GEOG 6160 Spatial Modeling

HALLIE JOHNSON & SUSAN PETHERAM

Modeling Urban Sprawl along the Wasatch Front

Introduction

The purpose of the model is to evaluate the influence of certain factors on the pattern of urbanization along the Wasatch Front, focusing on the two most populous counties in Utah: Salt Lake County and Utah County.   

Urban development is influenced by several factors, including proximity to transportation, the presence of existing infrastructure (roads, utilities), the density of the surrounding population, and land uses. Additional factors that impact the spread of urbanization include the type of transportation infrastructure (major roads vs. local roads), slope, cost of living, household size, and existing urbanized areas. The intent is to determine which factors have influenced the growth patterns of the populations of Utah and Salt Lake Counties along the Wasatch Front.

The Wasatch Front provides a unique study area for urban growth, as it exhibits several geographic constraints that function as physical urban growth boundaries. These include mountain ranges to the east and west (the Wasatch Mountains and Oquirrh Mountains) and two large lakes to the north and south (the Great Salt Lake and Utah Lake). Over the past 20 years, increasing attention has been given to regional planning efforts that are focused on supporting a less sprawling, more sustainable pattern of development focused on the strategy of developing more densely populated centers across the region as illustrated in Figure 1 (Envision Utah et al. 2010). Despite these efforts, road-building efforts continue to link dispersed populations to the urban areas, local planning still favors low-density development patterns, and efforts to densify centers are tempered by decreasing affordability in those areas. Our two primary purposes are to first, include the influence of housing affordability in an urban growth simulation model, and second, evaluate the influence of major road infrastructure by comparing model results with road influence toggled on/off. We will then compare our model results to the development pattern envisioned by the Wasatch Choice for 2040, see Figure 1 (Envision Utah et al. 2010).

The foundation for this model is the SLEUTH model of urban development, developed by Dr. Keith C. Clarke at UC-Santa Barbara, which integrates both an urban growth model and a land use/land cover model. The SLEUTH model is a cellular automata (CA) model, with the urban areas behaving like a living organism. SLEUTH has been used successfully to assess historical urbanization and predict future growth patterns over the past 20 years, in at least 66 different cities and regions. SLEUTH uses inputs for (S)lope, (L)and Cover, (E)xcluded areas, (U)rban, (T)ransportation, and (H)illshade to establish initial conditions and then enters into a set of growth cycles governed by rule sets (Chaudhuri and Clarke 2013).

Data

Input data

The four environments illustrated in Figure 2 are manipulated renderings of data acquired for the model. Road and elevation data were obtained from Utah’s Automated Geographic Reference Center (AGRC). Considering the large study area, it was decided the entire layer would not be utilized. The road layer was queried to select only those roads with state or federal designation, as these would have a greater impact on urban expansion than local roads. A 100-foot buffer was then applied to the major roads to create a larger area of influence. This vector layer was then converted to a raster and clipped to the study area. The elevation data was taken from DEMs with 30-meter resolution. Four tiles were merged to create a single elevation model. Once these had been merged the slope was calculated across the area.   

The housing data was derived from the Center for Neighborhood Technology’s (CNT) Housing + Transportation index. CNT compiled multiple variables to create an index value, indicating the percentage of Area Median Income (AMI) spent on housing and transportation at the census block group scale. The data is converted to a single digit number based on the overall percentage of annual income spent on housing and transportation, numbers range from 1 to 10. A raster was created from the polygons to reflect the value of the range of values as shown in Figure 2, Top Right.

Demographic data comes from two sources, the US Census Bureau and from the SILVIS lab for Spatial Analysis for Conservation and Sustainability from University of Wisconsin-Madison. Population density data from 1990, 2000 and 2010 was derived from SILVIS. The block groups and urban designations were provided by the US Census Bureau. After each decennial census, the Census Bureau delineates urban areas by applying specific criteria to decennial census data and other data. Urban areas are intended to represent densely developed areas, and encompass residential, commercial, and other non-residential urban land uses (Ratcliffe 2016; U.S. Census Bureau, n.d.).

Entities, State variables, and Scales

This model primarily focuses on the spread of urbanization in the Wasatch Front. The variables are meant to ascertain which factors are key in encouraging urbanization. The primary variables being measured are the influence of existing urbanization, roads, slope and housing cost to new urban growth.

Slope influences urbanization through several dependent variables and coefficients. Critical slope is a threshold value that is involved in assessing the suitability of rural blocks for urbanization. This value ties directly into the suitable variable which based on the slope represents a 0 or 1 if the block is suitable for urbanization. The patches are also assigned a value of urban or not urban, represented by a 1 or 0 respectively. This is achieved through the initialization or through the model’s process. At the initial set up, urban values are derived from 1990 demographic data. Within the model, the variable of new urbanized is the value assigned to cells that change from rural to urban. The housing cost variable is assigned to each patch at initialization. This housing value relates to the housing based suitability variable which is a factor in the spontaneous growth simulation.

The fourth patch value is Boolean expression for road on each block, represented as a 0 or 1. Within the road influenced growth simulation there are several integral variables. A maximum search radius defines the neighborhood the model will search for urbanized blocks around the road network. In the interface the dispersion coefficient is set from 1 to 100 and influences the road based urbanization spread. Also within the road influenced growth simulation is a breed coefficient which affects the seeking of road adjacent or occupied cells. The road influenced growth is controlled by a road influence variable which can be set to true or false. When true the road influenced growth will occur, when false it will not impact the model. For the edge spreading simulation of urban growth, the spread coefficient determines the weight and speed of this method of development. The final variable in the model is a max coefficient which generates random numbers throughout the model. Figure 3 shows the interface of the model within Netlogo.

The geographic scope of this model is centered in Salt Lake and Utah County and includes the areas immediately east of the Wasatch mountains. Patches that make up the model environment represent the demographic data of census block groups. The roads are a generalized representation of major highways that cross through the study area. Each time the model ticks, it represents a year passing. Below in Figure 4 the work area is illustrated with city names for reference.

Methods

Process Overview and Scheduling

Each tick the blocks run through several processes. There are two processes to determine the suitability of each block for development. First of the processes is a slope suitability analysis. This is based on the relative proportion of the block’s slope and the variable critical slope. If the slope is less than the critical slope then the patch is considered to be suitable for development. The next suitability analysis is the housing cost where the block is judged to have a favorable housing cost. In order to warrant spontaneous development, outside of the confines the other urbanized growth, the value has to represent less than 30% of the income of the residents. These are weighed against the status of the block (whether it is urban or not) and the "if this" criteria is met a block has the potential to become urban (See Figure 5). The selection among these blocks is random.

 The next operation is checking the status of the patch for roads. If the patch contains part a road the patch is given a value of 1, if it is not then the patch is given a value of 0. Patches with a value of 1 are included in the road based urban spread. At each iteration patches with a road can influence adjacent patches to become urban.

The final analysis completed in each iteration is analyzing each cell to determine if it is adjacent to an urban cell and suitable for development (based on slope) then the cell has the potential to be chosen for urbanization. With each iteration the urban boundary continues to influence adjacent cells to urban cells that do not have urban neighbors or are considered suitable for development are excluded from this process.

 Design Concepts

The basic principle of this model is primarily based on the SLEUTH model of urbanization. The concept was to create a model that accounts for factors such as infrastructure, population, terrain, housing affordability and quality of life in the pattern of urbanization. The resulting landscape of urbanized development can then be compared to existing and planned development patterns for the Wasatch Front to evaluate differences and similarities. 

The objective of this model is to illustrate the spread of urbanization along the Wasatch Front. In the model interface we track the percentage of newly urbanized blocks. The model environment also provides a realistic view of the urbanized blocks and their relative location along the Wasatch Front.  

Patches within the environment have values assigned for the following categories; slope, housing costs, major highways, and urban. The environment is set up to sense adjacent cells for their urban or rural status, their slope and the value of the housing cost. Patches with value of road also sense their adjacent cells for urban or rural status and slope. 

Stochasticity is accounted for in the spontaneous growth simulation in this model. The random growth of urban blocks that do not follow the usual prescribed pattern are dependent on the slope and housing cost suitability. This particular simulation was included to account for random outliers of growth already observed in the Salt Lake Valley, such as Eagle Mountain, that are neither related to infrastructure or current urbanization (See Figure 6). There is not a particular agent or individual represented in this model, however the blocks do interact. The main driving force of the model is the interaction of non-urban and urban blocks with their adjacent neighbors, determining if they will become urban.

It is apparent from the initialization that the urban blocks do aggregate. According to the Census Bureau when a block is categorized as an urbanized area, the adjacent blocks are considered in that classification. The minimal area classified is about 1 square mile, and it requires several of these blocks to constitute that area. In the documentation the Census Bureau explains that this creates a more realistic representation of urbanized areas, rather than solely relying on population density of an individual block. The nature of the growth patterns within the model perpetuates this clustering. First, in the instance of edge growth urbanization there will be greater numbers of urban blocks in areas that are already urbanized. This comes into effect with the road influenced growth and spontaneous growth, as these simulations create new initial urban areas, the edge growth will then create larger urbanized aggregates.

Results

A sequence of experiments was run on the model to simulate the weighted influence of each of the four types of urbanization; spontaneous, diffusion, sprawl and road influenced spread.  From range of 1 to 100 each variable was tested at an interval of 10. When testing a single variable, others were maintained at a value of 10 to minimize their influence. Each tick represents a year of time, the model was allowed to run in each simulation for 50 intervals.  Much research has been conducted on the future of development in the Wasatch front, and a particular model simulating the growth by the year of 2040 was the inspiration for this cutoff.

Figure 7 is a sequence of tests with each coefficient tested at 40.  The first image is a baselined experiment with all the coefficients set equally at 20, the other 5 images are the result of the weighted influence of the different variables.  Several things become apparent upon  visual inspection of the results.  First is the importance of slope in an regional setting such as the Wasatch Front.  In other urban settings where spread can distribute more evenly, the Salt Lake Valley is hemmed in on both sides by mountain ranges.  As a result of this growth limitation many of the models behave in similar ways and consistently by 2040 (in this simulation) the entire west valley area is urbanized, and the urbanized areas south of Thanksgiving point are completely connected.  The other most evident result is the role of urban sprawl, what Clarke and Gaydos (1998) refer to as organic growth, has the potential to increase urbanized areas significantly more than the other variables. By the 50th iteration, with a coefficient value of 40, the spread scenario had tripled the number of urbanized blocks.  The other models either doubled or nearly doubled the number of urbanized blocks. Running the model with spread coefficient showed continual growth, with the coefficient set to 100 the model achieved almost four times the growth of urbanized cells.   

Conclusions

It appears from the various model runs that the most impactful elements in this scenario are the existing slope and urban development.  The exponential growth of the urbanized areas relative to the existing urban areas create the largest amount of growth in the modeled 50 year scenarios.  With unimpeded growth, much of the land between the Wasatch and the Oquirrh mountains will the developed in the next 50 years, according to this model. 

The biggest restrictive and impactful variable is the slope.  As shown in Figure 3, urbanization does not extend into the areas where the slope exceeds that of the valley floor.  Also in consideration is the repetitive new areas of spontaneous growth. In this model we included in the spontaneous growth model the housing cost as a third variable, and it is evident this has impacted the areas or urbanization outside of the core.  The same cluster of areas, with low housing cost and low slope, continue to develop into new urbanized areas, primarily small areas close to Park city and along the I-15 corridor. In comparison with the Wasatch choice from 2040, this simulation model lines up quite well when all variables are at an equal threshold.

For more information, read the full report here

MGGIS Program Skills

• GIS Analysis

• Spatial Data and Algorithms

• GIS Workflow

• Model Building

• Spatial Analysis

• Communication Skills

• Basic Programming or Scripting