Environment and Planning B: Urban Analytics and City Science

In this paper is described a model of residential mobility, built to simulate individual households, their perception of and reaction to varying conditions across different scales of interaction, and their movements to occupy housing in a physical, social, and economic environment. The methodology underpinning the model is based on an automata core, which leverages the advantages it offers in terms of representing individual entities and their rule-based interactions. This methodology is extended, however, to incorporate geography-specific functionality, with advantages for the modeling of human systems. The applicability of the methodology is demonstrated through the development of a rich model of residential mobility, in which individual households interact with other households and real-estate infrastructure, dynamically in space and time, to form synthetic communities and artificial property submarkets. Use of the model for what-if experimentation is demonstrated with synthetic economic and sociodemographic simulation scenarios.

In this paper we present and test the functionality of a parcelization algorithm, implemented in our spatially explicit, agent-based land-use-change model which we call the Multi Agent-based Behavioral Economic Landscape (MABEL) model. In order to test the best possible spatial configuration of the algorithm and its efficiency compared with historically observed land-use changes, we employed a Monte Carlo simulation approach with a series of replication experiments across time, and compared observed changes between 1970 and 1990, and across two different landscapes in Michigan, USA. We compare the simulated parcel shapes with historically observed land-use changes using the landscape-ecology metric program, FRAGSTATS.

Geographic complexity—the explicit integration of complexity research with space and place-based research—faces interrelated methodological, conceptual, and policy challenges. The rubric of model evaluation is central both to understanding and to meeting these challenges. They include methodological issues such as sensitivity and complex scaling; the conceptual challenges of conflating pattern and process, and reconciling simplicity and complexity; and policy issues posed by the science–policy gap and postnormal science. The importance of these challenges and the centrality of model evaluation in meeting them are demonstrated through examples drawn from human-environment systems, with particular reference to global environmental change and land-use and land-cover change. Specific model-evaluation strategies are also offered.

This research develops an agent-based land-use model for shifting cultivation, where the spatial distribution of crop cultivation is dynamically determined by the relationship between demand and supply of crops. We apply and evaluate the model using statistical and geographic data from Luangprabang Province, Laos, where rice and other crops are cultivated in various forms, including shifting cultivation. Our model explicitly incorporates socioeconomic dimensions of shifting cultivation in which villages are assigned the role of decision makers. We evaluate the model by comparing the simulation results with the existing statistical data and remote sensing images from the 1990s. Our model provides reasonably satisfactory estimates of aggregate area and volume of each crop type at the provincial level. We also evaluate the model across differing spatial resolutions for shifting cultivation areas. We find that the model has limited explanatory power at higher spatial resolutions of 0.5 km to 2.5 km grid cells, but can account for the spatial patterns fairly well at more aggregate levels with the resolutions of 5 km to 10 km.

Agent-based models offer a promising framework for analyzing interactions between agents and a heterogeneous landscape. Researchers have identified a complex of factors that influence exurban development, including demographic shifts and location attractiveness of natural amenities as a magnet to amenity-seeking migrants. Attractiveness is often defined in terms of local or on-lot amenities, including scenic views, the availability of natural features, and low levels of noise. However, exurban-growth models have not fully incorporated a fundamental insight of this literature, that the location behavior of exurban residents is sensitive to fine-grained variations in their biophysical environment. In this study we evaluate how agents and households operate in exurban environments and respond to biophysical features. We simulate household decisionmaking in terms of preferences for features such as site accessibility, two-dimensional amenities, and three-dimensional scenic views. Our results show that, as we build two-dimensional and three-dimensional landscape layers, our model captures the characteristics of landscape change with increasing accuracy. This approach has considerable potential to improve our ability to describe development dynamics in heterogeneous land markets.

In this paper we report the results of a study on secondary school planning made within the framework of Coimbra's Educational Charter. Coimbra is a medium-sized municipality of 320 km² and 150000 inhabitants located in the center-littoral region of Portugal. The planning problem addressed in the study consisted of defining the location, type, and size of the schools that should integrate Coimbra's secondary school network in 2015, given, first, the excess of aggregate school capacity that currently characterizes the municipality, and, second, the change in school typology that needs to be implemented as a consequence of a recent reorganization of the Portuguese educational system. This problem was analyzed with a discrete facility-location model and considers decisions both of closing existing schools and of opening new schools. The model is a variant of the well-known p-median model, which aims at maximizing the accessibility of students to schools, with constraints on maximum and minimum capacity occupation that make model solutions lose the single assignment and closest assignment properties. As these are desirable properties in a public-facility planning context, they are enforced with explicit constraints. The results obtained through the model are discussed from the standpoint of the trade-off between school accessibility improvements and school network changes.

Agent-based modeling and simulation (ABMS) has been a part of geospatial sciences for over a decade. Most research activities so far have concentrated on either extending complexity theory to spatially explicit phenomena, or on designing computational models and software tools. Only a few of these activities have focused on using ABMS for spatially explicit modeling of real-world policy scenarios. In this paper we present a realistic application of ABMS to simulating alternative futures for a small community in Washington State, USA. We develop an ABMS assessment benchmark that comprehensively covers diverse aspects of a good operational agent-based model. Using an ABMS software tool-CommunityViz Policy Simulator-we generate future development scenarios in the municipality of Chelan, WA based on the County and the City Comprehensive Growth Plans. Simulation results are compared with Washington State projections for growth-management planning. The indication of the highest probability locations of urban growth in the studied community is crucial for environmental and economic planning and decisionmaking. Endangered salmon protection and recreational and retirement influxes of people from the Puget Sound metropolitan area have a direct impact on future growth of the region. The bottom-up microsimulation allows for interposition of individual decisions and actions into forecasting option generation. The ‘heterogeneity, adaptability, and tractability’ benchmark is instrumental in evaluating CommunityViz Policy Simulator and outlining possible challenges for future development of applied agent-based models.

In this paper I present an empirical analysis of spatial patterns in land markets in Mecklenburg County, North Carolina, between 2000 and 2003. It is well known that land markets reflect a variety of spatial factors that collectively influence market value, yet it is empirically difficult to sort out the relative contribution of overall and localized, the spatial and aspatial determinants of sales prices. Some of the classical assumptions about urban form that feed into hedonic analyses of land markets are explored. Then, three analyses are presented: simple visualization of single-family residential sales prices with regard to factors likely to influence land value, univariate and bivariate measures of spatial autocorrelation, and, finally, spatial econometric hedonic modeling.

GIS has been welcomed worldwide for its potential production of accessible, timely, and accurate information for decision making. In recent discussions it has been presented as a condition for governance. The implementation of GIS in large and complex institutions like ministries or planning institutes, however, is still problematic. Therefore, attention has shifted from the mere ‘technical factors' to organisational and institutional problems of GIS implementation, looking for the right ‘fit’ of GIS for the organisation in question. This paper will sketch the importance of organisational culture for understanding GIS implementation, and draw some conclusions on the implications for the use of GIS in promoting good governance. The paper discusses a Costa Rican case study on the implementation of GIS for forest monitoring. This case shows the strengths and weaknesses of different organisational cultures in adopting GIS, and illustrates that the (inter)organisational complexity of monitoring forest resources requires more attention to the potential variety in data demands from different organisations.

Habitat fragmentation and habitat loss are causing many species to become locally extinct. The need to restore or conserve biodiversity has initiated plans in Europe—and, more specifically, in the Netherlands—to create networks of nature areas for the improvement of spatial coherence. In this paper spatial coherence is measured by the total boundary length of all nature areas. Spatial optimization is used to optimize the spatial coherence, in order to evaluate the so-called Dutch 2003 plans. The goal of the optimization is therefore to minimize the total sum of boundary lengths for all nature areas, while increasing the total surface area. Potential configurations have been calculated and compared with the 2003 plans, with respect to boundary length and the size of nature areas. The calculations have been performed by allocating increasing amounts of new nature to test if smaller numbers of hectares could yield a comparable perimeter decline. Optimization has been carried out separately (per province) and collectively (for all the Dutch provinces together). So-called genetic algorithms were used to tackle this strongly nonlinear problem. An interesting feature of the method used is its ability to optimize more than half a million variables—that is, potential nature cells. Allocating only half the area allocated in the 2003 plans resulted in a boundary-length decline of 30% compared with the 2003 plans, and yielded many more coherent areas than are found in the 2003 plans. Compared with the optimization per province, the collective optimization resulted nationwide in larger nature areas, across provincial borders. Considering the original aims, we can conclude that carrying out the 2003 plans will not improve the overall spatial coherence in terms either of the boundary length or of the number of areas and total surface area in the largest surface class. An additional advantage of the heuristic method discussed is that slightly different configurations can be found with almost the same characteristics in a relative short time. This will allow the provinces to choose the most suitable configuration.