SEARCH

SEARCH BY CITATION

Keywords:

  • Artificial neural network;
  • biotemperature;
  • Caribbean;
  • digital elevation models;
  • Holdridge;
  • land use;
  • model comparison;
  • potential evapotranspiration ratio;
  • St Lucia

Abstract

Aim

The purpose of this study is to apply geographical information and artificial neural network (ANN) technologies in assessing ecosystem distribution on the island of Saint Lucia, as well as to develop an improved ecological classification using Holdridge’s system of natural life zones.

Location

Saint Lucia is a Caribbean island state located at 14°N and 61°W and of a land area of 616 km2.

Methods

The main inputs for classifying life zones were a 25-m × 25-m digital elevation model of Saint Lucia (DEM), mean annual temperature and annual total precipitation. The DEM was initially obtained by digitizing contour lines on a topographic map. Elevation–temperature regressions developed for Puerto Rico were used to generate point-estimates of mean temperature across the island of Saint Lucia. A generalized (trained) ANN was employed to create an annual total rainfall surface for the island. The variables of longitude, latitude and elevation were used to construct the rainfall model. Comparison of predicted and observed total precipitation revealed that the ANN explained over 95% of variability exhibited in the observed data, within a standard error of estimate of 123 mm (~6% of the total precipitation).

Results

Three complete and three transitional life zones were identified as occurring on Saint Lucia. Twelve per cent of the island was classified as tropical premontane moist/wet, 20% as tropical premontane wet, 6% as subtropical dry/moist, 29% as subtropical moist, 26% as subtropical moist/wet and 7% as subtropical wet.

Conclusion

Quality of life zone delineation depends on an objective application of universally accepted criteria and available terrain analysis technologies.