Impact of lightning-NO and radiatively- interactive ozone on air quality over CONUS, and their relative importance in WRF-Chem M a t u s M a r t i n i.

Impact of lightning-NO and radiatively-

interactive ozone on air quality over CONUS,

and their relative importance in WRF-Chem

M a t u s M a r t i n i

Dale Allen, Kenneth Pickering, Amanda Hansen, Barry Baker

Atmospheric and Oceanic ScienceUniversity of Maryland, College Park, MD

WRF Users’ Workshop, Boulder, CO June 28, 2012

Why is lightning NOx important

Indirectly affects our local air quality and global climate

NOx = NO + NO2

• primary pollutant found in photochemical smog• precursor for tropospheric O3 formation

O3

• is the third most important greenhouse gas• impacts the Earth’s radiation budget• (can cause changes in atmospheric circulation patterns)• is toxic to humans, plants and animals

2

3

Increase in 8-hr O3 due to LNOx

Impact of lightning-NO on Colorado air quality is large (sunny conditions,

good mixing). Mean contribution of LNOx to surface layer ozone during July 2007 was 9 ppbv as diagnosed

by WRF-Chem.

Recently EPA has proposed tightening the air quality standard even further, current NAAQS for ozone is 75 ppbv.

What is the lightning contribution to policy relevant background ozone?

8-hr ozone [ppbv] at surface

New approaches

• WRF-Chem simulations are driven by NASA’s MERRA reanalysis.

• Initial and boundary conditions for chemical species are taken from NASA’s global chemical transport model GMI with combined stratospheric and tropospheric chemistry (two separate GMI simulations, also driven by MERRA, with and without lightning-NO emissions).

• The most recent segment altitude distributions of VHF sources from the Northern Alabama Lightning Mapping Array to best represent the vertical distribution of lightning-NO (with “N-region” peak at height of -15°C isotherm which averages 7.3 km AGL for the eastern U.S. and 5.5 km AGL for mountains).

• A look up table that utilizes convective precipitation and mixed phase depth (indicative of lapse rate) to estimate total flash rates over the CONUS.

• Interactive ozone in longwave and shortwave radiation schemes. 4

Assumed O3 profiles in longwave and shortwave schemes

Longwave scheme (RRTM) uses the average of midlatitude summer and midlatitude winter profiles (WRF default RRTM).

Shortwave scheme (Goddard) uses midlatitude summer for latitudes 30–60°N and tropical O3 profile south of 30°N.

5

Goddard SW: 5 profiles (tropical, midlatitude summer/winter, polar summer/winter)

RRTM schemes: 1 profile prescribed

CAM schemes: 12 monthly profiles

Assumed O3 profiles in radiation schemesvs.

WRF-Chem calculated profiles

O3 generated from LNOx emissions (blue shaded area)

Enhancement of 22 ppbv between 5–12 km

Longwave scheme (RRTM) uses the average of midlatitude summer and midlatitude winter profiles (WRF default RRTM).

Shortwave scheme (Goddard) uses midlatitude summer for latitudes 30–60°N and tropical O3 profile south of 30°N.

6

Single column experiments with offlineRRTM longwave scheme for clear-sky conditions

Midlatitude summer profile

Tropical profile

7

Ozone bandIntegrated LW

Single column experiments with offlineRRTM longwave scheme for clear-sky conditions

Midlatitude summer profileTropical profile

8

Single column experiments with offlineRRTM longwave scheme for clear-sky conditions

Vertical sensitivity of heating rate due to changes in vertical ozone distribution. Adding ozone at an atmospheric layer causes an increase of the heating rate at that level.40-ppbv ozone increments were added to each atmospheric layer. The peak increases of heating rate are normalized to 1 Dobson unit ozone increment.

The most sensitive is the upper troposphere at ~12 km (20% increase if 1 DU ozone added).

Midlatitude summer profileTropical profile

9

WRF-Chem tests

Version 3.2.1

Ten 4-day simulations, reinitialized every 3rd day for July 2007

36 km horizontal resolution, 40 vertical levels up to 50 hPa

Four sensitivity simulations:

1. Standard (no LNOx)

2. With LNOx

3. No LNOx, interactive ozone

4. With LNOx, interactive ozone 

Interactive ozone simulations pass ozone from the chemistry array to both the longwave and shortwave radiation schemes at each radiation time-step.

Non-interactive ozone uses one prescribed ozone profile for each grid point.

No data assimilation

10

Ten cases = one month

WRF-Chem configuration

Atmospheric process WRF/Chem option Reference

Longwave radiation RRTM Mlawer et al. [1997]

Shortwave radiation Goddard Chou and Suarez [1998]

Surface layer Monin-Obukhov Janjic [2002]

Land surface model Noah Chen and Dudhia [2001]

Boundary layer Yonsei University (YSU) Hong et al. [2006]

Cumulus Grell 3D ensemble Grell and Devenyi [2002]

Cloud microphysics Lin Chen and Sun [2002]

Photolysis Fast-J Wild et al. [2002]

Meteorological IC/BC MERRA Bosilovich et al. [2006]

Chemical initial and LBC GMI-CTM Duncan et al. [2008]

Gas-phase chemistry CBM-Z Zaveri and Peters [1999]

Aerosol chemistry MOSAIC 4 bins Fast et al. [2006]

11

Stage IV NCEP WRFNARR WRFMERRA

6 hour Accum [mm]

Impact of different meteorological IC/BC

(July 8, 2007 snapshots)

T =

54

ho

ur

T =

42

hou

r

Precipitation: 6 hour accumulation [mm] 12

13

Time series of daily flash rate over CONUS

Look-up table (Hansen et al. 2011) that uses mixed phase depth (measure of

lapse rate) and convective precipitation is prone to model biases (precipitation

and temperature) and vertical resolution.

Advantage: no need to scale the modeled flash rate to observed.

We use combination of look-up table and scaling – slightly better day-to-day

variation (correlation of 0.51 vs. 0.49) compared to approach of

Allen and Pickering [2002].Overestimation with respect to OMI NO2 colums (DOMINO).IC/CG ratio likely overestimated.

Flash rate based on Hansen et al. [2011] look up table

14

Ozone enhancements from LNOx

(Difference between two simulations with LNOx and without LNOx emissions)

O3 [ppbv] at 300 hPa

GMI CTM –> WRF-Chem

2° x 2.5° –> 36 km

Both driven by MERRA reanalysis.

Pre

ssur

e [h

Pa]

WRF-Chem

15

No LNOx

LNOx

Impact of interactive ozone on OLR

(Difference between interactive O3 simulation and climatological O3 simulation)

Need to compare to observed OLR.Fast et al. [2006] showed that simulated SW radiation was 30–40 Wm-2 closer to observations when aerosols were coupled, a factor of 10 smaller effect of ozone.

16

Outgoing longwave radiation due to LNOx

(Difference between two simulations with LNOx and without LNOx emissions)

Ozone generated from LNOx reduces the OLR by 0.22 Wm-2 during the month of July 2007 on average, 0.43 Wm-2 for clear sky in a fully coupled framework.

Martini et al. [2011] showed values of 0.20–0.50 Wm-2 for summer 2004 in offline calculations for clear sky.

17

Temperatureat 200 hPa

Ozoneat 200 hPa

Impact of interactive ozone

(Difference between interactive O3 simulation and climatological O3 simulation)

We see temperatures cooler by 0.1°C, because there is much less modeled O3 (tropopause at higher altitude) than assumed climatology in longwave scheme.

18

Ozone Temperature Temp Bias

When LNOx emissions are in place interactive O3 improves temperatures above 200 hPa.

Observations with s bars

19

Ozone TemperatureTemp Bias

Observations with s bars

When LNOx emissions are in place interactive O3 has improves temperature above 200 and below 600 hPa.

Summary

• The heating rates are most sensitive the upper troposphere around 12 km, with no impact below 5 km.

• Back of envelope calculation: Impact on O3 can be 1 ppbv. Climate penalty factor ~2.2 ppbv/K [Bloomer et al. 2009] and temperature response, determined from climate simulations, of 0.3 K per 18-ppbv O3 enhancement [Mickley et al. 2004].

• Impact of using modeled O3 in LW scheme on OLR: 3.2 Wm-2 on average with 4.0 Wm-2 for clear sky. Fast et al. [2006] showed that simulated SW radiation was 30–40 Wm-2 closer to observations when aerosols were coupled, a factor of 10 smaller effect of ozone.

• Ozone generated from LNOx reduces the OLR by 0.22 Wm-2 on average, 0.43 Wm-2 for clear sky in a fully coupled framework. Martini et al. [2011] showed values of 0.20–0.50 Wm-2 in offline calculations for clear sky. Compared to aerosols, a factor of 100 smaller effect of ozone from LNOx.

• Need realistic ozone in radiation schemes for longer simulations.

20

• Mean O3 enhancement from LNOx is 22 ppbv, but more important is where the tropopause is located. Impact on UT temperatures is immediate (0.1 K).

• Comparison with Beltsville ozonesonde is encouraging (good agreement with LNOx simulation), temperature biases are slightly decreased when interactive ozone is used.

• Initial and boundary conditions are important. Convection was better represented in simulation driven by MERRA reanalysis than in simulation driven by NARR. Moreover, convective system entering WRF domain is captured only by the simulation driven by MERRA.

• Hansen [2011] look-up table that uses mixed phase depth (measure of lapse rate) and convective precipitation is prone to model precipitation biases and vertical resolution. Advantage: no need to scale the modeled flash rate to observed flash rate. Day-to-day variations are slightly improved (correlation of 0.51 vs. 0.49) compared to approach of Allen and Pickering [2002].

21

Summary 2