A comparison of downscaled modeled land use scenarios and their methodologies

Introduction 

Land use and land change models are growing in abundance and complexity due to technological advances and the mating of methodologies from distinct disciplines. A common thread among these models is the need to downscale global, continental, or regional land requirements to more local, spatially explicit, and visually satisfying outputs. This need stems from the necessity to communicate and inform policy and decision makers of the consequences of different development pathways [1]. This research aims to compare the spatial allocation of global agricultural lands from scenarios modeled using the CLUMondo [2], GCAM [3], and GLOBIOM [4] models, and then explain the differences in agricultural land allocation by comparing their downscaling methods.

Methodology

Harmonization of spatial resolution and land use classes for GCAM's RCP4.5, GLOBIOM, and CLUMondo outputs is performed to enhance the comparability of the scenarios across the models. An initial spatial pattern cluster analysis is performed using the Global Moran's I index to determine the degree of clustering in each dataset. The global clustering values for each time step for each model output are plotted to evaluate the temporal differences in agricultural clustering trends. Next, regional clustering of high and low concentrations of agricultural land is evaluated using the Local Moran's I index as well as the Getis-Ord Gi* statistical measures. A directed evaluation of the downscaling procedures is performed at locations where prominent discrepancies between models exist.

Results and Conclusions

High levels of clustering of agricultural lands are found in all three modeled outputs. The regional clustering analysis reveals that consistently high concentrations of agricultural land are located in Southeast China, Central North America, Central Africa, and India, for example. Discrepancies among the different datasets exist, and are mostly located along the fringes of clusters of high and low concentrations of agricultural land. It is concluded that discrepancies can be explained by differences in the downscaling procedures. Moreover, while discrepancies between datasets exist, a discussion of these serves to elucidate potential methodological improvements and enhance confidence in and  utility of their results.

Please note these Proceedings have received limited or no review from supervisors and IIASA program directors, and the views and results expressed therein do not necessarily represent IIASA, its National Member Organizations, or other organizations supporting the work.

Last edited: 19 August 2015