Application of Support Vector Machine and Boosted Tree Algorithm for Rainfall-Runoff Modeling (Case Study: Tabriz Plain)

Abstract

This research focused on the application of Support Vector Machine (SVM) and Boosted Trees (BT) algorithms for simulating precipitation and runoff in two stations, Akhula and Pole Senikh, in the Tabriz Plain, Iran. Meteorological and hydrometric data were collected from 24 stations in the Tabriz watershed, obtained from the Regional Water Company and East Azerbaijan Meteorological Organization. Precipitation and runoff values were used as input to the model with a one-day time lag, and monthly runoff values were estimated and compared with monthly observations using evaluation criteria. The results showed that for both study periods, SVM model performed better than BT model for Akhula station, while BT model performed better than SVM model for Pole Senikh station. Additionally, the cross-correlation coefficient for the two study periods was found to be 0.83 and 0.82 for Akhula station, and 0.83 and 0.77 for Pole Senikh station, respectively. In the time series results, there was no clear trend in precipitation over the observation period. However, river flows at the Ahvaz and Pole Senikh stations, particularly after 1995, showed a significant decline, mainly due to factors such as runoff, agricultural expansion, and industrial development

Keywords

• aji chai
• boosted tree
• tabriz plain
• support vector machine
• rain-runoff modeling