A DETAILED DESCRIPTION OF THE UPLAN MODEL

  Robert A. Johnston and David R. Shabazian

Dept. of Environmental Science and Policy

University of California, Davis, CA 96161

rajohnston@ucdavis.edu

July, 2001

TABLE OF CONTENTS
   
(Headings are links)

A. UPLAN
     1. STRICT COMPLIANCE
     2. LIMITED COMPLIANCE
     3. INDUSTRIAL COMPLIANCE ONLY
     4. NO COMPLIANCE
B. UPLAN DATA
     1. ATTRACTION GRIDS
     2. EXCLUSION GRIDS
          a. slope discouragement

             UPLAN is a GIS-based expert decision rule urban growth model that was conceived by Robert Johnston at UC Davis and built by David Shabazian.  Its original purpose was to disaggregate the zonal output of urban models such as TRANUS and MEPLAN to specific locations within each analysis district.  However, UPLAN was eventually fitted with a demographic/land consumption module that generates demand for land endogenously from user-specified parameters.  Its structure is most similar to that of CUF, but employs user-specified attraction weights rather than potential profit as the main factor driving the allocation routine.  UPLAN also differs from CUF in data structure: it uses 50-m2 raster data because it requires less disk space, and model run time can be substantially reduced as cell size increases (data sets also become smaller).  There is a tradeoff however, because larger cells mean less detailed allocation results.  This issue of resolution adds another dynamic to the model that can be adjusted to fit the needs of the user.  The model runs on a PC in ArcView, a common desktop GIS application that is used by most planning and natural resource agencies.  The data required to run the model are generally available from regional or state resource and planning agencies.  Because UPLAN uses complementary software and data, it can easily become a “spatial process model” for agencies that already have “spatial data models” in place.

UPLAN uses 1990 as its base year and allocates the increment of additional land consumed in future years.  This increment of urban growth is in discrete land use categories: Industrial, High Density Commercial, Low Density Commercial, High Density Residential, Medium Density Residential, and Low Density Residential.   One method of estimating the incremental growth uses the land consumption projections from TRANUS or MEPLAN.  These integrated models provide estimates of land consumed in 53 zones that are then fed into UPLAN to be allocated to specific locations within each zone.

County or regional land consumption can also be calculated endogenously on user-specified assumptions thereby allowing one to run the model independently.  UPLAN is built with a module that allows the user to input demographic and land use factors that are converted to acres of land consumed in each activity class.  The conversion starts with population projections for counties or the entire region.  The user is then prompted to specify the demographic and land use characteristics he or she would like to test.  To determine acres needed for future housing, the user specifies persons per household, percent of households in each density class, and average parcel size for each density class.  A similar conversion is used to determine acres of land consumed for industry and commerce and uses workers per household, percent of workers in each employment class, and average area per worker (in acres).  Both methods of deriving future land consumption produce a table from which the model operates its land allocation routine.

It is assumed that development occurs in areas that are attractive due to their proximity to existing urban areas and transportation facilities, such as highway ramps or light rail stations.  It is also assumed that the closer a vacant property is to an attraction, the more likely it will be developed in the future.  For example, a property that is a quarter mile away from existing development (or any attraction for that matter) is more desirable then one that is a mile away from the same location.  Following this crude bid-rent assumption, each development attraction (described below) is surrounded by user-specified buffers.  The user can designate the number and size of the buffers and assign an attractiveness weight to each buffer.  Buffer specifications are applied to each of the attraction grids and then the grids are overlaid and added together to make a composite Attraction Grid.  The composite Attraction Grid is a single grid of the sum of the weights specified for each individual attraction grid.  Each cell in this grid has a value resulting from the summation.  Grid cells with the highest value are considered the most attractive areas for development while cells with a value of zero are considered the least attractive areas.

Naturally, there are areas where development cannot occur, which are called exclusions.  Exclusions include features such as lakes and rivers, public open space, existing built-out urban areas, and other such areas where development cannot occur.  There are also features that the user can choose to exclude, such as sensitive habitats, 100-year floodplains, and farmland.  Once the user decides which features are to be excluded, the model adds together the various exclusion grids to generate a “Mask.”  Like the composite Attraction Grid, the Mask Grid is a sum of the individual exclusion grids.  In this case, however, grid cell values are not important per se; rather, simply that a cell has a value makes it part of the Mask.

Once the Attraction Grid and the Mask Grid are generated, the model overlays the two grids and attraction cells that fall within the mask are converted to “no data” cells, thereby removing them from possible development allocations.  This process creates the Suitability Grid, which becomes the template for the allocation of projected land consumed in the future.  The Suitability Grid is overlaid with a grid of the 20-year General Plan land use map for the region enabling the model to further isolate areas which are suitable for each of the various land use categories that are allocated.  The model is then ready to allocate projected acres of land consumed in the future.  There are four ways in which this allocation occurs
| TOP |

1.       Strict Compliance – Each activity can only be allocated to its corresponding designations in the General Plan as follows:
Activity Type                                           General Plan Category
Industry                                               Industry  

High-Density Commercial                   Commercial

High-Density Residential                     Multi-Family Residential

Low-Density Commercial                    Commercial

Medium-Density Residential               Single Family Res., Urban Reserve

Low-Density Residential                      Single Family Res., Urban Reserve and Rural Residential

                        This is called “two-way zoning,” legally.  
| TOP |

2.       Limited Compliance – Each activity is allocated to its corresponding General Plan designation and to the designations available to the previously allocated land use.  Activities are listed in order of allocation.

Activity Type                                            General Plan Category 
Industry                                                 Industry
High-Density Commercial                      Industry and Commercial
High-Density Residential                       Ind., Comm., and MF Residential
Low-Density Commercial                      Industry and Commercial
Medium-Density Residential                 Ind., Comm., MF Res., SF Res., UR
Low-Density Residential                       Ind., Comm., MF Res., SF Res., UR, and Rural Res.

              This is called “one-way zoning,” legally

3.       Industrial Compliance Only - Industry must go to industrially designated areas, and all other land uses can go anywhere.  

4.       No Compliance – All land uses are allowed to go into any land use designation.

The model allocates future development starting with the highest valued cells.  As the higher valued cells are consumed, the model looks for incrementally lower valued cells until all acres of projected land consumption are allocated.  The model does this for each of the discrete land use categories and the user can decide the order in which land uses are allocated and to which general plan categories they can be allocated.  By default, the model starts with industry, then proceeds to high-density commercial, high-density residential, low-density commercial, medium-density residential, and finally low-density residential.  This order is chosen to represent the way in which the land market typically operates - higher valued land uses are more competitive in acquiring the most desired properties thereby outbidding the less valuable uses.  The allocation routine converts future acres consumed to the number of cells needed.  It then determines how many cells are available in the highest valued category and if this is less than what is needed, simply converts all those cells to the designation of the land use it is allocating at that time.  It then subtracts the number of cells it just allocated and moves on to the next highest cell value and again determines how many cells are available.  When the model reaches a point where the cells available are greater than needed, the model completes its allocation of that particular land use by randomly allocating the remaining development to cells within the current value class. 

This allocation method does not apply to low-density residential, which is randomly allocated throughout rural areas to represent the prevalent noncontiguous patterns of exurban rural residential development, such as hobby farms.  Because the allocation is random, low-density residential does not use the Attraction Grid to find the best locations; however, the Mask Grid does apply.  The low-density residential allocation routine starts by making a generic grid of random values.  It then makes a list of the values and allocates, in descending order, to the random cells until all acres of low-density residential land are used.  While all other land uses are allocated in 50 meter grid cells (≈ ½ acre), 200 meter grid cells are used for low-density residential to represent average parcel size (≈ 10 acres).

After a land use is allocated for all zones, counties, or for the region, the model makes a new grid of that allocation.  This grid is saved in the working directory, but also added to the Mask Grid so that the next land use being allocated does not overlap previous allocations.  Once the model has allocated all the land uses, it merges all of the allocation grids it has created to make the final Allocation Grid: a grid that has the allocation of all land use types in all zones, counties, or the region as projected out to the year tested.  
| TOP |

 A. UPLAN Data

Most of the data listed below would typically be used in any region where UPLAN is applied.  For example, any application of the model would require basic attractions, such as cities and sphere of influence or freeway ramps, or basic masks, such as waterbodies or existing urban development.  There are also attraction and exclusion characteristics that are specific to a region.  For example, in the Sacramento region there is an international airport that is used as an attraction, and there are unique geological features called vernal pools that can be used as exclusions. The data used for the application of UPLAN to the Sacramento region are described below. 
| TOP |

1. Attraction Grids

As mentioned above, there are a number of attraction grids that are used to create the Composite Attraction Grid that identifies where development will most likely occur in the future.  The user can specify a number of buffers, the size of each buffer and the weight or attractiveness of each buffer.  Generally, higher weights are assigned to the buffers that are closer to the source.  This section provides a brief description of each of these grids.

1.       Freeway Ramps – This grid represents the location of on- and off-ramps on major freeways, rather than the entire facility.  Since major freeways have limited access, ramps are the most desired locations for development along these facilities.  This grid allows one to test the effect of changes in the facility by adjusting the weighting of the ramp buffers.  For example, if the facility were improved with carpool lanes, one would increase the weights on ramps to reflect the additional capacity, which would in turn make property around ramps more valuable/attractive.

2.       Minor Highways – These highways are generally two lane roads that are intersected by county roads and other major arterials.  Therefore, they are considered to have enough access to use the entire facility as an attraction.  As with other facility improvements, if minor highways are changed, the user can adjust the weights and buffers to reflect to these changes.

3.       Major Arterials – These are also two lane roads, but generally have lower speed limits and are much shorter in length.  These roads have even more access points than minor highways; therefore the entire road is used as an attraction.  Buffers and weights can be changed at the user’s discretion as with other attraction grids.

4.       Minor Arterials – This is the least important roadway type used in the model.  Speeds are generally slower than major arterials and there are numerous intersections with other arterials and local roads (which are not used in the model).

Because the roadway classes are in separate grids and then added together, intersections between roads receive a higher weight because of the occurrence of two or more roads.  This provides a fairly realistic representation of the real world in terms of where developers look for property to develop. By default, roadways with more volume (i.e., Highways) have higher weights than arterials, however, the user may choose to specify different buffers and weights.  All road data were derived from U.S. Geological Survey (USGS) digital line graphs (DLG) for transportation.

5.       Cities and their Spheres of Influence – This grid is a representation of the existing boundary that defines a city.  The boundaries used in the model are not city limits, but rather the city and the area around it where development will most likely occur in the future – the Sphere of Influence.  The data were provided by SACOG (Sacramento Area Council of Governments, the region’s Metropolitan Planning Organization).

6.       Light Rail Stations – This grid represents the location of existing and or future light rail stations.  The rail line itself is not used as an attraction since development activity would most likely be concentrated around the stations.  Using this grid one is able to test scenarios with varying degrees of emphasis on Transit-Oriented Development and transportation policies favoring rail transit.  The data were generated by the author by digitizing SACOG maps of existing and proposed light rail lines.

      The following grids are used in addition to the aforementioned data, for industrial allocation only.  These grids are not used for the allocation of the other land use types because these ports are surrounded by industrially designated land.  This land use policy may not apply to all regions where UPLAN might be used and the model can be modified to reflect any such changes.

7.       Sacramento International Airport – The airport is currently in the process of expansion; therefore, the model accounts for the likely buildout of the surrounding areas by including this grid as an industrial attraction.  The data were digitized by the author using California State Automobile Association maps of the region.

8.       Port of Sacramento – The port and its surrounding area has traditionally been an industrial district.  While it is more developed than the airport, there are still areas that can accommodate future industrial development. This grid was also generated by the author using California State Automobile Association maps of the region.
| TOP |

2. Exclusion Grids

Once the attraction grids are combined to create the Composite Attraction Grid, the following exclusion grids are combined to create the Mask Grid:

1.       Regional General Plan – This grid is a composite of all the general plan land use maps for cities and counties in the study region.  SACOG constructed this grid by merging all the individual general plans and then generalizing the land use designations across the region.  The model uses this grid to find locations that are appropriate for each land use category it allocates.  This restriction depends on the type of compliance chosen by the user and converts all cells that are not “in compliance” to “no data” in order to remove them from possible allocation.  For example, if the user chooses to have strict general plan compliance, then the allocation of commercial land is restricted to commercially designated locations as determined by the regional general plan.

2.       Rivers – The model uses the “Naturally Occurring Waterways” grid as the river exclusion for this model.  This grid was generated from the EPA River Reach files by the Information Center for the Environment (ICE) at UC Davis.  The EPA data originates from the USGS hydrography DLGs.  The grid includes all major and minor waterways for the study region, but does not include ephemeral streams or man-made waterways such as channels and canals.  A user-specified distance buffers the rivers before they are added to the Mask Grid.  This precludes development from occurring too close to waterways.

3.       Lakes – This grid contains all of the water bodies in the study region.  As with rivers, all lakes and ponds are buffered by a specified distance prior to being included in the Mask Grid.  The data were generated from the USGS hydrography DLGs as modified by the U.S. Department of Fish and Game.

4.       Vernal Pools – Since many vernal pools are small and difficult to represent in a digital format, a grid of vernal pool “complexes” created by the U.S. Fish and Wildlife Service is used.  There are four types of complexes that the user can choose as exclusions: High Density, Medium Density, Low Density, and Cultivated.

5.       Floodplains – This grid contains the 1996 100-year floodplains for the study region as determined by FEMA.  This grid can be used to test how development patterns might change due to the implementation of land use policies that prohibit development in floodplains.  This grid can also be used to evaluate the potential damage and financial impacts of building in the floodplain.

6.       Slope – Slope is used to mask out areas that are too steep to develop, and as a discouragement factor for areas that remain.  Slope is derived from Digital Elevation Models (DEM) provided by the USGS.  The discouragement factor works by dividing the Attraction Grid weights by the values shown in the table below.  This is done for each land use type before the allocation begins.  The slope and mask restrictions are as follows:
| TOP |

Slope Discouragement Factor by Land Use Type:

            Ind., Comm. High,

% Slope                Comm. Low, Res. High            Res. Med.            Res. Low

0-5                                1                                          1                            1

5-10                              2                                          1                            1

10-15                            3                                          2                            2

15-25                        mask                                        2                            2

25-50                        mask                                        3                            3

50-100                      mask                                        mask                      3

100 +                        mask                                       mask                   mask

           

7.       Public Lands – This grid includes all of the areas designated as publicly owned land in the SACOG General Plan grid for the region.  The author extracted the “Public-Owned” designated grid cells from the General Plan to create this grid.  Public lands include areas such as schools, military bases, and forestlands.  Generally, these lands should be excluded from development, except where military bases are converted to private uses.

8.       Fire Severity – There are three levels of potential fire severity that can be used to exclude development: Very High, High, and Moderate.  The ratings were generated by the California Department of Forestry and are based on characteristics such as fuel loading, slope, fire weather, and other factors.  As with the Floodplain grid, the fire severity grid can be used to assess the potential loss due to fires.

9.       Existing Urban – This grid was constructed using data from the Department of Fish and Game and the Department of Forestry.  Each agency has 30 meter grid data of existing urban development that is generated from satellite imagery.  Urban designations are distinguished from water and vegetation by their spectral value.  While the interpretation of the satellite data does produce some false positive cells (an urban designation where there is no urban) and false negative cells (no urban designation where there should be), the grid allows us to test infill and redevelopment policies at a detailed scale.

10.   Open Space – There are “public open space” areas designated on SACOG’s Regional General Plan that are used as an exclusion.  This designation includes parks, greenbelts, wildlife preserves, and other areas set aside for recreation or preservation.

11.   Farmland – Currently the “agriculture” designation in the regional general plan is used as the farmland exclusion grid.  The user is allowed to exclude these agricultural lands from development if desired.  This coverage can also be used to determine how many acres of farmland are being converted for future urban uses.  A data set provided by the Department of Conservation, Farmland Mapping Project that delineates four types of prime farmland can also be used.  This grid allows the user to specify which types of farmland to exclude, and/or conduct more detailed assessment of farmland conversion for urban uses.

12.   GAP Vertebrate – The GAP data is used to determine which areas in the region have the richest concentration of species.  Currently, the model uses the top 10% richest areas for each species type: amphibians, birds, mammals, and reptiles; however, the user may change the percent richness to test various policies.  This grid can be used as an exclusion or as an overlay to determine where and to what extent urban development encroaches on sensitive habitat.  The data were originally generated by the UCSB GAP Analysis Project.
| TOP |