Testing TM and SPOT data as input to a canopy reflectance model for mapping density and size of trees in forest stands

A new method, based on region-growing segmentation and a variance-dependent geometric-optical canopy model, is tested to map forest stand attributes such as crown size and tree spacing. The principal objective was to test whether SPOT´s smaller 20-m pixel leads to improved inversions of tree size and spacing relative to what has been experienced earlier with Landsat TM´s larger 30-m pixel. Theory gives good reason to expect that a finer spatial resolution-smaller pixels-will improve the results, especially for tree size. However, the results show that 20-m pixels is not significantly better than 30-m pixels as input to the model. It turns out that many stands are outliers far off the field-measured values, but when eliminating outlier stands, the results are promising