Analisis regresi dengan metode support vector machine IMERG downscaled di Karanganyar

Abstract

Rainfall is the accumulation of rainwater in a uniform rain measurement that is flat and does not absorb, drip, or flow. The rain occurs on a wide scale and spreads out, but in Indonesia, the measurements are only conducted at a few points, so it is hard to describe the rainfall temporally or spatially. When predicting rainfall information on a regional or global scale, alternative data is needed, namely remote sensing in the form of Integrated Multi-satellite Retrievals for GPM (IMERG) with large-scale spatial resolution. Furthermore, IMERG needs validated downscaling to be used as alternative data. IMERG downscaling uses the support vector machine (SVM) regression method and is validated using the R-squared (R2), root mean square (RMSE), and bias methods. IMERG downscaling data is validated using rainfall data taken with a rain gauge. The results of the downscaled and calibrated IMERG validation show relatively high R2 and bias values, namely 0.901 and 0.2, respectively. However, the model exhibits a relatively small prediction error for rainfall, as evidenced by the average error value of 2.227 mm/day each year. Based on these validation results, this method holds great promise for utilization in areas with limited spacing between rainfall stations, as the model is capable of predicting rainfall data accurately in regions without rainfall stations. Further research is highly needed for the upcoming years.