DayCent Model Predictions of NPP and Grain Yields for Agricultural Lands in the Contiguous U.S.

Accurate estimation of crop net primary production (NPP) and yields is fundamental for regional analyses of agroecosystem dynamics using process‐based models. In this study, we simulated croplands in the contiguous U.S. using the DayCent ecosystem model with new production algorithms. Crops were divided into crop variety groups based on regional varieties of three major crops (corn, soybeans, and winter wheat) and generic parameter values that were generated for each group. These varieties have been developed through crop breeding programs and enhance production of major crop types in different temperature and precipitation regimes. NPP and yields for the three major crops were evaluated at the county level with reported yields from the National Agricultural Statistics Service (NASS). The predictions of the multiyear average yields in all counties were more accurate than most other published results using process‐based models. DayCent predictions of yields produced an overall R2 of 0.54, 0.54, and 0.38 for corn, soybean, and winter wheat, respectively, with predictions for most counties within ±20% of the NASS reported yields. Our estimations of the total annual NPP for the three crops in the contiguous U.S. are 0.24, 0.09, and 0.06 Pg C yr(−1) for corn, soybean, and winter wheat, respectively. Together, they contribute 7.3% to 14.8% of the total NPP for all vegetation in the contiguous U.S. We conclude that crop variety groups capture heterogeneity in NPP for major crop types and can improve biogeochemical model predictions of NPP for croplands.