Perbandingan Variasi Resolusi Spasial DEM terhadap Model Kerawanan Longsor dengan Regresi Logistik di Kabupaten Magelang

Vikarinda Virginia Taniu, Dr. Eng. Guruh Samodra, M.Sc

2025 | Skripsi | GEOGRAFI DAN ILMU LINGKUNGAN

Model elevasi digital (DEM) merupakan sumber data yang paling penting dalam penilaian kerawanan longsor. Banyak faktor pengontrol longsor yang harus dihasilkan dari data DEM. Sebagian besar penelitian kerawanan longsor di Indonesia memanfaatkan DEM yang tersedia secara bebas atau open source. Namun demikian, masih minim penelitian yang membandingkan pengaruh variasi resolusi spasial DEM terhadap penilaian kerawanan longsor. Kabupaten Magelang merupakan wilayah yang memiliki status rawan longsor hampir di seluruh wilayahnya dan banyak studi selalu memanfaatkan DEM sebagai  basis data topografi. Penelitian ini bertujuan untuk membandingkan distribusi spasial kerawanan longsor di Kabupaten Magelang dan tingkat akurasi hasil pemetaan kerawanan longsor berdasarkan variasi resolusi data DEM open source, yaitu DEMNAS (8,2 m), ALOS PALSAR (12,5 m), SRTM (30 m), dan SRTM (90 m). Pemodelan dilakukan dengan pendekatan statistik Regresi Logistik untuk menghasilkan peta probabilitas kejadian longsor. Uji akurasi model menggunakan dua pendekatan statistik yaitu Area Under Curve (AUC) dari Receiver Operating Characteristic (ROC) dan Pseudo R-Squared (Nagelkerke R²).

Hasil penelitian menunjukkan bahwa distribusi spasial kerawanan longsor di Kabupaten Magelang umumnya banyak ditemukan pada area yang berada di lereng dengan kemiringan sedang, serta dekat dengan jaringan jalan dan aliran sungai (<50>

Digital Elevation Model (DEM) is one of the most important data sources in landslide susceptibility assessment. Many landslide-controlling factors such as slope, aspect, elevation, plan and profile curvature, are derived from DEM data. The majority of landslide susceptibility research conducted in Indonesia use open- source, freely accessible DEMs. Research that systematically evaluates the impact of differences in spatial resolution in DEM on mapping landslide vulnerability is still lacking, nevertheless. Magelang Regency is known as a highly landslide-prone area, and numerous studies have highlighted the crucial role of DEM in spatial analysis within the region.

This study aims to compare the spatial distribution of landslide susceptibility and the accuracy of susceptibility mapping results using various open-source DEMs with different spatial resolutions, namely DEMNAS (8.2 m), ALOS PALSAR (12.5 m), SRTM 30 m, and SRTM 90 m. Logistic Regression was employed to model landslide susceptibility and generate probability maps of landslide occurrence. Model validation was conducted using two statistical approaches: Area Under the Curve (AUC) from Receiver Operating Characteristic (ROC) analysis and Pseudo R-Squared (Nagelkerke R²).

The results show that landslide-susceptible areas in Magelang Regency are generally concentrated on moderate to steep slopes and located near roads and river networks. A comparison of models based on different DEMs reveals that DEMNAS yields the highest AUC value, followed by ALOS PALSAR. Similarly, the highest Nagelkerke R² value is obtained from the DEMNAS-based model, followed by ALOS PALSAR, with only minor differences between them. These findings indicate that DEMNAS is the most suitable digital elevation model for landslide susceptibility mapping the study area.

Kata Kunci : kerawanan longsor, digital elevation model, resolusi spasial, regresi logistik, Kabupaten Magelang