Evaluating the suitability of NLDAS-2 climate data for watershed modeling: A case study of AnnAGNPS in Goodwin Creek Experimental Watershed in Mississippi

Authors: RÉBILLOUT, LUC - University Of Mississippi; POPHET, NUTTITA - University Of Mississippi; AL-HAMDAN, MOHAMMAD - University Of Mississippi; Ozeren, Yavuz; Bingner, Ronald

Submitted to: Journal Hydrologic Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/22/2025
Publication Date: 1/20/2026
Citation: Rébillout, L., Pophet, N., Al-Hamdan, M., Ozeren, Y., Bingner, R.L. Evaluating the suitability of NLDAS-2 climate data for watershed modeling: A case study of AnnAGNPS in Goodwin Creek Experimental Watershed in Mississippi. Journal of Hydrologic Engineering, 31(2). https://doi.org/10.1061/JHYEFF.HEENG-6596. 2026.
DOI: https://doi.org/10.1061/JHYEFF.HEENG-6596

Interpretive Summary: This study evaluates the effectiveness of the Annualized Agricultural Non-Point Source (AnnAGNPS) model in simulating water and soil movement in the Goodwin Creek Experimental Watershed in north-central MS. To overcome the challenge of obtaining direct climate measurements, the researchers utilized the North American Land Data Assimilation System Phase 2 (NLDAS-2) dataset, which provides comprehensive information on precipitation, temperature, and wind patterns as input to AnnAGNPS. By comparing the AnnAGNPS model's performance using NLDAS-2 data against actual measurements, the study found that the model produced satisfactory results even without direct measurements. This finding has significant implications for understanding hydrological processes and sediment transport in the watershed, enabling informed decision-making regarding water resource management and the environmental impact of agricultural practices. By leveraging advanced modeling techniques and reliable datasets like NLDAS-2, this study enhances our understanding of hydrological processes in watersheds, particularly when direct climate measurements are challenging. The findings can support effective land management strategies and inform watershed planning efforts, ultimately contributing to sustainable water resource management and promoting environmentally conscious agricultural practices.

Technical Abstract: Watershed modeling is a critical component in the management of water resources and the understanding of the impact of agricultural practices on the environment. One of the essential inputs required for such simulations is climate data. However, obtaining accurate and reliable climate data can be challenging, particularly when field observations are unavailable. In such scenarios, modelers rely on data from various sources, such as regional climate models or reanalysis datasets, to simulate the watershed response. This paper presents a study to compare the performance of Annualized Agricultural Non-Point Source (AnnAGNPS) using various sets of input climate data from the North American Land Data Assimilation System Phase 2 (NLDAS-2) against field observations. The main objective is to assess the suitability of NLDAS-2 climate data for AnnAGNPS when local gauge data is not available. The advantage of NLDAS-2 is that it offers a spatially and temporally continuous dataset, which can be used to generate long-term simulations. The relevant variables obtained from NLDAS-2 come from the forcing dataset A that includes: precipitation, temperature (daily minimum and maximum), specific humidity (used to compute the dew point), wind speed, wind direction, solar radiation, and potential evapotranspiration. Goodwin Creek Experimental Watershed, located in North Mississippi is used as a test case with a study period from January 1982 to January 2002. A direct comparison of input NLDAS-2 climate data against field observations is first performed. Then, a sensitivity analysis of AnnAGNPS outputs to changes in specific input climate variables is performed. The outputs of each simulation using different combinations of input climate data are compared against the best available simulation that used field observations using the Normalized Nash-Sutcliffe coefficient as a metric of performance and accuracy. The results of this sensitivity analysis reveal that precipitation data is the primary driver of the simulation results. The simulation results are then compared against measured total runoff, effective rainfall, and total sediment load. Comparisons against field observations will be over the first ten years of the study period (based on available data). The results indicate that using NLDAS-2 precipitation data with the bank erosion option of AnnAGNPS turned on provides a viable source of climate data when field observations are unavailable.