic errors. J. Hydrology, vol. 239, pp.
179–202.
Hamill T.M. (1999), Hypothesis tests for evaluating numerical precipitation
forecasts. Wea. Forecasting, vol. 14, pp. 155–167.
Kalkstein L.S. & Watts J. (2003), The Development of a Warm Weather Rel-
ative Stress Index for Environmental Applications. University of Delaware,
Newark.
Kalman R.E. (1960), A new approach to linear filtering and prediction problems.
Trans. ASME. J. Basic Engrg, Series D, vol. 82, pp. 35–45.
Kalman R.E. & Bucy R.S. (1961), New results in linear filtering and prediction
problems. Trans. ASME. J. Basic Engrg, Series D, vol. 83, pp. 95–108.
300 Forecasting and Warning

Krishnamurti T.N., Kishtawal C.M., LaRow T.E., Bachiochi D.R., Zhanh Z.,
Williford E., Gadgil S. & Surendran S. (1999), Improved Weather and Sea-
sonal Climate Forecasts from Multimodel Superensemble Science, vol. 285,
pp. 1548–1550.
Krishnamurti T.N., Kishtawal C.M., Zhang Z., LaRow T., Bachiochi D., Williford
E., Gadgil S. & Sajani Surendran (2000), Multimodel Ensemble Forecasts for
Weather and Seasonal Climate, J. Climate, vol. 13, pp. 4196–4216.
Loglisci N., Cassardo C., Balsamo G.P. & Qian M.W. (2001), A technical descrip-
tion of the Land Surface Process Model (LSPM version 2000). University of
Turin DFG Report.
Lorenz E.N. (1963), Deterministic non-periodic flow. J. Atmos. Sci., vol. 20,
pp. 130–141.
Milelli M., Oberto E., Bertolotto P. & Pelosini R. (2003), Verification of QPF
over Piedmont and Northern Italy using high-resolution non-GTS data. Proc.
of Ecam, 15–19 September, Rome.
*Molteni F., Buizza R., Palmer T.N. & Petroliagis T. (1996), The new ECMWF
ensemble prediction system: methodology and validation. Q.J.R. Meteorol.
Soc., vol. 122, pp. 73–119.
Oberto E., Milelli M., Bertolotto P. & Pelosini R. (2003), High Resolution QPF
Verification of LM Using Non-GTS Data over Piedmont and Central-Northern
Italy. COSMO NEWSLETTER n. 3.
Orlansky I. (1975), A rational subdivision of scales for atmospheric processes.
Bull. Amer. Meteor. Soc., vol. 56, pp. 527–530. NP; Unterteilung in macro-,
meso-, mikro-Skala.
Steadman R.G. (1984), A universal expression of apparent temperature. J. Appl.
Meteor., vol. 23, pp. 1674–87.
Stanski H.R., Wilson L.J. & Burrows W.R. (1989), Survey of common verification
methods in meteorology. World Weather Watch Tech. Rept. n. 8, WMO/TD
n. 358, WMO, Geneva, p. 114.
*Trenberth K. (1993) (ed.), Climate System Modelling. Cambridge University
Press, p. 818.
Turco M., Oberto E. & Bertolotto P. (2005), Progresses on LAMI, LM-DWD, aLMo
verification over Northern Italy. Submitted to COSMO NEWSLETTER n. 5.
Wilks D.S. (1995), Statistical Methods in the Atmospheric Sciences. An Introduc-
tion. Academic Press, San Diego, p. 467.
Wu J. (1982), Wind-stress coefficients over sea surface from breeze to hurricane.
J. Geophy. Res. C87, pp. 9704–9706.