Application of Fuzzy Modeling to Estimate Soil-Water Evaporation

摘要

The simplified energy balance method for estimating soil-water evaporation using infrared thermometry requires the use of assumptions often found to be invalid. Without these assumptions, the method is too expensive in labor or equipment to be applied at remote locations, and eliminates the possibility of using remote sensing to collect the necessary data. In this study, a method is described that eliminates the necessity of using these assumptions, while retaining model simplicity and versatility. The proposed method uses an optimized fuzzy learning from example (OFLFE) algorithm to develop the model. The OFLFE algorithm develops a fuzzy rale base using a learning from example approach, fine tunes the rule base using a recursive least squares approach, and independently optimizes the rale base using constrained nonlinear optimization. K-fold cross-validation was used to test the model. The results indicate that the fuzzy model can predict evaporation better than the existing simplified energy balance method, with a reduction in root mean squared error (RMSE) of nearly 50%.