Abstract Archive Select a year below to view:
Model performance and uncertainty analysis of APEX for hydrologic predictions in the Mississippi Delta region
Proceedings of the 2022 Mississippi Water Resources Conference
Year: 2022 Authors: Mendez J., Ramirez Avila J.J., Locke M.
An uncertainty analysis assesses the effects of knowledge gaps with respect to understanding the model output and performance. In order to study the impact of using different weather stations and soil conditions (inputs) on model outputs, the Agricultural Policy/Environmental Extender (APEX) model was used to predict runoff from agricultural fields in the Mississippi Delta. Two scenarios were proposed: (i) varying climate inputs from five stations located within a radius of 40-km from the study site (Minter at 37.8km-N, Greenwood at 15.9km-NE, Belzoni at 36km-SW, Lexington 37km-SE, and Moorhead at 25km-W); and (ii) assuming a homogeneous soil type in the fields from the actual four soil series present in the study site. An APEX model (v.1501) was calibrated and validated using four years (1996 to 1999) of available information for runoff, water quality, soil characteristics and in-situ climate dataset from two fields (DH1-18ha and DH2-12ha) within the Deep Hollow Watershed (DH) in Leflore county, MS. The APEX baseline scenario used the actual soil series distribution in the fields (Alligator, Dubbs, Dundee and Tensas) and climate information collected in-situ. For all the scenarios, APEX model performance of runoff predictions was assessed using three objective functions: Nash Sutcliffe Efficiency (NSE), percentage of bias (PBIAS), and Kling-Gupta efficiency (KGE). The baseline scenario yielded satisfactory results in the prediction of monthly runoff with NSE values of 0.67 and 0.90; PBIAS values of 29% and 21%; and KGE values of 0.61 and 0.73 for DH1 and DH2, respectively. For the DH2 field, APEX runoff predictions were only satisfactory when using information from the nearest weather station to the study site (Greenwood). For DH1 the same station yielded a better performance than the other three stations, but none yielded satisfactory. For both fields the use of the Greenwood dataset represented an underestimation of runoff. When assuming a homogeneous soil series for the entire fields' area, the use of the Alligator and Tensas soil series (silty clayly and clayly soils with hydrologic groups C and D, respectively) improved the APEX runoff estimations and performance. When assuming the soil series Dubbs and Dundee (very fine sandy loam and loamy with hydrologic groups C and B, respectively) runoff underestimation increased with respect to the baseline. Overall, results offer valuable information to modelers of watershed in the Mississippi Delta, in their decision-making regarding model setup and output interpretation for areas with limited climate and soil information.