MODELING OF ANISOTROPIC SEMIVARIOGRAM USING GENETIC ALGORITHM AS PARAMETER ESTIMATION METHOD (CASE STUDY: LAND COVER DATA OF EAST JAVA PROVINCE)

• Main Author: Idayanti, Arini
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/73033
• Institution: Institut Teknologi Bandung

Land cover is the expanse of objects that cover the Earth's physical surface. East Java Province has the largest land cover area on Java Island, which is 4,837,400 hectares. In this study, land cover data of East Java Province in 2018 was used with the slope variable. The slope of land cover is a measure of the steepness of a land surface. Several locations were taken and analyzed using semivariogram to see the spatial correlation. East Java land cover can be affected by distance and direction so an anisotropic semivariogram will be used. This study aims to determine the Cressie-Hawkins experimental semivariogram, then construct an anisotropic semivariogram model based on land cover data of East Java Province in 2018, particularly for built-up land. In addition, the best model was applied to interpolate unobserved locations using ordinary kriging method. In general, parameters are estimated using the Least Square or Maximum Likelihood methods. However, this study will use Genetic Algorithm as parameter estimation method. The advantage is that the solution is obtained randomly. In the implementation of this method in a semivariogram, individuals are defined as vectors of nugget effect, sill, and range parameters and the fitness value is calculated based on Mean Squared Error (MSE). Based on the simulation by changing the parameters above, 7 types of experiments were carried out. Parameter estimation results are taken from individuals with the smallest fitness values. The best model is anisotropic spherical zonal in two pairs of sector directions, namely zonal 1 direction TL-BD and U-S and zonal 2 directions T-B and Tg-BL with MSE 349.024.