A framework for analyzing and designing scale invariant remote sensing algorithms

The land surface exhibits heterogeneity across a range of spatial scales. Remote sensors provide integrated information at the pixel scale, however, there is important spatial variability at scales smaller than the scale of the sensor. On the other hand, large scale models that use remotely sensed data do not require them at the same spatial resolution at which remote sensors are required to operate. In this paper, a framework for testing aggregation-disaggregation properties of remote sensing algorithms is presented. The proposed framework provides a systematic approach for parameterizing the land surface heterogeneity effects. For the estimation of the pixel scale response, the lumped response should be modified by the variance and covariance terms. This representation of land surface heterogeneity could lead to substantial savings in remote sensing data storage and management. Using simulated land and vegetation scenarios, the authors have successfully parameterized subpixel scale heterogeneity effects for the estimation of vegetation index, by modeling the variances and covariance terms with the pixel scale values