Home
As transportation planners and engineers have begun to focus more on freeway commute reliability, it has become necessary to analyze large time spans of data over continuous lengths of road.
The Caltrans PeMS website provides good graphical tools for analyzing individual vehicle detector stations, but lacks many features:
- Ability to measure variability at a sufficiently granular level to calculate reliability metrics.
- Robust data quality filtering methods.
- Convenient bulk download options.
The PeMS_Tools project addresses these issues. I have written everything with the intention of working with the 5-minute rollup data as the “raw” source. A big goal is to move analysis from a simple linear trend line (e.g. average hourly speed) to something that captures variability.
For example, below we see the density of speeds observed at a single location over all Tuesday, Wednesday, and Thursdays in 2015. Although the mean trend line (dark red section) would indicate this link performs well during morning and evening peaks, we see that on many days, the speed actually approaches zero.
The end users for the data output are planners and engineers across various agencies and companies. These users may have different internal database systems and / or mainly use Excel. As such, PeMS_Tools produces nested folders of CSV files that follow a strict structural schema, but are accessible to any user.
For now, instructions.docx is the only written guidelines. It only covers the processing of batch downloading. Hopefully I will fill this in (AAC 16/08/03).