Incorporating land-use mapping uncertainty in remote sensing based calibration of land-use change models

Building urban growth models typically involves a process of historic calibration based on historic time series of land-use maps, usually obtained from satellite imagery. Both the remote sensing data analysis to infer land use and the subsequent modelling of land-use change are subject to uncertainties, which may have an impact on the accuracy of future land-use predictions. Our research aims to quantify and reduce these uncertainties by means of a particle filter data assimilation approach that incorporates uncertainty in land-use mapping and land-use model parameter assessment into the calibration process. This paper focuses on part of this work, more in particular the modelling of uncertainties associated with the impervious surface cover estimation and urban land-use classification adopted in the land-use mapping approach. Both stages are submitted to a Monte Carlo simulation to assess their relative contribution to and their combined impact on the uncertainty in the derived land-use maps. The approach was applied on the central part of the Flanders region (Belgium), using a time-series of Landsat/SPOT-HRV data covering the years 1987, 1996, 2005 and 2012. Although the most likely land-use map obtained from the simulation is very similar to the original classification, it is shown that the errors related to the impervious surface sub-pixel fraction estimation have a strong impact on the land-use map's uncertainty. Hence, incorporating uncertainty in the land-use change model calibration through particle filter data assimilation is proposed to address the uncertainty observed in the derived land-use maps and to reduce uncertainty in future land-use predictions.