Module 6, Allocating land use
learning objective: How to use the GeoDMS to allocate land use
The Land Use Scanner Student Edition provides an example model configuration and geographic data. It allows you to experiment with land-use modelling. It contains an assignment that serves as a first introduction to the modelling environment used at various research institutes worldwide to support spatial planning.
This exercise is created by the SPINlab of the Vrije Universiteit Amsterdam (Eric Koomen) and is adapted for this Academy.
Looking into the future has fascinated humanity for ages. With the advent of sufficiently powerful personal computers, we no longer need a crystal ball to foresee what our country will look like in, say, 30 years from now. All that is now required to simulate future spatial developments is a set of spatial data, modelling software and some thoughts on which spatial developments you expect to happen in the coming decades. The latter is, of course, the tricky part. Looking at current developments, we may discern trends likely to continue for 5 to 10 years, but looking much further is risky. Many uncertainties prevail: Will population growth come to a halt? Will economic development be slow? Will society prefer ecological sustainability or economic prosperity? What will be the role of the government in steering socio-economic development?
A much-favoured approach to dealing with the uncertainties relating to future spatial developments is using scenarios. By describing several opposing views on the future, we can simulate a broad range of spatial developments, thus offering a complete overview of possible land-use alterations. Each outlook on the future will not necessarily contain the most likely prospects, but as a whole, the simulations provide the bandwidth of possible land-use changes. The individual scenarios should, in fact, not strive to be as probable as possible but should stir the imagination and broaden the view of the future. Essential elements are plausible unexpectedness and informational vividness (Xiang and Clarke, 2003).
Furthermore, the projection of land use and land-use change can help to identify, mitigate or accommodate conflicts between different sectoral land claims and needs. While predicting what will happen is difficult, we can identify what cannot happen and what is likely to happen, given geographic restrictions and interactions. Land use and land-use change and their environmental effects on spatial quality, habitat, emissions, etc., result from (planning for) natural processes and the geographic allocation of land-occupying activities. The analysis and simulation of land-use change have resulted in developing various land-use change models, i.e. computational representations of land-use dynamics. Land use is modelled as the characteristics and intensity of spatially located activities or processes, such as residential, agricultural, natural, industrial or infrastructural areas. This relates to, but differs from, the physical state of land and what was built on it. For modelling urban developments, a connection is often made between infrastructure and accessibility. Such models are often labelled as Land Use/Transport Interaction (in short, LUTI) models.
In short, the focus and purposes of land use modelling are:
- check for the geographic feasibility of sectoral claims for land, taking into account geographic and planning restrictions and interactions. What needs can be combined?
- project different types of land use on land units, given sectoral claims and needs and the geographic characteristics and restrictions of these land units.
- assess the effects of projected land-use change on critical indicators, such as food production, housing supply, energy transition, carbon and nitrogen-emissions, natural habitats, and other spatial quality indicators.
Current dominant land use per cell at the 100m (left) and 500m raster resolution.
In this assignment, we will use the Land Use Scanner, an integrated land-use model that has been used for various policy-related research projects. Early applications include, among others, the simulation of future land use following different scenarios (e.g. Borsboom-van Beurden et al., 2007) and the evaluation of alternatives for a new national airport (Scholten et al., 1999). The model has been used in many outlooks on the future relating to a range of planning themes, such as water management (Dekkers and Koomen, 2007; De Moel et al., 2011); climate change (Koomen et al., 2008); the prospects of agricultural land use in the Netherlands (Koomen et al., 2005); and the development and evaluation of regional spatial strategies (Koomen et al., 2011). Apart from these Dutch applications, the model has also been applied in several European countries (Hoymann, 2010; Te Linde et al., 2011; Schotten et al., 2001). The GeoDMS framework underlying Land Use Scanner is also used in a modelling framework developed for the European Commission and that now covers the 27 member states of the European Union at a 100-meter resolution (Lavalle et al., 2011). A complete account of the original model is provided by Hilferink and Rietveld (1999), whereas recent applications are documented in a book by Koomen and Borsboom-van Beurden (2011). For an extensive overview of all publications related to Land Use Scanner, visit https://spinlab.vu.nl.
Land Use Scanner is a GIS-based model that simulates future land use. Unlike many other land-use models, its objective is not to forecast the dimension of land-use change but to integrate and allocate future land-use claims from different sector-specific models. The model offers an integrated view of all types of land use. It deals with urban, natural and agricultural functions, distinguishing ten or more different land-use categories. The model is grid-based, and the Student Edition comes in two versions: the version recommended for slower computers uses 351,000 cells of 500 by 500 meters to cover the Netherlands. The second version of the model covers the country at a 100 by 100-meter resolution (8,775,000 cells). Because this version has much more spatial detail, calculation times are longer. However, most contemporary computers will calculate the result in seconds or a minute or two.
Example composition of a 500m cell
In the provided model version, each cell describes the relative proportion of all land-use types present at a location. So, a cell can contain more than one land use type, presenting a highly disaggregated description of the whole country. For example, an individual cell with a 500m resolution can have the composition on the right.
To provide a simple, single visualisation of this disaggregated description of land use, maps of predominant land use are created. However, remember that this representation may be a bit misleading. In the example cell introduced on the right, residential land use is dominant, with only 11 hectares (less than half) of the total surface area of the cell (25 hectares). With nine land-use classes, the predominant land use may cover as little as 3 hectares in a cell.
In addition to land-use data, the model contains many different spatial data sets that are organised in three main folders:
- Current situation: a collection of spatial data sets that describe the current situation;
- Scenario components: the definition of the future scenarios that will be simulated;
- Simulations: the land-use maps and other outcomes that result from simulation.
The following sections describe the content and relevance of these elements.
In Land Use Scanner, the current situation refers to current land use, and other data sets describe the current situation of essential drivers of land-use change. These data sources can be used to define local suitability. They are grouped into the following subfolders:
- Current land use describing nine aggregate land-use classes for 2017;
- Policy maps showing operative (government) policies based on restrictions (yes or no);
- Thematic maps showing location characteristics of the grid cells in terms of, for example, the attractiveness of the urban environment, physical characteristics (e.g. landscape quality or altitude), accessibility or safety issues;
- Distance decay maps that describe the presence of specific land-use types and two nature policy areas within different distance ranges.
The model scenarios of possible future developments can be constructed using the following components:
- local suitability
- regional demand
- exogenous developments
These aspects are further introduced below.
Suitability is a crucial component in the allocation of future land use. This suitability of a location (grid cell) can be interpreted as the net benefits that a land-use function derives from that specific location and are expressed in Euros per square meter. A high suitability for a specific land-use type in a particular area leads to a high probability that this land-use type gets allocated there. The value of a grid cell in a suitability map can also be harmful, indicating that the cell is highly unsuitable for a particular land use. For every location, the suitability or attractiveness of the different land-use types is described based on several site-specific characteristics. The factors influencing this suitability are divided into the aforementioned four groups:
- Current land use;
- Policy maps;
- Thematic maps; and
- Distance decay maps.
Reference to current land use is added because the allocation process does not automatically retain this. In contrast to many other land use models, the Land Use Scanner starts the simulation with an empty map. To account for the fact that land use has a certain inertia (e.g. due to the cost of converting land from one use to another), we include a reference to current land use in the definition of suitable locations. The importance of each contributing factor depends on the scenario and is weighted in a script, as explained in the exercise.
Sector-specific models of specialised institutes, such as housing and employment models, provide the regional projections of land-use change used as input for the Land Use Scanner model. These are called regional demand or claim sets in the model.
The spatial developments of several specific land-use types (infrastructure, water and exterior) are not simulated through the allocation module because they are either absent or have a very confined, local appearance. These usually only constitute planned additions to the infrastructure network directly (exogenously) inserted in the simulation outcomes. The remaining land-use types are simulated within the model and are referred to as endogenous land-use types.
The demand for land is matched with the supply of suitable land, resulting in simulated future land use for each grid cell. This process is called allocation. The Land Use Scanner can use two different descriptions of land use per cell in the allocation process:
- a continuous description using heterogeneous cells to indicate the amount of land for each land-use type present at a specific location (as was explained previously).
- a discrete description using homogenous cells to indicate the single dominant land use at a particular location.
Basic layout Land Use Scanner model
Each description of land use has its allocation procedure. The continuous allocation uses a logit function to calculate the most probable land use at that location, whereas the discrete allocation uses a spatial optimisation procedure solved through linear programming.
This assignment uses the continuous allocation procedure, but the discrete allocation procedure is also included in the provided model version to offer the opportunity to compare the different approaches. The outcomes of the continuous model version can be interpreted as the expected proportions of land to be used for the various types of land use. The predominant land-use maps depicting a single dominant land use per cell from these calculation results are visually similar to the outcomes of the discrete allocation process that - by definition - results in one land use type per grid cell.
Go to next module: 6a: Getting started
- 0: Install GeoDMS GUI and setup a configuration
- 1: Learning the basic concepts of GeoDMS
- 2: Loading and storing data sources
- 3: Basic analyses with vector data (WORK IN PROGRESS)
- 4: Basic analyses with grid data (WORK IN PROGRESS)
- 5a: Working with networks over a road network
- 5b: Working with networks in a public transport setting
- 6: Allocating land-use