Location: Soil and Water Management ResearchTitle: Downscaling of Land Surface Temperature Maps in the Texas High Plains with TsHARP Method) Author
Submitted to: GIScience and Remote Sensing
Publication Type: Peer reviewed journal
Publication Acceptance Date: 8/15/2011
Publication Date: 8/25/2011
Citation: Ha, W., Gowda, P., Howell, T.A. 2011. Downscaling of Land Surface Temperature Maps in the Texas High Plains with TsHARP Method. GIScience and Remote Sensing. 48(4):583-599. Interpretive Summary: Evapotranspiration (ET) consists of evaporation of water from soil surface to the atmosphere and transpiration by plants. Estimating ET is critical for managing irrigated crops and groundwater management in arid and semiarid regions. Remote sensing-based ET estimation methods require land surface temperature data derived from thermal infrared images. Remote sensing data acquired daily provide only coarser thermal images with pixels larger than individual agricultural fields. These data also contain simultaneously acquired high resolution visible and near infrared images. A downscaling method known as TsHARP was evaluated with synthetic remote sensing data to downscale 960 m coarser land surface temperature images to finer 240 and 120 m resolution using visible and near infrared images. Results indicate that the TsHARP method is relatively easy to implement and it has the potential to be used as downscaling method in an operational ET remote sensing program.
Technical Abstract: High spatial resolution daily evapotranspiration (ET) maps would significantly improve assessing crop water requirements in arid and semi-arid regions of the world such as Texas High Plains (THP) where water demand exceeds supply for irrigation. Remote sensing-based models that use energy balance equations for mapping ET require land surface temperature (LST) images derived from thermal-infrared (TIR) images. Due to technical and cost constraints, satellite sensors that acquire TIR data either have a low spatial resolution with high temporal resolution (e.g. Moderate Resolution Imaging Spectroradiometer (MODIS) and Geostationary Operational Environmental Satellite (GOES)) or high spatial resolution with low temporal resolution (e.g. Landsat 5 Thematic Mapper (TM)). Low spatial resolution TIR data acquired daily by MODIS and GOES sensors are inadequate as their thermal pixel sizes (1,000-4,000 m) are larger than individual agricultural fields. High spatial resolution TIR data acquired with Landsat 5 TM (120 m) is available only once in every 16 days. However, there exists an opportunity to use simultaneously acquired high resolution visible, near-infrared, and shortwave-infrared images on the same satellite platform, or from other high resolution satellite sensors to improve spatial and temporal resolution of LST images. Numerous image downscaling methods are available in the literature of various scientific disciplines for downscaling of images by employing relationships between simultaneously acquired coarser and finer datasets. However, limited or no testing of these methods has been done to improve spatial resolution of LST images using coarser thermal and finer non-thermal datasets, especially in semi-arid irrigated regions. In this study, the TsHARP image downscaling technique was evaluated for its capability of downscaling coarser LST images to finer resolutions. A subset of Landsat 5 TM image covering the southern part of the THP was utilized for this purpose. The TsHARP algorithm was implemented at seven different spatial resolutions (240, 360, 480, 600, 720, 840, and 960 m) using synthetic images derived from Landsat 5 TM-based 120 m LST image. Also, the TsHARP was evaluated to downscale 960 m LST image to 240 m to mimic MODIS datasets. Comparison of original and TsHARP-derived 120x120 m LST images downscaled from 240, 360, 480, 600, 720, 840, and 960 m synthetic LST images gave root mean square errors of 1.0, 1.3, 1.5, 1.6, 1.7, 1.8, and 1.9 deg C, respectively. Results also indicated that TsHARP has the potential to be used to downscale LST images with simultaneously acquired high resolution NDVI image derived from MODIS data.