Air quality forecasts using the NASA GEOS model

Transcription

1 Air quality forecasts using the NASA GEOS model K. Emma Knowland USRA/GESTAR NASA () In collaboration with: : Christoph Keller, Eric J. Nielsen, Clara Orbe, Lesley Ott, Steven Pawson, Emily Saunders Atmospheric Chemistry and Dynamics Lab: Bryan Duncan, Melanie Follette-Cook, Junhua Liu, Julie Nicely 7 June

2 Why we care about atmospheric chemistry 1. Climate & Dynamics 2. Air Quality 2

3 Air pollution is the single largest environmental risk factor The Lancet,

4 Air pollution is the single largest environmental risk factor The Lancet,

5 Air quality is a global problem NO 2 CO VOCs 1 of every 9 death is related to air pollution (WHO) $5 Trillion in welfare losses every year (World Bank) Locally up to 50% crop loss due to ozone 5

6 Surface observations are sparse! Tropospheric Ozone Assessment Report TOAR (Schulz et al., 2017) 6

7 Point source measurements Surface observations of pollutants are point source measurements which can be sparse. O 3 PM 2.5 NO 2 7

8 And highly variable! 8

9 Current data coverage from space is limited Terra MOPITT (CO) Aura OMI (NO 2, O 3 ) 9

10 NASA s GEOS Model 10

11 NASA s Composition Forecast GEOS - FP GEOS - Chem 11

12 GEOS-Chem is a state-of-the science chemistry model Tropospheric and Stratospheric full chemistry 220 reactive species, 720 reactions 100+ user/developer groups worldwide Updated version is released about every year 12

13 Chemistry is not cheap! GEOS - CF GEOS - FP GEOS - Chem One 5-day forecast per day since March day analysis 5-day forecast c360 (0.25, ~25x25 km 2 ) resolution Currently no data assimilation of constituents in GEOS-CF 13

14 NASA s Composition Forecast O 3 GEOS - FP GEOS - CF NO 2 GEOS - Chem PM

15 NASA s Composition Forecast Research product GEOS - FP GEOS - CF GEOS - Chem 15

16 GEOS configuration for GEOS-CF CAP ExtData GCS (Root) History GCM IAU Atmosphere Ocean Physics Dynamics Data Ocn Data Sea I RAD Moist Sfc GWD Turb Chem FV3 SW LW GAAS TR Lake Land Salt LandIce GoCart Veg Catch GeosChem 16

17 GEOS is a modular system Run a version of the GEOS model with both GOCART and GEOS-Chem Big differences are 1) GEOS-Chem are fully coupled to the gas-phase chemistry, especially for SOA 2) GOCART assimilates AOD Chem GAAS TR GoCart GeosChem 17

18 GEOS is a modular system Run a version of the GEOS model with both GOCART and GEOS-Chem Big differences are 1) GEOS-Chem are fully coupled to the gas-phase chemistry, especially for SOA 2) GOCART assimilates AOD Future work to couple GOCART and GEOS-Chem Chem GAAS TR GoCart GeosChem 18

19 Contributors to Air Pollution Aerosols Reactive gases Particulate matter (PM): Organic Carbon Nitrate Black Carbon Sulfate Sea salt Dust GOCART GEOS-Chem Ozone (O 3 ) Nitrogen dioxide (NO 2 ) Sulfur dioxide (SO 2 ) Volatile organic compounds (VOCs): e.g., Formaldehyde, Benzene, Toluene, and many more 19

20 OpenAQ surface observation data base OpenAQ is a non-profit compiling publically available air quality data in near-real time into an open-source data base 20

21 OpenAQ surface observation data base North America Europe Asia Southern Hemisphere 21

22 GOCART vs GEOS-Chem PM 2.5 GOCART GEOS-Chem In all the following analysis showing PM 2.5 from GEOS-Chem 22

23 High resolution critical to resolve features relevant to air quality 23

24 GEOS-OMI OMI GEOS National Aeronautics and Space Administration Global evaluation of NO 2 : comparison against OMI tropospheric columns 24

25 Comparison of GEOS-CF O 3 against ozone sondes Trinidad Head, CA Sapporo, Japan Payerne, Switzerland Hohenpeissenberg Legionowo, Poland Goose Bay, Canada Edmonton, Canada Churchill, Canada Sondes GEOS-CF 25

26 Ozone forecast against surface observations for Mexico City Observation GEOS-CF 26

27 Local evaluation of PM 2.5 from wildfires Observation GEOS-CF 27

28 Local evaluation of NO y from wildfires NO y = NO+NO 2 +HNO 3 +HNO 4 +HONO+2*N 2 O 5 +PAN+OrganicNitrates+AerosolNitrates Observation GEOS-CF 28

29 Local evaluation of NO 2 : model captures diurnal and weekly variations Observation GEOS-CF 29

30 What comes out depends on what goes in Working with local governments Rio De Janeiro Jakarta to improve emissions in the model to improve local forecasts Fires from Indonesia impact urban centers in Singapore and Malaysia Observation GEOS-CF 30

31 Stratospheric intrusions (SI) Several peaks in O 3 at monitoring stations reported in AZ, CO and MD April 16-18, 2018, likely caused by SIs With several exceeding the NAAQS O 3 > 70 ppb regulatory limit 32

32 Application: Health Air Quality Index (HAQI) HAQI is a multi-pollutant index The Canadian HAQI is a function of O 3, NO 2, and PM 2.5 (Stieb et al., 2008, J. Air & Waste Manage. Assoc.) 33

33 Application: Health Air Quality Index (HAQI) HAQI is a multi-pollutant index Ozone O 3 influences Background levels Nitrogen Dioxide NO 2 is Short-lived Extreme gradients Particulate Matter PM 2.5 driver of spatial gradients 34

34 Application: Health Air Quality Index (cont.) (based on Stieb et al., 2008) NYU and UNICEF will use GEOS-CF to refine HAQI for children 35

35 Summary GEOS-CF produces daily global air quality forecasts at 25km horizontal resolution Output available to public in late Visualization tool 36

36 Summary GEOS-CF produces daily global air quality forecasts at 25km (16 miles) horizontal resolution Output available to public in late Visualization tool

37 What s next: Data assimilation system for tropospheric constituents Impacts of joint assimilation of O 3, NO 2 and CO: Reduction of CO bias Better spatiotemporal representation of NO 2 Further increase of tropospheric ozone Weak observational constraint in current configuration

38 What s next: Machine learning mechanism for chemistry solver (random forest) Black: GEOS-Chem Red: Machine Learning Mat Evans York University 39

39 Summary GEOS-CF produces daily global air quality forecasts at 25km (16 miles) horizontal resolution Output available to public in late Visualization tool 2. Data access through NASA GES DISC and OpenDAP Under development: 1. A 2-5 year simulation to collect statistics 2. Assimilation system for trace gases (O 3, NO 2, CO) 3. Machine Learning for chemistry solver 4. GOCART and GEOS-Chem coupling in GEOS 40

40 Collaborations & Opportunities Government, Public, NGO, Industry, Research / Mitigation Flight campaign planning 41

41 Thank you! :: 42

42 Compared to other global modelling centers GEOS - CF 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 0 E 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 80 S 60 S 40 S 20 S CAMS - European 0 N 20 N 40 N 60 N 80 N Monday 23 April UTC CAMS Forecast t+000 VT: Monday 23 April UTC Surface ozone [ ppbv ] WACCM - NCAR 43

43 Surface O 3 observations compared to GEOS CF 44

44 Diurnal cycle of surface O 3 is reproduced at the higher resolution 0.25º 2º 45

45 Closeness plot At each station, monthly long hourly data N=744. Calculate the (N G5 )frequency of GOCART with lower biases: G5 t obs t < GCC t obs t -1*N G5 /N*100 (negative shown with blue) Similar calculation of N GCC for GOES-Chem with lower biases. N GCC /N*100 (positive shown with red)

46 Closeness plots GOCART Better GEOS-Chem better

47 GOCART Better GEOS-Chem better Mean biases (G5-GCC) RMSE(G5-GCC) Correlation(GCC-G5)