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

• Main Author: Nurfauziah, Azmi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/73034
• Institution: Institut Teknologi Bandung

Anisotropic semiovariogram is a geostatistical analysis used to describe the spatial correlation between an observation that depends on distance and direction. The spatial data used is land cover data, namely the physical appearance of a land surface. Land cover data for East Java Province forest class in 2013 and land class built in 2018 which were observed from the variable land height were used to construct the Cressie-Hawkins experimental semivariogram and its modifications. The modifications made are combining semivariogram values at 2-3 successive intervals in the middle or at the end of the lag distance. Then modeled with exponential, Gaussian, and spherical semivariogram. The model is built in the East-West (E-W), Southeast-Northwest (SE-NW), North-South (N-S), Northeast-Southwest (NE-SW) directions with an angle tolerance of 22.5º. Based on the parameter estimation method of Generalized Least Squares, the best model is the spherical geometry anisotropic semivariogram model with a sum of squared errors of 2.175×108. Then Ordinary Kriging interpolation is performed to estimate the value of an unobserved point and predict future land cover changes. The benefit is that it can assist the government in making decisions regarding forest land cover plans. The interpolation results of kriging at the height of the land were entered into lowlands (< 400 masl), medium (400-700 masl), and high (> 700 masl).