A Multipollutant Low-Cost Sensor Network in Bengaluru, India: Long-Term Performance Evaluation and High-Resolution Mapping of Particulate and Gaseous Pollutants

Hybrid air quality monitoring integrating reference-grade instruments, low-cost sensors (LCS) and satellite data is revolutionising the air quality monitoring standards globally. The current study is from Bengaluru, a metropolitan city in southern India, where a multipollutant (particulate and gaseous) sensor network comprising 60 LCS (5-25 nodes each from five different Indian integrators) has been deployed across the city. Before deployment, these sensor nodes were initially collocated with reference-grade instruments at the India Sensor Evaluation and Training (Indi-SET) centre in Bengaluru. Correction models for PM_2.5, PM₁₀, NO₂, and O₃ were developed using regression and machine learning methods to improve sensor data reliability. The sensors used include electrochemical sensors (Alphasense (A4 or B4 series) or EC Sense TB600 series) for the gases and optical sensors (Plantower (PMS7003 or PMS5003), Tera Sense New Gen OEM, Sensirion SPS30 and/or Alphasense (OPC-R2 or OPC-N3)) for particulate matter (PM). One node from each integrator was collocated for a year to assess the long-term performance of these multipollutant sensors. During this period, an Aerodyne Time-of-Flight Aerosol Chemical Speciation Monitor (ToF-ACSM) is operated to evaluate the real-time performance of PM sensors with varying aerosol chemical composition. The localised correction of sensors reduced errors to 10-40 %, achieving a correlation with reference instruments greater than 0.7 and ensuring uniform performance across different integrators. Initial analysis of the deployed sensors indicated that PM_2.5 levels exhibit significant monthly temporal variation but minimal spatial variation. In contrast, NO₂ showed both spatial and diurnal variations across different nodes, with peaks in the morning and evening traffic rush hours. Additionally, spatial maps of various particulate and gaseous pollutants are being developed using the Land-Use Regression (LUR) model to estimate population exposure. The presentation will cover the long-term performance of multipollutant sensors, the performance of PM sensors with varying aerosol chemical compositions, the most effective long-term correction model, and the high-resolution mapping of pollutants using low-cost sensors and reference-grade instruments.