Soil moisture modeling for agricultural needs in Brazil, France, and the U.S.A.

Sammanfattning: The purpose of this master’s thesis project is to aid Thomson Reuters to forecast soil moisture, enhancing the agricultural productivity of its clients. This thesis focuses on testing the company’s HBV model (HBV-TR) to accurately produce daily soil moisture values in Brazil, France, and the U.S.A. The problem is that, up to this day, the most reliable monitoring of this soil water content is also through a model (the one-layer “Leaky-Bucket model”), whose results are only published as monthly hindcasts. This delay prevents most stakeholders from real-time information and planning. By using input series similar to the Leaky-Bucket inputs, the HBV-TR simulated the target soil moisture for the last thirty-seven years. This project adapted the HBV-TR to calculate soil water with a one-layer and a three-layer version. The HBV-TR daily results are then compared to the monthly target series at consistent dates by the Nash-Sutcliff parameter, volume error, visual aspects, field capacity and evapotranspiration. According to all these performance parameters, the results were sound, showing substantial evidence that this method can safely emulate the Leaky-Bucket model. Ultimately, this project concludes that soil moisture has potential to become the new feature in the company’s forecast portfolio, providing planning capacity to more stakeholders.