Traffic Predict
How Many Freeway Traffic Sensors Still Work? Southern California, New York City and Boston
In 2010, 68% of the traffic data from Los Angeles freeway sensors came from detectors that were working. In 2026 (January to September) it was 12%. Caltrans fills in the rest with estimates. Orange County went from 82% to 48%. San Diego went from 88% to 48%. In New York City, 86% of the expressway readings in 2018 were real; in 2026 so far, 65%. Boston's travel-time sensors had a reading 97% of the time in 2026 so far (97% in 2021).
Los AngelesOrange CountyBay AreaSan DiegoNew York CityBoston
Right now
Share of the latest readings (1:00 am PDT, updated every 5 minutes) that came from working detectors:
- Los Angeles and Ventura: 13% (366 of 2,809 stations)
- Orange County: 27% (1,219 of 4,520 stations)
- Bay Area: 54% (2,121 of 3,900 stations)
- Sacramento: 84% (1,564 of 1,873 stations)
- Inland Empire: 49% (1,614 of 3,318 stations)
- San Diego: 53% (2,502 of 4,730 stations)
What the numbers mean
Caltrans measures freeway traffic with detectors in the pavement: loops of wire under each lane that count cars and time how fast they pass. Their readings go to PeMS, the state's traffic database, every 5 minutes. When a detector is broken, unplugged or cut off by road work, PeMS doesn't leave a gap. It fills one in, from nearby detectors if any are working, from that detector's own past if it once worked, and otherwise from an average for a whole group of stations. Every reading is marked with how much of it was actually observed. The chart shows that share.
Filled-in numbers are fine for what PeMS was built for: statewide reports on miles driven and hours of delay, where errors average out. For a single stretch of freeway at 5 pm they can be badly wrong. One recent morning, PeMS reported the same 62.9 mph at 680 Los Angeles stations at once. Detectors that work never agree to the decimal.
This isn't new. In 2013, ABC7 reported that about a third of the 27,000 traffic sensors under Los Angeles pavement weren't working. Since then, the share of real data from the region's freeway detectors has kept falling.
How Traffic Predict handles it
- Filled-in readings that PeMS copies across many stations are left out entirely.
- On the live map, roads with no working sensor are drawn faded and marked estimated, from the working sensors on either side when they are close, or from that road's usual speed for the time (measured in the years its detector worked) when they aren't.
- Drive times on the trip pages say how much of each drive is measured right now.
- Typical speeds (the Typical view, drive-time tables and forecasts) come only from real readings, going back to the years each detector last worked.
Who can fix it
The detectors belong to Caltrans. Los Angeles and Ventura counties are Caltrans District 7; Orange County is District 12. If you want them repaired, the people to tell are Caltrans (District 7, District 12) and your state Assembly member and Senator, who set Caltrans' budget.
New York City
New York City's expressway, bridge and tunnel speeds come from NYC DOT and the agencies around it, through TRANSCOM: about 125 stretches of road, each reporting a speed every 5 minutes. When a stretch's sensors have nothing to report, the reading says so ("no data") instead of being filled in, so the New York line is the share of readings that had real data. It isn't the same measure as Caltrans' "% Observed", but it tells the same kind of story. NYC DOT publishes these speeds hours late, so Traffic Predict uses them for typical New York traffic and drive times, not live speeds.
Boston
Massachusetts measures its highways differently: MassDOT's GoTime system times vehicles between Bluetooth readers along I-90, I-93, I-95, I-495, Route 3, Route 24 and other highways, about 600 stretches of road, each updated every minute. When too few vehicles pass the readers, the stretch has no travel time for that minute. The Boston line is the share of minutes that had one, sampled at four hours each month (morning and evening rush hour, midday and overnight, when fewer cars pass). It isn't the same measure as Caltrans' "% Observed". Traffic Predict uses these travel times for live Boston traffic and drive times.
Year by year
| Year | Los Angeles and Ventura | Orange County | Bay Area | Sacramento | Inland Empire | San Diego | New York City | Boston |
|---|---|---|---|---|---|---|---|---|
| 2010 | 68% | 82% | 59% | 78% | 76% | 88% | ||
| 2011 | 64% | 79% | 52% | 72% | 72% | 88% | ||
| 2012 | 66% | 72% | 48% | 62% | 54% | 88% | ||
| 2013 | 65% | 73% | 50% | 65% | 41% | 87% | ||
| 2014 | 58% | 72% | 47% | 70% | 57% | 88% | ||
| 2015 | 59% | 74% | 48% | 58% | 48% | 88% | ||
| 2016 | 55% | 76% | 55% | 52% | 45% | 87% | ||
| 2017 | 55% | 69% | 63% | 53% | 39% | 82% | 74% | |
| 2018 | 60% | 66% | 62% | 67% | 46% | 81% | 86% | |
| 2019 | 58% | 76% | 65% | 68% | 49% | 85% | 86% | |
| 2020 | 53% | 76% | 57% | 68% | 52% | 84% | 85% | |
| 2021 | 47% | 76% | 53% | 63% | 53% | 83% | 84% | 97% |
| 2022 | 27% | 69% | 46% | 58% | 50% | 66% | 80% | 97% |
| 2023 | 29% | 58% | 57% | 65% | 46% | 58% | 75% | 97% |
| 2024 | 25% | 59% | 54% | 75% | 47% | 59% | 76% | 96% |
| 2025 | 21% | 60% | 38% | 80% | 44% | 52% | 72% | 97% |
| 2026* | 12% | 48% | 38% | 82% | 43% | 48% | 65% | 97% |
Source: Caltrans PeMS Data Clearinghouse, "Station Day" files: the average "% Observed" of mainline freeway stations, weighted by station-days. Los Angeles and Orange County use every month; the other districts January and July. New York City: NYC DOT traffic speeds on NYC Open Data, readings with data as a share of all readings; 2017 starts in September. Boston: MassDOT GoTime, link-minutes with a travel time as a share of all link-minutes, four sampled hours a month; 2021 is the first year GoTime's history covers. *2026 so far. Updated monthly.