Learning Materials·July 31, 2025
This practical, technology-forward handbook is written to help government agencies, cities, researchers, and civil society in Asia-Pacific build modern, scalable air-quality monitoring systems. It explains—in accessible language—the building blocks of an integrated monitoring “fabric”: satellite remote sensing (LEO and GEO), ground-based remote sensing (e.g., Pandora, AERONET), low-cost sensors (LCS), and machine-learning (ML) fusion that ties everything together for real-time insight, forecasting, and early warning. It opens with the public-health case: ~6.7 million premature deaths annually are linked to air pollution (WHO, 2024), with ~70% occurring in Asia-Pacific, underscoring the need for regionally tailored strategies and affordable monitoring options that work outside capital cities.
Chapter 1 — Remote sensing: the space & ground pillars
The handbook introduces remote sensing as essential for closing spatial gaps left by sparse regulatory monitors. It walks through core concepts (spatial/temporal/spectral/radiometric resolution) and contrasts active (radar/LiDAR) vs passive (radiometers/spectrometers) sensing, noting trade-offs in cost, coverage, and dependence on illumination/atmospheric conditions. Diagrams position common sensors by revisit rate and resolution, and flag which pollutants each can “see.”
Key satellite instruments and what they’re good for
TROPOMI (Sentinel-5P, LEO) – Daily global coverage at km-scale with retrievals for aerosol, NO₂, O₃, CO, SO₂, CH₄, HCHO; includes API pointers to the Copernicus Data Space (OData/OpenSearch/STAC) for automated access.
GEMS (GEO-KOMPSAT-2B, GEO) – The world’s first geostationary air-quality spectrometer for Asia with hourly columns of NO₂, O₃, SO₂, HCHO, aerosol over 5°S–45°N, 75°E–145°E; includes open-API steps (get key → create URL → query) and an image viewer for real-time exploration.
MODIS (Terra/Aqua, LEO) – Two-decade workhorse for AOD; explains aging satellites, continuity via VIIRS, and links to NASA LAADS + Worldview for easy browsing/downloading.
VIIRS (Suomi-NPP/NOAA-20, LEO) – Wider swath (3,000 km) and higher spatial detail for AOD and fire/smoke monitoring; complements MODIS with better equatorial coverage.
OMI (Aura, LEO) and OMPS (Suomi-NPP/NOAA-20, LEO) – Focus on trace gases (e.g., total ozone; BrO, NO₂, SO₂; volcanic ash) with daily global coverage; the handbook includes NASA Earthdata download steps and tooling.
Ground-based remote sensing: Pandora & AERONET
AERONET sun-photometers (up to 7 bands) provide long-term AOD and microphysical properties at 370+ global/98 Asia-Pacific sites (multiple quality levels: L1.0, L1.5, L2.0).
Pandora spectrometer (280–525 nm, ~0.6-nm resolution) retrieves high-cadence column NO₂, O₃, SO₂, HCHO for satellite validation and diurnal profiling. As of Dec 2024, 56 instruments are active in 15 Asia-Pacific countries; 20 were installed in 7 countries through ESCAP/NIER/KOICA’s PAPGAPI, which also established the Pandora Asia Network (PAN). Public data endpoints are provided.
Real-time viewers
Quick links help non-specialists start exploring: NASA Worldview (MODIS/VIIRS/OMI, etc.), Copernicus Dataspace Browser (Sentinel-5P CO example), and NIER’s GEMS viewer for hourly pollutant maps.
Chapter 2 — Low-cost sensors (LCS): types, pitfalls, best uses
LCS extend coverage cheaply and quickly—ideal for resource-constrained settings and community science—but need calibration, QA/QC, and thoughtful siting. The chapter introduces applications (community monitoring; supplementing reference networks; research), a Seoul case (S-DoT, >1,000 stations since 2019), and points to EPA’s Air Sensor Toolbox for siting/installation checklists. A table summarizes pros/cons, costs, and typical response times by sensor class.
Five LCS families explained
Electrochemical (EC-LCS) — for NO₂, NO, O₃, CO, SO₂ via redox currents. Strengths: moderate cost, sensitivity, low power. Caveats: strong temp/humidity dependence and cross-interference (e.g., O₃↔NO₂), with illustrative bias magnitudes and the need for meteorological correction and subtraction of interfering gases.
Optical Particle Counters (OPC) — for PM via light scattering. Good sensitivity from µg/m³; best for PM₂.₅; accuracy degrades at low concentrations/large particles; keep RH <85% to avoid hygroscopic bias; average over time for stability.
NDIR — for CO₂ (and sometimes CO) via IR absorption (Beer–Lambert). Pros: selectivity, short response, long life; cons: sensitivity to T/RH/pressure; high thresholds for non-CO₂ gases.
PID — for VOCs, using UV photo-ionization. Pros: fast, sensitive; cons: high power, poor selectivity across VOC classes, frequent calibration/maintenance.
Metal-oxide (MOx) — for NO/NO₂/O₃/CO/CO₂, via conductivity changes on heated oxides. Pros: very low cost; cons: T/RH sensitivity, slow response (5–50 min), drift/instability and memory effects; requires careful correction.
Bottom line: LCS shine when designed as a system—stable power/enclosures, co-located met sensors, representative siting, regular cal/val against reference instruments, and ML bias-correction. They are complements, not replacements, for reference monitors.
Chapter 3 — Machine learning: from data to decisions
The handbook positions ML as the glue that turns disparate measurements into decision-grade products:
Predicting air quality (nowcasting/forecasting) from satellites + ground + meteorology + traffic, using models from multivariate regression and Random Forest/GBM to deep learning (CNN/LSTM) and SVM/K-NN.
Enhancing sensor accuracy via calibration models that correct LCS drift and meteorological sensitivity; data fusion to blend AOD/columns with in-situ PM/gases; and optimising network design (e.g., reinforcement learning to place new sensors strategically).
Real-time monitoring & alerts, including episodic detection (smoke, dust, volcanic plumes), exceedance warnings, and user-facing mobile/wearable guidance. The chapter stresses uncertainty communication as essential to public trust.
Practical features and “how-to” touches you can use tomorrow
Step-by-step data access: API endpoints and examples for GEMS, TROPOMI, OMI/OMPS, MODIS/VIIRS, and links to an AERONET site browser and Pandora station data directories.
Comparative tables you can drop into project notes: satellite capability summary (orbit, resolution, cadence, pollutants) and LCS comparison (costs, response times, limitations, best-use notes).
Programme context: the handbook sits under RAPAP and documents PAPGAPI’s role in deploying Pandora for GEMS validation and regional data sharing—helpful for agencies writing proposals or MoUs that hinge on cooperation and open data.
What the authors want policy teams to remember
Think hybrid. Use reference-grade stations as your regulatory backbone; EO (satellites + ground-based) for coverage and transport/diurnal insights; LCS to densify; ML to fuse and validate.
Plan for operations, not pilots. Budget for O&M, spares, QA/QC, training, and open data plumbing (APIs, metadata, versioning).
Make it useful to people. Deliver real-time maps, plain-language advisories, and alerts that municipal teams and the public can act on—especially during seasonal burning, dust, or volcanic events.
Keywords
Remote sensing; GEMS; TROPOMI; OMI/OMPS; MODIS/VIIRS; Pandora; AERONET; low-cost sensors (EC/OPC/NDIR/PID/MOx); machine learning; data fusion; nowcasting/forecasting; open APIs; QA/QC; PAPGAPI/Pandora Asia Network.