SPATIAL INTERPOLATION OF RAINFALL INTENSITY IN JAVA ISLAND USING ORDINARY KRIGING

Indonesia, situated between two continents and two oceans, experiences a complex climate system influenced by global warming. Climate change has disrupted weather patterns, making it increasingly difficult to predict the rainy and dry seasons and rainfall intensity. However, neighboring regions often exhibit similar weather characteristics, which can be leveraged for prediction. As Indonesia’s economic center, Java Island displays distinct yet interconnected weather patterns, making accurate rainfall prediction crucial for various sectors. This study utilizes 10 years of average rainfall data from NASA’s Power database, covering 64 observation points across Java. Ordinary point kriging is the estimation of a value at a given point and is often used in spatial interpolation analysis in general. Through ordinary point kriging analysis, this study aims to find an accurate kriging equation for predicting rainfall in various regions of Java Island. To achieve this, semivariogram modeling was performed to determine the best theoretical model for spatial interpolation. From 53 sampled regions, 1,378 sample pairs were used to calculate the experimental semivariogram obtained using the R programming language. Next, the theoretical semivariogram was determined using the sill parameter derived from the variance of the sampled data. Three theoretical semivariogram models were considered: spherical, exponential, and Gaussian. The results indicated that the exponential model was the most suitable as it had the smallest SSE value. The results of this analysis enrich our understanding of climate patterns in Indonesia and will contribute to developing mitigation and adaptation strategies related to climate change in the future. The Kriging equation obtained can provide highly accurate prediction results on the test data with a MAPE (Mean Absolute Percentage Error) error measure of 4.85% and RMSE (Root Mean Square Error) of 18.17, which indicates that the prediction results obtained are highly accurate predictions.