Analisis Model Harmonisasi Citra Landsat-8 dan Sentinel-2 Berdasarkan Konsistensi Citra Indeks Spektral
Lissatu Qurrotil Ainiyyah, Dr. Ir. Catur Aries Rokhmana, S.T., M.T., IPU.
2026 | Tesis | S2 Teknik Geomatika
The use of multisensor remote sensing data such as Landsat-8 and Sentinel-2 in multitemporal analysis often encounters challenges due to differences in spatial, spectral, and radiometric characteristics between sensors, which can lead to inconsistencies in reflectance values and spectral indices. Without a harmonization process, these differences may introduce bias in interpreting surface dynamics. The integration of Landsat-8 and Sentinel-2 imagery is commonly applied in environmental monitoring, particularly in regions with limited data availability due to high cloud coverage. In addition, variations in atmospheric and climatic conditions can further amplify discrepancies between the two datasets when used simultaneously. The Harmonized Landsat-Sentinel (HLS) product developed by NASA provides a solution to improve data consistency at a 30-meter spatial resolution. However, studies examining the impact of harmonization on the consistency of derived spectral indices, particularly after resampling to finer spatial resolutions (15 m and 10 m), remain limited. Therefore, this study aims to evaluate the consistency of reflectance and spectral indices from harmonized data and to analyze the effect of resampling within a multisensor multitemporal analysis framework.
This study utilizes Landsat-8 Level-2 Surface Reflectance, Sentinel-2 Level-2A, and NASA HLS products within the 2023–2025 period, with a cloud cover threshold of less than 30%. Data were selected based on image pairs acquired on the same date or within a short temporal window (1–3 days) to minimize surface condition variability. Sentinel-2 imagery was used as the reference (master) due to its higher spatial resolution and finer spatial detail, serving as the basis for coregistration and spatial grid alignment. The methodological workflow includes pre-processing (reprojection, coregistration, and cloud masking), inter-sensor harmonization through band adjustment, and spatial resampling from 30 meters to 15 meters and 10 meters. Evaluation was conducted using statistical metrics, including Root Mean Square Error (RMSE) and coefficient of determination (R²), along with multitemporal analysis of spectral indices, namely NDVI, SAVI, NDWI, NDBI, and NDBSI. These indices were selected to comprehensively represent land cover characteristics, including vegetation conditions (NDVI and SAVI), surface water or moisture content (NDWI), built up areas (NDBI), and bare land (NDBSI).
The results of this study indicate that the harmonization process applied to Landsat-8 and Sentinel-2 imagery was able to improve the consistency of data between sensors by minimizing differences in their spectral characteristics, thereby making the data more suitable for consistent use in multitemporal analysis. The resampling process also contributed to enhancing visual clarity and spatial detail without significantly altering the main spectral patterns. Temporally, all spectral indices exhibited relatively consistent patterns across sensors and spatial resolutions, demonstrating that harmonization supports the analysis of surface dynamics over longer time periods. Among the tested resolutions, 15 meters provided the best balance between improved spatial detail and spectral stability, whereas the 10-meter resolution tended to introduce greater deviation. Overall, both harmonization and resampling were shown to preserve the consistency and integrity of spectral information in cross-sensor multitemporal analysis.
Kata Kunci : Harmonized Landsat-Sentinel, resampling, indeks spektral