Wikis.uit.tufts.edu



Jay Monty 12/17/10Intro to GISFinal ProjectPutting Back the Tracks: A suitability analysis for commuter rail expansion north of Boston.Project DescriptionThe purpose of this project was to see whether GIS tools could use census and infrastructure data to identify abandoned rail corridors which may be suitable for new commuter rail service. The basic methodology was to use a raster calculation that would assign values to land parcels in a manner which would highlight corridors that may be suitable for new rail service. Suitability for this exercise was based primarily on a land parcel’s housing unit density, proximity to an abandoned rail corridor, and distance from other public transportation facilities (other commuter rail and rapid transit stops). I began this exercise with the belief that there are in fact many abandoned rail corridors that share land use characteristics of existing commuter rail corridors and would therefore be good candidates for re-instated train service. The intent of the GIS exercise was to prove or disprove these beliefs.The primary benchmark which is used to determine a neighborhood or town’s ability to support transit is its density, either in persons per square mile or housing units per acre. For this project, I chose to use housing units per residential acre. This method isolated the measurements of housing density strictly to residential areas, discounting areas with other land uses. This methodology was more critical in lower density areas with large census block groups which may have had dense residential development on one end near a downtown, but open space or industrial uses in outlying areas which would have given the block group a lower overall housing density. The downside of this approach is that it can overstate the total population of a census block group. If for example, a census block contains only 10% residential area, its density per residential acre could be very high, but overall density acre could be low. The other primary criteria used for the analysis was geographic location in relation to other transportation infrastructure including abandoned rail lines and existing rail lines and stations. In order to indentify suitable corridors, land parcels adjacent to abandoned corridors were ranked very high, so long as they were not within the catchment areas of other commuter rail or transit stations. As will be discussed later, the wide variations in geographic locations and densities created conflicts in the ranking system.Data SourcesAll of the data layers were obtained from MassGIS. Census data is based on the year 2000 US Census while data layers regarding physical infrastructure and land use was compiled and published solely by MassGIS and updated between 2005 and 2007. Metadata for each of these data layers is available on the MassGIS website.Data LayerFile NameAccuracy (feet)DateCensus Block GroupsCENSUS2000BLOCKGROUPS_POLY.shp2000Total Housing UnitsCEN2K_BG_HOUSING_UNITS.dbf2000Major RoadsEOTMAJROADS_RTE_MAJOR.shp502007MBTA StationsMBTA_NODE.shp502006Railroad LinesTRAINS_RTE_TRAIN.shp502008MBTA Transit LinesTRANSLINES_ARC.shp502007Land UseLANDUSE2005_POLY.shp2005In addition to the data layers themselves, this project also took advantage of several layer files which isolated various data subsets. Layer FileFile NamePurposeRailroads by Service TypeRailroads_by_Service_Type.lyrIsolates railroad lines by type of service provided: Passenger and Freight, Freight-only, muter Rail Lines and StopsCommuter_Rail_Lines_and_Stops.lyrIsolates only rail lines which have active commuter rail service.MBTA Rapid TransitMBTA_Rapid_Transit.lyrIsolates the MBTA’s rapid transit lines: Green, Red, Blue, Orange and SilverMethodologyThe focus area of this suitability analysis was the area of Massachusetts within the MBTA commuter rail service area north of Boston. This includes all of Essex County and most of Middlesex County, as well as a small portion of Suffolk County. Thus the first task was to isolate these three counties by selecting them by attribute. This would later allow the large land-use and census data-sets to be clipped to these areas.Calculating Housing Unit Density: Housing density in the units Housing Units per Residential Acre was calculated using the following steps.The Land-use and Census Block Group data layers were clipped to the 3-county area described above.Using the ‘intersect’ command, Census Block Groups and Land-use were combined in a table which isolated land parcels based on land-use in each census block group. Looking at Figure 1, we see how for block group 718 there are multiple parcels for each type of land use (low density residential, forested wetland, etc). Figure 1443230614045The next step was to summarize the table based on land use. This was done using the summarize command and the attributes LU_05DESC and LOGRECNO. Figure 2As we see in Figure 2, a total land area is given for each land use by the category LOGRECNO. Finally, as shown in Figure 3, a pivot table is created which creates one row for each LOGRECNO and lists the area of each land use within that LOGRECNO.Figure 3This table is then joined back to the census block groups such that the land uses are summarized by block groups. A new attribute was created to combine all residential land types into one area. Finally, the Join command was used to join the layer for total housing units to the pivot table. In Figure 4, the layer “blockgroupden” reflects the total number of housing units divided by the total number of residential acres.Figure 4Station and Corridor Catchment AreasA catchment area is defined as the land area around a transit station from which transit riders are drawn. Catchment areas vary based on several criteria, mainly travel time, travel distance, and travel mode used to access the station. For example, a commuter rail station in a rural area where there is little congestion and most passengers drive to the station may have a catchment area of 5 miles or more. However, a rapid transit station located in a dense urban area where most passengers access the station by foot may only have a catchment area of ? mile. Since it was impractical to calculate the appropriate catchment area for each existing station, a distance of 2-miles was chosen for this analysis. This seeks a middle ground corresponding to areas in moderately dense suburban areas and also reflects the typical distance at which commuter rail stations are spaced (3-4 miles) along a route. Two methods were used to define the catchment areas, a network analysis using the Network Analyst tool and a straight line analysis using the Spatial Analyst tool. Figure 5 shows the network analysis overlaid on the straight line analysis as to show the difference between the two methods. The network analysis gives a much clearer picture as to the densities and total populations around each station by the density of the street network.Figure 5 A similar analysis was also done for the MBTA rapid transit stations closer to Boston using the straight line distance method only. This applied a smaller catchment area of 1-mile due to the density of the physical network and the population it serves. Next, a similar 2-mile catchment area was applied to the abandoned rail corridors. Because stations no longer exist along these lines, a straight line distance was used to the actual rail line giving the graphic shown in Figure 6. It is interesting to note how if the catchment areas of Figures 5 and 6 are combined, nearly the entire land area would be within 2 miles of a rail line or station, demonstrating the density of the historic rail network.Figure 6Finally, in Figure 7, a straight line analysis was used to generate distances from North Station in Boston, where all of the northern commuter rail lines terminate. This was necessary in order to give preference to abandoned rail corridors close to Boston.Figure 7Raster AnalysisThe next step was to reclassify each of the data layers described above into 5 categories with 1 being the least preferable and 5 being the most preferable. Land parcels were ranked higher if they were near an abandoned rail corridor, and outside of an existing station catchment area. An exception to this was a lower ranking the farther a parcel was from North Station. Land parcels were generally ranked higher proportional to their housing unit density however, as will be discussed later, it became necessary to rank the highest densities below the more moderate densities. This can be seen in Figure 8 where the core of the cities of Boston, Lawrence and Lowell rank lower than their immediate suburbs.Figure 8Raster Calculations and Final AnalysisThe final analysis was done using the Raster Calculator tool and the rubric shown below with the resulting suitability scores shown in Figure 9.Data LayerCoefficientDistance from Abandoned Rail Line0.25Housing Unit Density0.25Distance from North Station0.15Distance from Existing Commuter Rail Station0.25Distance from Existing Rapid Transit Station0.1Sum1.0There were several caveats to the raster calculator giving the desired results. The biggest issue was dealing with a very large range of values both in terms of density and proximity to infrastructure. For example, during an initial attempt when the highest housing densities were given the highest ranking, all of the downtowns of Boston, Lawrence and Lowell were given the highest suitability scores for ability to support new commuter rail, even though they were crisscrossed by rapid transit and existing commuter rail. The first attempt at solving this problem was to give more weight to the existing rapid transit stations however this skewed the results by allowing less flexibility with the other criteria, and did not affect Lawrence and Lowell which do not have rapid transit. The most feasible solution to this problem became to rank the highest densities as a 3 out of a total score of 5. Figure 9Other DifficultiesInitially, the intent had been to use a network analysis for both the catchment areas of the existing commuter rail stations, but also as a means of weighting various rail corridors based on their rail mileage from Boston. There were two issues with using the Network Analyst which prevented using it for either of these purposes in the Raster Calculator. The first was in regard to the existing station catchment areas. Unlike the straight line analysis using the Spatial Analyst which calculates a maximum distance based on the extents of the map area, the Network Analyst only calculates buffer distances defined by the user. The result in the Raster Calculator is that a suitability score will only be given to land parcels within those buffers. Thus, if the existing station catchment area is defined by a 2 mile radius, only these catchment areas will be scored by the Raster Calculator.The second issue with regard to the Network Analyst was that there were no junctions defined for the rail network, making it impossible to analyze the rail network in the same manner as the street network. However this may have been a moot point. Even if these junction points were present, they would not necessarily reflect the realistic movements of a train at a junction since most rail junctions do not permit a train to turn in any direction where two lines converge.Final ThoughtsTo my surprise, the analysis actually worked fairly well. It showed with relatively clarity, several rail corridors which have surrounding densities similar to existing commuter rail corridors and are in areas under-served by transit. It took little imagination to take the suitability analysis and create a new map, by hand, showing possible extensions of the existing system. That said, the analysis is not necessarily what one would use to determine which rail lines are best suited for new service. It favors areas which are already built-out and does not address the issue of creating inter-city connections which would connect isolated urban areas with rural lands in between. It is important to note as well that I started this analysis with a desired result. I suspected beforehand that most of the rail lines identified by the GIS analysis were in fact suitable for commuter rail service. In that regard, it could be argued that I structured the analysis to favor a certain result. Not that this is totally unrealistic. Many if not most planning studies and transportation models are created with a certain desired result in mind whether it is smart growth, transit expansion or economic development. Everything from the densities needed to support transit, to catchment areas vary based on local conditions such as other available travel options, culture and other behavioral and economic patterns. They are not set in stone and relatively impossible to calculate with certainty.Annotated BibliographyMatisziw, Timothy C.; Murray, Alan T.; Changjoo, Kim. 2006. Strategic Route Extension in Transit Networks. European Journal of Operations Research. 171 (2). June 2006. pp. 661-673This article discusses how GIS analysis can be used to determine the best routing for new transit. It is focused primarily on bus transit, which provides an added level of complexity since bus routes are flexible and can be placed anywhere or moved at any time (some would argue this is a disadvantage not an advantage). However, many of the criteria and methods used in evaluating the best corridors on which to extend service are the same ones I chose to use in this project including: extracting existing service locations, demand (though oddly, they did not use density as one of their criteria), and network structure. The article discusses finding the best route based on capturing the most riders using the shortest route distance, which is the ideology I attempted to use in this project.Cervero, Robert. 2007. Transit-oriented development’s ridership bonus: a product of self-selection and public policies. Environment and Planning 39, pp. 2068-2085.Unlike the previous article, this article discusses the physical attributes required to make transit successful including transit oriented development and the “catchment area” of transit stations. It discusses how physical attributes of a neighborhood or development influence a person’s decision to use transit. While I could not evaluate physical attributes of all of the communities north of Boston, I would argue that the network analysis performed on the street network surrounding existing stations is one measure of the physical attributes of a neighborhood.McGuckin, Nacy A.; Srinivasan, Nanda; 2003. Journey to Work in the United States and its Major Metropolitan Areas – 1960-2000. U.S. Department of Transportation.This lengthy report details the commuting characteristics of several dozen major US cities including data on transit availability and ridership.Urban Transportation Planning, 2nd Edition. Michael Meyer and Eric Miller. McGraw-Hill, New York, 2001. This is a text book used by several transportation planning courses. For the purposes of this project, it provides hard data on what population densities are required to support different modes of transit. I used this reading primarily to determine what an appropriate catchment area would be for the existing commuter rail stations, existing rapid transit stations and potential new corridors. ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download