Modeling Spatially and Temporally Complex Land-Cover Change: The Case of Western Honduras*

*This research was supported by the National Science Foundation (NSF) (SBR-9521918) as part of the ongoing research at the Center for the Study of Institutions, Population, and Environmental Change (CIPEC) at Indiana University. We thank Dr. Gerald Nelson for technical assistance in the estimation and presentation of model results and Laura Carlson for helpful GIS suggestions. We also acknowledge the useful comments and suggestions of three anonymous reviewers that have greatly improved the paper. Finally, we are grateful to Dr. Elinor Ostrom for her comments on earlier drafts.

Abstract

This article presents an econometric analysis of land-cover change in western Honduras. Ground-truthed satellite image analysis indicates that between 1987 and 1996 net reforestation occurred in the 1,015-km² study region. While some reforestation can be attributed to a 1987 ban on logging, the area of reforestation greatly exceeds that of previously clear-cut areas. Further, new area was also deforested between 1987 and 1996. Thus, the observed land-cover changes represent a complex mosaic of changing land-use patterns across time and space. The analysis contributes to the literature on land-cover change modeling in that: (1) it compares two econometric approaches to capture complex and often bidirectional changes in land cover from 1987 to 1996 as a function of agricultural suitability and transportation costs, and (2) it addresses techniques to identify and correct for spatial autocorrelation in a categorical regression framework.

Key Words:

• reforestation
• satellite image analysis
• econometrics
• land-use/land-cover change
• Honduras

Notes

*Source: Mertens and Lambin (2000, 474).

**Denotes significance at 95% level and

***at 99% level; standard errors in parentheses.

**denotes significance at 95% level and

***at 99% level.

¹Another alternative would be to specify a binary logit with panel data; for example, with some of the independent variables varying with time as well as with space. See Munroe, Southworth, and Tucker (2002) for an example of a more rigorous panel formulation.

²Because of the sheer size of the image, anything below a 9 × 9 window was so computationally intensive that it would have been very difficult to derive a weights matrix for each observation. However, at this window size, everything was spatially autocorrelated, so it is evident that an ideal window must be bigger to correct for spatial dependence.

*This research was supported by the National Science Foundation (NSF) (SBR-9521918) as part of the ongoing research at the Center for the Study of Institutions, Population, and Environmental Change (CIPEC) at Indiana University. We thank Dr. Gerald Nelson for technical assistance in the estimation and presentation of model results and Laura Carlson for helpful GIS suggestions. We also acknowledge the useful comments and suggestions of three anonymous reviewers that have greatly improved the paper. Finally, we are grateful to Dr. Elinor Ostrom for her comments on earlier drafts.

Additional information

Notes on contributors

Darla K. Munroe

An assistant professor

Jane Southworth

An assistant professor

Catherine M. Tucker

An assistant professor