Abstract
The goal of this paper is to evaluate the DisTrad disaggregation technique for deriving land surface temperatures over dry-semi arid urban areas using impervious indices of the Gaza Strip.
The DisTrad technique is originally developed for downscaling land surface temperature (LST) for vegetated areas based on the correlation between LST and NDVI, however, DisTrad is adapted by using LST – impervious index relationship in humid climate areas (e.g. Dublin). DisTrad shows applicability for easy and parameter-free implementations. However, the impervious indices - LST relation is a city-based due to urban land use activities and land cover materials are different. Moreover, recent studies (Erbil-Iraq) have shown that arid and semi-arid regions have different spectral characteristics from other climatic regions and thus lead to a different statistical relationship than in humid areas.
Many spectral built-up indices recently (~ 35 indices) become readily available and considered promising to be adapted in DisTrad framework. In this study, 18 impervious indices (e.g. UI, ISA, NDBI, NDISI, IBI, EBBI, ENBI , BBI, NDII, NDBaI, BUc, BRBA, BUI, NBI, VgNIR-BI, VrNIR-BI) will be evaluated for characterizing LST in Gaza Strip. Other vegetation and bare soil indices will be also evaluated for comparison (for bare soil e.g. BI, DBSI, NDBaI, SAVI. for Vegetation e.g. FC, NDVI). The performance of the indices will be tested as inputs for DisTrad over two land covers scenarios; all classes (including urban, sand dunes, vegetation, bare soil and urban class), and only urban class. land cover classification will be produced using the support vector machine (SVM) method.
The LST and impervious indices will be derived from the Landsat 8 image of 2017 in the summer-time season. Landsat thermal band at 100m resolution will be sharpened to the spectral indices of 30 m resolution. Sharpening will be evaluated using R2 and RMSE error analysis with the observed LST products and the sharpened LST products at 100 m resolution.
Keywords: land surface temperature, urban, thermal sharpening, impervious indices, DisTrad, Landsat 8.
The DisTrad technique is originally developed for downscaling land surface temperature (LST) for vegetated areas based on the correlation between LST and NDVI, however, DisTrad is adapted by using LST – impervious index relationship in humid climate areas (e.g. Dublin). DisTrad shows applicability for easy and parameter-free implementations. However, the impervious indices - LST relation is a city-based due to urban land use activities and land cover materials are different. Moreover, recent studies (Erbil-Iraq) have shown that arid and semi-arid regions have different spectral characteristics from other climatic regions and thus lead to a different statistical relationship than in humid areas.
Many spectral built-up indices recently (~ 35 indices) become readily available and considered promising to be adapted in DisTrad framework. In this study, 18 impervious indices (e.g. UI, ISA, NDBI, NDISI, IBI, EBBI, ENBI , BBI, NDII, NDBaI, BUc, BRBA, BUI, NBI, VgNIR-BI, VrNIR-BI) will be evaluated for characterizing LST in Gaza Strip. Other vegetation and bare soil indices will be also evaluated for comparison (for bare soil e.g. BI, DBSI, NDBaI, SAVI. for Vegetation e.g. FC, NDVI). The performance of the indices will be tested as inputs for DisTrad over two land covers scenarios; all classes (including urban, sand dunes, vegetation, bare soil and urban class), and only urban class. land cover classification will be produced using the support vector machine (SVM) method.
The LST and impervious indices will be derived from the Landsat 8 image of 2017 in the summer-time season. Landsat thermal band at 100m resolution will be sharpened to the spectral indices of 30 m resolution. Sharpening will be evaluated using R2 and RMSE error analysis with the observed LST products and the sharpened LST products at 100 m resolution.
Keywords: land surface temperature, urban, thermal sharpening, impervious indices, DisTrad, Landsat 8.
Original language | English |
---|---|
Title of host publication | EARSeL Liege 2020 |
Publication status | Published - 26 May 2020 |