Electronic Theses and Dissertation

Perubahan tutupan lahan merupakan fenomena yang terus terjadi akibat aktivitas manusia dan proses alam sehingga diperlukan pemantauan secara berkala. Penelitian ini bertujuan untuk mengevaluasi kinerja Convolutional Neural Network (CNN) dalam klasifikasi tutupan lahan multitemporal serta menganalisis perubahan tutupan lahan di Kabupaten Aceh Tamiang. Data yang digunakan berupa citra Sentinel-2 tahun 2019 dan 2025 dengan penambahan indeks spektral Normalized Difference Vegetation Index (NDVI), Normalized Difference Built-up Index (NDBI), Modified Normalized Difference Water Index (MNDWI), dan Normalized Difference Moisture Index (NDMI), serta data Esri 10 m Annual Land Cover sebagai label referensi. Model CNN dibangun menggunakan dua convolutional layer, dua batch normalization, max pooling, dropout, dan fully connected layer. Evaluasi model dilakukan menggunakan confusion matrix, accuracy, precision, recall, dan F1-score. Hasil evaluasi menunjukkan bahwa model mampu mengklasifikasikan tujuh kelas tutupan lahan dengan akurasi 78,33% pada tahun 2019 dan 74,10% pada tahun 2025. Analisis perubahan tutupan lahan menunjukkan adanya peningkatan luas lahan pertanian sebesar 428,41 ha, sedangkan luas vegetasi dan semak belukar mengalami penurunan masing-masing sebesar 246,40 ha dan 162,84 ha selama periode 2019–2025. Hasil penelitian menunjukkan bahwa CNN dapat digunakan untuk mengidentifikasi perubahan tutupan lahan menggunakan citra Sentinel-2 dan memberikan informasi yang mendukung pemantauan kondisi lingkungan serta perencanaan wilayah.

Land cover change is a continuous phenomenon due to human activities and natural processes, requiring regular monitoring. This study aims to evaluate the performance of Convolutional Neural Network (CNN) in multitemporal land cover classification and analyze land cover change in Aceh Tamiang Regency. The data used are Sentinel-2 images from 2019 and 2025 with the addition of spectral indices, namely Normalized Difference Vegetation Index (NDVI), Normalized Difference Built-up Index (NDBI), Modified Normalized Difference Water Index (MNDWI), and Normalized Difference Moisture Index (NDMI), as well as Esri 10 m Annual Land Cover data as reference labels. The CNN model was built using two convolutional layers, batch normalization, max pooling, dropout, and a fully connected layer. Model evaluation was carried out using a confusion matrix, accuracy, precision, recall, and F1-score. The evaluation results showed that the model was able to classify seven land cover classes with an accuracy of 78.33% in 2019 and 74.10% in 2025. Analysis of land cover changes showed an increase in agricultural land area by 428.41 ha, while the area of vegetation and shrubs decreased by 246.40 ha and 162.84 ha, respectively, during the 2019–2025 period. The results showed that CNN can be used to identify land cover changes using Sentinel-2 imagery and provide information that supports environmental monitoring and regional planning.