Main "driving forces" cited in the literature

A large body of literature exists on the proximate causes and so-called "underlying driving forces" of deforestation in the region. The forest clearing agents (i.e. individuals or groups of individuals) are studied in a large number of case-studies at a more or less theoretical level, analysing the variables that influence their behaviour and decisions (i.e. proximate causes or proxies). It is implicitly or explicitly shown that these causes are scale dependant. Three main points limit the usefulness of existing deforestation studies to our work:

• detailed case studies focus on small areas and highlight the local validity of their results. Moreover, results are often biased by the specific (often disciplinary) approach of the teams, when our study requires a more holistic conceptualisation;
  
• the spatial aspects of the process are not taken into account by the numerous deforestation studies at regional or global scale. Moreover, the willingness to explain the process by identifying principal "driving forces", puts different factors into competition whereas it is their time and space dependant interrelations that should be emphasised;
  
• there is no generally accepted definition of what a "cause" of deforestation means, which may well explain much of the disagreement among specialists. Barraclough and Ghimire (2000), studying agricultural expansion, point to this core problem when they state that "... even if agricultural expansion did account for half or more of deforestation, this tells little about why this was occurring."

Still, the results obtained by the numerous studies realised over the last two decades allow to identify a set of factors that contribute to the deforestation process in the Neotropics, as well as their estimated level of influence.

This study represents an important step towards filling the gap between on the one hand national, regional and global estimates on the magnitude of the process, and on the other hand case-studies addressing the causes of the process.

An extensive review of literature on deforestation allowed to list the driving forces that have been mentioned to contribute to deforestation, in sub-areas or in the study region as a whole. These will next be referred to as "land use change location factors ", considered as a more appropriated denomination. The list of indicators in the table has been obtained after grouping "driving forces" with a similar definition. The construction of these indicators requires the combination and processing of several spatial data sets. In most cases, these data sets are available at a high spatial resolution (about 1km²), but the varying spatial precision and definition (spatial sampling) imposes a decrease to at least 3x3km resolution for the indicator set to be spatially consistent. The table also shows how scale is considered: either through downscaling of location factors or by defining specific location factors. A first set of factors, named “local factors”, describes the local situation in each point in space independently from its surroundings. For example, edaphic and climatic factors only describe the local situation, and are not affected by the neighbouring pixels.

The second set of factors, named “contextual factors”, qualifies the pixel for which they are defined, in relation with its surroundings. These are used to describe the landscape in which the pixel is situated, as well as its relative position within the country. Contextual factors can be the result of a simple downscaling, as for population growth and density (15×15 km and 45×45 km neighbourhood averages of the original 3×3 km resolution). They also can result from specific calculations.

As the table indicates, some factors change in time. These changes are considered by the model through the updating of the corresponding input maps after each yearly step model run.