Subpixel spatiotemporal pattern analysis of remote sensing observations for predicting grassland ecological and biophysical parameters

One of the most important limiting factors in reliable estimation of grassland ecosystem parameters from remotely sensed data is "too coarse" resolution. Areal averages obtained from readily available imagery do not match well with ecological field data. To address this discrepancy the authors have collected simultaneous spectral, other biophysical and ecological data with similar ground resolution between May and August 1995 in the Grassland National Park, Saskatchewan, Canada. Beyond calibration for a wide variety of conditions, the applied nested sampling design facilitates scaling of measured and estimated properties from 0.5 m to 100 m by partitioning overall variability, and by hierarchical scene simulation. Analysis of the spatiotemporal pattern reveals potential problems with using areal averages in long-term monitoring and assessment of ecological status of grasslands.