Land use change in California, 2001-2100

The SLEUTH urbanization and land use change model was used to produce century-long forecasts of California’s land uses to the year 2100. We describe how data were assembled and conflated for the model, how the model was applied to the very large and high resolution dataset, and how calibration of the model was performed using a genetic algorithm. The calibration results showed that the model accuracy was high, and suitable for simulations, which used Monte Carlo methods to capture both future land uses and their modeling uncertainty. The simulations showed the amount and location of anticipated futures changes in land use, dominated by changes to urban and a few other pairs of land use transitions. The proportion of urban land was 6.7% in 2001, 8.8% in 2011, and is forecast to reach 14.5% in 2050 and 17.6% in 2100, converting some 4,456,160 hectares of land. Farmland lost 224,135 Ha. between 2001 and 2011, but is projected to lose another 1,346,912 Ha by 2100, a decline over the century from 10.05% to 6.76% of the total land area. Of the 15 top land transitions by land area, accounting for 88% of the expected change by area from 2001-2100, only 6 of them are expected to be transitions to urban, the remainder are combinations of changes among shrublands, grassland and forest that will also have major consequences for California’s future. The authors hope that the projections of land use change will be of use to other scientists, land managers and policy makers.

Methods
The data set consists of the full set of inputs for the SLEUTH land use change model for California for the period 2001-2100. The primary data source is the 2001 and 2011 land use data from the National Land Cover Database. The model was calibrated using methods described in the paper, and forecasts of future land use produced for each decade from 2020 to 2100. Also included are the SLEUTH calibration results and individual forecasts by data tile, along with uncertainty estimates for the model forecasts. Forecast data sets are in GeoTIFF format, recificed to the Albers Conformal Conic projection and NAD83.