Comparison of model estimates of background ozone concentrations based on zero-out and source apportionment modeling

Transcription

1 Cmparisn f mdel estimates f backgrund zne cncentratins based n zer-ut and surce apprtinment mdeling Pat Dlwick, Kirk Baker, Nrm Pssiel, Heather Simn, Tyler Fx 7 th biannual AQAST Harvard University June 17, 2014

2 Intrductin / Gal It is well-established that backgrund zne (BGO3) surces and prcesses can influence surface zne cncentratins ver parts f the U.S. Accurate estimates f BGO3 are needed t craft effective cntrl prgrams Clear example f AQAST success in infrming air quality management There are several cmplexities assciated with backgrund zne: Definitins f backgrund can vary Impacts vary greatly acrss space/time Generated by a variety f prcesses Difficult t measure. Typically requires air quality mdeling. Nn-linear zne chemistry cmplicates identificatin f backgrund influence Tw separate methdlgies fr estimating backgrund zne are cmpared in this analysis: Zer ut mdeling Surce apprtinment mdeling 2

3 Tw separate methdlgies fr estimating backgrund EPA has characterized BGO3 fr a 2007 simulatin using tw separate mdel methdlgies which define backgrund differently. Zer ut / emissins perturbatin analyses (ZO): What wuld mdel zne be in the absence f certain sets f zne precursr emissins? Apprach is frequently used t estimate BGO3. When cupled with a glbal mdel, allws flexibility in defining backgrund (NB, NAB, USB). Cunterfactual scenari has limited plicy applicatin. Surce apprtinment analyses (SA): Hw much f the current (mdeled) zne is frmed due t emissins frm certain surces? Apprach is widely used in estimating manmade surce cntributins; less zne used fr BGO3. Assessment f BGO3 is limited t influence f bundaries and ther specified surces within the dmain. Ptentially mre plicy-relevant but substantial pressure n mdel tagging technique. Cmbining results frm tw methdlgies may help expand characterizatin f BGO3. 3

4 2007 Mdeling Cnfiguratin Tw sets f reginal-scale (12km) analyses were cnducted t estimate BGO3 levels at lcatins acrss the U.S. Mdel cnfiguratins were similar between the tw sets: Meterlgical inputs frm WRF mdel. Emissins based n 2008 NEI versin 2; climatlgical-average wildfire emissins. Bundary cnditins based n GEOS-Chem 2.0 x 2.5 degree glbal mdeling. Fur CMAQ zer ut runs were cmpleted using fur separate bundary cnditins Base simulatin Natural backgrund: glbal anthrpgenic precursrs remved + pre-industrial CH4 levels ** N. American backgrund: anthrpgenic precursrs remved in N. America U.S. backgrund: anthrpgenic precursrs remved in United States One CAMx surce apprtinment run was cmpleted using the base bundary cnditins. 11 categries were tracked: 5 bundaries, U.S. anthrpgenic emissins, and 5 in-dmain sectrs (Can/Mex, fires, bigenics, C3 marine, GM platfrms) The CAMx mdeling used the APCA apprach which attributes O3 prductin t manmade surces frm cmbinatins f anthrpgenic & bigenic emissins Mdel perfrmance was evaluated and bth mdels reprduced seasnal mean MDA8 zne levels with little bias (CMAQ: +3.5, CAMx: ** CH levels ppb) were nly mdified in the glbal simulatin, nt the reginal simulatin 4

5 Difference in USB estimates: (as a functin f base mdel zne level) ZO ratis f USB / ttal O3: Prprtin f BGO3 drps as base mdel zne increases. Because f nn-linear zne, is pssible t have ratis > 1.0 SA ratis f nn US / ttal O3: Prprtin f BGO3 drps as base mdel zne increases, even mre steeply than ZO. Nt pssible t have ratis > 1.0 In cases where MDA8 O3 > 70 ppb, typically < 50% f zne is due t emissins ther than US anthr 5

6 Mdel estimates f seasnal mean MDA8 zne frm USB Zer ut mdeling (CMAQ) estimates f USB USB ~ ppb in WUS USB ~ ppb elsewhere Surce apprtinment (CAMx) estimates f apprtinmentbased USB USB ~ ppb in WUS USB ~< 30 ppb elsewhere 6

7 Difference in USB estimates: (unadjusted vs. bias-adjusted) Unadjusted differences between ZO and SA: ZO mdel estimates f BGO3 are 5-10 ppb higher than SA estimates ver SJV and Great Plains/Midwest. Practical impacts f differences are small. Bias-adjusted ** differences between ZO and SA Accunting fr mdel seasnal mean MDA8 O3 biases results in even clser estimates f the influence f nn-u.s. surces ** Bias in the seasnal mean MDA8 is calculated and equally assigned t USB and nn-usb prtins f ttal zne based n USB/ttal rati. 7

8 Difference in USB estimates: (significant gradients acrss an urban area) Once the bias adjustment is applied t accunt fr base CMAQ and CAMx predictins, the greatest differences between the tw estimates ccur in urban areas with peridic NOx suppressin f zne (e.g., Ls Angeles, Bakersfield). 8

9 Cnclusins Tw separate mdeling analyses were cnducted t estimate the influence f surces ther than U.S. anthrpgenic emissins acrss the U.S. Defining backgrund as the O 3 that wuld remain after U.S. anthrpgenic emissins are remved (zer ut) typically results in a higher estimate f USB than when the influence f nn-u.s. surces is estimated via surce apprtinment. Adjusting the results fr differences in base mdel predictins narrwed the differences between the tw methdlgies. Even after bias adjustment, there was a tendency fr the zer ut apprach t estimate higher USB backgrund levels in urban areas with significant NOx. Zer ut apprach reveals latent backgrund The use f tw separate methds blsters cnfidence in the BGO3 estimates, given their generally similar results. 9

10 BGO3 recmmendatins fr AQAST (natinal AQM perspective) Many recent mdeling analyses have characterized influence f BGO3: Zhang et al., (2011): 2006 GEOS-Chem zer ut Emery et al., (2012): 2006 CAMx zer ut Lin et al., (2012): 2010 GFDL AM3 zer ut and tracers WRAP (2013): 2008 CAMx surce apprtinment Lefhn et al., (2014): 2006 CAMx surce apprtinment USEPA (2014): 2007 CMAQ zer ut and surce apprtinment All cnfirm the general nature f BGO3 and have infrmed plicy discussin: Mean values f USB range frm ppb, higher in discrete events BGO3 is particularly prblematic in the intermuntain WUS in spring/summer BGO3 cmprises a larger prtin f ttal zne as U.S. emissins are (have been) reduced Suggestins fr fcus f next generatin analyses: Effrts t quantify individual cmpnents f BGO3 (fire, strat, Asia, methane, etc.) Effrts t quantify BGO3 in future years (influence f emissins/climate changes) Lk fr pprtunites t grund truth mdel estimates f BGO3 w/ bservatins Develpment f mdel tls that allw fr unified hemispheric/reginal mdeling 10