Studi Komparasi Kinerja Estimasi Debit Sungai Menggunakan Metode Calibration Measurement Hierarchical Classification (CMHC) Berbasis Citra Sentinel-2 (Kasus: Sungai Progo, Sungai Oyo, Sungai Serang)
Anissa Zuhrita, Dr. Nur Mohammad Farda, S.Si., M.Cs. ; Dr. Sandy Budi Wibowo, S.P., M.Sc.
2026 | Tesis | S2 Penginderaan Jauh
River discharge is a key hydrological parameter in water resources management, infrastructure planning, and flood and drought mitigation. The availability of discharge data is increasingly limited due to the uneven spatial distribution of gauging stations and the declining operational sustainability of these stations. This condition has encouraged the development of remote sensing–based discharge estimation methods as a more efficient and sustainable alternative. One such method is the Calibration Measurement Hierarchical Classification (CMHC), which utilizes the spectral characteristics of satellite imagery through a hierarchical discharge classification process. However, the application of this method to tropical rivers in Indonesia remains limited, particularly with respect to the influence of seasonal separation and differences in river morphological characteristics on its performance. This study aims to evaluate the performance of river discharge estimation using the CMHC method based on Sentinel-2 imagery, taking into account wet–dry seasonal separation and variations in river typology. The study was conducted on three river segments in the Special Region of Yogyakarta—Progo River, Oyo River, and Serang River—which represent diverse morphological characteristics. The data used include Sentinel-2 Level-2A (surface reflectance) imagery with 10 m spatial resolution, with the near-infrared (NIR) band as the primary input for the CMHC method, as well as daily discharge data measured at hydrological stations. The analytical procedures consisted of generating CMHC spectral features, classifying discharge classes using the Random Forest algorithm under both non-seasonal and seasonal scenarios, and performing regression-based discharge estimation. Model performance was evaluated using the Nash–Sutcliffe Efficiency (NSE), Root Mean Square Error (RMSE), and relative RMSE (rRMSE). The results indicate that separating wet and dry seasons does not consistently improve classification or discharge estimation performance compared to the non-seasonal scenario. The performance of the CMHC method is more strongly influenced by river morphological characteristics—particularly channel width—than by seasonal factors. The Progo River, with an approximate channel width of ±85 m, produced relatively stable discharge estimates, whereas the narrower Oyo and Serang Rivers (±30–35 m) exhibited lower performance due to the dominance of mixed pixels and the influence of riparian vegetation. These findings emphasize that the spatial representativeness of the water body at the given image resolution is a critical factor in applying the CMHC method to tropical rivers.
Kata Kunci : debit sungai, Sentinel-2, Random Forest, musim, morfologi sungai