VIEWS: 3 PAGES: 11 POSTED ON: 11/14/2011
1) There is a student organization attempting to start a Bike Share program at Tufts University. They are requesting the installation of a bike rack to house the bikes for the share and want to analyze Tufts’ campus to find where best to put a new bike rack. For this project it is important to map existing bike infrastructure (i.e. roads, paths, pavement, existing bike racks etc). There are two potential uses for this mapping project: 1) to get a general feeling for existing infrastructure so as to pick a building or area of campus that would be well suited as a bike storage hub; 2) to locate the specific site to Grounds and Maintenance can install the rack. Consequently the use of the map by the student organization would change how accurate the map should be. If the former, it is important for the buildings building footprints and existing bike rack data to be geospatially accurate and current, but it is not important for the land use data to be particular accurate or current. The pathways and roads must have 100% connectivity to make sure the rack is connected by paved and accessible surfaces but do not have to be from current data sources. The geographical position of the geocoded and GPS bike racks should be within a few meters so we can understand their relationship to the built environment around them so relative positional accuracy is important, but absolute accuracy is not. The bike rack data must be current (this academic year) because in the past few years the racks have been moved and many have been edited and must be complete. 2. The three road center lines used were: TIGER 2000 census road center lines, ESRI’s StreetMap USA and street center lines from Medford and Somerville. The street center lines that were the closest representations of the areal imagery were from Medford and Somerville. The Medford and Somerville data would be the best layer to display the roads for my project. These lines had very few instances including roads that do not exist unlike the Street Map USA and especially the TIGER file which have several roads that cut across buildings (2B), lacked major campus roads (Talbot Ave and much of lower campus) (2C), have incorrect simplifications seen in map 2A. The Medford/Somerville layer has curves when the road curves rather than angular connections while the other two files only use straight lines. Overall there are many roads that can be seen on the areal imagery layer that are not represented by any of the road layers, but the Medford/Somerville layer most frequently and most accurately remains true to the image. The TIGER road layer would not be sufficient for my project because it would limit the placement of a bike rack significantly and misrepresents the connections of roads. 2B 2C 2A The TIGER road file and the Street Map USA are very similar and share many of the same issues. From the measurements I did, they only deviate about 10feet on average, but differ from the Medford/Somerville streets around 50feet. One measurement showed that the TIGER and Street Map USA were 138ft from the Medford/Somerville line and 160ft off of the actual areal imagery. This is at the high end of positional error but this alone would dramatically impact the quality and accuracy of my map. 3. The two land use layers used, USGS 2001 National Land Cover and MassGIS’ land use layer, do not show the same information and so do not relate to each other generally. There are certain areas where the land use and land cover reflect each other (i.e. the athletic fields), but the Land Cover layer has a much larger resolution (lower accuracy) than the land use file. 2a 2b The land use data from MassGIS takes into consideration parcels, but the land cover file was recorded as a raster file so it doesn’t use parcel data to assign different uses but instead uses imagery. So the comparison between the raster land cover and the vector land use layers yields general similarities, but the shapes of the use cannot be compared even though the relative positions generally can be. A big critique of the land use data is that Tufts is only classified as two different uses: urban open space and recreational space. Although perhaps many students dedicate most of their time to those uses, the data itself is inaccurate and does not representational of the thousands of students who live on campus and minor commercial space. Comparing the land use data to the areal images, some errors are visible. In the following pictures, some areas are miscoded like in 2c with residential space designated as commercial space or 2d where a parking lot is labeled as barren space. The most flagrant issues yield measurements where: the barren lot designation seen in 2d is at maximum 87m off; the commercial zoning behind Carmichael in 2c is 387m off; The dorms zoned as commercial are 452m from the intended commercial zones. Also the entire surrounding neighborhoods are designated only as multi-family residential but in my experience living in this neighborhood, some of these properties (although few) house only one family. The Land Cover map does a better job distinguishing between building density (which likely would correspond to different renting styles (single-family, multi-family etc). 2c 2d The land cover data compared to the aerial photos has substantial error (or rather lack of accuracy). Qualitatively, the football field is recorded as forest. This data only vaguely corresponds to the physical position and geometry of the area because it is meant to be viewed at a smaller scale. There are vague representations of the library, for instance, that has the right general placement and shape but it is difficult to really know what high, medium, or low density shapes correspond to which buildings. The position is also a bit twisted counter- clockwise. The tennis court is about 84m off the areal imagery, the library is 97m off, and the president’s lawn is about 160m off. Many of the other builds are so simplified that it is difficult to do measurements to quantify the positional accuracy. 4. To make a quantitative assessment of the positional accuracy for the layers metadata of the scale or comprehensive measuring between the layer points and the “true” points would be necessary. -For the GPS points, it is possible to quantify the accuracy. I took the GPS points and made sure that when averaging the waypoints, the estimated accuracy according to the GPS device was less than 10 meters. Many of the points’ estimated accuracy levels were higher more around 5 meters, but the GPS device struggled to get below 5 meter accuracy. Although the estimated accuracy was below 10m, there were three points that were much more inaccurate. In the picture 4a there a green GPS point seen in the tennis courts but it should be sited 98ft away outside Ginn Library. There is a point for the Campus Center that is 89ft away from the actual location. I believe that two points were issues of mixing up data points with mistakenly recorded points, while a point that is 34ft from East Hall was likely due to a lack of positional accuracy of the point taken. Other points I measured were very accurate: 3ft Dewick, 5ft from the tennis courts, 6ft from Carmichael hall. -For the geocoded bike rack points it is possible as well to quantify the accuracy because the distance between the geocoded points and the actual position of the bike racks (known through areal image and personal experience). But this too would be an estimate accuracy assessment because the areal imagery has a 1m resolution according to the ESRI ArcGIS metadata and so it is difficult to see the bike racks and so reliance on personal knowledge must be used and although I do park my bike on campus and did visit each of the racks, using personal knowledge is not consistently accurate (or rather is consistently inaccurate). Some of the distances between the position of the bike racks and the geocoded addresses of buildings that have bike racks points are as follows: 333ft from the covered Fletcher rack 234ft from upper campus center rack 402ft from bookstore rack 120ft from front Halligan rack 564ft from Carmichael rack 4a -For the building block prints, I could not find the necessary metadata from ESRI so it is not possible to know the positional accuracy overall, but the buildings seemed to match with the areal image almost flawlessly so would have a similar quantitative positional accuracy as the areal imagery (1m resolution). -The walkways layer was found from Medford, but the needed metadata was no longer linked to the layer (if ever linked). I did measurements and the positional accuracy was consistently about 15 meters off when it was off. But this does not give overall accuracy and all the points would have to be analyzed compared to the imagery to verify the largest discrepancy of position to quantify the largest possible variation (e.g. +/-15ft cannot definitively be claimed as the accuracy without checking all points). -USGS 2001 National Land Cover, I couldn’t find the scale or metadata that the map was created from so a quantitative assessment is impossible with that data. -MassGIS land cover: The standards for the land cover layer used a minimum mapping unit (MMU) of on average 1 acre, but it went as low as ¼ of an acre especially in urban areas in which parcels are delineated especially in dense multi-family residential areas (e.g. Somerville). The positional accuracy should not be more than 1 acre and most likely is at most closer to ¼ acre. -The Street Maps USA layer was created at a 1:5000 scale and so the accuracy should be close to +/-13feet (which corresponds to a 1:4800 scale according to the US National Map Accuracy Standard for paper maps. But this is for a paper map and it is likely that when it was digitalized the accuracy decreased. I took some measurements between the Street Maps USA line features and the areal imagery and found that the roads ranged from 10-156ft off from the actual road and the average was 50ft off. The 156 was a bit of an outlier and so excluding this from the measurements yields an average of about 25ft. -There was not any metadata for Medford/Somerville road center lines so it would be difficult to quantify the accuracy without measuring all the points in relation to the areal imagery. The only major issue that I saw is seen in 4b. This road was the only problematic road did not correspond entirely with the areal imagery and so the most the data was off is 54ft so at the least the accuracy is +/-54ft. 4b -The US Census 2000 TIGER road layer had many issues of positional accuracy visible and the metadata yielded that the map was created at a 1:100,000 scale which corresponds to 166.7ft accuracy. In fact the road by Eaton was measured at 167ft away from the actual road. Some other readings were 18ft, 62ft. 432 ft, 31 ft. 5. GPS bike rack points: Generally the positional accuracy is fairly good with these points. There is a point by Ginn library that is in the tennis courts which can be seen in 4a. There is a point that intersects the building of East Hall (5a), but the point that seems to intersect the Campus Center should intersect it (5b). A Few points are on the street rather than on the building grounds (see 5c and 5d). 5a 5b 5c 5d Geocoded bike rack points: The position of these points are generally very off from the actual position of the bike racks. Many of the points are in the street rather than on the sidewalk or against the buildings (see 5c). No points actually reflect the position of the bike rack and many do not even accurately represent the position of the nearest building to the bike rack (there are three points all in the same, incorrect position that should represent the library, campus center and bookstore, see 5e). Building footprints: The building footprints reflect the physical space according to the areal imagery layer very accurately with only very minor differences seen 5f. There are no issues of positional accuracy with this data, but issues with other data become apparent when this feature is turned on. Roads from the TIGER road file run through the library, Anderson, Bromfield-Pearson and go on the side walk behind the Campus Center. The Medford and Somerville road lines are much more accurate but even those intersect a building (East), but they follow the areal imagery layer much better than the other two road files—curves are curved unlike in the TIGER layer. 5f Walkways: The walkway layer is very accurate as well, but it does have an angled path from Ginn to the tennis courts that doesn’t exist according to the imagery layer (as well as from personal experience). But that was the only major discrepancy that I noticed. The geometry of the data is correct even when the actual positional accuracy is slightly off as seen in 5h. It also has slight issues with its connectivity seen in 5g. 5g 5h 5i 6. Walkways: The positional accuracy for the walkway data is not perfect, but it is appropriate with only minor issues where a few minor walkways that do not exist are included. The walkways are accurate enough where they add valuable data to show a much more representative view of potential bike rack site locations because much of our campus isn’t directly connected to city streets. Seen in the following images, the layer is necessary and appropriately accurate. 6a 6b Building footprints: The building foot prints match flawless to the areal imagery. It is crucial that these be accurate so we can figure out open space and over hangs for the site. The building footprint data matches almost flawlessly to the areal imagery and so its positional accuracy meets the demands of the project even if the map were to be used for the actual construction of the rack. GPS bike rack points: The GPS points have some issues of positional accuracy—the bike rack on the patio porch of the Campus Center (see 6c) is said to be on the other side of the building, and there is a mismatched point by Fletcher that is located on the tennis courts instead of outside Ginn (see 4a). But the positional accuracy overall fits the needs of the project. 6c Geocoded bike rack points: The positional accuracy is not appropriate for this project. Many points are hundreds of feet off and it does not give an accurate portrayal of bike space on campus. The main racks of campus (Campus Center and Library) are even grouped basically on top of one another a block away from either location (see 6d). 6d 7. Completeness: The road files have a few problems with completeness. The Medford/Somerville layer is almost entirely complete except for the below break in which all the road layers have a gap of connectivity. The attribute information of road names is complete for the road files as well, but the address file for the TIGER is missing several addresses (all the Academic Quad addresses, address for the tennis courts, address for Olin etc.) Many of the campus address data is condensed into a few addresses on campus and does not reflect the mailing addresses of the buildings. The land cover and land use have data for the entire area (no unknown land use or cover) but both are simplifications and lack specific designations of use: the land cover layer even is lacking many buildings entirely, and the land use data does not make distinctions between residential, academic, and open space coding it all as urban open space. So although these have information for the entire area, the lack of accuracy and simplification makes these data sets incomplete for the uses of this project. The building footprint data is complete with the exception of newly constructed buildings. The walkways have some minor problems with connectivity seen in 7b. 7a 7b The GPS bike rack points are not entirely complete. After data was collected I found another bike rack on the side of the Crafts House. It is possible that there are other bike racks on campus that are not present. Also, although the attribute data table for the bike racks is complete, some of the GPS points did not translate onto the map (see 7c). Also in certain areas where there were two bike racks touching each other, they are counted as one bike rack although more accurately it should be one bike rack site. 7c The geocoded bike rack points are not a complete data set. Many points were not able to be geocoded properly and could not be matched to an address and many of the current bike racks were not present in the outdated information from which the addresses were compiled. 8. The time of the data collection for certain things is ok to be old (80’s) if there have been no changes in that data (i.e. streets). So one way to know if the information is current is just from the date of collection. Census data from 1990’s is not current but road lines from the 1990’s can be. Another way is to judge currency from the areal imagery which is at least as current as 2006. Sophia Gordon was built in 2006 and is in the areal imagery but not in the building foot prints. The areal imagery is up to date, the building foot prints are very close, but have small currency issues with Sophia Gordon Hall and Granoff Music Center seen in 8a. The TIGER road files are from 2000, and are current just inaccurate. The Street Maps USA data is also inaccurate, but current because it matches the imagery. The Medford and Somerville road centerline information reflects accurately the 2006 imagery so it is up to date even if the information is old (although due to a lack a metadata the date of data collection is unknown). The land use data is from census 2000 so should also be up to date. The land cover data is inaccurate but from 2001, so up to date because the land cover at tufts hasn’t changed significantly since then. The walkways are up to date as seen from the accuracy against the areal image. The Geocoded points are not current because I used a map from the Tufts Sustainability office to find the building names that according to their transportation map had bike racks (8b). This map is out of date and several points no longer exist and several new racks are not reflected. The GPS points are up to date because I collected that data last week. 8b 8a 9. Attribute accuracy: The land cover attributes were not very accurate. The football field was coded as forest and many of the attributes do not correspond to the project’s need. The land cover distinctions between building, pavement and open space important for this project but coding several different types of green space is not needed like in this data set’s attributes. The land use data did not have appropriate attribute accuracy for this project either. The entire campus is labeled as urban open space or recreational space. Having better attribute accuracy that broke land use down into residential, commercial, academic, recreational, and open space would clarify much more what types of space have bike racks already and where people would most likely use them. The road attribute accuracy for names is not really needed because this would be an internal map used by people familiar with campus and these names exist in the data sets anyway. But the address attribute accuracy in the TIGER road file which was used to geocode the bike rack points had incorrect information about Talbot St. Apparently there is a 40 Talbot street in Medford and the TIGER road file either didn’t have or didn’t properly label Talbot St. at Tufts with the correct addresses. Consequently the bike points for Sophia Gordon at 40 Talbot were mapped as being in Medford.
Pages to are hidden for
"Bike Rack proposal"Please download to view full document