Blog

Field notes: the collector road no count program had touched since the last census

A quiet residential road crossing with stop signs, snow banked along the shoulders.

Every region has a handful of these. Not the arterials, those get swept on a rolling three-year cycle whether anyone asks or not. Not the freeways, ATR stations handle those continuously. It's the collector in between, the one that connects a subdivision built in 2015 to a state route that was already there, carrying trips nobody budgeted a count for because it wasn't on the program list when the list got made.

I pulled volumes for one last month for a signal warrant check at the collector's intersection with a minor arterial. The traffic counting database had exactly one entry, a short-duration tube count from a decennial census year, factored forward twice since then by whatever regional growth rate the MPO was using at the time. Two growth factors stacked on a count that was already seasonally thin to begin with. Nobody could tell me if the subdivision on the north end had built out to plan. Nobody could tell me if the connector had picked up cut-through traffic once the arterial got its new signal timing plan two years back, which is exactly the kind of thing that happens and exactly the kind of thing a stale number won't show you.

This kind of link is the normal condition for anything below minor arterial on the functional classification chart in a region that isn't flush with count crew hours. The count program has a list, the list has a cycle, and the cycle has to prioritize the roads that feed the model's biggest TAZs. A collector serving eleven hundred trips a day loses that argument every time, until somebody needs a real number for a capacity study or a signal timing update and discovers there isn't one.

What you're left with is a judgment call dressed up as data. You take the census-year count, apply a growth factor pulled from the travel demand model or a nearby corridor with better instrumentation, and you write it into the study with a footnote nobody reads. It's defensible. It's also a guess wearing a decimal point, and if the subdivision build-out ran ahead of the model's land-use assumptions, which it usually does, your synthetic ADT is wrong in a direction that matters for a capacity study.

The honest fix has always been to send a crew out with tube counts or a radar unit, and that's still the right call for a link where you need peak hour factors, directional splits by fifteen-minute bin, and vehicle classification. But a lot of these collector checks don't need that resolution. A capacity study scoping memo wants a defensible current ADT to replace a twelve-year-old factored number, not a full seven-day classification count. For that narrower question, pulling a current tally off a single high-resolution pass over the corridor covers the gap without scheduling a crew for a link that was never going to make this year's count program anyway. That's the whole premise behind Vehicle Counting: one image, one pass, a count by link and direction you can drop straight into the study instead of another stacked growth factor.

I'm not arguing the census-year count was wrong when it was taken. It probably wasn't. The problem is what happens to a reasonable number over a decade of compounding assumptions on a road that was never going to get re-counted on its own. If you're sitting on a link like this, worth asking whether what you actually need is a fresh tally rather than a third growth factor stacked on the last one.

Start a pilot