COMPARISON OF LOGISTIC REGRESSION AND MAXENT IN SPATIAL MODELING OF LAND COVER CHANGES IN CIKAMAL GRASSLANDS, PANANJUNG PANGANDARAN NATURE RESERVE

• Main Author: Sakenia, Nindita
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/61328
• Institution: Institut Teknologi Bandung

Cikamal Grassland is the only grassland left from the three original grasslands in The Pananjung Pangandaran Nature Reserve. Initially, the area of this grassland was about 20 ha until continues to decrease due to the secondary succession and shrubs invasion. It affects the function of grassland itself as the feeding ground for the herbivores, especially the timor deers (Cervus timorensis). Furthermore, it has led the changing of timor deers behaviour, especially its feeding habits. Timor deers also wander around outside the nature reserve area. The aim of this study is modeling the changing of grassland cover in Cikamal grassland by using logistic regression and MaxEnt. For modeling purposes, 50 Presence points and 96 absence points were used and taken randomly. Presence points were taken in the loss grassland cover while absence point are taken in the unchanged grassland cover based on the land cover classification by using SPOT 7 imageries of 2018 and 2019. The explanatory variables were elevation, aspect, slope, hillshade, distance to the edge of non-grass land cover, and annual incoming solar radiation. Logistic regression modeling provides an accurate result with an Area Under ROC Curve (AUC) has value of 0.8925. The environmental variables that affect the model were elevation, hillshade, annual incoming solar radiation, slope, and the distance from the forest edge. MaxEnt model has lower accuracy with the value of AUC was 0,776 as the AUC value. The environmental variables that contribute in this model were are the distance from forest edge, aspect, annual incoming solar radiation and elevation. Based on logistic regression, it was estimated that 2.16 ha of the grassland area will be invaded by shrubs or encounter secondary succession while based on MaxEnt. only 0.92 of grassland area.