A Random Forest-Cellular Automata modelling approach to explore future land use/cover change in Attica (Greece), under different socio-economic realities and scales

Abstract

This paper explores potential future land use/cover (LUC) dynamics in the Attica region, Greece, under three distinct economic performance scenarios. During the last decades, Attica underwent a significant and predominantly unregulated process of urban growth, due to a substantial increase in housing demand coupled with limited land use planning controls. However, the recent financial crisis affected urban growth trends considerably. This paper uses the observed LUC trends between 1991 and 2016 to sketch three divergent future scenarios of economic development. The observed LUC trends are then analysed using 27 dynamic, biophysical, socio-economic, terrain and proximity-based factors, to generate transition potential maps, implementing a Random Forests (RF) regression modelling approach. Scenarios are projected to 2040 by implementing a spatially explicit Cellular Automata (CA) model. The resulting maps are subjected to a multiple resolution sensitivity analysis to assess the effect of spatial resolution of the input data to the model outputs. Findings show that, under the current setting of an underdeveloped land use planning apparatus, a long-term scenario of high economic growth will increase built-up surfaces in the region by almost 24%, accompanied by a notable decrease in natural areas and cropland. Interestingly, in the case that the currently negative economic growth rates persist, artificial surfaces in the region are still expected to increase by approximately 7.5% by 2040.

Assessing and mapping multi-hazard risk susceptibility using a machine learning technique

“…The RF can precisely handle heterogeneous inputs of different nature and scalability from different sources 55,57 . Another important benefit of the RF model is that there are significant criteria that indicate the importance of each predictor variable 55,58 . However, it has some sources of uncertainty that are frequently unacknowledged or even unrecognized.…”

Using a hybrid demand-allocation algorithm to enable distributional analysis of land use change patterns

“…Such a procedure could be done independently by county, and it could possibly be used as an explanatory variable when determining initial groupings to estimate transition probability models. Finally, we used the random forests approach to estimate our transition probability models because random forests can easily and robustly combine disparate data types into a common predictive framework for land use transitions [21]. In this study, we used one transition probability model per county grouping (Fig 3), effectively limiting the variation in projected landscape pattern due to choice of the underlying probability model.…”

“…While there are a wide variety of land use change models available as the result of broad interest and research, which can be categorized along economic/non-economic, spatially-explicit/aggregated, and empirical/process-based lines [ 15 , 16 ]; demand-allocation methods [ 17 – 21 ] are particularly well-suited to situations where a projected land use change is available at a coarse scale and the intent is to downscale the projection. Demand-allocation models tend to input exogenously generated transition quantities (or quotas) for a given area (e.g., county), for example via socioeconomic models [ 10 , 18 ]; then allocate those quotas within the study area [ 22 ].…”

Delineation of Urban Growth Boundaries Using a Patch-Based Cellular Automata Model under Multiple Spatial and Socio-Economic Scenarios

“…These randomly-sampled observations were used to train the RF model, which was implemented using the "Scikit-learn" package in Python 3.6 [41]. The two primary hyperparameters in the RF model, whifh are the number of features for splitting and the number of trees in the "forest," were calibrated through a trial-and-error process in this study [42]. Finally, the number of variables for tree splitting was 5 (half of the candidate variables), and the size of the forest was 100.…”