SCA2014-080 1/12

PARAMETER BASED SCAL

- ANALYSING RELATIVE PERMEABILITY FOR FULL FIELD
APPLICATION
Einar Ebeltoft, Frode Lomeland, Amund Brautaset and Åsmund Haugen
Statoil ASA, Stavanger, Norway

This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in Avignon, France, 8 -11 September, 2014.

ABSTRACT
A method is proposed where the flow properties from SCAL experiments are parameterized and
implemented into saturation functions for full field applications. Parameterization of the flow
properties facilitates analysis of end point saturations and curve shapes individually for each
SCAL experiment, and storing the parameters in a SCAL database. A number of these
parameterized SCAL experiments are subsequently combined to generate representative
saturation functions for full field application that are consistent with wettability as well as rock
properties.
The method starts by verifying each SCAL experiment by simulation followed by
parameterization of the appropriate flow properties, including residual oil saturation (Sorw or Sorg),
end point relative permeabilities (krw(Sorw) or krg(Sorg)) and the shape parameters of the curves.
These flow parameters are then stored in a SCAL database and interrelated with plug specific
data, including experimental conditions and geo-references. Each flow parameter is analysed
individually as functions of initial water saturation, porosity and permeability and supported by
wettability consistent trend models. The trend models for each flow parameter have been
developed based on the underlying theory of wettability, reservoir physics and observed
behaviour of a large set of SCAL data from the Norwegian Continental Shelf. The trend models
and application are presented in this paper.
The advantage of parameterization, using trend models and possible analogues from the
database, are smooth and well defined saturation functions that are consistent with wettability.
The resulting relative permeability is determined as a base case with optimistic and pessimistic
bounds. These are easily implemented as saturation functions in full field reservoir simulators for
sensitivity and field performance analyses. This approach is demonstrated for a real field case
from the Norwegian Continental Shelf, taking the underlying theory behind trend models, rock
quality and wettability into consideration.
INTRODUCTION
Relative permeability is important input for describing dynamic behaviour of fluid flow in
reservoir simulators and they are required properties in the saturation equation describing multi-
phase flow in porous media. These flow properties are represented as saturation functions and
will significantly influence the simulated and history matched production profiles. The industry
has determined these flow properties by SCAL experiments since the late 30’s [1] and it is still
the main approach to access saturation functions for full field applications. The design of proper
SCAL programs have improved the last couple of decades as there has been significant focus on
selecting and preparing representative core samples to obtain “true” wettability, performing the
experiments at reservoir conditions and using live reservoir fluids. Several types of SCAL
SCA2014-080 2/12

experiments are also combined with proper rates and fluids to mimic the reservoir behaviour.
Experimental equipment has improved by high precision equipment (pumps, transducers,
separators, etc.) and by monitoring of in situ saturations. Each SCAL experiment is routinely
verified by core flow simulations for verification and quality assurance, and in recent years the
use of pore scale models has also been increasingly used as support to experimental data. Hence,
there are numerous publications regarding how to determine and properly understand multi-
phase flow in core plugs. The majority of these publications focus on determining flow
properties from single core flow experiments and less is found in the literature on combining the
results from several experiments to provide saturation functions for full-field application.
Nevertheless, some publications have addressed the benefits of using bulk data and studied
trends between various properties. There are well-known trends between porosity and
permeability. There are also clear trends between saturation end-points (Swi, Sorw, Sorg) and curve
shapes in fluid/rock-systems at strongly wetted conditions – mainly water-wet – as the flow
properties are mainly influenced by pore geometry and topology [2] However, when wettability
is taken into account, the clear trends tend to vanish [3]. The effect of wettability on oil recovery
and relative permeability curves have been well established during the last couple of decades and
are emphasized by Hamon [3] and its references. The importance of accounting for field
wettability variations was suggested by Hamon [3, 4] and also used by Ghedan et al. [5, 6] to
give proper sets of relative permeability curves suitable for reservoir simulators. Several papers
also address the importance of introducing trends between dynamic and static properties to
improve dynamic flow characterization, see [5] and its references.
Trend models for relative permeability parameters have been developed based on a physical
understanding of multi-phase flow in porous media and how shape and end points of the relative
permeability curves should behave with respect to permeability, Swi and wettability. This paper
suggests a method for anchoring experimentally measured relative permeability curve shapes and
end points to these trend models. In addition, a database for storage of parameterized relative
permeability curves was developed. The database allows easy comparison of field specific or
analogue SCAL data. We have observed that SCAL data from the Norwegian Continental Shelf
(NCS) align with trends based on physics, all though some trends are more obvious than others.
We observe that trends can be used to propagate bulk-data into a suitable format for saturation
functions provided as a base case, with optimistic and pessimistic bounds. Using this approach
elevates the uncertainty from focusing on the accuracy in each experiment to analysing the
scatter of each parameter from a number of experiments.
The methodology presented here can be extended to model the Capillary Transition Zone (CTZ),
in which wettability, and hence the curve shape and end-points of the relative permeability
curves vary with height above oil-water contact [3-6]. This paper is based on clastic reservoir
rock, but it can also be extended to other rock types such as carbonates, when sufficient bulk data
is available and properly parameterized and stored in a database.

RELATIVE PERMEABILITY EXPERIMENTS AND SENSITIVITIES

The most common experiments used to determine relative permeability and residual saturations
are steady state, unsteady state and single speed centrifuge. The three types of experiments
provide information in different saturation ranges as indicated in Figure 1.
SCA2014-080 3/12

The steady state method provides two-phase

relative permeabilities in a wide saturation range,
but is generally not recommended as a reliable
measure of residual saturation.
The unsteady state method provides two-phase
relative permeabilities in a narrow saturation range
at high saturations, and is generally not
recommended as reliable measure of residual oil
saturation.
The single speed centrifuge method provides tail-
end relative permeability for the displaced phase
toward its residual saturation, and is well suited for
determining residual saturation. Figure 1: Relative permeability experiments and the
saturation range in which they provide information
In addition to these experiments for determining relative permeability, multi-speed centrifuge is
used to determine capillary pressure and wettability. This method is also well suited for
determining residual saturation. In order to cover the widest possible saturation range of the
relative permeability curves, it is recommended to combine these experiments when designing a
SCAL program [7, 8].

LET FLOW PARAMETERS AND RELATIVE PERMEABILITY

All experiments are generally verified by Sendra core flood simulator [9] where the relative
permeability curves are parameterized using the LET correlation [10], Eq.1-3 for oil/water.

Equation 1

Equation 2

Equation 3

k0ro is the end point relative permeability to oil

at irreducible water saturation (Swir) and k0rw is
the end point relative permeability to water at
residual oil saturation (Sorw). The respective
LET parameters for oil- and water relative
permeabilities dominate different parts of the
curves as illustrated in Figure 2: L describes
the shape of the lower part of relative
permeability curve, E describes the shape of
the intermediate height of the curve and T
describes the upper (or top) part of the curve.
The L, E and T parameters in the LET
correlation replace the Corey exponent,
allowing smooth and flexible relative
Figure 2: Dominating parts of the LET parameters.
SCA2014-080 4/12

permeability curves. All the individual LET shape parameters and end points are stored in a
SCAL database with plug data, experimental conditions and geo-references (well name, plug
depth, fluid viscosities at reservoir conditions, formation name, etc.).
TREND MODELS FOR LET FLOW PARAMETERS
Each flow parameter in the LET correlation has assigned individual trend models as a function of
Swi. The trend models have been developed based on physics and wettability considerations, and
they are anchored to SCAL-data from the Norwegian Continental Shelf (NCS). The trend models
were developed to create relative permeability curves and to capture the scatter normally seen in
experimental data. They are consistent with wettability and physical considerations of multi-
phase flow in porous media. Each trend model is calibrated to fit the field specific SCAL-data.
The paper presents trend models for water-oil imbibition.

Initial water saturation, residual oil saturation

Initial water saturation (Swi) defines starting point in
most water-oil SCAL experiments. Wettability has
been shown to be strongly affected by Swi as
indicated in Figure 3; lower Swi tends to produce
less water-wet states [11, 12]. As the Swi is low, a
relatively large fraction of the pore wall surface has
been contacted by oil for wettability alteration
towards less water-wet or even oil-wetting
conditions. Low Swi is typically experienced for
high permeable samples. The residual oil
saturations after waterflood is a function of
wettability, as the minimum residual oil saturation
is usually reported near to neutral-wet conditions
[12-15], an example of this is shown in Figure 4
(left). The residual oil saturation is also a function Figure 3: Amott Wettability index as function of
of the initial oil saturation [14, 16, 17]. This is various initial water saturations [12].
exemplified for various wettabilities by contact angles and pore scale modelling shown in Figure
4 (right). There is thus an interrelationship between Sorw, Swi and wettability.

Figure 4: Left: Residual Oil saturation as function of USBM wettability Index from [15]. Right: Network modelling
of trapped oil or residual oil saturation as function of initial oil saturation for different oil-water contact angles from
[14].
SCA2014-080 5/12

Modified from Lomeland et al. we have derived the

correlation between Swi and Sorw [18].

Equation 4

The typical trend for this relationship is shown in Figure

5 and coincides with Figure 4 (left). The trend is
consistent with increasing Sorw by increasing Swi until a
maximum is reached where Sorw decreases with
increasing Swi coinciding with Figure 4 (right) for all
wettabilities. Figure 4 is hence used to exemplify the
Figure 5: Typical trend model for
relationship between low water saturation and wettability.
Sorw versus Swi by use of Equation 4.

End point water relative permeability

Relative permeability and wettability are correlated [19].
Increased water-wetness tends to result in low end-point
relative permeability to water [20] and a more depressed
relative permeability to water in the entire mobile
saturation range. The interrelation between Swi and
wettability thus provides the basis for the trend model
relating the end-point water relative permeability to Swi,
derived by Lomeland et al [10, 18] given by Equation 5,
but here generalized with maximum parameter Awko and
minimum parameter Cwko.

Equation 5
Figure 6: Typical trend models for end-point
relative permeability to water as function of
Swi by se of Equation 5.
Typical trend models for end-point relative permeability to
water (krw(Sorw)) as function of Swi is shown in Figure 6. The model is consistent with increasing
Swi leading to increased water-wetness, acknowledging that more water-wet behaviour leads to
lower relative permeability to water. The trend model for krw(Sorw) is closely related to the trend
model for Sorw versus Swi in Figure 5. Low Swi leads to low Sorw, and consequently the krw(Sorw) is
high. High Swi also leads to low Sorw, however, it also leads to strongly water-wet conditions and
consequently the krw(Sorw) is low.

LET shape parameters

The shape of the relative permeability curves are captured through the LET parameters for oil
and water. It is important that the trend models for these also reflect the change in wetting
behaviour with changing Swi. Figure 7 shows typical trend models for LET shape parameters for
oil. Increasing Swi leads to more water-wet behaviour and thus increased oil relative permeability,
which is consistent with the trend models for Lo, Eo and To. Similarly, trend models for the LET
parameters for water are shown in Figure 8. The models are consistent with increasing Swi
causing a more water-wet system with lower relative permeability and thus a more depressed
SCA2014-080 6/12

relative permeability shape of the water. Reversely, more oil-wet behaviour results in higher
water relative permeability.

Figure 7: Typical trend model for Lo, Eo and To as functions of Swi.

Figure 8: Typical trend model for Lw, Ew, and Tw as functions of Swi.

Field-specific trend models

Wettability and relative permeability are not only functions of Swi, they are also a function of
chemical interactions between crude oil/brine/rock (COBR) [21-27] and the geological features
of the reservoir rock, mineralogy, clay content, pore size distribution etc. It is thus necessary to
calibrate the trend model to incorporate the observations made by the special core analysis
measurements for the particular field in question. This is achieved by interpreting all relevant
SCAL experiments and parameterizing all relative permeabilities to the LET formulation. All
LET parameters are stored in a SCAL database together with plug data and geo-references (well-
name, depth, formation name, etc.) Each individual trend model can then be expressed and
calibrated to the field specific properties.
A SCAL program consists of several experiments, all with variations in the actual properties due
to the variability within core plugs and the inherent uncertainty of experimental measurements.
There will be varying Swi, Sorw, and LET parameters within a selection of relative permeability
measurements, and a method for calibrating trend models to a selection of relative permeability
measurements has been developed. The individual LET flow parameters are loaded into a
graphical interface overlain by the trend models shown in Figure 5-8. All trend models are given
with an expected base curve visually bounded by high and low curves to capture the scatter in
each flow parameter Sorw, krw(Sorw), Lo, Eo, To, Lw, Ew and Tw. Each individual trend model is then
calibrated by changing the trend model inputs to better capture the field specific SCAL-data.
SCA2014-080 7/12

This is illustrated in Figure 9. The left plot shows a general trend model for Sorw versus Swi. Plug
measurements from an actual field is the blue squares and may deviate from the general trend.
The right plot shows the calibrated trend models. Three models have been introduced, a base
model bounded by high and low models. This captures the scatter in the SCAL data and gives a
field specific trend model for Sorw versus Swi including a span of uncertainty. Note that calibration
is done by visual inspection, but based on the described trend models and field specific SCAL
data. The advantage of the graphical method for calibrating the trend models are:
 Easy identification of outliers which are not representative.
 Easy determination and evaluation of the scatter in the data selection.
 Fast calibration of trend models to capture base case trend models with high and low
bounds for uncertainty.

Figure 9: Left: Typical trend model for Sorw versus Swi, blue squares represent plug measurements. Right: Three
calibrated trend models are introduced; low, base and high. The trend models have been calibrated to better capture
plug measurements (blue squares). Red is the expected model, bounded by high and low trends in blue and green to
capture uncertainty.

This process is performed for all the flow

parameters, including, Sorw, krw(Sorw), Lo, Eo, To, Lw,
Ew and Tw, Figure 9-12. All trend models are given
with an expected base curve bounded by high and
low curves to capture the uncertainty. It should be
addressed that all the standard experimental
methods (steady state, unsteady state and single
speed centrifuge) for waterflood relative
permeability measurements have low sensitivity
near Swi. This implies that there is often observed
more scatter in the Lw and To parameters, and there
is in general more uncertainty associated with these
two parameters and also more uncertainty in the
trend models for these two parameters.
Figure 10: Calibrated trend for end point
relative permeability to water, krw(Sorw),
with low base and high.
SCA2014-080 8/12

Figure 11: Calibrated trend models for Lo, Eo and To with low, base and high.

Figure 12: Calibrated trend models for Lw, Ew and Tw with low, base and high.

Generating relative permeability curves for full field simulations

When all LET parameters models have been calibrated to the field specific models (Figure 9-12),
full field relative permeability curves are generated using the field specific trend models. A
common way of implementing relative permeability curves in full field applications is the use of
end-point saturation scaling of both relative permeability curves [28]. This introduces errors in
the oil relative permeability curve shape that is not consistent with wettability. Increased Swi
leads to higher curvature of the oil relative permeability curve shape when end-point saturation
scaling is implemented. This is inconsistent with wettability. The proposed technique to reduce
the effect of end point scaling in full field simulations is to use the specific trend model to
generate relative permeability over a smaller saturation interval or bin. Figure 13-16 show how 4
initial water saturations were selected for a field at the NCS. Each Swi represents a saturation bin
with given LET parameters which represent the oil and water relative permeability curves with
optimistic and pessimistic bounds.
The example shown below has four saturation bins at Swi = [0.0 – 0.15, 0.15 – 0.25, 0.25 – 0.35,
0.35 – 1.0]. Each bin represents a unique rock type where permeability increases with decreasing
Swi, i.e good quality reservoir rock is associated with the low Swi, and poorer reservoir rock is
associated with higher Swi. Figure 13 shows the relationship between Swi and absolute
permeability, indicating that curve shape and end point parameters also may be correlated to
permeability rather than Swi in the region above the capillary transition zone.
SCA2014-080 9/12

Figure 13: Relationship between Swi and absolute permeability.

Figure 14: Left: Trend models for Sorw as function of Swi with upper and lower bounds. Right: Trend models for
endpoint relative permeability to water with upper and lower bounds. The dark blue squares are the individual
values from SCAL plug experiments from Field A.

Figure 15: Left: Trend models for Lo as function of Swi. Middle: Trend models for Eo as function of Swi. Right: Trend
models for To as function of Swi. Blue squares represent values from plug experiments from Field A.
SCA2014-080 10/12

Figure 16: Left: Trend models for Lw as function of Swi. Middle: Trend models for Ew as function of Swi. Right:
Trend models for Tw as function of Swi. Blue squares represent values from plug experiments from Field A.
Throughout the rest of this paper we refer to Swi that is established at high capillary pressure
using porous plate or centrifuge, resulting in a Swi that is pragmatically similar to irreducible
water saturation (Swir). Hence, we emphasize the reservoir region above the capillary transition
zone where the production well starts by producing oil and practically no water. Swir varies with
absolute permeability across the reservoir, and it defines the lowest water saturation and it is
anchoring the relative permeability curves. The resulting relative permeability curves for the first
bin with base, optimistic and pessimistic curves are shown in Figure 17 (left). The resulting base
case relative permeabilities for all 4 bins are shown in Figure 17 (right). Note that both shape and
end points of the relative permeability changes as Swi increases. The advantage of this method is
relative permeabilities that are easily implemented in full field simulations, and they are
consistent with wettability and permeability in a large saturation range. The analysis of each
individual parameter provides a good measure of the uncertainty and scatter in experimental data
from SCAL.

Figure 17: Left: Relative permeability for highest permeability rock type, i.e. the lowest Swi with base, optimistic and
pessimistic bounds. Right: Comparison of the base relative permeabilities for the 4 different saturation bins
calculated along the trend lines as indicated in Figures 13-16.

CONCLUSION
 Parameterizing of flow properties enables simpler use of large sets of SCAL data when
defining saturation functions for full-field applications
 Honouring wettability and physics of flow through porous media enables the construction
of trend models of end points and shape parameters of relative permeability.
 The trend models are used to create wettability consistent flow properties for full-field
simulations
SCA2014-080 11/12

 Admitting the general scatter of SCAL data, the trend models and LET correlation guides
and defines smooth curves for full-field simulation
 The SCAL model/saturation functions are generated as a base case with optimistic and
pessimistic bounds allowing proper evaluation and uncertainty analysis of recoverable
reserves for specific fields.

ACKNOWLEDGEMENTS
We acknowledge Jon Knut Ringen and Egil Boye Petersen Jr. for valuable discussions. Statoil is
acknowledged for giving permission to publish.

NOMENCLATURE
kro = oil relative permeability k0rw = water relative permeability at Sorw
krw = water relative permeability Xo = parameter in kro X = L,E,T
Swi = initial water saturation Xw = parameter in krw X = L,E,T
Swir =irreducible water saturation Xorw = parameter in Sorw X = L,E,T,M,A,B
Sorw = residual oil saturation Xwko = parameter in k0rw X = L,E,T,A,C
k0ro = oil relative permeability at Swir

REFERENCES
1. Muskat, M., et al., Production engineering research - Flow of gas-liquid mixtures
through sands. Transactions of the American Institute of Mining and Metallurgical
Engineers. 123: p. 69-96, 1937.
2. Guo, G., et al., Rock Typing as an Effective Tool for Permeability and Water-Saturation
Modeling: A Case Study in a Clastic Reservoir in the Oriente Basin, in SPE Annual
Technical Conference and Exhibition. SPE97033, Society of Petroleum Engineers:
Dallas, Texas, USA, 2005.
3. Hamon, G., Two-Phase Flow Rock-Typing: Another Perspective, in SPE Annual
Technical Conference and Exhibition, SPE84038, Society of Petroleum Engineers:
Denver, Colorado, USA, 2003.
4. Hamon, G., Field-Wide Variations of Wettability, SPE63144, in SPE Annual Technical
Conference and Exhibition Society of Petroleum Engineers Inc.: Dallas, Texas, 2000.
5. Ghedan, S.G., Q. Huang, and Y. Wu, An Improved Approach for Generating Saturation
Functions for Simulation Models Using Dynamic Rock Types, SPE 160683, in Abu Dhabi
International Petroleum Conference and Exhibition, Society of Petroleum Engineers:
Abu Dhabi, UAE, 2012.
6. Ghedan, S.G., Dynamic Rock Types for Generating Reliable and Consistent Saturation
Functions for Simulation Models, SPE 111295 in SPE/EAGE Reservoir Characterization
and Simulation Conference 2007, Society of Petroleum Engineers: Abu Dhabi, UAE.
7. Sylte, A., E. Ebeltoft, and E.B. Petersen. Simultaneous Determination of Relative
Permeability and Capillary Pressure Using Data from Several Experiments, SCA2004-
17 in International Symposium of the Society of Core Analysts. Abu Dhabi, UAE, 2004.
8. Maas, J. and A.M. Schulte. Computer Simulation of Special Core Analysis (SCAL) Flow
Experiments Shared on the Internet, SCA-9719, in International Symposium of the SCA.
1997. Calgary, Canada.
9. Sendra, Sendra User Guide, 2013, Weatherford.
SCA2014-080 12/12

10. Lomeland, F., E. Ebeltoft, and W.H. Thomas. A New Versatile Relative Permeability
Correlation, SCA2005-32, in International Symposium of the Society of Core Analysts.
2005. Toronto, Canada.
11. Morrow, N.R. and N. Mungan, Wettability and Capillarity in Porous Media. 1971,
Petroleum Recovery Research Inst. Calagary.
12. Jadhunandan, P.P. and N.R. Morrow, Effect of Wettability on Waterflood Recovery for
Crude-Oil/Brine/Rock Systems. SPERE, Vol 10 01 (p.40-46), 1995.
13. Johannesen, E.B. and A. Graue, Systematic Investigation of Waterflood Reducing
Residual Oil Saturations by Increasing Differential Pressures at Various Wettabilities,
SPE108593., SPE, Offshore Europe, 4-7 September, Aberdeen, Scotland, U.K, 2007.
14. Spiteri, E.J., et al., A New Model of Trapping and Relative Permeability Hysteresis for All
Wettability Characteristics. SPE Journal, Vol. 13 03 p.277-288, 2008.
15. Lorenz, P.B., E.C. Donaldson, and R.D. Thomas, Use of Centrifugal Measurements of
Wettability to Predict Oil Recovery, Report no. 7873, U.S Bureau of Mines, Bartlesville
Energy Technology Center, 1974.
16. Pentland, C.H., et al., Measurement of Nonwetting-Phase Trapping in Sandpacks. SPE
Journal, Vol. 15, 02 p. 274-281, 2010.
17. Land, C.S., Comparison of Calculated with Experimental Imbibition Relative
Permeability. SPEJ, Vol. 11, 04 p. 419-425, 1971.
18. Lomeland, F., et al., A Versatile Representation of Upscaled Relative Permeability for
Field Applications, SPE154487, SPE Europec/EAGE Annual Conference, 4-7 June,
Copenhagen, Denmark, 2012.
19. Anderson, W.G., Wettability Literature Survey Part 5: The Effects of Wettability on
Relative Permeability. Journal of Petroleum Technology, 39(11): p. 1453-1468,1987.
20. Craig Jr., F.F., The Reservoir Engineering Aspects of Waterflooding. SPE Monograph
Series. Vol. 3. 1993: Society of Petroleum Engineers.
21. Anderson, W.G., Wettability Literature Survey- Part 1: Rock/Oil/Brine Interactions and
the Effects of Core Handling on Wettability. Journal of Petroleum Technology, 38(10): p.
1125-1144, 1986.
22. Buckley, J.S., K. Takamura, and N.R. Morrow, Influence of Electrical Surface Charges
on the Wetting Properties of Crude Oils. 1989. SPERE, Vol.4 03, p. 332-340, 1989.
23. Buckley, J.S., C. Bousseau, and Y. Liu, Wetting Alteration by Brine and Crude Oil: From
Contact Angles to Cores,. SPEJ Vol.1 03 p.341-350, 1996.
24. Yu, L. and J.S. Buckley, Evolution of Wetting Alteration by Adsorption From Crude Oil.
1997. SPE Formation Evaluation, Vol.12 01, p. 5-12
25. Buckley, J.S. and Y. Liu, Some mechanisms of crude oil/brine/solid interactions. Journal
of Petroleum Science and Engineering, 1998. 20(3–4): p. 155-160.
26. Buckley, J.S., Y. Liu, and S. Monsterleet, Mechanisms of Wetting Alteration by Crude
Oils. 1998. SPEJ, Vol. 3 01 p.54-61,1998.
27. Loahardjo, N., et al., Oil Recovery by Sequential Waterflooding: the Effects of Aging at
Residual Oil and Initial Water Saturation, SPE154202, in SPE Improved Oil Recovery
Symposium. 2012, Society of Petroleum Engineers: Tulsa, Oklahoma, USA.
28. Schlumberger, ECLIPSE Technical Description, Version 2013.2. 1 ed. Vol. 1, p. 975-
996.