regression modeling

The aim of this study is to propose an approach for modeling spatiotemporal changes in rainfall that can be used as input for rainfall-runoff models.

To achieve this, rainfall data from four rain gauge stations in the Paskouhak catchment were used. Five parameters, including elevation, slope, aspect, longitude, and latitude, were identified. The different combinations of these five parameters were prioritized using the gamma test in WinGammaTM software. After the use of different regression models, the best model was selected based on evaluation criteria such as R2, RMSE, and the Taylor diagram. A raster map of a selected rainfall event was drawn in the Arc GIS environment. Finally, using the proposed approach of relative equations, the spatiotemporal changes in rainfall were modeled.

The results showed that using a second-degree nonlinear model and parameters of elevation and latitude, it is possible to accurately obtain the spatial distribution of rainfall in the form of a regular pixel grid (100 square meters) with high precision (R2=0.917 and RMSE=0.2277).

In different rainfall events in small catchment areas, the variation in rainfall in each pixel is almost constant relative to other pixels, including the rain gauge station, the proposed approach in this study can model the spatiotemporal changes of each rainfall event as a three-dimensional matrix in the study area. The approach can be valuable in predicting potential flood events and in water resource management and planning. However, further research is required to validate the results and test the approach in other areas.