Pantropical calibrated gross primary productivity (GPP) of cropland and grassland at 30 m (2020-2024)

This dataset provides annual calibrated gross primary productivity (GPP) for cropland and grassland at 30 m spatial resolution for the pantropical belt (48°S-36°N) for the years 2020-2024, and the rate of change of GPP for both land-cover classes. The calibrated GPP is derived from the Global Pasture Watch uncalibrated GPP (uGPP) product (Isik et al., 2025), which uses a uniform light-use efficiency (LUE) of 1 gC m-2 d-1 MJ-1 across all land cover types. Here, uGPP is multiplied, pixel by pixel and year by year, by a land-cover-specific LUE factor taken from the Global Harmonized Land Cover (Moreno et al., 2026) map of the same year. The result is then masked to the cropland and grassland classes of that year. Layers Layer Value Unit Data type NoData Annual GPP (cropland) calibrated GPP of the year gC m-2 yr-1 UInt16 65535 Annual GPP (grassland) calibrated GPP of the year gC m-2 yr-1 UInt16 65535 Rate of change 2020-2024 (cropland) [mean(GPP2023, GPP2024) - mean(GPP2020, GPP2021)] / 4 gC m-2 yr-1 per year Float32 -9999 Rate of change 2020-2024 (grassland) [mean(GPP2023, GPP2024) - mean(GPP2020, GPP2021)] / 4 gC m-2 yr-1 per year Float32 -9999 Methods Input alignment. The annual uGPP and GHLC land cover maps for 2020-2024 were placed on one common grid (0.00025°, EPSG:4326) using virtual rasters (VRT). Both products share the same pixel grid, so no resampling was applied. Calibration. For every pixel and year: GPP = uGPP × LUE(land-cover class of the same year), using the LUE factors below. uGPP values ≥ 65000 and land-cover nodata (255) are treated as nodata. Values are rounded to 1 gC m-2 yr-1. Masking. The annual GPP was retained only where the land cover for that year belongs to the cropland (classes 2-9) or grassland (classes 10-11) group; all other pixels are nodata. Rate of change. Computed from the masked annual GPP as rate = [mean(GPP2023, GPP2024) - mean(GPP2020, GPP2021)] / 4, i.e. the difference between the late and the early two-year means, divided by the 2020-2024 period length (4 years). A pixel has a rate only if it belongs to the group in all four years used (2020, 2021, 2023, 2024) and has valid, non-capped GPP in all of them. Processing and quality control. Processing was done in 1° × 1° tiles in parallel. Each tile is tagged with the input version and a fingerprint of the LUE table. After processing, all tiles were scanned for block-shaped gaps caused by failed data transfers (areas missing in one year but valid in all other years); affected tiles were recomputed. LUE factors used (gC MJ-1) Class Land cover LUE Source 2 Annual crops - soybean 1.64 Xin et al. (2015) 3 Annual crops - maize 2.78 Xin et al. (2015) 4 Annual crops - rice 2.20 Xin et al. (2017) 5 Annual crops - other 1.80 Yan et al. (2015) 6 Perennial crops - coffee 1.20 Chen et al. (2014) 7 Perennial crops - cocoa 1.57 MOD17 User's Guide 8 Perennial crops - rubber 2.75 MOD17 User's Guide 9 Perennial crops - oil palm 1.80 MOD17 User's Guide 10 Cultivated grassland 0.50 MOD17 User's Guide 11 Natural grassland 0.50 MOD17 User's Guide Input data uGPP: Global Pasture Watch annual uncalibrated GPP, 30 m, v1.2 (2020-2024) - Isik et al. (2025) Land cover: Global Harmonized Land Cover, 30m, v1.2 (2020-2024) - Moreno et al. (2026) Tropical layer: Tropical data used to extract the extent - Pantropical Tree Cover Density Usage notes and limitations Land-cover changes. The rate-of-change layers are restricted to pixels that remain in the same group across 2020, 2021, 2023, and 2024. Rate period. The two 2-year means are centered on 2020.5 and 2023.5; the rate is divided by the 4-year span 2020-2024 as defined above. Extent. Only the pantropical belt (48°S-36°N) is covered; areas outside the land-cover or uGPP extent are nodata.