Regional simulation of water-carbon fluxes and winter wheat yield with assimilation of multi-source remote sensing data in the Haihe River Basin

• Coupling Noah-MP & MLDAS enhances ET, GPP, & WWY estimation. • LAI and SSM in DA_IRR scheme outperform OL scheme significantly. • Irrigation can improve simulation results, not as much as data assimilation. Accurately estimating evapotranspiration (ET), gross primary productivity (GPP), and winter wheat yield (WWY) is of great significance for climate change research, agricultural development, and water resources management in the cultivation regions of winter wheat. In this study, the coupling strategy of the Noah-MP land surface model and the multi-pass land data assimilation scheme (MLDAS) has been adopted to simulate the water-carbon fluxes and yield within the winter wheat-covered area in the Haihe River Basin (HRB). Utilizing the crop and irrigation modules embedded within the Noah-MP model, in conjunction with a double-variable assimilation approach encompassing leaf area index (LAI) and surface soil moisture (SSM), this study designed four schemes: Open Loop (OL), Open Loop with Irrigation (OL_IRR), Data Assimilation (DA), and Data Assimilation with Irrigation (DA_IRR). Validation results at both field and regional scales revealed that all schemes possess the capability to reasonably simulate spatiotemporal distributions for water-carbon fluxes (ET and GPP) and WWY to a certain extent. Through integrating multi-source observations, data assimilation schemes (DA and DA_IRR) outperformed open loop schemes (OL and OL_IRR). Notably, DA_IRR exhibited the most significant improvement, with reductions in root mean square deviation (RMSD) relative to the OL scheme as follows: 73.57 % for LAI (compared to a baseline RMSD of 0.88 m 2 /m 2 ), 41.60 % for SSM (compared to 0.05 m 3 /m 3 ), 54.99 % for root-zone soil moisture (RZSM) (compared to 0.08 m 3 /m 3 ), 14.98 % for ET (OL: 1.22 mm/day), 35.42 % for GPP (OL: 4.24 g/m 2 /day), and 67.23 % for WWY (OL: 518.10 kt). Meanwhile, this study found that while irrigation slightly improves simulation results (e.g., a 21.19 % RMSE reduction for RZSM with OL_IRR), its effect is less significant than data assimilation. This highlights the potential for additional utilization of the integration approach involving the Noah-MP model and the MLDAS methodology to winter wheat cropland ecosystems, as demonstrated in this investigation.