// <![CDATA[ANALISIS PERUBAHAN MORFOLOGI DAN DINAMIKA EROSI-DEPOSISI SUNGAI CITARUM MENGGUNAKAN GOOGLE EARTH ENGINE DAN CITRA SENTINEL-2]]> 0407096502 - Dr. Dewi Kania Sari, Ir., M.T. Dosen Pembimbing 1 OKTORIDO / 232021031 Penulis
Sungai Citarum mengalami perubahan morfologi yang dipengaruhi oleh dinamika alamiah dan aktivitas antropogenik di sekitar daerah aliran sungai. Penelitian ini bertujuan menganalisis perubahan bentuk dan dinamika erosi–deposisi dengan teknologi pengindraan jauh berbasis cloud computing. Lokasi penelitian mencakup ±10 km segmen sungai dari Kecamatan Klari, Kabupaten Karawang hingga Kecamatan Jatiluhur, Kabupaten Purwakarta, dengan luas area ±180 ha. Data yang digunakan berupa citra Sentinel-2 Level 1C tahun 2017 serta Level 2A tahun 2020 dan 2024 pada periode musim kemarau, diproses melalui Google Earth Engine. Prapemrosesan meliputi koreksi atmosfer, penyaringan awan <10%, dan komposit median. Klasifikasi badan air dilakukan menggunakan algoritma Random Forest dengan 100 pohon keputusan, diperkuat oleh indeks NDWI, MNDWI, AWEIsh, AWEInsh, serta band B2, B3, B4, B8, B11, dan B12. Area erosi dan deposisi diidentifikasi melalui operasi Boolean antar hasil klasifikasi multi-temporal, dan divalidasi dengan confusion matrix serta koefisien Kappa. Hasil menunjukkan tren penyempitan sungai secara berkelanjutan, dengan pengurangan luas 49.652,4 m² (2,45%) pada periode 2017–2020 dan 112.749 m² (5,7%) pada 2020–2024. Lebar rata-rata sungai menyempit 11,7 m (9%) dan 18,6 m (15,85%) pada periode yang sama. Analisis spasial mengidentifikasi erosi 218.064,9 dan 149.061,5 m² serta deposisi 267.716,8 dan 261.811,6 m². Erosi dominan terjadi di kelokan luar, sedangkan deposisi di kelokan dalam. Akurasi keseluruhan klasifikasi mencapai 94%–96% dengan Kappa 0,90–0,93 (Sangat Kuat), sementara identifikasi erosi-deposisi memperoleh akurasi keseluruhan 88% dengan Kappa 0,79–0,80 (Kuat). AWEInsh dan band 12 (SWIR2) teridentifikasi sebagai prediktor paling berkontribusi. Penelitian ini menunjukkan bahwa integrasi Google Earth Engine, Sentinel-2, dan Random Forest efektif untuk memantau perubahan morfologi sungai dengan resolusi 10 m, serta memberikan dasar penting bagi pengelolaan sumber daya air, mitigasi banjir, dan strategi konservasi berkelanjutan. The Citarum River has undergone morphological changes influenced by natural dynamics and anthropogenic activities around the watershed area. This research aims to analyze morphological changes and erosion-deposition dynamics using cloud computing-based remote sensing technology. The research location covers approximately a ±10 km river segment from Klari Subdistrict, Karawang Regency to Jatiluhur Subdistrict, Purwakarta Regency, with an area of ±180 ha. The data used consists of Sentinel-2 Level 1C imagery from 2017 and Level 2A from 2020 and 2024 during the dry season, processed through Google Earth Engine. Preprocessing included atmospheric correction, cloud filtering <10%, and median composite. Water body classification was performed using the Random Forest algorithm with 100 decision trees, enhanced by NDWI, MNDWI, AWEIsh, AWEInsh indices, as well as bands B2, B3, B4, B8, B11, and B12. Erosion and deposition areas were identified through Boolean operations between multi-temporal classification results and validated using a confusion matrix and the Kappa coefficient. The results show a continuous trend of river narrowing, with an area reduction of 49,652.4 m² (2.45%) in the 2017–2020 period and 112,749 m² (5.7%) in 2020–2024. The average river width narrowed by 11.7 m (9%) and 18.6 m (15.85%) in the same periods. Spatial analysis identified erosion of 218,064.9 m² and 149,061.5 m² as well as deposition of 267,716.8 m² and 261,811.6 m². Erosion predominantly occurred at outer bends, while deposition at inner bends. Overall classification accuracy reached 94%–96% with Kappa 0.90–0.93 (Very Strong), while erosion-deposition identification achieved an overall accuracy of 88% with Kappa 0.79–0.80 (Strong). AWEInsh and band 12 (SWIR2) were identified as the most influential predictors. This research demonstrates that the integration of Google Earth Engine, Sentinel-2, and Random Forest is effective for monitoring river morphological changes at 10 m resolution and provides an important foundation for water resource management, flood mitigation, and sustainable conservation strategies.