Date of Award
Master of Science (MS)
Earth System Science & Policy
Instantaneous heat fluxes were modeled using data obtained from Landsat 5 TM (Thematic Mapper), Landsat 7 ETM+ (Enhanced Thematic Mapper Plus) and Terra MODIS (Moderate Resolution Imaging Spectroradioineter) using the Surface Energy Balance Algorithm for Land (SEBAL) model for cloud-free days. The modeled results were compared with measurements of net radiation (both incoming and outgoing, shortwave and longwave), soil sensible and latent heat fluxes from two flux towers located in Brookings, SD, and Fort Peck, MT. Flux tower data consisted of 30 minute averages at every half an hour, and footprints of contributing movement of air within the period were estimated for each satellite overpass by taking into account the factors of observation height, atmospheric stability, and surface roughness, as well as wind speed and directions (Hsieh et al. 2000). It was found that footprints (considering 90% contributing areas) were normally larger than the size of one Landsat pixel (30 m) but smaller than that of one MODIS pixel (1 km). Therefore, for Landsat the data were averaged for pixels within the concurrent footprint, and for MODIS the data for the particular pixel covering the flux tower was used.
The R values between the modeled and the observed net radiation (Rn) for Landsat and MODIS were found to be 0.70 and 0.66, respectively. Relatively, comparisons between modeled and observed values were better at Brookings than at Fort Peck for both sensors. This may be because the former site has a relatively flat topography and larger fetch than the latter, minimizing the possible effects of terrain heterogeneity on incoming and outgoing radiation modeling. Both satellites performed poorly in modeling soil heat flux (G0) . Our results show that SEBAL provides a better modeling of sensible heat flux (H) with Landsat (R2= 0.62) than with MODIS (R2 = 0.11), even though the MODIS performance for estimating latent heat flux (lambdaE) improved (R2 = 0.37). The improvement found in estimating latent heat flux is probably due to the fact that in SEBAL cold pixels are used to estimate air temperature and then also used in computation for both Rn and H. The uncertainties associated with this assumption cancelled out in deriving lambdaE.
Overall, SEBAL performed better in modeling the heat fluxes when Landsat data were used. This may be due to the scaling issue, as the footprint areas were always significantly less than a single MODIS pixel. By simulating MODIS observations using Landsat, it was found that the R2 value for the aggregated Landsat pixels decreased from 0.62 to 0.25 with an increase of root mean square difference (RMSD) from 50.5 to 68.3 Wm'2. This suggested that the poor performance of MODIS in estimating heat fluxes was due to heterogeneity of the surface within a field of view. In addition, sensitivity analyses of the model to input parameters suggested that the model is more sensitive to surface-to- air temperature difference than to surface roughness conditions. Appendix A lists symbols mentioned in this thesis.
Berhane, Tedros M., "Remote Sensing of Heat Fluxes Validation and Inter-Sensor Comparison" (2007). Theses and Dissertations. 2965.