Abstract
Including the inter-annual variability of vegetation dynamics into land-surface models is necessary to account for land use/cover change effects on Global Climate Models. However, land-surface models use land-cover classifications dictated by structural attributes of vegetation that have little sensitivity to environmental change and are difficult to update and result in a delayed response. This rigid representation of vegetation reduces the ability of models to represent rapid changes including land-use shifts, fires, floods, droughts, and insect outbreaks. Functional attributes of vegetation describing its energy and matter exchange with the atmosphere, have a shorter response to environmental changes and are relatively easy to monitor with satellite data. We applied the concept of Ecosystem Functional Types (EFTs; patches of the land-surface with similar carbon gain dynamics) to characterize the spatial and inter-annual variability of vegetation dynamics across natural and agricultural systems in the La Plata Basin of South America.
Three descriptors of carbon gain dynamics were derived from seasonal curves of Normalized Difference Vegetation Index (NDVI) and used to identify EFTs based on annual mean (surrogate of primary production), seasonal coefficient of variation (indicator of seasonality), and date of maximum NDVI (descriptor of phenology). Results from two NDVI datasets were compared (AVHRR-LTDR version 2, 1982-1999, 15-day and 5 km resolution; and MOD13A2 MODIS, 2000-2006, 16-day and 1 km resolution). Both datasets showed greater spatial and inter-annual variability of the EFT composition in agricultural areas compared to natural areas.
During 1982-1999, the percentage of the La Plata Basin occupied by EFTs with low productivity, high seasonality, and spring and fall NDVI maxima tended to decrease, while EFTs with high productivity, low seasonality, and summer maxima tended to increase. We speculate that these trends may be due to a positive trend in precipitation. By contrast, during the period 2000-2006, the areas of EFTs with high productivity, low seasonality, and spring and fall maxima (mainly associated with natural forests) tended to reduce, while EFTs with low to medium productivity, high seasonality, and summer maxima (mainly associated with crops) tended to expand. The AVHRR-LTDRv3, covering 1981-2007, will help to assess whether the observed difference between the 1981-1999 and the 2000-2006 periods may be partially due to the change from AVHRR to MODIS datasets. Likewise, it will allow us to study how the yearly changes in the EFTs composition affect the performance of Climate Models during 1982-2007.
