Land Use Change Modelling in R

Abstract:

Land use activities, through the provision of natural resources, are essential to human existence. In many regions land use change is degrading biodiversity and threatening the sustainability of ecosystem services upon which communities and livelihoods depend. Spatially explicit land use change models are widely used to understand and quantify key processes that affect land use change and make predictions about past and future change. These models typically include a module to estimate the suitability of different locations to particular land use types based on biophysical and socioeconomic predictor variables and a module to allocate change spatially. They are commonly implemented in languages such as C/C++ and Fortran and made available as standalone applications or through proprietary GIS. In many cases the models are released under closed source licences, limiting the reproducibility of scientific results and making model comparison difficult. This work presents a new R package providing methods and classes to support land use change modelling and model development and comparison within the open source R statistical computing environment. The package makes use of existing R implementations of methods such as random forests and recursive partitioning and regression trees to estimate location suitability, as well as providing methods for statistical model building and evaluation. Currently two spatial allocation methods are provided: the first based on the widely used and tested CLUE-S algorithm and the second a novel stochastic procedure developed for large scale applications. Some common tools for evaluating allocation results are implemented. It is hoped that the package will provide a framework for the development of new routines that can be incorporated into future releases of the code.