Monitor Testing and Calibration

All of our air monitors are tested prior to deployment, and we work with the California Air Resources Board and San Joaquin Valley Air Pollution Control District to collocate low-cost PM2.5 monitors with regulatory PM2.5 monitors across the San Joaquin Valley. We also operate high performance PM2.5 monitors at schools in select communities, as described below.

Testing and Installing PurpleAir Monitors

Prior to deployment, all PurpleAir monitors are tested in Fresno. Batches of 30-40 monitors are set-up on a rack on the roof of the Root Access building and operated for 7-28 days, until the ambient PM2.5 mass range exceeds 35 ug/m3. Data from the PurpleAir monitors is then analyzed to generate intra-device (A vs B sensor) and inter-device (monitor 1 vs monitor 2) correlations and variance. Monitors must demonstrate intra- or inter-device correlations > 0.98 and average percent variance < 10%.

Monitors that pass the intra- and inter-device variability criteria are installed at homes, schools, businesses and public buildings in disadvantaged SJV communities. Monitors are normally installed on the side of a building (under the eaves), approximately 8-10 feet off the ground, and near a window that will allow for sufficient wireless connectivity.

Testing and Installing High Performance Monitors

Continuous PM2.5 beta attenuation monitors (BAM) are assembled and tested (leak test, flow calibration, zero calibration) at the school sites. BAMs are installed on a flat roof of a single-story building away from potential sources of PM (exhaust vents, HVAC units). BAMs are hardwired (ethernet) to a network and a Raspberry Pi microcontroller is used to access the BAM interface remotely. BAMs are maintained by SJVAir staff, including bi-monthly flow test, leak test, inlet cleaning, replacing and/or greasing O-rings and replacing filter tape (as needed). BAMs are recalibrated (zero test) annualy.

Acquiring and Analyzing PurpleAir and BAM data

Real-time data from PurpleAir monitors is accessed through ThingSpeak APIs and stored in a cloud database (AWS). Python scripts have been developed to automatically assess intra-device sensor correlations (A vs B particle counts) for every new entry. Eighteen (18) PurpleAir monitors previously tested by the SJVAir team (as described above) are collocated with high performance PM2.5 BAMs operated by the San Joaquin Valley Air Pollution Control District (Air District), the California Air Resources Board (CARB) and SJVAir at locations across the SJV, as shown on the map below. Circles indicate a 10-mile radius around collocation sites (blue are existing sites, red are planned sites).

Calibrating and Displaying PurpleAir and BAM data

Collocated PurpleAir and BAM data are used to generate correction equations (calibrations) by regressing hourly BAM PM2.5 concentrations with hourly average PurpleAir variables (0.3µm particle counts, 2.5µm particle counts, relative humidity, temperature). Regressions are run for varying time periods (past 24 hours, past 7 days, past 2 weeks, past month) to identify the strongest correlation (R2) and smallest standard error (SE). Python scripts allow calibrations to be updated daily and automatically applied to non-collocated PurpleAir monitor data based on proximity to each collocation site. Historical data from the collocated monitors is analyzed retrospecticely by SJVAir staff each quarter to assess the accuracy of the automated calibrations. Staff are currently evaluating other potential calibration methods, including ML algorithms like Random Forest, k-Nearest Neighbor, and Gradient Boosting. The graph below compares corrected PM2.5 concentrations for the same collocated PurpleAir monitor in Madera, CA. The SJVAir conversion is based on correlation coefficients generated from local data which is more spatiotemporally applicable than fixed conversion factors available on the PurpleAir website (e.g. US EPA, Woodsmoke).

Calibrated real-time data is displayed for all PurpleAir monitors (average concentration during previous 5 minutes). Near real-time is displayed for Air District and CARB BAMs (data (average hourly concentration during previous 80-139 minutes) and for SJVAir BAMs (average hourly concentration during previous 60-74 minutes). Trends for individual monitors are shown as a line graph for the past 72 hours (graph can be customized to display specific time periods).

BAM-anchored School Monitoring Networks

SJVAir BAMs are installed at centrally located school sites, with PurpleAir monitors installed at surrounding schools. SJVAir coordinates with school district Facilities and IT directors to install and connect monitors. The map below shows the network of regulatory and SJVAir BAMs in the SJV. The inset shows the SJVAir monitoring network installed in Delano, CA, in cooperation with Delano Union School District, including a centrally located BAM and collocated PurpleAir monitors at Cecil Ave Middle Schools that calibrates PurpleAir monitors at ten (10) surrounding schools throughout the community.