Annual cultivated extent and agricultural land use expansion across the central grasslands of North America, 1996–2021

Accurate monitoring of cropland dynamics in North American grasslands is essential for assessing biodiversity threats, guiding sustainable land management, and ensuring food security amid rapid environmental change. We developed a 30-meter resolution dataset capturing annual ’active’ and ’cumulative’ cropland (1996 to 2021) across the central grasslands of North America using an Attention U-Net convolutional neural network. Our bias-corrected estimates reveal that while active cropland showed no significant change (+3.51 ± 1.32 million ha, p = 0.17), the cumulative cropland footprint expanded by 17 % (+20.64 ± 0.93 million ha, p < 0.05) relative to the baseline (1996–2000), reaching 142.21 ± 4.84 million ha by 2021. This divergence indicates substantial new conversion or recultivation of previously restored grasslands, occurring at a consistent rate of 0.98 ± 0.04 million ha per year (p < 0.001). Mexico showed the largest relative gain in cumulative cropland area, expanding by nearly half (48 %, 1.69 ± 0.06 million ha, p < 0.05). By distinguishing between active and cumulative cropland extents, our dataset enables differentiation between short-term, intermittent cultivation and longer-term land-use legacies, allowing for more nuanced assessments of agriculture's cumulative effects on biodiversity and critical ecosystem services at the biome scale. This approach provides critical information for conservation planning and sustainable land management across North American grasslands.