Water-limited yield potential is sensitive to soil hydraulic properties that govern plant-available water retention characteristics, and landscape and soil properties that influence infiltration rate and runoff. Therefore it is necessary to determine the dominant soil units at the level of the RWS buffers and derive the parameters required by crop simulation models to simulate water-limited yield potential (Yw). Key issues discussed in this protocol are: soil data sources and the procedures to derive the dominant soil units and soil parameters.

1. Soil data sources

GYGA will use four complementary soil data sources:

1. li>For Sub-Saharan Africa the AfSIS soil database is used. The used product is the Root Zone Plant Available Water Holding Capacity of the Sub-Saharan Africa Soil. This product is the result of a collaboration between AfSIS and GYGA. See: http://www.isric.org/projects/afsis-gyga-functional-soil-information-sub-saharan-africa-rz-pawhc-ssa
    
2. Global soil databases (ISRIC-WISE soil database, see: http://www.isric.org/explore/wise-databases)
    
3. Sometimes, high-quality national databases with ‘functional soil properties', that is, quantitative soil profile data from which required soil properties for crop simulation can be retrieved (e.g., SSURGO database, SOTER Kenya (http://www.isric.org/projects/soter-based-soil-property-estimates-kenya), SOTER Southern Africa (http://www.isric.org/projects/soter-and-wise-based-soil-property-estimates-southern-africa ) , European soil data base (http://eusoils.jrc.ec.europa.eu/ESDB_Archive/ESDB/), the soil database for Brazil compiled by Cooper et al. (2005) (http://www.esalq.usp.br/gerd/), Argentine soils database (GeoINTA, http://geointa.inta.gov.ar/visor/?p=model_suelos_bsas).
    
4. Sometimes, expert opinion from country agronomists

 

2. Procedure to identify and estimate the area of the dominant soil units

In the case the gridded AfSIS soil database is used to select the soil properties for simulating Yw of rainfed crops, the procedure is as follows:

1. There are in total 28 soil classes which consist of 7 classes of available water holding capacity (i.e., 4, 5, 6, 7, 8, 9 and 10 %v/v, adjusted by gravel content) and 4 rootable soil depth classes (i.e., 40, 75, 115 and 150 cm);
2. We select soil classes until achieving 50% area coverage of crop harvested area within RWS, with at least 3 dominant soil classes and at most 5 dominant soil classes;
3. Water-limited yield potential is simulated for all selected soil classes;
4. We discard soil series in which simulated water-limited yield potential is extremely low and highly variable, hence, unlikely to be used for long-term annual crop production;
5. The mean water limited yield potential per buffer zone is calculated by weighing (based on the area fractions per soil class) the simulated water limited yields for each of the selected soil classes

In case the ISRIC-WISE database will be used to select the soil properties, the procedure is as follows:

1. We select the three most dominant soil mapping units (SUID) per RWS buffer. The dominance of a SUID is dependent on its coverage of the crop-specific harvested area within the RWS buffer based on SPAM harvested crop area maps (You et al., 2009).
2. Each SUID consists of soil units, with a maximum of eight soil units per SUID. The area fraction of each soil unit, within the SUID, is given in the WISE database. Each specific soil unit is defined by a particular combination of soil properties, such as soil texture, soil depth, drainage class, etc.
3. Soil properties required for crop growth simulations are retrieved from the soil unit descriptions. Local agronomists and experts might be consulted whenever there is uncertainty about the selected soil units and their properties and to verify if there is a special preference to grow a certain crop on specific soil units.
4. We select soils until achieving 50% area coverage within SUID, after discarding those soils that are not suitable for long-term annual crop production based on the following rules:
   1. rooting depth < 60 cm
   2. soil water holding capacity < 7 cm/m (averaged over the different layers)
   3. % sand > 75% (averaged over the different layers)
   4. if terrain slope data are available, from other sources or local experts, soils with slope > 10% will also be discarded
5. Water-limited yield potential is simulated for all remaining soil units after (4).
6. The mean water limited yield potential per buffer zone is calculated by weighing (based on the area fractions) the simulated water limited yields for each of the soil units in the three dominant SUIDs.
7. Finally, we assume that most soil characteristics are linked to the soil unit (i.e. from the WISE database). However, if also terrain characteristics are used as input for the crop growth model, like the mean terrain slope angle to determine surface runoff, other sources such as a digital elevation or physiographic maps should be used and the characteristics should be linked to the SUID.

3. Determination of soil properties

Soil input data used by different crop simulation models to simulate water-limited yield potentials (Yw) may differ to some extent. However, the basic required soil information by all models consists of (i) rooting depth, (ii) plant-available soil water (PASW, difference in water content between field capacity and permanent wilting point), either as direct input or estimated from soil texture using pedotransfer functions, and (iii) terrain slope and drainage class (for calculating runoff).

(i) Rooting depth: Based on published literature on crop root dynamics, we recommend using a crop-specific maximum rooting depth in soils with no physical or chemical constrains to rooting (typically ˜ 1.5 m for most grain crops in rainfed systems, with exception of flooded lowland rice). If there are known soil constraints to rooting depth (e.g. caliche layer, hardpan, salinity, alkalinity, acidity) the maximum rooting depth for that site is reduced according to the information. Data about maximum rooting soil depth can be retrieved from soil unit descriptions and sometimes from national and global soil databases. However, in some cases the depth of the bottom soil layer in these databases is not necessarily a proxy for the maximum rooting soil depth. In such cases local experts will be consulted to find the actual maximum soil rooting depth for a given soil unit in a given RWS buffer.

(ii) Plant-available soil water: plant available soil water (PASW) is determined by the upper and lower soil limits for water retention (i.e., field capacity and permanent wilting point, respectively). Actual measurements of soil water characteristics are rarely available, so these are typically estimated using pedo-transfer functions (PTF) based on soil texture. Many PTFs are available to derive soil moisture limits (Gijsman et al., 2002). Unless locally-calibrated PTF are available, we recommend using the Saxton and Rawls (2006) PTF, which was satisfactorily calibrated and evaluated based on USDA soil profiles data (http://hydrolab.arsusda.gov/soilwater/Index.htm). Some models (e.g., WOFOST) use directly the PASW value reported in soil databases, e.g. national (SSURGO) and global soil databases (WISE), as input for the simulations. Other models require users to set soil water limits for each soil layer (CERES-type models, CROPGRO, ORYZA, CropSyst, APSIM) while others estimate PASW internally based on soil texture specified by the users for each soil layer (Hybrid-Maize).

(iii) Terrain slope and drainage class: is required to estimate water loss by surface runoff. The assumption that all precipitation will infiltrate and be available for crop water uptake is not realistic as reported by Tittonell et al. (2008) for rainfed maize in Kenya. Hence, surface runoff should be accounted for in simulations of water-limited yields (Yw), especially at locations with slopes > 2% and/or poor drainage. While some models account internally for runoff (CERES-wheat, CERES-Maize, CROPGRO, Hybrid-Maize), others do not. For the latter, we recommend using effective rainfall (i.e. rainfall minus estimated surface runoff) as input for the simulations. The most widely used approach for simulating runoff is the original curve-number approach (USDA) or the simplified approach proposed by Campbell and Diaz & Soltani and Sinclair (see Appendix A). Other ‘ad-hoc' approaches to estimate surface runoff based on location-specific information, can be followed, e.g. by assuming that a fixed fraction of rainfall is lost by runoff, depending on slope angle, drainage class per soil unit, and local management practices (see Appendix B). Information on slope can be retrieved from the description of the ‘typical pedon', sometimes available in the soil unit descriptions. If this is not the case, we may use elevation (calculate slope from a digital elevation model, e.g. SRTM DEM) and/or topography maps supplemented with local agronomist expert opinion. In all cases, it is important that the reported terrain slope corresponds to the area where the crop is mainly grown. In any case, the terrain slope should be representative of the area of simulation, not for a particular field.

 

Appendix A. Estimation of runoff following the USDA curve-number approach

Following Campbell & Diaz (1988) and Soltani & Sinclair (2012), surface runoff can be estimated as:

IF RAIN <= 0.2 x S THEN RUNOFF = 0

ELSE, RUNOFF = [(RAIN – 0.2 x S) ²] / [RAIN + 0.8 x S]

where RUNOFF, RAIN and S (surface storage capacity) are in mm. S values (in mm) values (adapted from Ritchie et al. 1990 & Gijsman et al., 2007) [1]:

Slope (%)	Drainage
	Excellent	Good	Poor
0-2	162	94	48
2-5	143	80	38
5-10	120	64	25
> 10	104	52	16

 

Determination of drainage class (from Gijsman et al., 2007):

Excellent: somewhat excessively to excessively drained. Deep sandy soils (depth >150cm; loamy sand or sandy loam)

Good: moderate to moderately well drained. Shallow sandy soils (depth <150 cm; loamy sand or sandy loam) OR soils with intermediate textures (any depth; loam, silt loam) OR deep soils with light clay textures (depth >150 cm; silty clay, silty clay loam or clay loam).

Poor: somewhat poorly to very poorly drained. Heavy clay soils (any depth; clay) OR shallow soils with light clay textures (<80 cm; clay, silty clay, silty clay loam or clay loam)

For example, for a maize crop grown in a well-managed field (slope < 2%, see LEFT panel) with a deep silt loam soil (good drainage), runoff will be >10 mm when daily rainfall > 60 mm. For a daily rainfall amount of 100 mm, 38% will be lost as surface runoff. In the case of a field located in steep terrain (slope = 6%, see RIGHT panel) with a deep silt loam soil (good drainage), runoff will be > 10 mm when daily rainfall > 45 mm. For a daily rainfall amount of 100 mm, 50% of the water will be lost as surface runoff.

 

Appendix B: Alternative approach for site-specific runoff estimation:

A fixed fraction of precipitation during the whole growing season is assumed to be lost by surface runoff. This loss fraction is determined ad-hoc per region or country, depending on slope angle, drainage class per soil type and local management practices. A literature search on the fraction of total seasonal rainfall lost by surface runoff in Sub-Saharan Africa has been performed[2]. Based on that information we have compiled Tables A and B with the surface runoff fractions of total rainfall as dependent on slope angle, drainage class, and presence of mulching.

Table A. Surface runoff fraction of total rainfall (in %) for soil cultivated with cereals

Drainage class,   Slope angle, in %	Very poor	Insufficient	Moderate	Well drained	Extremely well drained
0-2	20	13.3	6.7	0	0
2-6	26.7	20	13.3	6.7	0
6-10	33.3	26.7	20	13.3	6.7
>10	40	33.3	26.7	20	13.3

 

Table B. Surface runoff fraction of total rainfall (in %) for soil cultivated with cereals and mulching

Drainage class,   Slope angle, in %	Very poor	Insufficient	Moderate	Well drained	Extremely well drained
0-2	10	6.7	3.3	0	0
2-6	13.3	10	6.7	3.3	0
6-10	16.7	13.3	10	6.7	3.3
>10	20	16.7	13.3	10	6.7

Batjes, N.H., 2012.  ISRIC WISE derived soil properties on a 5 by 5 arc-minutes global grid (ver. 1.2). Report 2012/01. ISRIC - World Soil Information, Wageningen, The Netherlands, 57 pp.

Campbell, G.S., and R. Diaz. 1988. Simplified soil–water balance models to predict crop transpiration. p. 15–26. In F.R. Bidinger et al (ed.) Drought research priorities for the dryland tropics. ICRISAT, Pantancheru, India.

Cooper, M., Mendes, L.M.S., Silva, W.L.C., Sparovek, G., 2005. A national soil profile database for Brazil available to international scientists. Soil Sci. Soc. Am. J. 69, 649-652.

Gijsman, AJ, Jagtap, SS, Jones, JW, 2002. Wadding through a swamp of complete confusion: how to choose a method for estimating soil water retention parameters fro crop models. Europ. J. Agronomy 18, 77-106. 

Gijsman, AJ, Thornton, PK, Hoogenboom, G. 2007. Using the WISE database to parameterize soil inputs for crop simulation models. Computers and Electronics in Agriculture 56, 85-100.

Ritchie, J.T., Godwin, D.C., Singh, U., 1990. Soil and weather inputs for the IBSNAT crop models. International Benchmark Sites Network for Agrotechnology Transfer (IBSNAT) Project. In: Proceedings of the IBSNAT Symposium: Decision Support System for Agrotechnology Transfer. Part I. Symposium Proceedings, Department of Agronomy and Soil Science, College of Tropical Agriculture and Human Resources, University of Hawaii, Honolulu, Hawaii, Las Vegas, NV, October 16–18, 1989.

Saxton, K.E., Rawls, R.J., 2006. Soil water characteristic estimates by texture and organic matter for hydrologic solutions. Soil Sci. Soc. Am. J. 70, 1569-1578. See also: (http://hydrolab.arsusda.gov/soilwater/Index.htm)

Soltani, A., Sinclair, TR. 2012. Modeling physiology of crop development, growth and yield. CAB, Wallingford, UK. 322 pp.

Tittonell P, Corbeels M, van Wijk MT, Vanlauwe B, Giller KE, 2008. Combining organic and mineral fertilizers for integrated soil fertility management in smallholder farming systems of Kenya—explorations using the crop–soil model FIELD. Agron J 100:1511–1526

You, L., Z. Guo, J. Koo, W. Ojo, K. Sebastian, M.T. Tenorio, S. Wood, U. Wood-Sichra. Spatial Production Allocation Model (SPAM) 2000 Version 3 Release 1. http://MapSPAM.info.