Predictive modelling of tropical deforestation

CEES project

Predictive modelling of tropical deforestation

This project aims to generate a predictive model of tropical deforestation, able to predict the rate of deforestation in different regions of different countries under various economic and policy scenarios.

There are least three good reasons for wanting such a model. First, the effects of policies aiming to reduce deforestation can only be meaningully measured against a 'business as usual' projection of the deforestation that would have been expected to occur without such policies. Second, these business as usual projections could help to identify 'hotspots' where policy effort might best be targeted in the near future. Third, developing a predictive model of deforestation necessarily requires that we uncover the underlying causes of deforestation -- information which in turn can feed into policy development.

This project currently involves two halves, the first aiming to uncover the factors that predict deforestation at global to local scales (e.g. climate, population density, agricultural yields, crop prices), the second aiming to understand and simulate the development of road networks in forested regions (there is already strong evidence that roads are an important predictor of deforestation). In both cases, an important aspect of the work will be developing methods to capture and convey the uncertainty in model predictions. In come cases this uncertainty is likely to be very large (after all, deforestation is driven by humans, which are notoriously hard to predict) which itself has major implications for policy development.


  Dr Drew Purves (Microsoft Research Cambridge)

  Dr Rob Ewers (Imperial College London)

  Sadia Ahmed (PhD student, Imperial College London)

  Isabela Rosa (PhD student, Imperial College London)


Related Publications