The goal of the Global Yield Gap Atlas (GYGA) project is to estimate the yield gap for major food crops in all crop-producing countries based on locally observed data. Unlike past efforts to estimate Yg that rely on gridded weather data as described above, GYGA seeks to use a "bottom-up" approach with location-specific observed weather data. To aggregate results from location-specific observed data to larger spatial areas, the GYGA approach utilizes a hierarchal climate zonation scheme based on a matrix of climate zones (see Wart et al., 2013; van Bussel et al., 2015) as described below.

We expanded the CZ spatial framework to facilitate technology transfer in crop production systems by delineating technology extrapolation domains (GYGA-TEDs) that are defined by a unique combination of CZ and soil water storage capacity to support crop growth. Within a GYGA-TED it is expected that crop and soil management technology options would perform similarly because biophysical conditions governing crop and cropping system response are sufficiently homogeneous.  Hence, the extrapolation domain for field research evaluating crop and soil management options, or comparison of different cropping systems, conducted at a given location can be can spatially delineated by the GYGA-TED in which the study was conducted.

The power of this approach contributes to the effectiveness of agricultural R & D in three ways:

1. to identify field research locations with greatest potential for impact in terms of crop production area with similar climate and soils;
2. to identify regions with greatest potential impact for scaling up adoption of new technologies;
3. to improve both ex-ante and ex-post quantitative assessment of impact from potential or actual adoption of new crop and soil management technologies or alternative crops and cropping systems.

 

Data sources and delineation of GYGA-CZs

The GYGA-CZ scheme is constructed from three categorical variables:

1. growing degree days (GDD)
2. temperature seasonality
3. an annual aridity index (AI)

Grid cell size for the underpinning weather data was 5' grid (roughly 100 km² at the equator).

The GDD were calculated as in Licker et al. (2010) with:

 

in which  is the temperature (°C) for each time step and is  the base temperature (0 °C for our calculations).  Licker et al. (2010) used mean monthly temperatures for the period 1961-1990 from the CRU CL v. 2.0 dataset at 10' grid (http://www.cru.uea.ac.uk/cru/data/hrg/tmc/, (New et al., 2002)) and downscaled it to a 5' grid.

Temperature seasonality was taken from WorldClim (http://www.worldclim.org/current , data for current conditions (~1950-2000), Bioclim4 at 5' grid, (Hijmans et al., 2005)), calculated as the standard deviation of the 12 mean monthly temperatures × 100 (note that mean monthly temperatures are in °C × 10).

The annual aridity index values were taken from CGIAR-CSI (http://www.cgiar-csi.org/data/global-aridity-and-pet-database, at 30'' grid, (Trabucco et al., 2008; Zomer et al., 2008)), calculated as:

in which MAP is the mean annual precipitation (mm × 100) and MAE the mean annual potential evapotranspiration (mm × 100). We aggregated these AI values to a 5' grid. in which MAP is the mean annual precipitation (mm) and MAE the mean annual potential evapotranspiration (mm). We aggregated these AI values to a 5' grid, taking the spatial average of the 100 cells at 30 arcsecond resolution within each 5 arcminute gridcell. Next, we multiplied the spatially averaged AI with 10000.

Following Mueller et al. (2012), only terrestrial surface covered by at least one of the major food crops (maize, rice, wheat, sorghum, millet, barley, soybean, cassava, potato, yam, sweet potato, banana and plantain, groundnut, common bean and other pulses, sugar beets, sugarcane) was considered in this zonation scheme. To avoid inclusion of areas with negligible crop production, only grid cells with sum of the harvested area of major food crops > 0.5% of the grid cell area were accounted for, based on HarvestChoice SPAM crop distribution maps (You et al., 2006; You et al., 2009), which update geospatial crop distribution data of Monfreda et al. (2008).

The resulting range in values for GDD and aridity index were divided into 10 intervals, each with 10% of grid cells with harvested area of the major food crops, and combined in a grid matrix with 3 ranges of temperature seasonality to give a total of 300 classes. Of these, only 265 occur in regions where major food crops are grown.

This classification of the variables resulted in the following ranges:

GDD (°Cd)	GYGA-CZ Value
0 - 2670	1000
2671 - 3169	2000
3170 - 3791	3000
3792 - 4829	4000
4830 - 5949	5000
5950 - 7111	6000
7112 - 8564	7000
8565 - 9311	8000
9312 - 9850	9000
> 9851	10000

 

AI (-)	GYGA-CZ Value
0 - 2695	000
2696 - 3893	100
3894 - 4791	200
4792 - 5689	300
5690 - 6588	400
6589 - 7785	500
7786 - 8685	600
8686 - 10181	700
10182 - 12876	800
> 12877	900

 

Temperature seasonality	GYGA-CZ Value
0 - 3832	01
3833 - 8355	02
> 8356	03

 

Values of the GYGA-CZs

Value for each cell indicates the unique combination climate for that cell. The value of the GYGA-CZs is constructed by the sum of the three GYGA-CZ variables. A few examples:

GYGA-CZ value 6801 (6-8-01) =	GYGA-CZ Value GDD	6000 +
	GYGA-CZ Value AI	800 +
	GYGA-CZ Value Temperature seasonality	01
GYGA-CZ value 10402 (10-4-02) =	GYGA-CZ Value GDD	10000 +
	GYGA-CZ Value AI	400 +
	GYGA-CZ Value Temperature seasonality	02

 

Data sources and delineation of GYGA-TEDs

Each GYGA-TED is a unique combination of a GYGA-CZ and water storage capacity in the rootable soil depth, the latter defining the root zone plant-available water holding capacity (RZPAWHC). GYGA-TEDs were created for Africa and US.

Africa GYGA-TED

Plant-available soil water holding capacity in the root zone was taken from the Africa Soil Information Service (AfSIS; http://www.isric.org/projects/afsis-gyga-functional-soil-information-sub-saharan-africa-rz-pawhc-ssa version April 2015, af_agg_ERZD_TAWCpF23mm__M_1km.tif, with a resolution of 1 × 1 km). RZPAWHC is determined by evaluation and spatial interpretation of the AfSIS soil profile database:

1. the so-called "legacy" soil point data (Africa Soil Profiles database v1.2;18,500 points) and
2. all AfSIS sentinel site soil point data (approx. 9,600 points, including spectral data and 10% wet chemistry reference data), which were provided by AfSIS for this collaboration,
3. SoilGrids1km layers (www.isric.org/explore/soilgrids) produced at ISRIC using global models; updated and fine-tuned fitting a continental model, with finer resolution satellite data and above mentioned soil data, resulting in AfrSoilGrids250m (www.isric.org/projects/africa-soilgrids-soil-nutrient-maps-sub-saharan-africa-250-m-resolution).

As a component of the GYGA-TED spatial framework, RZPAWHC values are classified into nine 25 mm classes, with =< 50 mm and >250 mm as lower and upper classes, respectively. 

This classification of the variables resulted in the following ranges:

RZWHC (mm)	RZWHC Value
0 - 50	100000
50 - 75	200000
75 - 100	300000
100 - 125	400000
125 - 150	500000
150 - 175	600000
175 - 200	700000
200 - 225	800000
225 - 250	900000
> 250	1000000

US GYGA-TED

Plant-available soil water holding capacity in the root zone was taken from gSSURGO database (Soil Survey Staff; Resolution of 250 × 250 m). There are two classifications, both based on the RZPAWHC values. For the fine TEDs the RZPAWHC values are classified into thirteen 25 mm classes, with 0-25 mm and >300 mm as lower and upper classes, respectively. For the coarse TEDs the RZPAWHC values are classified into seven 50 mm classes, with 0-50 mm as the lower class and >300 mm as the upper class. 

This classification of the variables resulted in the following ranges: