Estimation of monthly and seasonal precipitation: A comparative study using data-driven methods versus hybrid approach

The purpose of this study is to estimate the monthly and seasonal precipitation using historical measured data. Firstly two data-driven methods such as Support Vector Regression (SVR), Gaussian Process Regression (GPR) along a distance-based k-nearest neighbor method (KNN) were used. Secondly, the hybrid Support Vector Regression with the Fruit fly optimization Algorithm (SVR-FOA) was used for comparing the estimated results. The historical measured meteorological data were entered into the model after preprocessing and determination of suitable input combinations based on the feature selection algorithms and sensitivity analysis. The results showed that the hybrid SVR-FOA approach performs better than the other methods, especially during seasonal precipitation estimation. In conclusion, the SVR-FOA model with r of 0.75, NS of 0.56, RMSE of 13.58 mm, and MAE of 10.01 mm during monthly precipitation estimation and r of 0.90, NS of 0.80, RMSE of 6.37 mm and MAE of 5.04 mm during seasonal precipitation is able to accurately estimate precipitation in the study area.

Automated summary

The experimental results showed that the hybrid SVR-FOA method outperformed other methods in seasonal precipitation estimation and also performed well in monthly precipitation estimation. The performance of the hybrid SVR-FOA model demonstrates considerable reliability in estimating precipitation.