ODD description

The MEDLUCC model

Overview

Purpose

The MEDLUCC model main goal is study the interaction between drivers of land use change at landscape level to replicates the observed spatial patterns of LUCC and predicts new landscape compositions governed by scenarios of global change. The model is designed to mimic the most relevant LUCC processes that have transformed the mosaic of the Mediterranean landscape in the last decades, those are: afforestation (i.e.: rural abandonment), cropping (i.e.: agriculture conversion) and urbanization.

State variables and scales

The state variables are spatial variables that describe the landscape composition and the landscape features related to the land use changes. All the spatial variables cover the full extent of the study area at 100m resolution. The temporal scale is fixed all along the simulation period and the user defines the time step at one up to ten years.

The landscape composition is described in terms of land-cover types (hereafter LCT). The dynamic land-covers are divided in natural (forests and shrubs), croplands (arable land and cereals cropping), and urban, while the static land-covers represent grasslands – pastures, water bodies – wetlands and rock – sands. The state variable time since change (TSC) records the number of time steps since the last LUCC event. Related to each LUCC process the probability of change (PC) estimates the probability at cell-level to change to that LCT. The variable PC is function of spatial constant factors describing the main drivers of change and the dynamic factor density that account for the percentage of the target land-cover in a 1 km radius neighbourhood. In the current version the only driver of change is the elevation.

Process overview and scheduling

The afforestation process is the result of cropland becoming natural, being scrubland in the first stage (the natural succession is not deal in this modelling framework). The cropping process is the transition from natural covers to cropland. The urbanization consists in the transformation of natural covers and croplands to urban. These processes occur simultaneously in the space and in the time during every simulation step.

The MEDLUCC model performs two fundamental tasks each time step. First, the spatial variable probability of change is assessed for each land-cover transition once the dynamic spatial variable density is updated. Second, the model allocates the target demand (that could be time-dependent) modifying the value of the specified number of cells from the state variable land-cover types to the target LCT. The MEDLUCC use a new allocation approach that allows spatial aggregation of change depends on the target land-cover. The details are given in the following ODD section.

At the end of every simulation the model returns a set of non-spatial outputs that describes the statistics related to the patches of change for each time step and each replicate. The dynamic spatial variables are in the same way recorded.

Design concepts

The MEDLUCC model adopted a modular framework that distincts but at the same time links the Demand and the Spatial allocation components of any LUCC model.

The Demand component specifies yearly (or n-yearly) proportions (or areas) that should become one of the dynamic LCT, so per each time step and each LCT a demand value must be provided. In LUCC modelling two main approaches have been used for construction of demand: extrapolations of historical trends or exploratory scenarios based on expert knowledge or reflecting hypothetical futures.

The Spatial allocation component spatially distributes the prescribed LUCC demand in a given region. The core drivers for identifying the best locations of LUCC are probability surfaces based on empirical relations between existing LUCC patterns and a wide array of spatially explicit biophysical and socioeconomic data. In addition, the probability of change depends on the LCT spatial distribution directly increasing according to the LCT spatial density. To spatialise the changes, the model uses a new spread-based allocation methodology.

For each dynamic LCT and according to its probability of change surface, a waiting time value is assigned to as many cells as target demand specifies, creating a time-sorted set of cells that potentially will change. The waiting time is negatively exponential distributed and is LCT-depending. The cell (or cells) with the lowest waiting time changes to the target LCT and new waiting times are computed for the four neighbours on condition of they have not already changed and are dynamic LCT. The waiting time of the spreading cells could depend on the variable probability of change, be a complex function aggregating drivers of spreading change or as in the current version, follow an exponential distribution with parameter differing from the initiation distribution. The next cell or cells to change will have the current lowest waiting time allowing the spatial aggregation of LUCCwhen controlling the relative interaction between the waiting time of potentially change cells and spreading change cells. Dynamic LCT cells will undergo a change until the demand is consumed.

Details

Initialization

The land-cover types spatial variable is derived by reclassification from the Land Cover Maps of Catalonia. The variable time since change is set to zero at the beginning of the simulation. Based on the initial landscape composition, the model assesses the density variable for each LCT as the percentage of cells of that LCT in a 1 km radius neighbourhood. The model counts for each pixel the number of neighbours cells in a 1 km radius according to the spatial variables resolution. The probability of change to a target LCT is function of the initialized density and the constant spatial drivers of change (e.g.: elevation, population density, distance to road, …).

Input

The model requires the externally defined target demand that means, the amount of hectares (or the percentage over the study area) which will undergo a LUCC process.