We present a procedure for estimating temporal persistence in Advanced Very High Resolution Radiometer (AVHRR)-Normalized Difference Vegetation Index (NDVI) data. Our algorithm provides for the estimations of the tendencies of NDVI to increase or decrease starting from selected reference conditions by means of linear fits that cover time periods increasing at a 1-year rate. The times at which such trends change their sign (first return times) are estimated, and the distribution that best fits their histogram is determined. Such a distribution reflects the character, stationary or not, of NDVI changes and allows for estimating persistence probability and for quantifying possible characteristic time scales. In order to check the algorithm's performances, we estimated persistence probability at a 1.1-km resolution over the test site of Southern Italy and selected subareas. Histograms of return times aggregated according to vegetation classes were also analyzed. In all of the cases, we found the exponential decay taw that is typical of processes that attain to long-term stationary patterns. Estimated mean life times of decreasing processes were generally shorter than those characterizing the increasing ones. Our findings express a good ability of vegetation to recover from disturbances and to efficiently exploit resources. Overall, rather stable conditions seem to characterize the test site and local exceptions can be easily explained with independent ground data. The good agreement between our inferences and results from independent studies suggests that dynamics estimated from AVHRR data can provide interesting information on vegetation resilience and response to human/climate forcing. In particular, quantitative information can be a key tool for the characterization and parameterization of vegetation changes and for the development and validation of dynamical models.
Temporal persistence in vegetation cover changes observed from satellite: Development of an estimation procedure in the test site of the Mediterranean Italy
Lanfredi M;Simoniello T;
2004
Abstract
We present a procedure for estimating temporal persistence in Advanced Very High Resolution Radiometer (AVHRR)-Normalized Difference Vegetation Index (NDVI) data. Our algorithm provides for the estimations of the tendencies of NDVI to increase or decrease starting from selected reference conditions by means of linear fits that cover time periods increasing at a 1-year rate. The times at which such trends change their sign (first return times) are estimated, and the distribution that best fits their histogram is determined. Such a distribution reflects the character, stationary or not, of NDVI changes and allows for estimating persistence probability and for quantifying possible characteristic time scales. In order to check the algorithm's performances, we estimated persistence probability at a 1.1-km resolution over the test site of Southern Italy and selected subareas. Histograms of return times aggregated according to vegetation classes were also analyzed. In all of the cases, we found the exponential decay taw that is typical of processes that attain to long-term stationary patterns. Estimated mean life times of decreasing processes were generally shorter than those characterizing the increasing ones. Our findings express a good ability of vegetation to recover from disturbances and to efficiently exploit resources. Overall, rather stable conditions seem to characterize the test site and local exceptions can be easily explained with independent ground data. The good agreement between our inferences and results from independent studies suggests that dynamics estimated from AVHRR data can provide interesting information on vegetation resilience and response to human/climate forcing. In particular, quantitative information can be a key tool for the characterization and parameterization of vegetation changes and for the development and validation of dynamical models.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
