Analyzing Time Series of Satellite Imagery Using Temporal Map Algebra

This paper presents the application of a set of spatio-temporal satellite image analysis functions based on temporal map algebra. The temporal map algebra functions treat time series of imagery as three-dimensional data sets where two dimensions encode planimetric position on the earth’s surface and the third dimension encodes time. Whereas the conventional local, focal, and zonal map algebra functions take as input one or more grids and output a grid (or, in the case of zonal functions, a table), temporal map algebra functions take one or more three-dimensional data ‘cubes’ as input and output a data ‘cube’ (or, in the case of temporal zonal functions, a table). As a demonstration of the utility of temporal map algebra for time series satellite image analysis, we investigate the impact of El Niño/Southern Oscillation (ENSO) on the 1982-1993 monthly time series Advanced Very High Resolution Radiometer (AVHRR) Normalized Difference Vegetation Index (NDVI) signal in southern Africa. Results suggest that vegetation intensity is suppressed, and its spatial and temporal variability enhanced, during ENSO warm and cold phase months, particularly over forest, woodland, and wooded grassland. These effects are generally diminished by the following growing season, particularly following an ENSO cold phase. For shrubland, vegetation intensity tends to increase during the growing season following an ENSO cold phase.