Accurate land-use information is one of the most important inputs in eco-hydrologic and water-quality models. In the U.S. Corn Belt, agricultural landscapes are dominated by corn–soybean rotations and extensive subsurface tile drainage systems. These features strongly control how water and nutrients, especially nitrate, move from fields to streams. However, most commonly used national land-use datasets do not explicitly show where tile drainage exists or where crops are grown continuously versus in rotation. This can reduce the spatial accuracy of model predictions.
To address this gap, we developed a new 30-meter-resolution dataset, the Tile-drainage and Rotation-Enhanced Cropland (TREC) dataset, for the conterminous United States. Our dataset combines three major public data sources: the USDA Cropland Data Layer, long-term crop frequency layers, and a national tile-drainage map. Using a geospatial workflow, we classified cropland pixels based on both drainage condition (tiled vs. non-tiled) and crop rotation intensity (continuous vs. non-continuous crops). The result is a land-use layer that better reflects real agricultural management patterns.
We evaluated this dataset using the SWAT+ watershed model in two Midwestern watersheds with different tile-drainage patterns. Without changing any model calibration parameters, we compared model results using the traditional cropland dataset versus our new TREC dataset. The overall water balance remained nearly the same, showing that our dataset does not distort total hydrology. However, the spatial distribution of tile flow was much more realistic with TREC. Tile discharge was concentrated in areas that are actually tile-drained, instead of being spread uniformly across cropland.
Model performance metrics also improved when using the TREC dataset, especially at upstream and midstream locations. In addition, simulated crop yields remained consistent with observed values, confirming that our classification approach did not introduce bias into crop simulations.
We made the dataset publicly available along with the code and model files. It can be used not only in SWAT+ but also in other hydrologic and water-quality models. By improving how tile drainage and crop rotation are represented spatially, our dataset supports better identification of nutrient hotspots and more targeted conservation planning.
In short, our work provides a practical, scalable approach to making watershed modeling more spatially realistic in tile-drained agricultural regions.