Framework for Calibration, Sensitivity, and Uncertainty Analyses of an Agro-Hydrological Model

Investigators: Mahesh L. Maskey and Amanda M. Nelson

Date: 2022

Project Summary

Introduction

Researchers are constantly improving hydrological models in response to climate change stressors in crop systems. However, these models need accurate parameterization (i.e., tweaking certain parameters and equations for more accurate representation) prior to implementation. In addition, it is crucial to examine sensitive parameters and their range of uncertainty since they may be highly dependent on model output. For these tasks, this study considered the Agricultural Policy Environmental Extender (APEX) - a widely used farm and small watershed-scale process-based hydrological model. The objective of this study was to develop a generalized approach to calibrate the APEX model and perform sensitivity and uncertainty analyses to assess influential parameters and identify how grazing impacts farmscale runoff from grassland and annual cropping systems.

Materials and Methods

We used APEX to compare water quality and quantity from a watershed managed with a native prairie grassland and another under an annual system (wheat and oats) based on a dataset with 20-years of measured data, including planting, tillage, fertilizer, pesticide, and surface runoff near El Reno, Oklahoma (Figure 1a).

Model input files were initially developed using the Nitrogen Tracking Tool and then modified to suit a new version of APEX that accounts for grazing. During the development of the model, only parameters relevant to hydrology and sediment were selected based on the literature. Still, 20 parameters needed to be optimized, requiring many model-runs (1060 runs) to determine the possibilities of parameter value combinations. Running 100k simulations requires significant computational resources, therefore we utilized the high-performance computing facilities provided by USDA-SCINet's Office of Scientific Computing for calibration.

To improve the non-linear behaviors of parameters in the existing methods, the proposed method used a normal distribution during calibration. We changed the calibrated parameter set by 5% of the difference in parameter bounds in increments of 0.05 for sensitivity analysis. For uncertainty analysis, we chose the range of 20 parameters between +-3 standard deviations from the average parameter sets.

Results and Discussion

For both watersheds, the calibrated model produced reasonable representations of monthly runoff, while optimized at daily scale (Figure 1b). As seen, the model's monthly values (blue; Figure 1b) of aggregated runoff closely follow the observed ones (red). Further, during calibration, the reasonable values of performance metrics corroborate the model's goodness (Nash-Sutcliffe efficiency and coefficient of determination were higher than 0.70 and 0.69, respectively, but these metrics became less favorable at the validation level).

Results of the model revealed increased biomass and deep percolation in grassland systems (Table 1), but less in cropland. Grazing operations, therefore, resulted in reductions in runoff, sediment yield, and nutrient loading (nitrogen and phosphorus). Grazing reduced forage production from grasslands with slight biomass production increases from croplands.

Conclusion

In this project, we have developed a conceptual framework for calibration, sensitivity, and uncertainty analyses of a hydrological model and validated its ability to quantify runoff dynamics from grazing systems. We predict that the proposed framework could be possible to calibrate an agro-hydrological model capable of simulating different cropping patterns, climate conditions, and management regimes to support NCAAR research in the Mississippi Delta. Additionally, the model can be used to simulate deep percolation and lead to the development of a hydroeconomic model. As a result, we will use this tool to analyze the impact of climate change and population growth on the Mississippi River Valley Alluvial Aquifer.

Project Photos

• Topic:
• Irrigation