Hussain et al. (2026) Development of high-resolution land surface temperature and paddy area estimation technique using multi-source satellite image-based downscaling

Identification

• Journal: Paddy and Water Environment
• Year: 2026
• Date: 2026-03-25
• Authors: Kiramat Hussain, Jeongho Han, Kyoung Jae Lim, Jonggun Kim
• DOI: 10.1007/s10333-026-01064-9

Research Groups

• Interdisciplinary Program in Earth Environmental System Science & Engineering, Kangwon National University, Chuncheon, South Korea
• Agriculture and Life Sciences Research Institute, Kangwon National University, Chuncheon, South Korea
• Department of Regional Infrastructure Engineering, Kangwon National University, Chuncheon, South Korea

Short Summary

This research developed a technique to downscale Land Surface Temperature (LST) from 30 meters to 3 meters using multi-source satellite imagery and multispectral indices, enabling accurate estimation and monitoring of paddy field areas and their temporal changes for precision agriculture. The study successfully revealed fine-scale LST variations crucial for understanding agricultural heterogeneity and resource management.

Objective

• To downscale Landsat-8/9 LST data from a resolution of 30 meters to 3 meters, providing a more detailed understanding of LST variations across different crop fields, with a specific focus on the unique surface conditions of paddy fields during the early growing season.
• To estimate the area of paddy fields using the downscaled 3-meter LST data, enabling an analysis of changes in paddy field area over time.

Study Configuration

• Spatial Scale: Heung-up Reservoir basin, Wonju, Gangwon-do Province, South Korea (approximately 17.5 square kilometers). LST downscaled from 30 meters to 3 meters.
• Temporal Scale: Early growing season (May) for 2019 and 2024. Post-harvest (October) for 2019 and 2024 for temporal filtering.

Methodology and Data

• Models used:
  • Multiple Linear Regression (MLR) model (modified from Kustas et al. 2003 and Bonafoni et al. 2016) for LST downscaling.
  • Otsu’s thresholding method for paddy area estimation.
  • NDVI threshold method for Land Surface Emissivity (LSE) estimation.
• Data sources:
  • Satellite: Landsat-8/9 (Collection 2 L1 imagery, OLI and TIRS bands), PlanetScope (surface reflectance data, red and near-infrared bands), Sentinel-2 (MSI level-2A imagery, green and shortwave infrared bands).
  • Observation/Ground Truth: GPS-based field survey data (May 2023) for paddy boundaries, electronic map of agricultural land, hourly air and ground temperature data from Wonju meteorological station (Korea Meteorological Administration).

Main Results

• Successfully downscaled LST from 30 meters to 3 meters (DLST 3m) using a Multiple Linear Regression model incorporating multispectral indices (NDVI, MSAVI, MNDWI).
• DLST 3m data revealed significantly greater thermal variability and fine-scale LST variations compared to 30-meter LST. For instance, in 2019, LST 30m ranged from 24.1 °C to 33.2 °C, while DLST 3m extended from 22.6 °C to 37.5 °C. In 2024, LST 30m ranged from 27.6 °C to 35.5 °C, while DLST 3m captured values between 26.01 °C and 39.37 °C.
• A multi-seasonal temporal filtering approach using October imagery effectively adjusted early-season vegetation indices by identifying and masking perennial vegetation (e.g., orchards), improving the reliability of LST analysis.
• Otsu’s thresholding method applied to DLST 3m accurately estimated paddy field areas, identifying optimal thermal thresholds of 30.39 °C in 2019 and 32.61 °C in 2024.
• The estimated paddy areas were 85.08 hectares in 2019 and 79.55 hectares in 2024, showing a reduction over the five-year period. These estimates closely aligned with the 2023 ground-validated reference of 74.84 hectares.
• Lower LST values were consistently observed in paddy fields due to the presence of standing water during the early growing season, which was validated against ground survey data.

Contributions

• Achieved an unprecedented LST downscaling resolution of 3 meters in agricultural areas, surpassing most existing studies (typically 10 meters or coarser) and enabling highly precise monitoring of field-scale thermal dynamics.
• Integrated a novel combination of multi-source satellite data (Landsat-8/9, PlanetScope, Sentinel-2) and multispectral indices (NDVI, MSAVI, MNDWI) to enhance LST downscaling and paddy field characterization.
• Developed a robust and practical method for paddy area estimation and change detection using downscaled 3-meter LST data and Otsu's thresholding, demonstrating its applicability for agricultural monitoring.
• Introduced a multi-seasonal temporal filtering approach using post-harvest imagery to refine early-season vegetation indices, effectively reducing misclassification caused by perennial vegetation and improving the accuracy of thermal-based land cover classification.
• Provided critical insights into the observed reduction of paddy field area in South Korea between 2019 and 2024, highlighting the implications for irrigation management and resource allocation in response to evolving agricultural challenges.

Funding

• Korea Institute of Planning and Evaluation for Technology in Food, Agriculture and Forestry (IPET) through the Agricultural Foundation and Disaster Response Technology Development Program, funded by the Ministry of Agriculture, Food and Rural Affairs (MAFRA) (grant number 322081-3).
• “Cooperative Research Program for Agriculture Science and Technology Development (Project No. RS-2023-00232079)” Rural Development Administration, Republic of Korea.

Citation

@article{Hussain2026Development,
  author = {Hussain, Kiramat and Han, Jeongho and Lim, Kyoung Jae and Kim, Jonggun},
  title = {Development of high-resolution land surface temperature and paddy area estimation technique using multi-source satellite image-based downscaling},
  journal = {Paddy and Water Environment},
  year = {2026},
  doi = {10.1007/s10333-026-01064-9},
  url = {https://doi.org/10.1007/s10333-026-01064-9}
}