Graduate students in the University of Utah’s Metropolitan Planning Department worked in a semester-long urban design studio to propose a rethinking of how the university engages with its urban surroundings through a dramatic expansion of transit accessibility. For an institutional setting like a university campus, pedestrian mobility through the landscape is not realistically modeled using the 1.x dimensional space of a street centerline network, but nor is such a space uniform in its two-dimensional resistance to travel. Here I utilized a digital terrain map and street centerlines from the State of Utah, sidewalk centerlines and building footprints from Salt Lake City, and campus building footprints, sidewalks, and land cover data from the University of Utah, along with crosswalks that I generated procedurally, to produce a “friction of distance” surface representing a pedestrian’s perspective. I identified Tobler functions from the research literature – these describe the relationship between walking speed and slope in the direction of travel. Using these data inputs I computed the estimated cost in pedestrian travel time from starting points (transit stations for UTA’s TRAX light rail service) to each point within the study area.
An analysis like this could be used to quantify the user benefits or costs of proposed changes to transit service such as the addition or removal of new stops or lines. The method could also inform planners who wish to gauge the impacts to pedestrian accessibility stemming from other landscaping or development changes in a campus-like environment, where by “campus-like” I mean an urbanized landscape that is relatively porous to foot traffic off the grids of streets or formal sidewalks.
For a tour of the conceptual logic of the analysis, read on below.
The University of Utah campus is situated at the foothills of the Wasatch Range, on the edge of the Salt Lake Valley.
Graduate students in the University of Utah’s Metropolitan Planning Department worked in a semester-long urban design studio to propose a rethinking of how the university engages with its urban surroundings through a dramatic expansion of transit accessibility. In the figure above, existing light rail TRAX stations are shown in blue, and proposed new stations for two new services are shown in red. Studio participants wanted to know how situation of new stations at these locations would influence how far away each point on campus would be from transit, in the currency that most matters to pedestrians: travel time.
The pleas of these students (and the professor running the design studio) did not fall on deaf ears! We defined the study area for the analysis as the region within a quarter-mile of any proposed or existing transit stop (shown above as the yellow line) – a standard rule of thumb for the zone of transit accessibility in transit planning, although as this analysis demonstrates, a problematic oversimplification in many environments.
If the ground were uniformly smooth and flat, and people could walk through buildings in straight lines, the time it would take to get anywhere on campus would be a simple function of crow-flies distance, as shown here. For the more than half of humanity that now lives in urban environments (this figure is about four-fifths of Americans), geographic space is far more constrained and complex.
The network of street centerlines is a typical starting point for modeling mobility in urban landscapes. Here high-volume and high-speed roads are shown in red, with calmer (i.e. easier to cross) roadways are shown in black. This network is clearly incomplete in representing routes actually followed by foot traffic on and around campus, some of which will penetrate the “empty spaces” surrounding network elements, and some of which will violate the routing structure of the network by cutting perpendicularly across roadways, particularly where or when vehicular traffic is light. Perhaps sidewalks tell the rest of the story?
The Salt Lake City municipal government and University of Utah Facilities Management group each maintains a layer of sidewalk centerlines, which would form a more realistic basis for modeling pedestrian movements in this landscape, but there are several problems with these data. First, the City’s layer models sidewalks narrowly and doesn’t connect sidewalk facilities themselves with crosswalks at street intersections. If considered alone this yields a transport network that is balkanized into many unconnected blocks. I scripted a method for automatically generating new crosswalk features to connect sidewalks at intersections.
Even if the entire infrastructure of roads and sidewalks is considered together in a multi-modal network, we miss much of the reality of walking around on campus. Large portions of campus are traversable “off-piste” as it were, and a combined street/sidewalk/plaza network doesn’t represent this sort of possible movement through the space of campus. In an open urban environment like this campus, walking cross-country is an efficient and preferred way to go for many foot trips.
If we wish to model the whole land surface within the study area as potential walking routes, we’re confronted with the fact that when any of us steps out for a stroll, a step is not simply a step in any direction. The key to producing a more accurate model of pedestrian travel time is to account for the multiple factors that influence the cost of travel in every direction at every step. Geographers refer to this cost as the “friction of distance”.
Would-be transit riders walk between and around campus buildings, over a physical landscape that rises dramatically from southwest to northeast – a gain of 540′ in elevation across the mile and a half span of this study area. And of course from any given point the time it takes to walk to some other point is influenced by intervening buildings, which often if not always function as barriers to travel.
The already significant overall slope of campus has been sculpted at the scale of individual buildings. Here slope is shown shading from flatter ground (green) to steeper (red). Any realistic model of how people walk about on this landscape will have to take into account the fact that, from a pedestrian’s perspective, walking down a gentle slope is faster than walking up a gentle slope – and that walking up a gentle slope can be faster than navigating down a really steep slope. Movement in some directions is faster and easier than in others, all other factors held constant.
As you must know if you’re a walker, all other factors don’t hold constant. A terrain surface is a typical starting point for GIS analyses examining travel cost over a landscape, but in this environment the 3D geometry of terrain isn’t the only significant variable influencing walking pace in any given direction: the nature or quality of the surface you walk on is another.
To begin with, to a pedestrian walking around in an auto-centric environment, streets themselves can be barrier as much as conduit to movement. Shown in red here are regions (buildings and high-speed/high volume roads) that are especially strong barriers to walking. The daunting friction of a busy road is significantly eased at crosswalks, which are represented in this data layer although invisible at this scale. You may note that I haven’t represented passages through buildings in this analysis, which could be done with building floor plans for each building. In practice, I find that most people are familiar with the interior of only a subset of buildings on a large campus, and a few attempts at short-cuts resulting in costly detours teaches most pedestrians to route around, not through, buildings they don’t know the interiors of.
The University’s Facilities Management group maintains another polygon layer representing paved areas dedicated to foot traffic. Here I’ve combined all sidewalk centerlines from both city and university sources, together with pedestrian plazas from the university and street crosswalks generated algorithmically by me. Pavement is actually the lowest-friction land cover in an environment like this campus, with its uniform, smooth surfaces. The sidewalks are less noticeable in this figure than previous figures above because they are shown here rasterized to their actual width, as recorded in feature attributes of the vector data.
A substantial fraction of campus and its surroundings is covered by turf grass. Conditional upon impediments like slope or fences, these regions are also a part of the “transportation infrastructure” for a pedestrian in this space. Turf-grass is a pretty low-friction surface for a pedestrian too. Like other campus land cover layers, grass coverage is from the University’s Facilities Management group.
Open pavement in the form of parking lots also comprises an significant fraction of campus. In practice parking lots also function as “pedestrian infrastructure”, although walking can be impeded by parked cars and by passing vehicular traffic, and so has a higher mean friction than do paved surfaces mandated as pedestrian infrastructure.
Considered collectively then land cover represents a variable surface of resistance to walking, all aside from questions of terrain. Impassable buildings or busy roads; comfortable sidewalks, soft grass, congested parking lots, or off-limits landscaped plantings and athletic fields (not shown above but included in the analysis) – each affords a pedestrian walking over them a slower or a faster pace per unit of distance. I calibrated the relative resistance of each of these surface types and combined them into this layer quantifying the overall friction of distance attributable to land cover.
Taking into account differences in the “cost of distance” due to land cover and slope in the direction of travel, I can compute the walk time to every point in the study area from any abitrary starting point, in this case from the TRAX light rail station serving the hospital and health sciences complex of upper campus. Cooler colors represent shorter walk-times; warmer colors represent longer walk-times. Note that this model of walking time is not reversible due to the differential impacts of walking uphill vs downhill along a given route.