Publications·November 30, 2024
This UNESCAP working paper assesses how Mongolia can use satellite and ground-based remote sensing to strengthen air-quality management. It sets the pollution context, maps the institutional and monitoring landscape, reviews what EO (Earth Observation) data are already used (fires, AOD, trace gases), documents the rollout and current status of Pandora spectrometers and GEMS (Geostationary Environment Monitoring Spectrometer), and finishes with a practical action agenda to move from pilots to operational, decision-ready products.
Mongolia’s air-pollution picture & policy response
Mongolia’s urbanisation and harsh winter inversions make particulate matter the dominant risk—Ulaanbaatar routinely experiences the worst air in the country. Ger-area coal heating historically drove winter spikes; estimates cited in the report attribute ~80% of UB’s ambient pollution to ~200,000 ger households (pre-briquette era). National annual mean PM₂.₅ exposure ≈ 38 µg/m³ (~7.6× the WHO guideline), with ~2,245 deaths in 2019 attributable to fine particles.
Policy instruments include the 2019 raw-coal ban in UB and switch to briquetted coal, the National Programme on Air & Environmental Pollution Reduction (2017–2025) targeting large emission cuts, a Clean Air Fund, and climate-aligned measures via the NDC (updated 2019) that also curb black carbon and methane. Implementation capacity and financing remain challenges; despite significant expenditure (≈US$52 m, 2008–2016), air-quality outcomes improved unevenly.
Monitoring & institutions (who does what)
The ground network (run by NAMEM with the Air Pollution Reduction Department in UB) covers ~42 sites measuring SO₂, NO₂, PM₁₀, PM₂.₅ (O₃/CO at a few UB stations). MET sets policy and standards; IRIMHE operates remote-sensing and environmental data systems (MODIS, NOAA feeds, Mongolian Data Cube) and contributes to AQ modelling and EO ingest. The National Remote Sensing Centre (NRSC) and Information Computer Centre (ICC) provide fire products (MODIS thermal anomalies) and aerosol parameters, while local authorities implement bans, fuel switches and enforcement in “improvement zones.”
What remote sensing is already doing
Long-range transport & dust. Studies use MODIS/MISR AOD to map PM and Gobi dust episodes; analyses show strong spring peaks and cross-border transport into and out of Mongolia. WRF–CMAQ regional forecasts (with ADAM3 dust) are being tested; plans include fusing real-time satellite inputs.
Fires. ICC publishes daily MODIS fire products (hotspots, diurnal occurrence, monthly composites) and aerosol parameters; NRSC processes NOAA/AVHRR for fire mapping and dispatch to provincial agencies during emergencies. The paper encourages shifting from hotspot counts to burn-area + FRP to improve emissions estimates.
Policy evaluation, vulnerability, and composition. The report highlights EO’s role in: filling unmonitored areas with monthly/annual PM₂.₅ maps (AOD + CTM + bias-correction); tracking SLCPs and ozone precursors (NO₂, SO₂, HCHO, O₃ via TROPOMI/GEMS/Pandora); building equity maps (pollution × population/vulnerability), and, over time, linking to composition initiatives like MAIA for health-relevant aerosol components.
Pandora & GEMS — current status and early lessons
Pandora (PGN / Pandora Asia Network). Two sites are operating: P216 Ulaanbaatar (urban) and P217 Dalanzadgad (suburban/southern). Both target NO₂, HCHO, SO₂ columns for science and for satellite validation. Data-quality diagnostics show NO₂ has a large share of “assured high-quality” records, while SO₂ is mostly “not-assured low quality,” indicating the need for QA/QC, stable power, and tracker maintenance. Cross-checks show:
Pandora AOT vs AERONET: strong agreement (R² ≈ 0.77, slope ≈1.02), with Pandora resolving sharper temporal variability.
Pandora vs TROPOMI NO₂: broadly coherent trends with Pandora’s finer temporal detail.
Pandora vs UB03 ground NO₂/SO₂: correlated peaks, capturing short events and diurnal structure.
The report recommends expanding Pandora, co-locating with reference PM/NO₂/SO₂/O₃, and institutionalising SOPs and spares.
GEMS (geostationary). GEMS’ swath covers the south of Mongolia; the plan is to use AOD, NO₂, SO₂, O₃ for spatio-temporal analysis where ground monitors are sparse, validated against Pandora, LIDAR and models. Early comparisons of Pandora (P217) vs GEMS NO₂ show Pandora capturing higher, more variable signals than the smoother GEMS series (diurnal/visibility effects and retrieval differences likely contributors). ADAM dust forecasts vs GEMS show low-to-moderate correlation (Pearson ≈ 0.34) with a tendency to overestimate event intensity, underscoring the need for bias-correction and blended products.
Where Mongolia is still stuck (gaps & constraints)
People & processes. Too few trained operators/analysts; gaps in SOPs/QA-QC for Pandora and regulatory monitors; intermittent power/connectivity.
Platforms. No unified EO–AQ data layer (ingest, storage, versioning, APIs) that routinely blends GEMS/TROPOMI/AERONET/Pandora/ground/fires/models and publishes reproducible, documented products with uncertainty metadata.
Productisation. Analyses have not consistently become operational tools for advisories, enforcement targeting, or policy evaluation.
Comparability. Differences between satellite columns, Pandora columns and surface stations complicate validation; standardised methods and bias-correction are needed.
Recommendations — a practical action agenda
Build the national EO–AQ platform. Automate ingest of GEMS/TROPOMI/OMI, AERONET, Pandora, AQMN, MODIS/VIIRS fires, and WRF–CMAQ/ADAM; publish near-real-time maps, archives and open APIs with clear data-quality flags and bilingual documentation.
Harden Pandora operations. Co-locate with reference PM₂.₅/NO₂/SO₂/O₃; deploy SOPs, tracker maintenance, humidity control, and spares; expand to additional sites to cover northern airsheds as well as the south.
Operationalise GEMS workflows. Bias-correct against Pandora/AERONET/ground; stand up episode tools (dust and fire plume tracking, local vs. transported attribution); add health-relevant indicators.
Make products decision-ready. Municipal “episode rooms” for alerts, enforcement and post-event forensics; policy-effect dashboards to evaluate the 2019 fuel switch, stove/heating programmes, and seasonal burn controls.
Grow skills & partnerships. Recurring training (EO ingest, ML bias-correction, WRF/CTM, nowcasting/forecasting); shared code repos and help-desk; deepen ties with ASEAN/EANET/SERVIR and neighbouring NMs for transboundary haze coordination.
What success looks like (12–24 months). A stable Pandora–GEMS–ground “fabric” delivering routine PM₂.₅/NO₂ maps and episode analytics nationwide; validated dust/fire emissions based on burn-area + FRP; and public policy dashboards tracking progress toward standards and health gains.
Key words
Remote sensing; GEMS; Pandora (PGN); AOD; PM₂.₅ mapping; NO₂/SO₂/HCHO/O₃; dust storms (Gobi); fires & FRP/burn-area; airsheds & transboundary transport; WRF–CMAQ/ADAM; QA/QC & SOPs; IRIMHE/NAMEM/MET.