Search

This study examines the correlation between deforestation, carbon dioxide emissions and potential causal factors of land-use change within an area of 2.7 million ha in Chiapas, southern Mexico between 1975 and 1996. Digitized land-use maps and interpreted satellite images were used to quantify land-use changes. Geo-referenced databases of population and digitized maps of roads and topography were used to determine which factors could be used to explain observed changes in land-use. The study analyzed the relationship between carbon emissions during this period and two types of possible causal factors: “predisposing” factors that determine the susceptibility of a particular area of forest to change (slope, distance to agriculture and roads, land tenure) and “driving” factors representing the pressures for change (population density, poverty). The correlated factors were combined in risk matrices, which show the proportion of vulnerable carbon stocks lost in areas with defined social, economic and environmental characteristics. Such matrices could be used to predict future deforestation rates and provide a verifiable evidence-base for defining baseline carbon emissions for forest conservation projects. Based on the results of the analysis, two matrices were constructed, using population density as the single most important driving factor and distance from roads and distance from agriculture as the two alternatives for the predisposing factors of deforestation.

We present a methodology for testing and applying a regional baseline for carbon (C) emissions from land-use change, using a spatial modelling approach (hereafter called the Climafor approach). The methodology is based on an analysis of causal factors of previous land-use change (Castillo et al. 2005). Carbon risk matrices constructed from the spatial correlation analysis between observed deforestation and driving factors (Castillo et al. 2005), are used to estimate future carbonemissions within acceptable limits for a forest conservation project. The performance of two risk matrices were tested by estimating carbon emissions between 1975 and 1996 from randomly selected sample plots of sizes varying from 1,600 to 10,000 ha and comparing the results of the observed emissions from these sample plots with the model estimations. Expected emissions from continued land-use change was estimated for the community applying the risk matrices to the current land cover. The methodology provides an objective means of constructing baseline scenarios including confidence intervals, using the sum of variances of the various data sources, such as measured carbon densities, classification errors, errors in the risk matrices, and differences between the model prediction and observed emissions of sample plots due to sample size. The procedures applied in this study also give an indication of the impact of the variance in the various data sources on the size of the confidence intervals, which allows project developers to decide what data sources are essential to improve his baseline. The modelling approach to estimate the deforestation pattern is based on readily available cartographic and census data, whereas data on carbon densities are required to assess the potential for forest conservation projects to offset carbon emissions.