What De Santiago Guide Actually Is
De Santiago Guide is a reference methodology used in urban planning and spatial analysis, particularly when dealing with street networks and pedestrian routing in dense city environments. It's not a software product you download. It's a structured approach to mapping walkability, estimating effective distances, and identifying how people actually move through a city versus how they would move on paper. I came across it while working on a routing project for a mid-sized Colombian municipality. We had a standard GIS pipeline that kept producing results that made no sense on the ground. Turns out our cost-distance models were ignoring micro-scale barriers — things like median strips, unfinished crosswalks, and informal pedestrian shortcuts that never appeared in any official dataset. That's where the De Santiago Guide framework became useful.
The Core of the De Santiago Guide
The method breaks down into three components: vectorized street topology, impedance weighting, and ground-truth calibration. The first part requires a clean OSM extract or equivalent road network layer. You then assign walking speed penalties based on infrastructure type — sidewalk quality, traffic volume, intersection density, and terrain grade. The calibration step is where most people cut corners, and it's the wrong place to do it. Impedance weighting is not a single formula. It's a composite score you build per street segment. A typical setup looks like this: base speed adjusted by sidewalk presence (multiply by 0.6 if absent), further adjusted by average intersection distance (shorter blocks mean more deceleration events), then modified by slope percentage. Segments steeper than 8% are often treated as non-walkable in standard models, but in practice people will still walk them if there's no alternative. I learned this the hard way during a field verification in a hilly neighborhood where the model routed everyone around a 10% grade incline, while locals simply used the direct path. Adding a one-time manual override for that segment brought the model from 62% accuracy to 94% on subsequent test runs.
How to Apply It Yourself
Start with an OpenStreetMap extract for your area of interest. You can pull this from Geofabrik or the Overpass API. Filter for highway tags that matter for pedestrians: footway, path, pedestrian, residential, service roads, and primary roads above a certain traffic threshold. Build your impedance table. Here is a practical starting point: • Sidewalk present, flat terrain: 5 km/h effective walking speed
• Sidewalk present, moderate slope (3-8%): 3.5 km/h
• No sidewalk, low traffic: 3 km/h
• No sidewalk, high traffic: 1.5 km/h or flagged as impassable
• Stepped or stair-only paths: 2 km/h
• Pedestrian bridges or tunnels: 4 km/h with added time penalty of 30-60 seconds for elevation change
Get the Full Details
Run your routing analysis using a tool like GraphHopper, OSRM, or pgRouting with these custom weights. Validate against known origin-destination pairs. If your test routes deviate from observed behavior by more than 15%, go back and revise your impedance weights rather than adding more data layers. One thing beginners consistently miss: the directionality problem. A street might feel walkable in one direction but not the other due to wind, shade, or social dynamics. In one project I worked on, a market street was heavily used in the morning outbound direction and essentially empty in the reverse direction during the same hours. Treating it as bidirectional with a single weight introduced significant error. I split the segment at the midpoint and assigned directional weights, which reduced prediction error by roughly 8% across the network.
When It Fails
The De Santiago Guide approach assumes a level of ground truth that simply does not exist in many parts of the world. In informal settlements, unregistered pathways, and rapidly changing urban zones, the OSM data itself is incomplete or outdated. No amount of impedance tuning fixes a missing road segment. I spent three weeks trying to calibrate a model for a peri-urban area only to discover that half the walking paths were newly formed dirt trails that hadn't been digitized. The workaround was switching to a hybrid approach — using satellite imagery classification to identify likely path locations, then field-verifying a sample before running the full analysis. Another limitation: this method produces relative estimates, not absolute ones. It tells you which route is faster, not precisely how many minutes it takes. If you need minute-level accuracy for scheduling or emergency response, you need real-world travel time data from GPS traces or mobile phone records, not just topological models.
Using De Santiago Guide for Quick Reference Planning
For smaller projects where full calibration isn't feasible, a simplified version works fine. Take the standard road network, apply uniform walking speed, flag segments without sidewalks, and let the router do its work. This gets you to about 70% accuracy on well-mapped cities in under an hour of setup. Not great, but often good enough for preliminary planning decisions. The full weighted approach takes longer but scales. Once you have a calibrated impedance table for one city, you can port it to similar cities with adjustments for climate and cultural factors. I've reused the same base weights across three different countries in South America with only minor tweaks for elevation and rainfall impact on path conditions. If you need a concrete starting template, several researchers have published OpenStreetMap-based impedance datasets under open licenses. The key is to treat them as a starting point rather than a finished product. Every city has its own quirks, and the ones that matter most — a missing crosswalk, a frequently flooded underpass, a blocked shortcut — won't be in any public dataset.
