AbstractMitigation of greenhouse gas emissions from agriculture requires an understanding of spatial‐temporal dynamics of nitrous oxide (N₂O) emissions. Process‐based models can quantify N₂O emissions from agricultural soils but have rarely been applied to regions with highly diverse agriculture. In this study, a process‐based biogeochemical model, DeNitrification‐DeComposition (DNDC), was applied to quantify spatial‐temporal dynamics of direct N₂O emissions from California cropland employing a wide range of cropping systems. DNDC simulated direct N₂O emissions from nitrogen (N) inputs through applications of synthetic fertilizers and crop residues during 2000–2015 by linking the model with a spatial‐temporal differentiated database containing data on weather, crop areas, soil properties, and management. Simulated direct N₂O emissions ranged from 3,830 to 7,875 tonnes N₂O‐N yr⁻¹, representing 0.73%–1.21% of the N inputs. N₂O emission rates were higher for hay and field crops and lower for orchard and vineyard. State cropland total N₂O emissions showed a decreasing trend primarily driven by reductions of cropland area and N inputs, the trend toward growing more orchard, and changes in irrigation. Annual direct N₂O emissions declined by 47% from 2000 to 2015. Simulations showed N₂O emission variations could be explained not only by cropland area and N fertilizer inputs but also climate, soil properties, and management besides N fertilization. The detailed spatial‐temporal emission dynamics and driving factors provide knowledge toward effective N₂O mitigation and highlight the importance of coupling process‐based models with high‐resolution data for characterizing the spatial‐temporal variability of N₂O emissions in regions with diverse croplands.