University of Twente Student Theses


Modeling surface temperature under clouds using geo-stationary satellite observations

Lu, Lei (2009) Modeling surface temperature under clouds using geo-stationary satellite observations.

[img] PDF
Abstract:Land surface temperature (LST) has a high impact on energy exchanges between surface and atmosphere, climate change, and climatic and hydrological modelling. However, clouds make it difficult to obtain LST from satellite observations of thermal infrared bands directly. The application in polar-orbiting satellite observations proved that the neighboring-pixel approach (NP) based on surface energy balance (SEB) is potential in retrieval of LST under clouds. This study utilized the method to geostationary satellite data MSG/SEVIRI to obtain LST under clouds with high temporal resolution which is more close to the dynamic weather. The four-channel algorithm was used to retrieve LST under clear skies, and the Heliosat-2 algorithm was applied to estimate solar irradiance. Both Navasha, Kenya and Burkina Faso were taken as study areas and in situ data from these areas were used for validation. The RMSE of LST under clear skies was 4.5 K for Naivasha and 5.55 K for Burkina Faso in daytime, and the RMSE of estimated solar irradiance was 234.05 Wm-2.for Naivasha, 388.414 Wm-2 for Burkina Faso. However, error derived from estimating LST under clear skies contributed more to the accuracy of estimated LST under clouds (RMSE=5.6 K for Naivasha). Both temporal interpolation and spatial interpolation were applied to interpolate LST under clouds by using neighboring-pixel approach, and the temporal interpolation performed best (RMSE=5.6 K for temporal interpolation, RMSE=20.14K for spatial interpolation, RMSE=11.76 K for a combination of the previous two methods). This indicates that the neighboring-pixel approach is sensitive to heterogeneity of land surface, and temporal interpolation can eliminate the effect of heterogeneity of land surface. The statistic analysis on LST under clouds shows that the hypothesis “there is no difference between estimated LST under clouds from satellite observations and in situ data” was rejected at the level of 0.05. Using neighboring-pixel approach based on SEB, LST under clouds in daytime was obtained; in order to obtain the diurnal LST in all skies, method for estimating LST under clouds in nighttime is necessary.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page