Synergistic Use of Multitemporal SAR and Optical Data with Random Forest Approach for Rice Mapping in Bangladesh
Abstract
Accurate in-season crop mapping is vital for national food security, yet traditional ground-based estimation methods in Bangladesh often lack precision. This study develops a robust machine-learning framework to map Boro rice in the Manikganj district by integrating Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 optical imagery. To overcome the challenge of seasonal cloud cover, Sentinel-1 VH polarization was utilized to identify unique backscatter reductions during the transplantation phase (late February to early March) caused by specular reflection from flooded fields. By combining these SAR signatures with high-resolution spectral indices (NDVI, EVI, and BSI) from Sentinel-2 during peak growth stages, the research captures the distinct phenological signatures of paddy rice. A pixel-based Random Forest algorithm was applied within the Google Earth Engine (GEE) platform, utilizing about 200 field-verified reference points per sub-district (1,385 in total) for training and validation. The results demonstrated high classification performance, with overall accuracies ranging from 88.1% to 94.9% across the seven sub-districts. Furthermore, a comparison with official government statistics revealed a minimal total area difference of only 2.34%, underscoring the reliability of this integrated satellite approach for operational crop monitoring.
Related
- Phase-noise tolerant fully-coherent optical coherence tomography angiography for high-contrast 3D microvasculature mapping
- The landscape of structural variants in male infertility identified by optical genome mapping
- Optical genome mapping identifies source-associated structural variant differences across early-passage human iPSCs
- Optical Genome Mapping of BCL11B::TLX3-Rearranged Pediatric T-ALL: A Recurrent Structural Subgroup and Its Co-Occurring Genomic Alterations
- Socio-economic status of pig rearers and management system of native pigs in Bangladesh
BibTeX
@misc{perveen2026synergistic,
title = {{Synergistic Use of Multitemporal SAR and Optical Data with Random Forest Approach for Rice Mapping in Bangladesh}},
author = {Mst. Farida Perveen and Sravan Shrestha and Abdullah Al Mehadi and Saima Kazmina and Amio Al Saem},
howpublished = {EarthArXiv},
year = {2026},
month = oct,
doi = {10.31223/x5f22w},
url = {https://eartharxiv.org/repository/view/15552/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Synergistic Use of Multitemporal SAR and Optical Data with Random Forest Approach for Rice Mapping in Bangladesh
AU - Mst. Farida Perveen
AU - Sravan Shrestha
AU - Abdullah Al Mehadi
AU - Saima Kazmina
AU - Amio Al Saem
PY - 2026
DA - 2026/10/18
PB - EarthArXiv
DO - 10.31223/x5f22w
UR - https://eartharxiv.org/repository/view/15552/
AB - Accurate in-season crop mapping is vital for national food security, yet traditional ground-based estimation methods in Bangladesh often lack precision. This study develops a robust machine-learning framework to map Boro rice in the Manikganj district by integrating Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 optical imagery. To overcome the challenge of seasonal cloud cover, Sentinel-1 VH polarization was utilized to identify unique backscatter reductions during the transplantation phase (late February to early March) caused by specular reflection from flooded fields. By combining these SAR signatures with high-resolution spectral indices (NDVI, EVI, and BSI) from Sentinel-2 during peak growth stages, the research captures the distinct phenological signatures of paddy rice. A pixel-based Random Forest algorithm was applied within the Google Earth Engine (GEE) platform, utilizing about 200 field-verified reference points per sub-district (1,385 in total) for training and validation. The results demonstrated high classification performance, with overall accuracies ranging from 88.1% to 94.9% across the seven sub-districts. Furthermore, a comparison with official government statistics revealed a minimal total area difference of only 2.34%, underscoring the reliability of this integrated satellite approach for operational crop monitoring.
ER -