Improved Calibration of Apex Model for a Small Watershed Managed with No-till System.

Surface runoff from unprotected agricultural soils can lead to transport of sediments, nutrients and agricultural chemicals. These pollutants can affect water bodies, especially increasing toxicity and eutrophication. It would be challenging and costly to monitor long-term agricultural management, conservation practices and climate change impacts on runoff on a wide spatial scale, thus, use of hydrological models to simulate land management and climate impacts on runoff is sometimes required. The Agricultural Policy and Environmental Extender (APEX) model, a process based, spatially- distributed hydrologic model is suitable for runoff simulation for small watersheds. However, the APEX model requires proper calibration and validation for making accurate predictions. The traditional trial and error calibration method is time consuming and its success can depend on the modeler’s experience with the model.  These disadvantages can be avoided by using automated calibration. In this study we applied parameter estimation (PEST) model, a method based on inverse modeling approach to estimate model parameters by minimizing the sum of squared weighted residuals when comparing simulated and measured runoff data.  In this study, the comparison of the trial and error method and PEST showed that the calibration performance criteria for PEST (R²=0.82 and NSE=0.82) were generally higher compared to the trial and error method (R²=0.65 and NSE= 0.82) for a land managed with corn (Zea mays L.)-soybean (Glycine max L.)-rye (Secale cereal L.) rotation during a period from 2006-2011. The PEST calibration performance criteria (R²=0.70 and NSE=0.69) were also higher compared to the trial and error method (R²=0.46 and NSE= 0.45) for a land managed with continuous corn under no-till. Better calibration was achieved in the growing season compared to the winter period possibly due to reduced accuracy in capturing winter processes and error in winter weather data measurement. In conclusion, PEST improved calibration and reduced calibration time for the small no-till watershed in this study.