Kerawanan Tanah Longsor Dengan Metode Regresi Logistik Di Sub Daerah Aliran Sungai Mata Allo, Daerah Aliran Sungai Saddang = Landslide Susceptibility Using Logistic Regression Method in Mata Allo River Sub-Watershed, Saddang River Basin

Latar belakang. Sub DAS Mata Allo, sebagai bagian dari DAS Saddang, merupakan kawasan dengan tingkat kerawanan tanah longsor yang tinggi akibat kondisi topografi curam dan perubahan tata guna lahan. Oleh karena itu, diperlukan pendekatan analisis spasial berbasis model regresi logistik untuk memprediksi potensi kejadian longsor berdasarkan faktor-faktor penyebabnya, guna mendukung upaya mitigasi bencana dan perencanaan penggunaan lahan yang berkelanjutan. Tujuan. Penelitian ini bertujuan untuk mengidentifikasi kejadian tanah longsor tahun 2018–2022, menentukan faktor dominan penyebab tanah longsor, dan memetakan kerawanan tanah longsor. Metode. Data kejadian tanah longsor diperoleh dari interpretasi citra Google Earth Pro dan survei lapangan, dengan sembilan parameter independen: kemiringan lereng, elevasi, arah lereng, kelengkungan bumi, curah hujan, jarak sungai, litologi, jarak patahan, dan penutupan lahan. Analisis frekuensi rasio untuk mengukur kontribusi faktor, dan regresi logistik untuk memodelkan kerawanan tanah longsor. Hasil. Tahun 2018 teridentifikasi 43 kejadian tanah longsor, tahun 2019 teridentifikasi 58 kejadian, tahun 2020 teridentifikasi 53 kejadian, tahun 2021 teridentifikasi 35 kejadian dan tahun 2022 teridentifikasi 78 kejadian. Faktor penutupan lahan, kemiringan lereng, curah hujan, dan jarak sungai memiliki nilai koefisien regresi (B) yang relatif besar. Kerawanan tanah longsor dibagi dalam lima kelas, dengan validasi model menunjukkan akurasi AUC kategori "baik". Kesimpulan. Inventarisasi tanah longsor selama lima tahun menunjukkan 267 kejadian longsor, dengan faktor dominan meliputi penutupan lahan, kemiringan lereng, curah hujan, dan jarak sungai. Peta kerawanan tanah longsor terbagi lima kelas, dengan kelas sangat rendah paling luas dan kelas sangat tinggi paling kecil. Pendekatan ini efektif mendukung mitigasi bencana berbasis spasial. Background. The Mata Allo sub-watershed, as part of the Saddang watershed, is an area with a high level of landslide vulnerability due to steep topography and land use changes. Therefore, a spatial analysis approach based on a logistic regression model is needed to predict the potential for landslide events based on the causal factors, in order to support disaster mitigation efforts and sustainable land use planning. Objectives. This study aims to identify landslide events in 2018–2022, determine the dominant factors causing landslides, and map landslide vulnerability. Methods. Landslide occurrence data were obtained from Google Earth Pro image interpretation and field surveys, with nine independent parameters: slope gradient, elevation, slope direction, earth curvature, rainfall, river distance, lithology, fault distance, and land cover. Frequency ratio analysis to measure the contribution of factors, and logistic regression to model landslide vulnerability. Results. In 2018, 43 landslide incidents were identified, in 2019, 58 incidents were identified, in 2020, 35 incidents were identified, and in 2022, 78 incidents were identified. Land cover, slope gradient, rainfall, and river distance factors had relatively large regression coefficient (B) values. Landslide vulnerability was divided into five classes, with model validation showing AUC accuracy in the "good" category. Conclusion. A five-year landslide inventory revealed 267 landslide incidents, with dominant factors being land cover, slope gradient, rainfall, and river distance. The landslide vulnerability map was divided into five classes, with the very low class being the largest and the very high class being the smallest. This approach effectively supports spatial-based disaster mitigation.