The spatial analysis of global spatial datasets is computationally intensive. We need ever more computation given the growth in availability of such datasets from a growing range of sensing devices, as well as the increase in interoperability that supports the amalgamation of national datasets. We present some of the methods used with open source GIS to support the analysis of large spatial datasets, including the use of Python libraries OGR, Shapely and Rtree.
The data analysed included a global urban forecast map that was generated by Seto et al. 2012, who applied a Monte Carlo approach in a two stage process to generate the probabilistic forecasts of global urban expansion to 2030. Using this data combined with data on global cultivated land and crop yield productivity, we were able to calculate the amount of cultivated land and crop yield that would be lost to the forecasted urban expansion at global and national levels. Given the size of the datasets involved, we parallelised parts of the analysis process such that the spatial analysis was applied per nation state per computing node, using 32 CPU cores of POWER755.
- Seto KC, Güneralp B, Hutyra LR (2012) Global forecasts of urban expansion to 2030 and direct impacts on biodiversity and carbon pools Proceedings of the National Academy of Sciences of the United States of America 109(40): 16083-8