Infodoc : Sciences, Techniques et Culture

The deforestation is increasing worldwide mainly due to human activities. This has aconsiderable impact on carbon dioxide levels and increases global warming. Suriname is still considered a High Forest cover, Low Deforestation (HFLD) country, but there are areas that undergo high deforestation pressure due to population growth and the gold fever.In this study the focus will be on predicting the deforestation risk in a shifting cultivation and gold mining site in Suriname using spatial factors. Gold mining is now suspected to be themain driver of deforestation in Suriname and the population growth in some villages in theinterior can lead to high deforestation pressure in this area. The spatial risk of deforestation due to shifting cultivation and gold mining will be predicted by creati[...]

The deforestation is increasing worldwide mainly due to human activities. This has aconsiderable impact on carbon dioxide levels and increases global warming. Suriname is still considered a High Forest cover, Low Deforestation (HFLD) country, but there are areas that undergo high deforestation pressure due to population growth and the gold fever.In this study the focus will be on predicting the deforestation risk in a shifting cultivation and gold mining site in Suriname using spatial factors. Gold mining is now suspected to be themain driver of deforestation in Suriname and the population growth in some villages in theinterior can lead to high deforestation pressure in this area. The spatial risk of deforestation due to shifting cultivation and gold mining will be predicted by creating a model using Random Forest, which is a powerful machine learning classifier tool.The explicative variables used in the deforestation location model were the distances toformer deforestation, villages, rivers and streams. The data were analyzed using open-source software’s : GRASSGIS and R. The generalized error of the model for shifting cultivation and gold mining were 7.43% and4.46% respectively. Distance to former deforested areas seemed to be the most important predictor for both study areas, which can be explained by former studies showing that locations surrounded by recently deforested land have a higher risk to be deforested. The distances to streams seemed to be more important for deforestation in the shifting cultivation study area, whereas the distances to villages were more important in the gold mining site. When the model was used with data from another period in the future, this resulted in a higher error rate meaning that there is more incorrect prediction. The ratio of forest gain to forest loss shows that the regeneration of deforested areas due to gold mining is slow compared to there growth in shifting cultivation, whereas the forest loss is much greater in gold mining than inshifting cultivation. Finally, the deforestation risk maps produced by the model give a better idea of the deforestation pressure on the shifting cultivation and gold mining study area.