We just published an interactive 15-minute walkshed layer for Boston’s draft comprehensive plan using vector tiles and ArcGIS Urban, and it’s already shifting how planners evaluate zoning scenarios in workshops this week… Who else is wiring network-based accessibility visuals directly into planning documents — are you caching tiles, and has anyone pulled live GTFS to keep the transit layer current?
We wired “live GTFS” into our 15‑minute map via OpenTripPlanner and cache vector tiles by time‑of‑day (AM/PM/off‑peak) behind CloudFront; for workshops we pin to the latest snapshot so the layer doesn’t jitter… Precomputing isochrones in 2–5 min bins and simplifying per zoom kept pans fast, but if your network rarely changes it’s simpler to cache nightly. If you do realtime, poll GTFS‑rt every 30–60s and rate‑limit — spec: GTFS Realtime Overview | Realtime Transit | Google for Developers.
Quick tip from our last comp plan: we version tiles by scenario and service period (weekday/Sat/Sun) and pin workshops to a fixed 8am snapshot so the map doesn’t ‘dance’ when GTFS updates. Live feeds are great for dashboards, but for public meetings we switch to nightly snapshots and append the feed version in the tile URL for cache-busting. In ArcGIS Urban, caching the street graph separately helped more than any tile tweak.
Found that snapping the 15‑minute walkshed to parcel front doors and marked crossings (not centroids) made Urban scenarios way more defensible. Tiles come off a PostGIS materialized view refreshed nightly; we stick to GTFS‑static with headway modeling because ‘live GTFS’ created odd gaps mid‑session — real‑time only for demos, @vmartin.
In our last comp plan, the biggest realism boost was baking signal delay into the ped network before we cut the vector tiles, so crossings of Mass Ave–scale arterials didn’t inflate access during ‘workshops’. In ‘ArcGIS Urban’ that kept scenario deltas believable around corridor rezones; small caveat: you’ll need to refresh timings when DOT updates plans, which added about 10 minutes to our nightly tile build.
We ended up filtering destinations by whether they’re open at the pinned “8 a.m.” snapshot so the access layer in ArcGIS Urban isn’t counting shuttered groceries or clinics. We parse OSM opening_hours into an open/closed flag during tile bake and keep a tiny vector-tile attribute for the workshop filter — small cost upfront, far fewer “why is that counted?” moments later. Caveat: opening_hours tags are messy, so expect a few manual overrides or a fallback rule.
We wired live GTFS through OpenTripPlanner and cache the access tiles behind CloudFront with a scenario-date version key; the only gotcha is we throttle refresh to every 5 minutes so maps don’t jitter like espresso. Adding a wheelchair profile (penalizing stairs/elevator wait) shifted a few hot spots compared to a generic walk cost and made tradeoffs clearer. If you’re testing GTFS, OTP’s isochrone tools are handy: https://www.opentripplanner.org/.