 |
|
|
Introduction
Location Map
Base Maps
Database Schema
Conventions
GIS Analyses
Flowchart
GIS Concepts
Results
Conclusion
References
|
Conclusion
Conclusion
Climate accounts for only a small amount of the variablity in current population distribution in Ethiopia. This is to be expected, given the complexity of human decision-making and settlement patterns. However, it also sheds light on previous studies that have positied a strong connection between population distribution and climate (Clegg et al. 1972, Roundy 1976, Egziabher 1998, Nyssen et al. 2004). Our model also builds on past research by demonstrating which specific climate variables are the best predictors for population density in Ethiopia. Because climate accounts for only a small amount of human population variability, our final population distribution is more valuable in terms of the spatial patterns that it shows than for its absolute population density numbers.
|
| |
|
|
| Ideas of how to improve on our work: |
- Stratify by elevation and create separate models for each elevation zone, since increased temperature in the highlands may be favorable, but not in the lowlands. This might give a more realistic and nuanced picture of where people would actually move.
|
- Consider a model that accounts for the potential of the response variable being categorical. Analysis of covariance (ANCOVA) may potentially be an improved model structure to use on the Ethiopian data. Logistic regression may potentially be used as well, but typically it is applied to cases where the response variable is binary (i.e. present/absent).
|
- Consider resmapling the layers to a coarser, instead of finer, resolution because the spatial scale may be too fine for our study. The pixel resolution used in this study was 1 kilometer squared, but human decision making is likely occurring over a broader spatial extent, particularly with regard to climate effects.
|
| Potential Problems : |
- The population data seems to have been sampled by Woreda, and therefore does not have fine enough resolution to allow for a very good climate model to be parameterized. It can account for very general differences in topography and climate, but it does not differentiate between population density over short distances, such as changes that would occur between river valleys and plains, or between steep slopes and flatter areas. Our model could be improved in terms of being able to explain more of the population variablitiy if we had averaged climate and topographic variables within each Woreda, but the results would have been correspondingly coarse.
|
| Caveats: |
- There may be a lag effect for processes that should respond more slowly to climate change, such as soil formation and properties (e.g., pH, nutrient availability). Our model predicts an immediate population response to climate change, but in reality it would occur more gradually.
|
| |
| Acknowledgements: |
| We would like to thank Jessica Salo, Greg Newman, and Paul Evangelista for their help on this project. |
|
|
|
