Comparison of IDW and GP models with application to spatiotemporal interpolation of rainfall in Bali Province, Indonesia

Abstract

Precipitation is a critical weather component in our daily life. Spatiotemporal variation in rainfall is very influential in agriculture, health, tourism, and many more. Low or large precipitation volume or intensity can be a problem. However, precipitation is not easy to control, so minimize the negative impact of rainfall is an essential thing that can be done. Precipitation volume or intensity in an area is generally reported based on meteorology stations at several location points, which are then used to describe the value of precipitation in the broader region. However, the number of weather observation stations is often minimal. So using this average of little station information can produce incorrect rainfall information. Various approaches have been developed to be able to provide accurate information based on several observation points. These approaches include Inverse Distance Weighted (IDW) and Gaussian Process (GP). This study aims to compare the accuracy of the IDW and GP methods in conducting rainfall interventions in Bali’s Indonesian province. Bali is one of the world’s tourist destinations where rainfall is very influential on tourist visits. We found IDW provides a good prediction for close location, and GP is much better for a distant place.