Global gridded crop harvested area, production, yield, and monthly physical area data circa 2015

Here we describe methods for the GAEZ+ 2015 Annual Crop Data, and the GAEZ+ 2015 Monthly Cropland Data. The Annual Crop Data was generated first, then the Monthly Cropland Data was calculated based on the Harvest Area results of the Annual Data (Fig. 1).

GAEZ+ 2015 Annual Crop Data Methods

The GEAZ+ 2015 Annual Crop Data updates the 2010 GAEZ v4 crop harvest area, yield, and production maps^6,7 (identified as Theme 5 in ref. ⁷) using national-scale data on the change in crop harvested area and livestock numbers from 2010 to 2015, based on statistics for 160 crop groups, and cattle and buffalo, from FAOSTAT⁵.

Three datasets were used to produce GAEZ+ 2015 Annual Crop Data:

1. 1.

   FAOSTAT crop production domain: annual, country-level data on crop harvested area (H) and crop production (P) for each crop from the FAOSTAT database (Table 1)

   Full size table

2. 2.

   GAEZ v4^6,7 gridded global annual harvested area, yield, and production by crop for the 26 FAOSTAT crops and crop categories at 5-minute resolution

3. 3.

   Global Administrative Unit Layer (GAUL 2012)¹³ data. GAUL 2012 reports the fraction of each global 5-minute grid cell that falls within a given country or disputed territory. There are 275 unique global administrative units.

Step 1. Calculate crop changes from 2010 to 2015 by country:

For each country, we extracted the harvested area (H) and crop production (P) for each of the 160 FAOSTAT crop categories, c, from the FAOSTAT database. We averaged three years (2009–2011) of annual national crop harvested area data to represent 2010 national crop harvest area, H₂₀₁₀, and three years (2014–2016) of annual crop harvested area data to represent 2015 national crop harvest area, H₂₀₁₅, then calculated a ratio, rH_c, of 2015 to 2010 harvested areas for each crop c in each country, and equivalently, for crop production:

This results in 160 rH and rP values per country. If harvest area and production values for a particular crop are zero or unreported in the FAOSTAT data, then rH_c and rP_c are both set to 1.0 (i.e., no change from 2010 to 2015). Three years of data are averaged (2009 – 2011 and 2014 – 2016) to account for missing data for some country/year combinations and to avoid emphasizing reported outliers.

Step 2. Aggregate FAOSTAT-based ratios to the GAEZ crop categories:

We followed the crop aggregation methods of the GAEZ model to aggregate the FAOSTAT crop list (160 unique crops as of 2019) to 26 crops (see Table 1). For each of the 26 GAEZ crop categories, if there is more than one matching FAOSTAT crop (see Table 1) then we applied an area-weighted average (based on FAOSTAT year 2015 harvested area) of the FAOSTAT crops within each country to the rH and rP values for that crop and country. This results in 26 rH and rP values per country. There was one exception to this: the GAEZ_2010 crop category ‘fodder crops’ was an aggregate of 17 FAOSTAT crops (see Table 1) for which harvest area data are no longer reported on FAOSTAT; i.e., GAEZ_2010 had obtained FAOSTAT data on fodder crops circa 2010, but FAOSTAT no longer provides any data on fodder crops for any year. We assumed that the 2010 to 2015 fractional change in fodder crop harvest area in each country was proportional to the change in the FAOSTAT reported national herd sizes for cattle and buffalo livestock data⁵ for that country, following the same methodology as for crop harvested area change (see Step 2 below). This method assumes a negligible international trade of fodder crops as indicated by bilateral trade matrices available from FAOSTAT.

Step 3. Apply country-level ratios to grid cells:

Calculated country-level ratios were then applied to each grid cell k, using the GAUL_2012¹³ definitions for which grid cells fall within which countries. Some grid cells are split between two or more countries. In this case, all model output variables for the grid cell are divided between the countries based on the fraction of grid cell area falling within the country i:

where ({H}_{c,2015}^{k}) is the year 2015 harvested area (or production) for crop c in grid cell k; ({f}_{i}^{k}) is the fraction of country i in grid cell k, and rH_c,i and rP_c,i are the ratios for crop c in country i as calculated in Eqs. 1 and 2. This results in 26 H and P values per grid cell. If the sum of all crop harvest areas exceeds 99% of the grid cell area, all crop harvest areas are reduced equally to fit within 99% of the area.

Special Case: Sudan

FAOSTAT data for years before 2011 report data for Sudan, and for South Sudan and Sudan after 2011. To compute the ratios for these grid cells, we split the 2010 data for Sudan into a virtual ‘North’ Sudan and ‘South_Sudan’, using the data for the year 2012, which was reported for both countries. We then used these generated 2010 data and applied the same methodology as described above to calculate changes in harvested areas and production in all grid cells in both countries.

Special Case: Small regions and islands

Forty-nine countries – generally small regions or islands – had no data reported for crop harvested area by FAOSTAT. We assumed that there was no change in crop harvested area for the grid cells in these countries. Note that many may have had zero ha as previously-reported crop area in GAEZ v4. These countries are (the number following each region is the region’s number in ADM0_CODE in the GAUL_2012 data¹³):

Anguilla (9), Aruba (14), Ashmore_and_Cartier_Islands (16), Azores_Islands (74578), Baker_Island (22), Bassas_da_India (25), Bird_Island (32), Bouvet_Island (36), British_Indian_Ocean_Territory (38), Christmas_Island (54), Clipperton_Island (55), Cocos (Keeling)_Islands (56), Europa_Island (80), French_Southern_and_Antarctic_Territories (88), Glorioso_Island (96), Greenland (98), Guernsey (104), Heard_Island_and_McDonald_Islands (109), Howland_Island (112), Isle_of_Man (120), Jarvis_Island (127), Jersey (128), Johnston_Atoll (129), Juan_de_Nova_Island (131), Kingman_Reef (134), Kuril_islands (136), Madeira_Islands (151), Mayotte (161), Midway_Island (164), Navassa_Island (174), Netherlands_Antilles (176), Norfolk_Island (184), Northern_Mariana_Islands (185), Palmyra_Atoll (190), Paracel_Islands (193), Pitcairn (197), Saint_Helena (207), Scarborough_Reef (216), Senkaku_Islands (218), South_Georgia_and_the_South_Sandwich_Islands (228), Spratly_Islands (230), Svalbard_and_Jan_Mayen_Islands (234), Tromelin_Island (247), Turks_and_Caicos_Islands (251), United_States_Virgin_Islands (258), Wake_Island (265), Gibraltar (95), Holy_See (110), Liechtenstein (146).

Special Case: Disputed Areas

Some grid cells in the GAUL_2012¹³ cell-table database are assigned to nine disputed areas, rather than to specific countries. We assumed that there was no change in crop harvested area or production from 2010 to 2015 for grid cells these disputed areas. These areas are (the number following each region is the region’s number of the ADM0_CODE in the GAUL_2012¹³ data):

Abyei (102), Aksai_Chin (2), Arunachal_Pradesh (15), China/India (52), Hala’ib_Triangle (40760), Ilemi_Triangle (61013), Jammu_and_Kashmir (40781), Ma’tan_al-Sarra (40762), Falkland_Islands_(Malvinas) (81).

Step 4. Compute 2015 crop yields:

Crop yields were computed for each crop, c, and grid cell, k, as the ratio of crop production to crop harvest area (if harvest area, H_c,k,2015, is zero, then yield, Y_c,k,2015, is set to zero):

The resulting gridded global data are:

1. A.

   GAEZ+ 2015 Crop Harvest Area¹⁴

2. B.

   GAEZ+ 2015 Crop Yield¹⁵

3. C.

   GAEZ+ 2015 Crop Production¹⁶

This new data product consists of 156 data files in geotiff format, one rainfed harvested area file and one irrigated harvested area file for each crop harvest area (1000 ha (10⁷ m²) per 5-minute grid cell), crop production (1000 tonnes (10⁶kg) per 5-minute grid cell), and crop yield (tonnes per ha (10⁻¹kg m⁻²) per 5-minute grid cell), for each of the 26 GAEZ crops or crop categories in Table 1.

GAEZ+ 2015 monthly cropland area methods

Two datasets were used to produce monthly cropland area by crop and by irrigated vs rainfed management. These are:

1. 1.

   GAEZ+ 2015 Annual Harvested Area¹⁴ (as developed above)

2. 2.

   MIRCA2000 cropland area⁴

Step 5. Harmonize the GAEZ+ 2015 and MIRCA2000 crop lists

The MIRCA2000⁴ cropland product provides monthly growing area grids (gridded physical cropland area) for 26 irrigated and rainfed crops and crop categories, as well as cropping calendars that identify the planting month and harvesting month for each crop (via ‘subcrops’ – see below). However, the MIRCA2000 crop list is not the same as the GAEZ+ 2015 crop list; we matched each crop type in the GAEZ+ 2015 crop list to a crop type in the MIRCA2000 crop list to enable the application of MIRCA2000 crop calendars to GAEZ+ 2015 crops (Table 2). Out of the 26 GAEZ+ 2015 crops, 18 had clear 1:1 matching crop categories within MIRCA2000. The remaining 8 crops were matched based on general crop characteristics, i.e., annual vs. perennial, or to unmatched MIRCA2000 cereals.

An essential component of the MIRCA2000 cropland dataset is the identification of subcrop categories within each crop category to split crops into areas grown in different seasons, or crops with different planting and harvesting dates within the same season. Up to 5 subcrops can be defined to represent such multi-cropping practices. Below, we use the following notation:

H_G = annual harvested area from the GAEZ+ 2015 product for a given crop

H_M =annual harvested area calculated from the MIRCA2000 data for a given crop

A_M,n = cropland area of MIRCA2000 crop, subcrop n, by month

A_G,n = cropland area of GAEZ+ 2015 crop, subcrop n, by month

A_G = cropland area of GAEZ+ 2015 crop, by month

Step 6. Apply MIRCA2000 monthly crop calendars to GAEZ+ 2015 annual data

To generate the monthly cropland physical area of GAEZ+ 2015 crops, we followed these steps for each GAEZ crop in each grid cell:

1. 1.

   For a given GAEZ crop in a given grid cell, is the area reported >0 for the matching MIRCA2000 crop?

   1. a.

      If YES, then use the MIRCA2000 data for the grid cell and crop considered.

   2. b.

      If NO, then find the closest grid cell with the matching MIRCA2000 crop category, and apply the MIRCA2000 crop rotation from that grid cell to the given crop/grid cell combination for the following steps.

2. 2.

   Does the matching MIRCA2000 crop category (Table 1) have more than 1 subcrop?

   1. a.

      If NO, then A_G = H_G for all months of the cropping season, as defined by the MIRCA2000 crop calendar.

   2. b.

      If YES, then for each subcrop category n, apply the ratio of A_M,n/H_M to H_G, then sum the subcrop areas within each month such that:

   $${A}_{G}=sum _{n}frac{{A}_{M,n}}{{H}_{M}}{H}_{G}$$

3. 3.

   For each month and each grid cell, check if the sum of all crops (irrigated and rainfed) is greater than the 99% of area of the grid cell. We assume that at least 1% of land must be retained as non-cropland for agricultural infrastructure such as roads, buildings, irrigation infrastructure, and other landcovers (e.g. rivers, wetlands).

   1. a.

      If NO, then no further processing is done.

   2. b.

      If YES, then reduce crop area by the excess value based on a removal order (Table 2). Rainfed crops have higher removal order numbers for the excess truncation (starting with 1) before removing irrigated crops, until the cell area is not exceeded. A large removal number (e.g., 20) indicates that the crop’s land is unlikely to be removed. Large priority numbers are given to the staple crops to ensure these important food producing lands are consistent with FAOSTAT country data.

The maximum monthly amount of physical cropland that was removed by step 3 is 711,543 ha, which is 0.05% of total global cropland physical area.

The resulting global gridded data from Step 6 are monthly time series of cropland physical area by crop, subcrop, and production system, called GAEZ+_2015 Monthly Cropland Data¹⁷. Combining the MIRCA2000 crop calendar and subcrop rotation information with the GAEZ+ 2015 annual data allows for the representation of crop seasonality; e.g., Fig. 2 shows the aggregate monthly cropland physical area for Rice 1 and Rice 2 (two sub-crops of rice) over the northern hemisphere, clearly illustrating the two main rice-growing seasons.

Source: Ecology - nature.com