See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/227862602

Effects of Afforestation on Water Yield: A Global Synthesis With Implications

for Policy

Article  in  Global Change Biology · October 2005

DOI: 10.1111/j.1365-2486.2005.01011.x

CITATIONS READS

1,023 3,361

3 authors, including:

Kathleen Farley Wolf Esteban Jobbágy

King County National Scientific and Technical Research Council
29 PUBLICATIONS   4,333 CITATIONS    317 PUBLICATIONS   24,974 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

“Flatness, Flooding and Farming (F3 ): Adapting to climatic and hydrological changes in the plains of Argentina and Paraguay” View project

PICT 3067- Expansión agrícola e inundaciones en la llanura Chaco-Pampeana: Bases para su comprensión, monitoreo y manejo View project

All content following this page was uploaded by Kathleen Farley Wolf on 19 December 2017.

The user has requested enhancement of the downloaded file.

Global Change Biology (2005) 11, 1565–1576, doi: 10.1111/j.1365-2486.2005.01011.x

Effects of afforestation on water yield: a global synthesis

with implications for policy
K A T H L E E N A . F A R L E Y *w, E S T E B A N G . J O B B Á G Y w z and R O B E R T B . J A C K S O N *w
*Center on Global Change, Duke University, Durham, NC 27708, USA, wDepartment of Biology and Nicholas School of the
Environment and Earth Sciences, Duke University, Durham, NC 27708, USA, zGrupo de Estudios Ambientales – IMASL,
Universidad Nacional de San Luis & CONICET, San Luis 5700, Argentina

Abstract
Carbon sequestration programs, including afforestation and reforestation, are gaining
attention globally and will alter many ecosystem processes, including water yield. Some
previous analyses have addressed deforestation and water yield, while the effects of
afforestation on water yield have been considered for some regions. However, to our
knowledge no systematic global analysis of the effects of afforestation on water yield
has been undertaken. To assess and predict these effects globally, we analyzed 26
catchment data sets with 504 observations, including annual runoff and low flow. We
examined changes in the context of several variables, including original vegetation type,
plantation species, plantation age, and mean annual precipitation (MAP). All of these
variables should be useful for understanding and modeling the effects of afforestation
on water yield. We found that annual runoff was reduced on average by 44% (
3%) and
31% (
2%) when grasslands and shrublands were afforested, respectively. Eucalypts
had a larger impact than other tree species in afforested grasslands (P 5 0.002), reducing
runoff (90) by 75% (
10%), compared with a 40% (
3%) average decrease with pines.
Runoff losses increased significantly with plantation age for at least 20 years after
planting, whether expressed as absolute changes (mm) or as a proportion of predicted
runoff (%) (Po0.001). For grasslands, absolute reductions in annual runoff were greatest
at wetter sites, but proportional reductions were significantly larger in drier sites
(Po0.01 and Po0.001, respectively). Afforestation effects on low flow were similar to
those on total annual flow, but proportional reductions were even larger for low flow
(Po0.001). These results clearly demonstrate that reductions in runoff can be expected
following afforestation of grasslands and shrublands and may be most severe in drier
regions. Our results suggest that, in a region where natural runoff is less than 10% of
MAP, afforestation should result in a complete loss of runoff; where natural runoff is
30% of precipitation, it will likely be cut by half or more when trees are planted. The
possibility that afforestation could cause or intensify water shortages in many locations
is a tradeoff that should be explicitly addressed in carbon sequestration programs.
Keywords: afforestation, land-use change, plantation, runoff, water yield

Received 21 December 2004; accepted 15 March 2005

market incentives to reduce atmospheric CO2 concen-

Introduction
trations. Through the clean development mechanism
The conversion of natural grasslands to plantations has (CDM), the Kyoto Protocol allows for developed
occurred over extensive areas of the southern hemi- countries to offset part of their CO2 emissions by
sphere and will likely continue with new policy and establishing carbon-sequestering projects, including
reforestation and afforestation. Afforestation has been
suggested as a way to simultaneously sequester carbon,
Correspondence: Kathleen A. Farley, Duke University, Box 91000, increase wood and paper supplies, and diversify rural
Durham, NC 27708, USA, tel. 1 1 831 241 1236; incomes (Vertessy, 2001). Not surprisingly, the focus of
fax 1 1 831 333 1736, e-mail: farley@duke.edu much of the research on this land-use change has been

r 2005 Blackwell Publishing Ltd 1565

1566 K A T H L E E N A . F A R L E Y et al.

on sequestering and storing carbon in the biomass and Although the effects of plantation age and rotation
soils of afforested areas. However, converting grass- length are important for predicting the consequences of
lands or shrublands to plantations will likely affect afforestation on water yield, these effects are lacking in
many other ecosystem processes, including water yield most studies (Best et al., 2003). A better understanding
from rivers and streams (e.g. Duncan, 1995; Dye, 1996; of the age–runoff relationship after afforestation will
Bashkin & Binkley, 1998; Paul et al., 2002; Jobbágy & allow managers to make predictions using more
Jackson, 2003, 2004; Farley et al., 2004). realistic rotation scenarios – in which a proportion of
Water yield is altered through changes in transpira- the landscape is in early growth stages, and full aging is
tion, interception, and evaporation, all of which tend to prevented by harvesting. In addition, the effect of
increase when grasslands or shrublands are replaced afforestation on low flow is an important component of
with trees. Transpiration rates are influenced by this framework. Changes in low flow may be even more
changes in rooting characteristics, leaf area, stomatal important than changes in annual flow, as the dry
response, plant surface albedo, and turbulence (Brooks season is when reduced water supply will have the
et al., 1997; Hoffmann & Jackson, 2000; Jackson et al., most severe effects for users, particularly in arid and
2001; Vertessy, 2001). Although transpiration is tradi- semiarid regions (Smith & Scott, 1992; Scott & Smith,
tionally considered the more important component of 1997; Sharda et al., 1998; Robinson et al., 2003).
forest evapotranspiration (ET), interception and subse- In this paper, we quantified the change in streamflow
quent evaporation from the canopy can also increase associated with afforestation globally. Our specific
substantially, particularly with conifers (Pearce & Rowe, objectives were to: (1) assess the direction, range, and
1979; Cannell, 1999). Evaporation of intercepted pre- extent of changes in total annual streamflow and low
cipitation is generally low in grasslands, but can flow associated with afforestation, (2) examine the
account for 10–20% of rainfall for broadleaf trees and interactions with original vegetation type, tree species
20–40% for conifers (Le Maitre et al., 1999). The sum of planted, plantation age, and climate, and (3) provide a
the changes in evaporation and transpiration in planta- predictive framework for modeling the effects of
tion catchments leads to an increase in ET (Holmes & afforestation on water yield for carbon sequestration
Sinclair, 1986); for example, ET from a catchment scenarios. To accomplish these objectives, we analyzed
planted with eucalyptus could be 40–250 mm higher 26 catchment data sets containing 504 annual observa-
than from a grassland catchment (Zhang et al., 1999). tions to assess the effects of afforestation on water yield.
Despite recognition of higher ET rates in plantations, These catchment studies included sites that were
the likelihood that this will reduce water yield has not converted from grassland, pasture, or shrubland to
always been acknowledged (Vertessy & Bessard, 1999), pines, eucalypts, or other species (primarily spruce).
particularly within the context of afforestation pro-
grams for carbon sequestration.
Methods
Studies of the effect of vegetation change on water
yield have focused primarily on the removal of woody
Data synthesis
vegetation (e.g. Bosch & Hewlett, 1982; although see
Scott et al., 2000). Using results from deforestation We compiled catchment data sets from peer-reviewed
studies to predict the effects of afforestation may be journals as well as reports from governmental and
problematic because they are not necessarily opposite nongovernmental research institutes, representing
and reversible processes (Robinson et al., 1991). The many parts of the southern hemisphere, as well as
changes in runoff induced by deforestation and India, the UK, and Germany in the northern hemi-
afforestation likely differ in magnitude, timing, and sphere (Appendix A). We examined data from affor-
relationship to site characteristics. Deforestation studies ested regions with a previous land cover of grassland or
are distinguished by factors such as soil disturbance shrubland where runoff was measured following
and deposition of slash and litter, which can affect planting, and included all the data sets we found with
streamflow patterns (Vertessy, 1999). The duration of these characteristics. Most of the data were from paired
most deforestation and afforestation studies is also catchment studies, in which streamflow from grassland
vastly different, and the short time period of the former or shrubland catchments was compared with that of
increases the chance that the effect of rainfall variability nearby afforested catchments.
will be difficult to separate from the catchment response We examined the data set for several variables. Most
(Vertessy, 1999). In addition, the timing of changes in studies reported changes in runoff for several years
runoff may differ significantly, with abrupt changes after afforestation, with some beginning at age 0 and
associated with deforestation and more gradual others beginning later in the rotation. The number of
changes with plantation age following afforestation. years of runoff data varied with each study, with some

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576

 E F F E C T S O F A F F O R E S TAT I O N O N WAT E R Y I E L D 1567

studies covering less than a decade and others as much control catchment minus runoff in the planted catchment.
as four decades (Appendix A). For afforested catch- We used the data as the authors presented them.
ments that were harvested after the full rotation length, In the results, we refer to changes in runoff as
we included only the data up to the time of harvesting. absolute changes (mm) and proportional changes (% of
In cases where data were available from more than one predicted or control runoff). Because changes in runoff
source for the same catchment, we used multiple vary from year to year with variations in rainfall,
sources if the information did not overlap (such as expressing changes as a proportion of expected flow is
covering different time periods or reporting different useful as a way to remove this climatic variability (Scott
types of flow data, such as annual vs. low flow). Where & Smith, 1997). Not all data sets provided both absolute
they reported the same data for the same time period, and proportional runoff values, so some of the data
we chose those that covered the longest time period points included in one analysis are absent in the other.
and, in some cases, used the additional data sets for In addition to the change in annual runoff, a number
supplementary information, such as area planted or of studies also reported change in low or base flow.
calibration period. In no cases were duplicate data from Low flow was typically defined in the studies as the
a single catchment used in the analyses. Our database flow rate during the driest 3–4 months of the year, or as
consisted of runoff data for each year reported for a dry weather summer flow (Smith & Scott, 1992; Scott &
given catchment; for example, where a study included Smith, 1997; Sharda et al., 1998; Robinson et al., 2003). In
data for plantation ages 1–8, we used each of the 8 years some cases, it was defined more precisely by using an
as a data point in our analyses. exceedance level (the flow exceeded for a certain
The percent of the catchment afforested varied among percent of the year, generally ranging from 75% to
data sets (Appendix A). In more than three-fourths of 95%) as a threshold (Fahey & Watson, 1991; Scott &
the cases, half or more of the catchment was planted, Smith, 1997; Robinson et al., 2003).
although in three cases it was only 20–40%, and in
several cases it was not reported. We based our analyses
Statistical analyses
on the original data sets, uncorrected for the proportion
of the catchment afforested, which means that our The effect of original (pre-afforestation) vegetation type,
estimates are conservative, as we are likely under- plantation species, plantation age, and mean annual
estimating the magnitude of the effects. An alternative precipitation (MAP) on the change in runoff were
used by some researchers is to scale the catchment tested using one-way ANOVAs followed by Tukey’s HSD
results by the percentage afforested. To satisfy those post hoc tests; where conditions of normality and
researchers, we performed a parallel analysis, scaling homogeneity of variance were not met, nonparametric
all the data to a minimum area planted (75%). For this Kruskal–Wallis tests were used as noted. In each case,
analysis, all catchments in which less than 75% of the the dependent variable was either the proportional
area was planted had the data scaled linearly up to 75% change in runoff (%) or the absolute change in runoff
– because it is not typical practice in forestry to plant (mm) following afforestation. The factors evaluated
100% of a catchment (Scott & Smith, 1997). The data included original vegetation type (grassland or shrub-
from the catchments for which we lacked information land), plantation species (pines, eucalypts, or other
on the area planted were included without scaling. The species), plantation age class (using 5-year intervals), or
figures and discussion are based on the unscaled data, MAP (o1000, 1000–1250, 1250–1500, 41500 mm). For
but we have included an overview of the analyses using the analysis of the relationship between change in
the scaled data in the results section. runoff and plantation age, linear, logarithmic, and
The studies included in the data set used one of two quadratic regressions were compared and the curve
general approaches to calculate the change in runoff with the best fit, based on adjusted least-squares
following afforestation. In  60% of the data sets, the regression, was selected. Because we knew, a priori,
change in runoff was reported as predicted runoff minus that there should be no change in runoff at age zero, we
observed runoff. The approach taken in each of these used regressions through the origin (Zar, 1999). This
studies to calculate predicted runoff was based on a alters the definition of the r2 from the more typical r2 of
calibration of runoff between the control and planted a regression that is not forced through the origin.
catchments before afforestation. Predicted runoff for a
given year was calculated based on runoff from the
Results
control catchment in that year and the relationship
between the control and planted catchments derived from Runoff decreased consistently and substantially with
the calibration period. In the remaining  40% of data afforestation across the entire data set (Fig. 1, Po0.001).
sets, the change in runoff was calculated as runoff in the More than one-fifth of the catchments experienced

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576

1568 K A T H L E E N A . F A R L E Y et al.

Table 1 Mean change in runoff (
SE) following afforestation, by original vegetation type and by planted vegetation type,
averaged across plantations  30 years old

Afforested Afforested to Change in Catchment Change in Catchment MAP (mm) Drunoff (mm)/
from runoff (%) n runoff (mm) n MAP (mm) (%)

Grassland Any species 44 (
3)** 13 170 (
13)ns 11 1241 (
16) 15 (
0.9)
Shrubland 31 (
2)** 8 162 (
8)ns 8 1262 (
10) 14 (
0.6)
Grassland or Pines 35 (
2)ns 14 165 (
8)ns 14 1236 (
10) 14 (
0.5)
shrubland Eucalypts 50 (
5)ns 4 173 (
20)ns 4 1336 (
23) 14 (
1.7)
Other species 39 (
7)ns 3 1415 (
33)
Grassland only Pines 40 (
3)* 9 167 (
13)ns 9 1260 (
18) 14 (
0.9)
Eucalypts 75 (
10)* 1 202 (
38)ns 1 1166 (
0) 19 (
3.2)
Other species 39 (
7)* 3 1415 (
33)
Shrubland only Pines 30 (
2)ns 5 163 (
9)ns 5 1226 (
9) 15 (
0.6)
Eucalypts 38 (
5)ns 3 159 (
23)ns 3 1414 (
24) 12 (
1.9)

In order to make differences between grasslands and shrublands and among plantation species comparable, all plantations  30
years old were included. Drunoff/MAP 5 change in runoff (mm)/mean annual precipitation (mm)  100. Catchment n 5 the
number of catchments represented in each category. Significance symbols refer only to comparisons of mean change in runoff among
the groups within a category (i.e. grassland vs. shrubland pooled across all tree species; pines vs. eucalypts vs. other species pooled
across grasslands and shrublands; and pines vs. eucalypts vs. other species compared within either grasslands or shrublands).
**Po0.001, *Po0.05, ns, not significant. Significance was determined using Kruskal–Wallis tests.

reductions of 75% or more during at least 1 year and above-average rainfall, after which runoff losses again
13% of the catchments experienced 100% runoff became more severe (Scott et al., 2000). In contrast,
reductions for at least 1 year (Fig. 1a, c). Both the shrublands showed a distinct recovery in runoff after
original vegetation type at a site and plantation species approximately 35 years of afforestation, both in
significantly influenced proportional changes in proportional and absolute amounts (Fig. 1c, d). Because
streamflow (Table 1, Po0.001 and Po0.05, respec- eucalyptus rotations are shorter than 35 years, none of
tively). When averaged across ages, annual runoff the eucalyptus sites in the database extended to the age
reductions were greater in grasslands (44
3%) than at which this recovery occurred. However, the data
shrublands (31
2%) (Table 1, Po0.001). Eucalypts had from the grassland sites demonstrated a more complete
a greater impact than pines in sites that were originally loss of runoff with eucalypts, with many reaching 100%
grasslands, with runoff reductions of 75% (
10%) and reductions in streamflow within 10 years (Fig. 1a),
40% (
3%), respectively (Table 1, Po0.001). suggesting that the trend toward recovery may only
Plantation age strongly affected runoff, whether apply to shrublands planted with pine.
expressed as absolute or as proportional changes (Table Afforestation reduced runoff across a broad range of
2; Fig. 1, Po0.001 in both cases). Runoff reductions in climates (Fig. 2). Reductions in runoff were significantly
afforested grasslands and shrublands were similar in related to MAP for afforested grasslands in both
the first 5 years after tree establishment (16% and proportional and absolute terms. For grasslands, the
15%, respectively), but diverged as the plantations wettest sites (MAP41500 mm yr1) had the largest
aged (Table 2). Afforested grasslands reached a 50% absolute reductions (287
44 mm) but the smallest
reduction in runoff by the tenth year, compared with proportional reductions (27
4%). In contrast, propor-
35% in afforested shrublands at the same age (Table 2). tional losses were far greater at the driest grassland site
In proportional terms, maximum reductions were (62
10%) (Fig. 2), suggesting that the effects of
reached  5 years earlier and were substantially larger afforestation on water yield will be more severe in drier
when grasslands were afforested (67% compared with regions. For shrublands, proportional and absolute
43% in shrublands, Table 2). reductions were largest at the driest sites
Decreases in streamflow were sustained through 30 (MAP 5 1000–1250 mm yr1, data not shown), but they
years in grasslands and, in absolute terms, showed no also were significantly older than the wettest sites, so age
sign of recovery with plantation age (Fig. 1b). In may be a confounding factor for the shrubland analysis.
proportional terms, there appeared to be some recovery Across the data set, proportional losses in low flow
for afforested grasslands after 20 years (Fig. 1a), but with afforestation were closely correlated with, but even
most of this is attributed to a single catchment where a larger than, proportional losses in annual flow (Fig. 3,
defoliation outbreak coincided with several years of Po0.001). These data suggest that dry-season losses are

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576

 E F F E C T S O F A F F O R E S TAT I O N O N WAT E R Y I E L D 1569

(a) 20 R 2 = 0.75; P< 0.001 (c) 40

R 2 = 0.71; P< 0.001
0 20

Change in runoff (%)

Change in runoff (%)

−20 0
Plantation type
−20
−40 Other Plantation type
−40
−60 Eucalyptus Eucalyptus
−60
−80 Pine Pine
−80
All species −100 All species
−100
0 10 20 30 0 10 20 30 40
Plantation age (yrs) Plantation age (yrs)

(b) 200 (d) 400

R 2 = 0.82; P< 0.001 R 2 = 0.74; P< 0.001
Change in runoff (mm)

Change in runoff (mm)

0 200

Plantation type 0
−200 Other Plantation type
−200
Eucalyptus Eucalyptus
−400
Pine −400 Pine

−600 All species −600 All species

0 10 20 30 0 10 20 30 40
Plantation age (yrs) Plantation age (yrs)

Fig. 1 (a–d) Proportional and absolute changes in runoff with plantation age, by original vegetation type. The complete data set was
used for curve-fitting, but to improve resolution of the figure grassland curves are only displayed to 30 years. Eight points 440 years are
not displayed, clustered around 30% and 500 mm (a, b; see Table 2). Regressions were through the origin. Regression equations: (a)
Y 5 05.636X 1 0.112X2, (b) Y 5 017.516X 1 0.115X2, (c) Y 5 04.398X 1 0.108X2, (d) Y 5 022.044X 1 0.520X2.

Table 2 Mean change in runoff (
SE) following afforestation as a function of plantation age, by previous vegetation type

Age (years) Grassland Shrubland

Drunoff (%) n Drunoff (mm) n Drunoff (%) n Drunoff (mm) n

1–5 16
5 35 45
17 34 15
3ab 36 81
20a 36

6–10 50
6 36 152
18 37 35
4c 40 158
17ab 40
11–15 67
5 30 216
18 29 39
4c 30 214
16b 30
16–20 58
5 29 247
28 27 43
4c 23 230
13b 23
21–25 42
6 12 304
62 10 35
4bc 20 168
22ab 20
26–30 54
4 4 456
48 4 32
4abc 20 193
20b 20
31–35 38
6c 17 203
26b 17
36–40 12
8a 8 80
56a 8
41–45 36
7 3 669
103 3
46–50 27
2 5 526
31 5
Po 0.001* 0.001* 0.001 0.001

Significance was determined using one-way ANOVAs followed by Tukey’s HSD post hoc tests, where conditions of normality and
homogeneity of variance were met; within each of those columns, means followed by different letters are significantly different from
each other at P  0.05.
*Kruskal–Wallis tests were used.
n 5 the number of runoff measurements in each age interval (taken from all catchments in which that plantation age range was
represented).

predicted to be even more severe than total annual region streams. In proportional terms, low flow declined
losses for afforestation scenarios, possibly leading to with plantation age through approximately 25 years and
shifts from perennial to intermittent flow regimes in dry- then began to recover somewhat (Fig. 4a, b). However,

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576

1570 K A T H L E E N A . F A R L E Y et al.

0 0 (a) 20
Change in runoff (mm)
mm** R 2 = 0.76; P< 0.001

Change in runoff (%)

−50 Plantation type

Change in low flow (%)

%*** 0
−100 −20 Pine
−150 −20 Eucalyptus
−40
−200 −40 Other
−250
−60
−300 −60

−350 −80 −80

<1000 1000 –1250 1250 –1500 >1500
−100
Mean annual precipitation (mm)
0 10 20 30 40
Fig. 2 Mean change in runoff (
SE) following afforestation as a Plantation age (yrs)
function of mean annual precipitation for sites that were originally
grasslands. The low precipitation sites were comprised of slightly (b) 50
R 2 = 0.42; P< 0.001
but not significantly younger stands than the wettest sites.

Change in low flow (mm)

0
*** Po0.001, ** Po0.01.
−50
20
R 2 = 0.82; P < 0.001 −100
0 Plantation type
−150
Change in low flow (%)

Pine
−20 −200 Eucalyptus
−250
−40 0 10 20 30 40
Plantation age (yrs)
−60
Fig. 4 (a, b) Change in low flow (% and mm) with plantation
age. Open symbols denote sites that were originally grasslands,
−80 closed symbols were originally shrublands. Regressions were
through the origin. Regression equations: (a) Y 5 05.694X 1
−100 0.142X2, (b) Y 5 04.799X 1 0.135X2.
−80 −60 −40 −20 0 20
Change in total annual runoff (%)
(e.g. 45% rather than 40%) for grasslands or shrublands
Fig. 3 Relationship between the change in low flow and the and for eucalyptus or pine. The relationship between
change in total annual flow. Regression equation: Y 5 6.161 1 plantation age and runoff reductions was not altered
0.990X. substantially by scaling, particularly for young planta-
tions; in the first 5 years after afforestation, runoff
the recovery may be species specific, as the loss of low reductions were only 1 percentage point higher. For
flow appears to be more complete for eucalyptus and afforested grasslands, scaling had little effect on max-
other species than for pines (Fig. 4a). The pattern of imum reductions, but maximum runoff reductions in
decline and recovery may also occur with the absolute afforested shrublands reached 50% after scaling (com-
change in low flow (Fig. 4b), although the effect in the pared with 43% without scaling). The pattern of runoff
first 10–15 years following afforestation was highly reductions across climatic zones also remained largely
variable, ranging from an increase of  10 mm to a unchanged with scaling, and it had a minor effect on the
decrease of almost 250 mm. magnitude of runoff reductions; scaling had no effect on
Scaling the data to 75% cover affected the magnitude either absolute or proportional runoff reductions for the
of the changes in streamflow somewhat, but did not alter wettest sites, but increased the average runoff reductions
the patterns with plantation age, climate, or vegetation in the driest sites from 62% to 66%.
type. The change in magnitude was most notable when
looking at single year runoff reductions; when data were
Discussion
scaled, one-third (rather than one-fifth) of the catch-
ments experienced reductions of 75% or more during at Our analysis clearly demonstrates that afforestation of
least 1 year. The effect of scaling was fairly uniform grasslands and shrublands will typically result in a loss
across vegetation types, as average reductions were of one-third to three-quarters of streamflow on average.
approximately 5 percentage points higher with scaling Runoff reductions are attained very rapidly after

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576

 E F F E C T S O F A F F O R E S TAT I O N O N WAT E R Y I E L D 1571

afforestation, with losses of more than 10% of stream- 0

flow occurring in the first 2–3 years after tree establish- R 2 = 0.45, P = 0.002
ment for most catchments. This indicates that the lag
time between planting and runoff response is usually −100

Change in runoff (mm)

short, although the full effect on runoff may not occur
for one or more decades.
Mean annual rainfall is one of the most important
−200
determinants of annual runoff and can have a strong
influence on change in runoff after vegetation change
(Vertessy, 2001; Zhang et al., 2001). In agreement with
previous analyses (Bosch & Hewlett, 1982), our data −300
show that vegetation change has the largest absolute
impacts on runoff in high-rainfall areas. However, it also
−400
reveals the opposite trend for proportional changes, 0 100 200 300 400 500
which may be a better measure of the effects on water Change in ET (mm)
supplies and are largest in dry areas (Fig. 2). The reason
for the more extreme reductions in drier regions may Fig. 5 Change in runoff as a function of change in evapotran-
simply be that there is less water in those systems; for a spiration (ET). All sites in this data set, which included plantation
given proportional increase in ET, the effect on runoff ages ranging from 0 to 20 years, were originally grasslands.
will be larger in drier regions because the fraction of
precipitation that reaches streams is already low. were afforested, runoff reductions occurred earlier,
Rooting depth may also be a factor, as it is expected to were larger, and were sustained for a longer period of
play a particularly important role in increasing ET in time than in shrublands. This may result from
dry climates (Zhang et al., 2001; Schenk & Jackson, 2002). differences in the underlying causes of the change in
As this source of increased water use is likely to persist ET that leads to lower runoff. The two primary causes
over the length of a rotation, it could result in larger of the increase in ET following afforestation are the
proportional runoff reductions overall in dry regions. greater capacity for water loss associated with higher
While runoff reductions occurred across many sites leaf area indexes (LAIs) of the higher stature vegetation
and species, afforestation had a greater effect on runoff in (Calder, 1986) and better access to water sources,
grasslands than in shrublands. The reason for higher through accessing of deep water or drawing on stored
runoff reductions in afforested grasslands compared with soil water (Calder et al., 1993; Zhang et al., 2001; Engel et
shrublands may be inherently higher runoff with al., 2005). When grasslands are afforested, deep water
herbaceous cover. Calder (1986) noted that transpiration access likely plays an important role, as there should be
losses from scrub vegetation in India tend to be relatively a large change in rooting depth (Jackson et al., 1996).
high, with such vegetation using twice as much soil This idea is supported by the fact that ET increases
water and drying the soil to twice the depth of annual more than runoff decreases in grassland sites. In the
crops. Contributing to this effect is the difference in the few studies in our data set (all originally grasslands)
depth and distribution of roots among vegetation types, where changes in ET were measured in addition to
which is altered by the shift from grasses or shrubs to change in runoff, a fairly strong relationship was
trees (Jackson et al., 2000). Shrubs have greater similarity revealed (Fig. 5, R2 5 0.45; Po0.001). The relationship
to trees, in terms of total root biomass and maximum was not 1 : 1, however, as ET increased more than runoff
rooting depth, than to grasses (Jackson et al., 1996); for decreased. While it is unclear whether this occurs in all
this reason, the change in access to water and the change plantation types or how long this pattern could be
in transpiration rates are not likely to differ as much maintained, it suggests the use of deep water to
between shrubs and trees as they do between grasslands subsidize the increase in ET in afforested grasslands.
and trees. In addition, shrubs may be characterized by a In contrast, when shrublands are afforested, the change
longer active transpiration period than seasonally dor- in rooting depth is not as large, so that the dominant
mant grasses, contributing to total annual transpiration mechanism behind increasing ET at those sites may be
that is higher than that of grasses and more similar to that the increased capacity for water loss by the trees
of trees. As a result, runoff reductions may be less severe relative to shrubs. This mechanism is also likely to be
when shrublands are afforested relative to grasslands. more a feature of younger plantations – occurring as
These differences between pre-afforestation vegeta- the LAI increases and declining as the tree canopy ages
tion types carried over to the age–runoff relationship of – so that with time the water use of the plantation may
afforested grasslands vs. shrublands. When grasslands approach the control and runoff could begin to recover.

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576

1572 K A T H L E E N A . F A R L E Y et al.

In addition to differences between afforested grass- the more important component of increasing ET (Holmes
lands and shrublands, there were significant differences & Wronski, 1981; Duncan, 1995), while in drier regions the
between pines and eucalypts in cases where the original ability of the vegetation to reach and exploit deep soil
vegetation was grassland. Eucalypts caused larger water stores to maintain transpiration is an important
proportional changes in annual runoff than pines did determinant of changes in water yield (Pearce & Rowe,
and also appeared to cause more severe and complete 1979). Therefore, in drier regions, where transpiration is
losses of low flow within the first 10–15 years after the more important contributor to absolute increases
afforestation. Differences in the growth patterns between in ET following afforestation (Scott & Lesch, 1997),
pines and eucalypts likely play a role in producing these eucalypts are likely to cause more severe runoff reduc-
differences. Decreases in runoff following afforestation tions. In wetter regions, where interception plays a more
are positively related to the growth rate of the planted important role, pines may cause more severe runoff
stands (Bosch & Hewlett, 1982), with evidence suggest- reductions. These differences can have important implica-
ing that the rate of increase in ET is more rapid under tions for decisions about where plantations are established
eucalypts because of their rapid early growth and and which tree species are used.
canopy closure (Dye, 1996). This rapid increase in ET
should correspond to larger reductions in runoff under
Implications for policy
eucalyptus in the early part of the rotation. Although
growth will begin to slow as the stands age, eucalyptus Our analysis shows that general relationships between
rotations tend to be relatively short compared with pine, plantation age and runoff responses exist. Streamflow
so that overall average water use per rotation is higher response to afforestation can be expected to be very
(Bosch & von Gadow, 1990), resulting in greater overall rapid (within 5 years of planting), maximum runoff
runoff reductions. While there is likely to be some reductions can be expected between 15 and 20 years
variation in species effects by region, generalizations after planting, and runoff reductions will likely be
regarding the effects of different plantation species on larger and more sustained when grasslands are
runoff should be useful for planning afforestation afforested than when shrublands are. These differences
projects and the tree species that will be used in them. among the areas in which afforestation is considered a
Important interactions may also exist between planta- potential land use are important in planning where
tion species and climate, with different plantation types plantations should be located, as well as which species
having more severe effects on runoff under different should be used. In addition, a better understanding of
precipitation regimes. In our data set, both eucalypts and the timing of the most extreme reductions in runoff
pines had their greatest relative impact in lower rainfall may help water managers in their planning. For
regions. However, the two types of plantations differed example, given that the effect of afforestation on low
markedly in terms of absolute reductions. Eucalypts flow is somewhat larger than on total flow, this may be
averaged across all ages up to 30 years in the data set an important variable to incorporate as a guide for
produced the smallest runoff reductions (90
14 mm) in afforestation zoning (Scott & Smith, 1997).
high rainfall zones (MAP41500 mm); for pines, the Our results also indicate that some past perceptions
largest runoff reductions (189
40 mm) occurred in about where afforestation projects should best be
higher rainfall regions. This pattern may be explained located in order to minimize effects on runoff may be
by the relative importance of increases in wet canopy misleading. Specifically, the assumption that changes in
evaporation vs. transpiration for different species and in runoff will be less severe in low rainfall areas does not
different climatic zones. Interception storage and evapora- hold true when proportional runoff reductions are
tion from the canopy are thought to be greater for needle- considered. While it has been suggested that some of
leaved than for broad-leaved trees (Zinke, 1967; Cannell, the negative hydrologic impacts of afforestation could
1999); the dense canopies of conifers allow for higher be minimized by establishing plantations in lower
canopy storage of rainfall and can lead to large intercep- rainfall zones (Vertessy, 2001), our data indicate that
tion losses (typically ranging from 15% to 24%, and in this prescription would be unlikely to ameliorate runoff
some cases reaching as much as 60%; Le Maitre et al., reductions and may actually result in more severe local
1999). For eucalypts, which tend to establish deep roots at impacts. This information may be critical for zoning of
a young age (Dye, 1996), higher transpiration is likely the afforestation projects, in particular in semiarid regions.
more important component of increasing ET following The ability to predict the likely effects of afforestation
afforestation (Vertessy, 2001). In addition to differing in specific locations with limited information will be the
among tree species, evaporation and transpiration also biggest challenge to zoning and planning for these
play different roles under different climate regimes. In projects. Catchment data are collected over decades
higher rainfall zones, evaporation of intercepted rainfall is and are unavailable for many regions of the world.

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576

 E F F E C T S O F A F F O R E S TAT I O N O N WAT E R Y I E L D 1573

However, some indicators, such as the change in runoff Best A, Zhang L, McMahon T et al. (2003) A critical review of paired
as a percent of MAP at a site, may provide a gauge of catchment studies with reference to seasonal flows and climatic
the probable severity of the loss of runoff. The average variability. CSIRO land and water technical report 25/03,
ratio tended to be around 14–15% of MAP for most CSIRO, Canberra, Australia, 30pp.
cases in our synthesis (Table 1), and was surprisingly Borg H, Bell RW, Loh IC (1988) Streamflow and stream salinity
conservative, regardless of whether the sites were in a small water supply catchment in southwest western
originally grasslands or shrublands (15% and 14%, Australia after reforestation. Journal of Hydrology, 103, 323–333.
Bosch JM (1979) Treatment effects on annual and dry period
respectively) and whether they were planted to pine or
streamflow at Cathedral Peak. South African Forestry Journal,
eucalyptus (14% for both) (this value also coincides well
108, 29–38.
with the difference in ET between forest and grassland
Bosch JM, Hewlett JD (1982) A review of catchment experiments
as a percent of MAP in the curves described in Holmes
to determine the effect of vegetation changes on water yield
& Sinclair, 1986). From this we can conclude that, on and evapotranspiration. Journal of Hydrology, 55, 3–23.
average, trees are able to use approximately 15% more Bosch JM, von Gadow K (1990) Regulating afforestation for
precipitation than grasses or shrubs. This suggests that, water conservation in South Africa. South African Forestry
in a region where natural runoff is in the range of 10% of Journal, 153, 41–54.
MAP, afforestation can be expected to result in a Brooks KN, Ffolliott PF, Gregersen HM et al. (1997) Hydrology and
complete loss of runoff; where natural runoff is 30% of the Management of Watersheds, 2nd edition. Iowa State
precipitation, it could be reduced by half or more when University Press, Ames.
trees are planted. The percent of precipitation used by Calder IR (1986) Water use of eucalypts – a review with special
trees may be higher than 15% in some regions (e.g. reference to South India. Agricultural Water Management, 11,
grasslands planted to eucalyptus; see Table 1), but this 333–342.
value can serve as a useful indicator for land managers Calder IR, Hall RL, Prasanna KT (1993) Hydrological impact of
and policy makers in guiding the location of plantations Eucalyptus plantation in India. Journal of Hydrology, 150, 635–648.
with respect to the demand for water resources. Calder IR, Newson MD (1979) Land-use and upland water
resources in Britain – a strategic look. Water Resources Bulletin,
Conclusion 15, 1628–1639.
Cannell MGR (1999) Environmental impacts of forest mono-
The environmental ‘co-effects’ of afforestation programs cultures: water use, acidification, wildlife conservation, and
have received much less attention than the carbon carbon storage. New Forests, 17, 239–262.
sequestration potential. However, one of the so-called Dons A (1986) The effect of large-scale afforestation on Tarawera
‘crunch issues’ that have been debated in determining river flows. Journal of Hydrology New Zealand, 25, 61–73.
how to implement land-use change and forestry projects Dons A (1987) Hydrology and sediment regime of a pasture,
within the CDM is their potential impact on local native forest, and pine forest catchment in the central North
livelihoods and environments (Pedroni, 2003). Our Island, New Zealand. New Zealand Journal of Forestry Science,
synthesis clearly indicates that a reduction in runoff 17, 161–178.
Duncan MJ (1995) Hydrological impacts of converting pasture
can be expected with afforestation of grasslands and
and gorse to pine plantation, and forest harvesting, Nelson,
shrublands, which will have ecological and socioeco-
New Zealand. Journal of Hydrology New Zealand, 34, 15–41.
nomic ramifications. In some locations, such as parts of
Dye PJ (1996) Climate, forest, and streamflow relationships in
Australia where lower runoff can ameliorate salinity
South African afforested catchments. Commonwealth Forestry
and groundwater upwelling, this will be a positive
Review, 75, 31–38.
change. In many other regions, reduced runoff will Engel V, Jobbágy EG, Stieglitz M et al. (2005) The hydrological
cause or intensify water shortages, a tradeoff that consequences of eucalyptus afforestation in the Argentine
should be explicitly recognized before land conversion. Pampas. Water Resources Research, in press.
Fahey BD, Jackson R (1997) Hydrological impacts of converting
Acknowledgements native forests and grasslands to pine plantations, South Island,
Funding for this work was provided by the Center on Global New Zealand. Agricultural and Forest Meteorology, 84, 69–82.
Change at Duke University, NSF, and the Biological and Fahey BD, Watson AJ (1991) Hydrological impacts of converting
Environmental Research (BER) Program, US Department of tussock grassland to pine plantation, Otago, New Zealand.
Energy, through the Southcentral Regional Center of NIGEC. We Journal of Hydrology New Zealand, 30, 1–15.
would like to thank two anonymous reviewers whose thought- Farley KA, Kelly EF, Hofstede RGM (2004) Soil organic carbon
ful comments were very helpful in improving this manuscript. and water retention following conversion of grasslands to pine
plantations in the Ecuadorian Andes. Ecosystems, 7, 729–739.
References Hoffmann WA, Jackson RB (2000) Vegetation–climate feedbacks
Bashkin MA, Binkley D (1998) Changes in soil carbon following in the conversion of tropical savanna to grassland. Journal of
afforestation in Hawaii. Ecology, 79, 828–833. Climate, 13, 1593–1602.

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576

1574 K A T H L E E N A . F A R L E Y et al.

Holmes JW, Sinclair JA (1986) Water yield from some afforested Scott DF, Prinsloo FW, Moses G et al. (2000) A re-analysis of the South
catchments in Victoria. Hydrology and Water Resources African catchment afforestation experimental data. WRC Report No
Symposium. River Basin Management, Griffith University, 810/1/00, Water Research Commission, Pretoria, South Africa.
Brisbane, 25–27 November 1986, Institution of Engineers, Scott DF, Smith RE (1997) Preliminary empirical models to
Australia, Barton, ACT, Australia. predict reductions in total and low flows resulting from
Holmes JW, Wronski EB (1981) The influence of plant commu- afforestation. Water SA, 23, 135–140.
nities upon the hydrology of catchments. Agricultural Water Sharda VN, Samraj P, Samra JS et al. (1998) Hydrological
Management, 4, 19–34. behaviour of first generation coppiced bluegum plantations
Jackson RB, Canadell J, Ehleringer JR et al. (1996) A global analysis of in the Nilgiri sub-watersheds. Journal of Hydrology, 211, 50–60.
root distributions for terrestrial biomes. Oecologia, 108, 389–411. Smith CM (1992) Riparian afforestation effects on water yields
Jackson RB, Carpenter SR, Dahm CN et al. (2001) Water in a and water quality in pasture catchments. Journal of Environ-
changing world. Ecological Applications, 11, 1027–1045. mental Quality, 21, 237–245.
Jackson RB, Schenk HJ, Jobbágy EG et al. (2000) Belowground Smith PJT (1987) Variation of water yield from catchments under
consequences of vegetation change and their treatment in introduced pasture grass and exotic forest, East Otago. Journal
models. Ecological Applications, 10, 470–483. of Hydrology New Zealand, 26, 175–184.
Jobbágy EG, Jackson RB (2003) Patterns and mechanisms of soil Smith RE, Scott DF (1992) The effects of afforestation on low
acidification in the conversion of grasslands to forests. flows in various regions of South Africa. Water SA, 18, 185–194.
Biogeochemistry, 64, 205–229. Vertessy RA (1999) The impacts of forestry on streamflows: a
Jobbágy EG, Jackson RB (2004) Groundwater use and saliniza- review. In: Forest Management for Water Quality and Quantity.
tion with grassland afforestation. Global Change Biology, 10, Proceedings of the Second Forest Erosion Workshop, May 1999,
1299–1312. Warburton, Australia. Report 99/6 (eds Croke J, Lane P), pp. 93–
Le Maitre DC, Scott DF, Colvin C (1999) A review of information 109. Cooperative Research Centre for Catchment Hydrology,
on interactions between vegetation and groundwater. Water CSIRO Land and Water, Canberra, Australia.
SA, 25, 137–152. Vertessy RA (2001) Impacts of plantation forestry on catchment run-
Mwendera EJ (1994) Effect on the water yield of the Luchelemu off. In: Plantations, Farm Forestry, and Water. Water and Salinity
catchment in Malawi of replacing indigenous grasses with Issues in Agroforestry no. 7, RIRDC Publication No. 01/20 (eds Nam-
timber plantations. Forest Ecology and Management, 65, 75–80. biar EKS, Brown AG), pp. 9–19. RIRDC, Kingston, Australia.
Nänni UW (1970) The effect of afforestation on streamflow at Cathe- Vertessy RA, Bessard Y (1999) Anticipating the negative
dral Peak: report no. 1. South African Forestry Journal, 74, 6–12. hydrologic effects of plantation expansion: results from a
Paul KI, Polglase PJ, Nyakuengama JG et al. (2002) Change in GIS-based analysis on the Murrumbidgee Basin. In: Forest
soil carbon following afforestation. Forest Ecology and Manage- Management for Water Quality and Quantity. Proceedings of the
ment, 168, 241–257. Second Forest Erosion Workshop, May 1999, Warburton, Australia.
Pearce AJ, Rowe LK (1979) Forest management effects on Report 99/6 (eds Croke J, Lane P), pp. 69–73. Cooperative
interception, evaporation, and water yield. Journal of Hydrology Research Centre for Catchment Hydrology, CSIRO Land and
New Zealand, 18, 73–87. Water, Canberra, Australia.
Pedroni L (2003) Ruling on the ‘crunch issues’ of land use, land- van Wyk DB (1987) Some effects of afforestation on streamflow in
use change and forestry: impacts on project viability. Interna- the Western Cape Province, South Africa. Water SA, 13, 31–36.
tional Journal of Global Energy Issues, 20, 75–94. Zar JH (1999) Biostatistical Analysis, 4th edition. Prentice-Hall,
Robinson M (1998) 30 years of forest hydrology changes at Englewood Cliffs, NJ.
Coalburn: water balance and extreme flows. Hydrology and Zhang L, Dawes WR, Walker GR (1999) Predicting the Effect of
Earth System Sciences, 2, 233–238. Vegetation Changes on Catchment Average Water Balance.
Robinson M, Cognard-Plancq AL, Cosandey C et al. (2003) Studies Cooperative Research Centre for Catchment Hydrology,
of the impact of forests on peak flows and baseflows: a European CSIRO Land and Water, Canberra, ACT, Australia.
perspective. Forest Ecology and Management, 186, 85–97. Zhang L, Dawes WR, Walker GR (2001) Response of mean
Robinson M, Gannon B, Schuch M (1991) A comparison of the annual evapotranspiration to vegetation changes at catchment
hydrology of moorland under natural conditions, agricultural scale. Water Resources Research, 37, 701–708.
use and forestry. Hydrological Sciences Journal, 36, 565–577. Zinke PJ (1967) Forest interception studies in the United States.
Samraj P, Sharda VN, Chinnamani S et al. (1988) Hydrological In: International Symposium on Forest Hydrology: Proceedings of a
behaviour of the Nilgiri sub-watersheds as affected by National Science Foundation Advanced Science Seminar held at the
bluegum plantations, Part I. The annual water balance. Journal Pennsylvania State University, University Park, Pennsylvania,
of Hydrology, 103, 335–345. August 29–September 10, 1965 (eds Sopper WE, Lull HW), pp.
Schenk HJ, Jackson RB (2002) Rooting depths, lateral root 137–161. Pergamon Press, Oxford.
spreads, and belowground/aboveground allometries of plants
in water limited ecosystems. Journal of Ecology, 90, 480–494.
Scott DF, Lesch W (1997) Streamflow responses to afforestation Appendix A
with Eucalyptus grandis and Pinus patula and to felling in the
Mokobulaan experimental catchments, South Africa. Journal of See Table A1 for catchment data sets used in the
Hydrology, 199, 360–377. synthesis.

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576

 Table A1 Datasets used in the synthesis

Source Name of site(s) Location Latitude/ MAP Original vegetation Plant species Plantation age Percent Data Notes
longitude (mm) (years) planted type

Borg et al. Padbury Reservoir Southwest Australia N/A 880 Crops and pastures Pinus radiata, 3–8 70 P–O
(1988) Eucalyptus
globulus
Bosch (1979) Cathedral Peak II Winterton, Natal Drakensberg, 29100 0 S/29115 0 E 1400 Grassland P. patula 1–26 52 Supp
South Africa
Calder & Wye and Severn Wales, UK N/A 2350 Pasture Picea sitchensis 43–50 100 C–P Planting from
Newson (1979) rivers, Plynlimon (80%) 1937 to 1964;
1937 used to
calculate
plantation age
Dons (1986) Tarawera North Island, New Zealand 1500 Scrub and native Pine Average for 28 C–P MAP
bush 1–18 estimated from
graph
Dons (1987) Purukohukohu North Island, New Zealand 38126 0 S/176113 0 E 1550 Pasture P. radiata Average for N/A C–P
8–11
Duncan (1995) C14 Moutere Gravel hill country, N/A 1020 Pasture P. radiata 1–21 N/A C–P
Nelson, New Zealand
Fahey & Glendhu Waipori River, Otago, New 45150 0 S 1350 Tussock grassland P. radiata 9–12 67 C–P
Jackson (1997) Zealand
Fahey & Glendhu Waipori River, Otago, New 45150 0 S 1355 Tussock grassland P. radiata 1–8 67 C-P Catchment
Watson (1991) Zealand was contour-

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576

ripped to
60 cm depth
prior to
planting
Mwendera Luchelemu River Malawi 11145 0 S/33150 0 E 1300 Montane grass and Pine and Mean 93 C-P Low flow only
(1994) scrub eucalyptus
Nänni (1970) Cathedral Peak II Winterton, Natal Drakensberg, 28100 0 S/29115 0 E 1660 Grassland P. patula 1–16 75 Supp Planting of CP
Cathedral Peak III South Africa 1545 Grassland P. patula 1–8 81 II began in
1951, but not
finished until
1961

Robinson Coalburn Northwest England N/A 1350 Grassland and bog Sitka spruce 24 N/A C–P
(1998)
Robinson et al. FM/N Chiemseemoors, Germany 47148 0 N/12126 0 E 1440 Bog formerly used Norway 4–25 N/A P–O
(1991) FM/S for agriculture spruce

Samraj et al. Glenmorgan, Nilgiri Plateau, South India 11128 0 N/76137 0 E 1535 Grassland and E. globulus 1–10 59 P–O
(1988) Ootacamund woodland
Scott & Lesch Mokobulaan A Lydenburg, Mpumalanga, 27117 0 S/30134 0 E 1135 Grassland E. grandis 1–17 100 Supp
(1997) Mokobulaan B South Africa 1170 Grassland P. patula 1–21 100
E F F E C T S O F A F F O R E S TAT I O N O N WAT E R Y I E L D

(Continued)
1575
 Table A1 (Contd.)
Source Name of site(s) Location Latitude/ MAP Original vegetation Plant species Plantation age Percent Data Notes
longitude (mm) (years) planted type

View publication stats

Scott et al. Westfalia D Tzaneen, Mpumalanga, South 23143 0 S/30104 0 E 1250 Scrub E. grandis 1–15 83 P–O Scrub forest
(2000) Africa cleared before
planting
Mokobulaan A Lydenburg, Mpumalanga, 27117 0 S/30134 0 E 1170 Grassland E. grandis 1–22 97
South Africa
Mokobulaan B Winterton, Natal Drakensberg, 28100 0 S/29115 0 E 1180 Grassland P. patula 1–20 95 Controlled
South Africa burn before
planting
Cathedral Peak II Jonkershoek, Western Cape, 33157 0 S/18115 0 E 1400 Grassland P. patula 1–29 75
South Africa
Cathedral Peak III 1515 Grassland P. patula 1–22 86
Bosboukloof 1125 Fynbos shrubland P. radiata 5–38 57
Biesievlei 1300 1–35 98
Tierkloof 1320
1576 K A T H L E E N A . F A R L E Y et al.

Lambrechtsbos B 1145 1–40 36

Lambrechtsbos A 1125 1–32 82
1–18 89
Sharda et al. Glenmorgan, Nilgiri Plateau, South India 11128 0 N/76137 0 E 1535 Grassland and E. globulus 1–10 59 P–O Second
(1998) Ootacamund woodland rotation
Smith (1987) Taieri River (Jura East Otago, South Island, New N/A 1000 Pasture P. radiata, P. N/A N/A C–P Catchments
Creek) Zealand nigra cleared of
native grasses
and shrubs,
planted with
pasture or pine
Smith (1992) Moutere Hills Nelson, New Zealand 41122 0 S/173104 0 E 1050 Pasture (ryegrass) P. radiata 5–9 20 P–O
Smith & Scott Westfalia D Tzaneen, Mpumalanga, South 22143 0 S/30104 0 E 1600 Scrub E. grandis 1–8 83 P–O Low flow data
(1992) Africa
Mokobulaan A Lydenburg, Mpumalanga, 24117 0 S/30134 0 E 1150 Grassland E. grandis 1–12 100
South Africa
Mokobulaan B 1150 Grassland P. patula 1–11 95
Van Wyk Bosboukloof Jonkershoek, Western Cape, 33157 0 S/18115 0 E 1300 Fynbos P. radiata 57 Supp
(1987) South Africa
Biesievlei 1425 98
Tierkloof 1800 36
Lambrechtsbos B 1475 82
Lambrechtsbos A 1415 89

Data type refers to the way in which the data were used in the synthesis: data were either reported and used as predicted–observed runoff (P–O) or as control catchment–planted
catchment runoff (C–P); some data sets were only used for supplementary information (e.g. planted area, year of plantation, etc.) (Supp).
Plantation age refers to the range of plantation ages reported for a given catchment.
MAP, mean annual precipitation; N/A, not available.

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 1565–1576