Loading the archive…
Not polls × fleet size; feed speed is unused (always 0)
Across days keeps cells that are slow at 08:30 and 17:30 on more than one collected day, and ranks the ones that repeat most. It is prepared from the archive, so it opens at once. This minute plots one clock time.
The list does not publish a centerline or the individual spot coordinates, so these roads are named and not drawn as crash lines.
Shown after the across-days peaks have been read. A route needs at least 12 observations and 30% of them under 12 km/h.
Click a route to draw it. A planning hypothesis for splitting or thinning departures. A depot or terminus can look slow. Not a new alignment and not a crash finding. Delhi GTFS has no shapes.txt.
A route is listed when at least 20 peak observations are at most 25% under 12 km/h, with a median of at least 18 km/h, and it was seen beside a repeating congestion cell. Those are the candidates for diverted buses.
Click a route to draw it. A planning hypothesis for where other buses could be diverted. Observed speed is not spare seats or road capacity. Not a metro alignment and not a crash finding. Delhi GTFS has no shapes.txt.
Archived service days. Count of service dates with an R2 archive manifest and/or hot GPS rows The open service day is not a closed Parquet day until after 03:10 IST the next morning
Unique GPS fixes. Count of unique (vehicle_id, gps_ts) rows Not polls × fleet size; feed speed is unused (always 0)
Archived bus movement. Haversine km/h between the same vehicle’s fixes in two archive snapshots about 90 seconds apart; colour uses the Records speed legend A sample of buses present in that minute, not every ping in the day. Feed speed is unused.
Congestion cells. About 500 m cells where at least 3 buses have moving speed under 12 km/h at the selected instant Standing buses count. A depot can look congested. This is not an accident layer.
Repeating congestion cells. The same 500 m cell is slow at 08:30 and 17:30 IST on collected service days; cells are ranked by how many of those peak samples repeat A short archive can make a depot look permanent. This is not an accident layer.
Published crash regions. OpenStreetMap district bounding boxes for the areas named by Gijre & Ram (2023) The paper does not publish crash coordinates. Central and North districts are the nearest official reading of their phrase “central north Delhi”, not their own boundary.
Gijre & Ram kernel bandwidth. Planar kernel bandwidth of 500 m around each crash point. They also considered 1,000 m and used 500 m Not a ring around a district. The crash points were not published, so this bandwidth is not drawn.
Named 2024 black spots. OpenStreetMap place for a site Delhi Traffic Police named, with a 500 m radius The police define a black spot as repeat crashes within 500 m of a midpoint. The circle is drawn on the OSM place, which may not be that midpoint. Not a Gijre kernel and not an individual crash.
Roads with many black spots. Corridors the 2024 list said held multiple black spots The list does not publish a centerline or the individual spot coordinates, so these roads are named and not drawn as crash lines.
Congested routes. Routes with at least 12 speed observations across the peak samples and at least 30% of them under 12 km/h, ranked by the slow observations. Drawn as the GTFS stop sequence A planning hypothesis for splitting or thinning departures. A depot or terminus can look slow. Not a new alignment and not a crash finding. Delhi GTFS has no shapes.txt.
Routes with room. Routes with at least 20 speed observations, at most 25% under 12 km/h, median at least 18 km/h, and at least 15% of observations (minimum 4) in or beside a repeating congestion cell. Drawn as the GTFS stop sequence A planning hypothesis for where other buses could be diverted. Observed speed is not spare seats or road capacity. Not a metro alignment and not a crash finding. Delhi GTFS has no shapes.txt.
Congestion concentration. Repeating cells ranked by slow-bus observations. The curve is the cumulative share of those observations against the share of cells Equal-sized cells among cells that repeated, not a share of Delhi’s area. Standing buses count. This is not an accident curve.
DTC crashes by year and severity. Stacked counts of minor, major and fatal DTC crashes for each year, as printed in Table 2 In six of the seven years the three severity rows sum to 1 or 2 fewer crashes than the printed total. This chart stacks the three rows and does not adjust them. Not computed from this GPS archive.
People hurt in DTC bus crashes. Counts of pedestrians, passengers, cyclists, two-wheeler riders and car riders by minor, major and fatal outcome, as stated in the results text for 2015–2019 Fatal deaths in that paragraph sum to 137. The same paper counts 145 fatal crashes in those years. Major injuries to car riders were not listed. Not a driver-age chart, and not this GPS archive.
Where fatal DTC crashes clustered. 31 of 145 fatal crashes in east Delhi, 16 in central north Delhi, and 98 as the remainder The remainder is 145 minus the two published counts. It is not a third district they mapped. The kernel surface and the ward population map are figures without a published value table, so they are not redrawn.
Gijre & Ram, Journal of Road Safety 34(4), 2023. District envelopes © OpenStreetMap contributors. Point crash files are not on this map.