EN Introduction

This is an English translation of the Dutch page Introductie, which is leading. Translated on 8 November 2026 from the Dutch text committed together with this page, on top of wiki commit 2eddf6a.

History

The RuimteScanner has a development history of more than thirty years, during which the model has gone through several fundamental transformations.

Origin: SimEurope and LUMOS (1994–1999)

The model has its origin in SimEurope, a European land-use model that was developed in 1994 at the Vrije Universiteit Amsterdam in collaboration with Geodan. In 1996 the model was taken over by the RIVM (sector 5, later transformed into the Milieu- en Natuurplanbureau, MNP) as part of the LUMOS consortium (Land Use MOdelling System). Within LUMOS, the Landelijk Grondgebruiksbestand Nederland (LGN) was combined with the Bestand Bodemgebruik (BBG) into an integrated land-use map.

In 1999 Hilferink and Rietveld (1999) published the formal basis of the model, in which the spatial resolution was raised from 500 to 100 metres. This publication described the logit approach to spatial allocation, based on discrete choice theory from transport modelling.

From MNP to PBL: maturation and audit (2000–2010)

The model was widely used at the Milieu- en Natuurplanbureau (later PBL Planbureau voor de Leefomgeving) for national scenario studies and spatial explorations. Two important developments took place in this period:

The first was the switch from probabilistic to discrete allocation: instead of probability distributions, each cell was allocated to exactly one type of land use. This made the results easier to interpret and directly comparable with observed land use. The discrete allocation required a new optimisation algorithm that maximises total suitability (Koomen et al., 2011).

The second was an international scientific audit (Timmermans et al., 2007), which led to a strengthening of the behavioural basis of the model. The recommendation was to link the model logic more closely to economic and policy logic.

Economic basis and broadening (2011–2020)

In response to the audit, an economic basis was introduced in the determination of suitability (Koomen et al., 2015). This allowed bid prices for land to serve as the basis for local suitability, and financial policy instruments (taxes, subsidies) to be included. This approach has been applied in studies of biomass crops (Kuhlman et al., 2013), agricultural change (Diogo et al., 2015; 2017) and peat meadow areas (Van den Born et al., 2016).

At the same time the model was applied regionally (Koomen & Jacobs, 2011) and used for a wide range of scientific studies, from the reconstruction of Roman land use (de Kleijn et al., 2018 ) to urban heat islands (Koomen & Diogo, 2017). The spatial allocation principles were also adopted in other models, including the global 2UP model (Koomen et al., 2023) and the LUISA platform of the Joint Research Centre of the European Commission (Lavalle et al., 2020).

RuimteScanner 2.0 and RSopen (2021–present)

The most recent transformation concerns the development of RuimteScanner 2.0, in which several innovations have been combined:

• Three-layer model: the one-dimensional land-use layer has been split into three layers: actors (employees), objects (dwellings, commercial buildings) and land use. This makes it possible to evaluate inner-city transformations on the basis of costs and benefits.
• Hybrid allocation: the allocation takes place sequentially (sector by sector) instead of integrally, with competition within sectors through iterations.
• Design component: alongside the empirical determination of suitability, a design component has been introduced (the rung concept), with which policy preferences can be explored explicitly.
• Open source: the model has been made available as RSopen via GitHub, so that external parties can inspect, use and adapt the model.

RuimteScanner 2.0 was first applied in 2023, in the Planmonitor NOVI 2023 (Kuiper, Rijken et al., 2023) and the Ruimtelijke Verkenning 2023 (Kuiper, Hamers et al., 2023). The system description is set out in Koomen, Rijken and Claassens (2024).

Purpose and positioning

The RuimteScanner model has been applied for more than 25 years to simulate and evaluate changes in land use. Recent applications range from depicting future scenarios (Kuiper, Hamers, et al., 2023; Wolters et al., 2018) and extrapolating trend developments (Koomen et al., 2018; Kuiper, Rijken, et al., 2023) to the prior (ex ante) evaluation of specific policy measures (for example in water management in peat meadow areas, see: Claassens et al., 2020; Van den Born et al., 2016). Besides these policy-related applications, the model is also used in a wide range of scientific studies that focus, for example, on the reconstruction of land use in Roman times (de Kleijn et al., 2018) or on the possible consequences of climate change for future urban heat islands (Koomen & Diogo, 2017) and agricultural land use (Diogo et al., 2017).

The large number of applications shows that the model is flexible in the type of application (trend extrapolation, scenario exploration), the themes studied and the related land-use types that are simulated, the time horizon and the study area. This flexibility is related to the relative simplicity and the open structure of the model. Basically it has only two types of input: a regional demand for space for different types of use (‘sectors’) and a suitability for each sector that is defined locally (per grid cell) (Figure 1).

The demand for space often consists of projections of regional developments within sectors such as housing, employment, recreation, agriculture and nature, some 30 years ahead (but shorter or longer time horizons have also been worked out). These projections usually come from regional models, but they can also be based on expert judgement or policy targets. Local suitability is determined on the basis of statistical analysis, expert judgement or combinations of these. Important elements here are restrictions under environmental and planning law and location factors that indicate which locations are attractive for specific land-use types.

The core of the model consists of the allocation of the regional demand for space of the various sectors to specific locations, on the basis of their suitability. Depending on adjustable preferences, the demand for space of these various sectors is allocated to locations simultaneously (in a mutual trade-off) or successively (sector by sector). Both approaches produce an integrated picture of possible or desirable spatial developments.

Figure 1 General set-up of the original RuimteScanner model (version 1.0).

In the original model, land use was allocated with a so-called ‘logit approach’ (Hilferink & Rietveld, 1999). This approach to describing choice behaviour has its origin in discrete choice theory (Rietveld & Koomen, 2004), which is also widely used in transport modelling (Zondag & Geurs, 2011). Where transport models focus on the choice of a route or mode of transport, the RuimteScanner models which type of land use is chosen at a specific location. In both cases a choice is made from a limited number of alternative uses of available options, and these alternatives are weighed against each other. This theoretical basis of the model makes it possible to simulate future developments on the basis of observed (changes in) land use. More specifically, the local suitability for different forms of land use can be based on statistical analyses of current land-use patterns or recent changes in them (through logistic regression analysis, see for example: Loonen & Koomen, 2009). This statistical approach links directly to the ‘logit approach’ (also called probabilistic allocation) with which the original RuimteScanner model allocates land use (for more explanation, see: Koomen et al., 2015).

Since the first set-up (Hilferink & Rietveld, 1999; Schotten et al., 1997), the model has been adapted many times for the applications for which it was used. An important and fairly fundamental adaptation was the introduction of a discrete representation of land use, in which each cell was allocated to only one type of use. This makes the model results easier to interpret, partly because of the new possibility of comparing simulated land-use maps one to one with observed land use in the base year. This change required an adapted allocation procedure for allocating the demand for space of land-use types to suitable locations, with a spatial optimisation algorithm that maximises the total suitability of the allocated land use (for more details, see: Koomen et al., 2011). Following an international scientific audit of the model (Timmermans et al., 2007), efforts were then made to improve the behavioural orientation of the model, through a stronger link between the model logic and the economic and policy logic that drive land-use changes in practice (Borsboom-van Beurden & Zondag, 2011). This allows the simulated patterns to be better substantiated, various behavioural and policy choices to be explored, feedback effects to be modelled more explicitly, and specific effects (for example linked to population densities or movements of people) to be determined better.

The first model improvement in this context was the introduction of an economic basis for determining the suitability of locations for different types of land use (Koomen et al., 2015). This has been used, for example, for studies of the introduction of new biomass crops (Kuhlman et al., 2013) or other changes within Dutch agriculture (Diogo et al., 2015; Diogo et al., 2017). In these studies, the choice of developments at a location is based on a weighing of local costs and benefits. The economic basis provides starting points for model input that is more recognisable, more comparable and easier to attribute to specific actors. This can be done, for example, by using bid prices for land as the starting point for local suitability. This monetary basis can also feed the model output (for example as the estimated value of land after development, or the expected value of agricultural production). It also creates possibilities to include specific financial policy measures (taxes, subsidies) in the simulation and so internalise them in the outcomes (as was done for the study of the future of the peat meadow area, see: Van den Born et al., 2016). This economic logic has also been applied in the simulation of urban developments. It has been used, for example, to describe the financial feasibility of urban transformation processes, and with it changes in inner-city land use (Claassens et al., 2019; Rijken et al., 2017).

The most recent model improvement concerns the splitting of the one-dimensional ‘land-use layer’ of the original model into three layers: actors (for example employees), objects (dwellings, commercial buildings) and land use. This makes the model more realistic and recognisable, and makes it possible, for example, to evaluate and simulate inner-city transformations, taking into account revenues and costs throughout the entire chain: from property revenues and acquisition costs (statistically estimated on the basis of hedonic price analyses) to construction costs, demolition costs and land production costs (based on key figures from, for example, bouw-kompas.nl). To this end, the allocation can be expressed in terms of decisions of virtual property developers, based on weighing the local costs and benefits of specific residential environments (called development packages in the model, which describe, among other things, the type of dwellings and the densities). Finally, the model has been extended with a number of design options, for more exploratory use. This report describes the new model version in which the various innovations have been combined: RuimteScanner version 2.0. This model version was applied in 2023 in the Planmonitor NOVI 2023 (Kuiper, Rijken, et al., 2023) and the Ruimtelijke Verkenning 2023 (Kuiper, Hamers, et al., 2023).