CONCLUSIONS
1. A limited number of topographic and climatic variables can be used to predict at least some of the major classes of potential vegetation in Mexico.
2. Replacing ‘current’ climate data with ‘projected’ climate data can be used to predict projected vegetation distributions.
3. Rzedowski’s classification for potential vegetation in Mexico, defined on the basis of expert judgement, appears to have some ecological significance because clustering identified several of Rzedowski’s major vegetation classes.
4. Comparisons between regional and global climatic data sets showed climate data resolution finer than 0.5 degrees did not significantly improve classification results, indicating that fine-scale control of vegetation prediction resides in the topographic data and that globally-available, climatic data sets can be successfully employed for such assessment.
5. Unique conditions of environmental variables that describe one vegetation type can be considered a ‘core’ region of vegetation; peripheral areas outside the core region described by slightly different environmental conditions that aren’t as unique as the environmental conditions that describe the core region should define peripheral regions, which may be (or become) zones of transition between adjacent vegetation types.
6. Clustering techniques provide a powerful tool for both potential vegetation classification and the classification of core and peripheral regions for each vegetation type.
7. Clustering within ecologically similar regions of landscape can improve potential vegetation classification.
8. Using the maximum scaled distance measure appears to better predict vegetation types characterized by environmental extremes.
9. Differences in results for classifying potential vegetation suggest that further study could improve both the methods developed and our understanding of the basis for expert classifications systems.
10. In view of promising results in developing quantitative links between
environmental factors and habitat classification, the methods developed
provide an additional approach to the problem of assessing the biogeographic
effects of global change.
RECOMMENDATIONS FOR FURTHER STUDY
The following avenues of research have been identified as potentially fruitful for future development:
- Use of supervised clustering or some other approach that allows the user to better control how the clustering algorithm clusters familiar or calibration data points.
- Better use of LoiczView features such as optimal cluster algorithms and visualization capabilities including color-coded similarity analysis and dimensional viewing of clusters.
- Further investigation of the scale or resolution dependence of the classification with respect to different variables.
- Use of different variables for modeling vegetation distribution such
as water budget, edaphic and geologic variables, and incorporation of variability
and the seasonality component of climate.