Scaling Transfer of Pixel Resolution Between Landsat and Modis Derived ET Imagery

Evapotranspiration (ET) is the largest component of the water balance in arid and semi-arid regions. In this study we will explore how the accuracy of ET maps is affected by their scale and the method used for scaling transfer. Remotely sensed satellite imagery of the Earth's surface provides information to estimate the spatio-temporal distribution of evapotranspiration (ET). In this study, the Surface Energy Balance Algorithm for Land (SEBAL) was used to derive daily ET from Landsat and MODIS images. The resolution of ET predictions depends on the sensor type and varies from Landsat scale (30m, 16 days) to MODIS scale (250m, daily) and there exists a trade-off between spatial and temporal resolution. Scaling transfer of the remote sensing imagery is useful technique because first, up-scaling fills the scale gap between satellite measurements and input requirements for large scale models and second, down-scaling combines the advantages of high temporal and spatial resolutions of surface condition observations.

The main objective of this study is to propose various up- and down-scaling schemes and investigate the potential and limitations of each scaling transfer process, especially how the relative accuracy of ET varies with different scaling transfer processes. Up-scaling results show that the output simple averaging method produces aggregated data with the most statistically and spatially predictable behavior among the 4 up-scaling methods compared; the input nearest neighbor method is the least predictable but is still acceptable. Output down-scaling with regression between two MODIS-based images is the best scheme among the 12 different down-scaling schemes, and input down-scaling with subtraction between MODIS and aggregated (nearest neighboring) Landsat is the worst scheme.