Electronic Theses and Dissertation

Tanah longsor merupakan bencana geohazard yang sering terjadi di wilayah tropis yang menyebabkan kerugian besar terhadap aspek sosial, ekonomi, dan lingkungan. kecamatan pintu rime termasuk dalam salah satu wilayah dengan tingkat kerentanan tinggi akibat kondisi topografi pegunungan, curah hujan tahunan yang tinggi, serta didukung dengan aktivitas manusia seperti pembukaan lahan dan pembangunan yang tidak mempertimbangkan kondisi fisik lingkungan. penelitian ini bertujuan mengembangkan model klasifikasi daerah rentan longsor menggunakan metode deep learning berbasis convolutional neural network (cnn) dengan memanfaatkan data radar sentinel-1, serta menghasilkan peta distribusi spasial tingkat kerentanan longsor di kecamatan pintu rime sebagai informasi pendukung mitigasi bencana. data penelitian mencakup 43 titik lokasi historis longsor di kecamatan pintu rime dan data dem 10 m. proses penelitian meliputi pengunduhan data sentinel-1, pra-pemrosesan (speckle filtering, kalibrasi radiometrik, koreksi topografi), ekstraksi patch, pembangunan dan pelatihan cnn menggunakan tensorflow/keras, serta evaluasi model dengan metrik accuracy, precision, recall, f1-score, dan roc-auc. hasil penelitian menunjukkan model terbaik diperoleh pada epoch ke 46 dengan akurasi 82,40%, presisi 89%, recall 76%, f1-score 80%, dan roc-auc 89,4%, serta berhasil menghasilkan peta kerentanan longsor yang diklasifikasikan ke dalam 3 kelas. hasil temuan ini menunjukkan potensi cnn dalam mendukung pemetaan kerentanan longsor secara akurat untuk mitigasi bencana di kecamatan pintu rime. kata kunci: tanah longsor, deep learning, convolutional neural network, sentinel-1.

Landslides are a geohazard disaster that often occurs in tropical regions causing major losses to social, economic, and environmental aspects. Pintu Rime District is one of the areas with a high level of susceptibility due to mountainous topography, high annual rainfall, and supported by human activities such as land clearing and development that do not consider the physical conditions of the environment. This study aims to develop a classification model for landslide-prone areas using the Convolutional Neural Network (CNN) based Deep Learning method by utilizing Sentinel-1 radar data, and produce a spatial distribution map of landslide vulnerability levels in Pintu Rime District as supporting information for disaster mitigation. The research data includes 43 historical landslide locations in Pintu Rime District and 10 M DEM data. The research process includes downloading Sentinel-1 data, Pre-processing (speckle filtering, radiometric calibration, topographic correction), patch extraction, building and training CNN using TensorFlow/Keras, and model evaluation with Accuracy, Precision, Recall, F1-Score, and ROC-AUC metrics. The results showed that the best model was obtained at epoch 46 with an accuracy of 82.40%, precision of 89%, recall of 76%, F1-score of 80%, and ROC-AUC of 89.4%, and successfully produced a landslide susceptibility map classified into three classes. These findings demonstrate the potential of CNN in supporting accurate landslide susceptibility mapping for disaster mitigation in Pintu Rime District. Keywords: Landslide, Deep Learning, Convolutional Neural Network, Sentinel-1.