Vegetation coverage retrieval scale effect analysis using multi-sensor data

Remote sensing provides a valuable tool and enables to derive vegetation coverage over a large area using either empirical statistical models or sub-pixel analysis. However, to build an empirical regressive model needs a considerate ground truth data, while the parameters for the dimidiate pixel model (DPM) are often difficult to determine in a specific area. We combine the advantages of these two types of inversion models, using the regressive model to determine and validate the DPM parameters. The validated DPM was then used to derive the vegetation coverage information in a large-scale area. The Landsat TM and Terra MODIS data of Shihezi area of Xinjiang, China were acquired and used to derive vegetation coverage using the proposed approach. The scale effect in the inversion of vegetation coverage from the TM and MODIS data were also addressed in the study area. The results indicate that the proposedstrategy is robust in the retrieval of vegetation coverage over a large area from remotely sensed data. The empirical regressive model can help determine the DPM parameters and the DPM can then be applicable when there is no ground data available. The analysis of scale effect shows that the non-linearity of the inversion models mainly causes the scale problem in non-water land covers, while the heterogeneity in a pixel is a major reason that causes the scale effect of water-land boundaries in the retrieval of vegetation coverage.