Multidimensional Map Algebra (MMA)

Case Study: Spatio-Temporal Variation in African Vegetation Intensity

(from Mennis, J., 2005. Spatial and temporal vegetation variability in Africa: An application of temporal map algebra. In Proceedings of the ASPRS Annual Conference, March 7-11, Baltimore, MD.) 

Introduction

Multidimensional map algebra (MMA) is used to analyze the impact of El Nino/Southern Oscillation (ENSO) phase on vegetation intensity over different land covers in three African regions: the western Sahel, southern Africa, and eastern Africa. The study regions are shown in the map below.

Data

Vegetation intensity is indicated using monthly 1982-1999 normalized difference vegetation index (NDVI) data derived from the Advanced Very High Resolution Radiometer (AVHRR).  These 8 km data were preprocessed to remove egregious errors and identify the anomaly from the long term monthly mean for each month..  The map below shows an NDVI image for the southern Africa study region for January 1982.

Eight km land cover data were also derived from AVHRR imagery.  The map below shows land cover for the southern Africa study region.

ENSO phase data indicated whether each month is associated with an ENSO warm, cold, or neutral phase.  Because of the potential lag in vegetation response to ENSO forcing, another ENSO time series was created that focused only on the growing season for each study region. 

Methods

Using a prototype version of MMA implemented in the scripting language IDL, mean NDVI anomaly is calculated for each land cover and simultaneous and growing season ENSO phase.

Results

The table below reports the mean NDVI anomaly for each ENSO phase using both the simultaneous and growing season ENSO phase data. 

The graphs below show mean NDVI anomaly (shown on the Y axis) over different land covers during different ENSO phases: eastern Africa (A and B), western Sahel (C and D), and southern Africa (E and F).  Purple bar is cold phase; red bar is neutral phase; and yellow bar is warm phase.  Note that graph sets A, C, E and B, D, F are generated using the simultaneous and growing season ENSO phase data, respectively.

Conclusion

In eastern Africa, ENSO cold phase is generally associated with enhanced vegetation intensity, particularly for woodland and open shrubland.  The exception is for wooded grassland, which exhibits suppressed vegetation intensity for an ENSO cold phase.  ENSO warm phase is associated with enhanced vegetation intensity in shrubland.  This pattern continues through the growing season, though notably cropland also emerges as having a positive vegetation response.to the ENSO cold phase.