Comparative Evaluation of Seasonal Patterns in Long Time Series of Satellite Image Data and Simulations of a Global Vegetation Model

A method has been developed and tested for comparing the complex spatio-temporal patterns present in two long time series of data of the seasonal cycles of vegetation for a large part of the global land surface. These two datasets are derived from global satellite observations (Advanced Very High Resolution Radiometer time series) and from a leading biogeochemical process model of global vegetation [Lund-Potsdam-Jena dynamic global vegetation model (LPJ-DGVM)], respectively. The datasets are completely independent of each other. The parameter compared is the fraction of photosynthetically active radiation. The comparison yields comparative parameters that quantify the differences between the two datasets. These comparative parameter images provide sufficient compression that they can be used for visual analysis so as to better understand the compatibility of, and the discrepancies between, the two datasets. The analysis shows that the LPJ model generally produces good correspondence with natural vegetation where the latter is primarily dependent upon climate. The correspondence was not as good where altitude, geomorphology, or hydrology of an area are the primary determinants of vegetation status. Due to a lack of representation of agriculture in the model, the correspondence with actual vegetative status was also poor where there is significant agricultural activity in an area.