1. Total Deposition (TDEP) Maps Gary Lear US EPA/CAMD Donna Schwede US EPA/AMAD 4/14/20141

2. Quick Review of the hybrid approach Use monitoring data where available –Air concentrations CASTNET (concentration HNO3, SO2, pNO3, pNH4, pSO4) AMoN (concentration NH3) SEARCH (concentration HNO3, SO2, NH3) –Wet Deposition NTN (PRISM adjusted wet deposition NO3 and NH4) Use modeled data from CMAQ (Dry deposition only) –Concentrations fused with measurements to create hybrid grid –Dry deposition velocities applied to hybrid/fused concentration grid to produce dry deposition of measured species –Dry deposition of unmeasured species from CMAQ alone 4/14/20142

3. CMAQ seasonal air concentration data Weekly average dry deposition for each measured species Average seasonal air concentration bias adjustment surface for each measured species Bias-adjusted grids of weekly CMAQ deposition for measured species Annual dry deposition of measured species Annual total deposition Annual NADP/PRIS M wet deposition CMAQ dry deposition Weekly observed air concentrations Weekly average deposition velocity from CMAQ data Weekly merged dry deposition Seasonal observed air concentrations CMAQ Weekly deposition of non-measured species Weekly observed NADP/NTN concentratio n in precipitation Weekly observed NADP/NTN concentratio n in precipitation PRISM- modeled precipitation amount 4/14/20143

4. Current status of maps and data An error was found in the bias adjustment for 2007-2012, so grids and maps were updated Version 2014.01 is available on the ftp site –ftp://ftp.epa.gov/castnet/tdepftp://ftp.epa.gov/castnet/tdep Atmospheric Environment paper (Schwede and Lear, 2014) describing the methodology and results accepted for publication in Atmospheric Environment 4/14/20144

5. Effect of the error in the bias adjustment Illustrative examples 4/14/20145 OLD NEW HNO3 - SppNH4 - FapNO3 - WiSO2 - Su

6. A Review of the Maps Some issues uncovered –Effects of model resolution (36 km vs 12 km) –Effects of temporal variations in availability of monitoring data (e.g. AMoN, SEARCH) –Effects of spatially limited data (e.g. SEARCH) –Effect of the IDW distances (e.g. NH 3 at RMNP) –Bias correction surface is limited by the number of network sites –Emissions inventory error in 2008 in NC 4/14/20146

7. Effects of CMAQ Model Resolution 4/14/20147 2006 (36 km) 2007 (12 km) Higher resolution model output starting in 2007 shows additional detail of local sources

8. Effects of Temporal Variations in Availability of Monitoring Data – NH3 4/14/20148 2008 was AMoN’s first full year Sites coming online may (or may not) cause artifacts in apparent trends 2007 (No AMoN Sites) 2011 (61 AMoN Sites)

9. Effects of Temporal and Spatial Variations in Availability of Monitoring Data – SO2 4/14/20149 2004 (No SEARCH Sites) 2005 (5 SEARCH sites)

10. Effects of Refining IDW Distance of Influence 4/14/201410 Correlation = 0.5; Distance = 200 km Correlation = 0.7; Distance (seasonal) Wi = 41.9; Sp = 109.6; Su = 84.6; Fa = 58.4 NH3_DD 2009 Original Approach New Approach

11. Sparsely Located Monitors Influence the Bias Correction Surface 4/14/201411 Fall pNH4 Summer pNO3 Spring SO2 Summer pSO4

12. Comparisons With Other Approaches These are comparisons, not evaluations –More measurement data needed for evaluations The comparisons have motivated some interesting analyses illustrated in the following slides Results of the analyses suggest additional research areas both in understanding deposition processes and for the hybrid methodology (summarized later) 4/14/201412

13. Comparison With Other Methods CMAQ - Nitrogen 4/14/201413 (d) (b) (c) (a)

14. Summary of Comparison With Other Methods CMAQ - Nitrogen 14 For HNO3, CMAQ concs tend to be higher than CASTNET concs so deposition for CMAQ will be higher than TDEP (slide 13a) NH3 – impact of AMoN in later years contributes to TDEP > CMAQ deposition; also some averaging time issues (sum of hourly values vs using weekly values) (slide 13b) pNO3 and pNH4 size distribution approach in TDEP needs to be improved (slide 13c,d) Compensating biases result in total nitrogen dry deposition values that are similar (this slide) 4/14/2014

15. Comparison With Other Methods CMAQ - Sulfur 4/14/201415 SO 2 Weekly Concentrations 36 km 12 km SO 2 Annual Dry Deposition (a) (b) (c) (e) (d)

16. Summary of Comparison With Other Methods CMAQ - Sulfur 16 Outliers tend to be individual sites and years (slide 15 c-e) Greater bias in CMAQ for years with 36 km grids (slide 15b) Sites such as ORL which are discontinued have a high bias – may indicate an interpolation issue (slide 15e) Highest bias tends to be in the west and the bias surface is strongly influenced by isolated sites (this slide) SO 2 Bias Ratio (Measured/Modeled) Average for Spring 2002-2006 4/14/2014

17. 17 Comparison With Other Methods MLM - Nitrogen

18. 4/14/201418 Deposition velocity is a big driver Differences in model algorithms between CMAQ and MLM Differences in meteorological inputs (on-site vs modeled) Differences in site characteristics For HNO 3 – aerodynamic resistance (Ra) is the controlling resistance (John Walker presentation) For particulate species, concentrations agree better than deposition – points to size distribution assumption MLM models all particulate species with the same vd which is based on the Wesely approximation for pSO 4 (fine) CMAQ uses a modal model, pNO 3 and pNH 4 are not all in the fine mode

19. Site characterization is important … and can be very different … Bondville, IL (BVL) CMAQ 36km (based on 1992 USGS) CMAQ 12km (based on 2006 NLCD) CASTNET (based on site survey in 1 km radius) Dryland Crop and Pasture84.09 Cultivated Crops92.18 Maize50 Urban or Built-up Land8.64 Developed, Open space3.87 Soybean50 Woodland/Cropland Mosaic3.73 Developed, Low Intensity2.62 Grassland/Cropland Mosaic2.05 Pasture/Hay0.94 Savanna1.44 Developed, Medium Intensity0.15 Connecticut Hill, NY (CTH) Broadleaf Deciduous Forest88.70 Deciduous Forest28.01 SUGAR MAPLE,250 Mixed forest4.04 Mixed Forest19.39 GRASS27 Water3.77 Pasture/Hay17.78 MAIZE10 Grassland/Cropland Mosaic3.45 Cultivated Crops11.79 BEECH,18 Urban or Built-up Land0.04 Shrub/Scrub7.25 HEMLOCK5 Rocky Mountain NP (ROM) Shrubland51.66 Evergreen Forest77.93 GRASS33 Grassland20.35 Shrub/Scrub9.70 ROCK*33 Savanna11.39 Grassland/Herbaceous5.44 POND/LODGP PINE,127 Evergreen Coniferous Forest8.17 Deciduous Forest1.61 ASPEN*7 Grassland/Cropland Mosaic3.41 Developed, Open space1.48 Dryland Crop and Pasture2.53 Woody Wetland1.10 Urban or Built-up Land2.31 Pasture/Hay0.83 4/14/201419

20. CASTNET – 1 km radius view CMAQ – 12 km grid view CMAQ – 36 km grid view (9 squares) Google Earth images – ROM site 4/14/201420

21. 4/14/201421 Comparison With Other Methods MLM - Sulfur SO 4 Weekly Concentrations

22. Comparison With Other Methods MLM - Sulfur 4/14/201422 Deposition velocity is a big driver Differences in model algorithms between CMAQ and MLM Differences in meteorological inputs (on-site vs modeled) For SO 2 cuticular resistance may be important –MLM – constant value, depends on pH of wetted surface Dry: rcut = 10000 Dew: rcut = 5000 Rain: rcut = 10000 –CMAQ – dry and wet pathways can operate at same time with fraction assigned by canopy wetness fraction Dry: Wet:

23. Comparison With Other Methods Throughfall Throughfall ≠ Modeled Total Deposition –Different processes are accounted for by the methods (e.g. modeled dry deposition includes stomatal uptake) –We expect that throughfall might be a lower limit of total deposition Limit comparisons to grids where dominant land use is the same as throughfall site (forest) 4/14/201423

24. Comparison With Other Methods Throughfall 4/14/201424 Grid cell dominated by lake Grid cell dominated by forest

25. Comparison With Other Methods Throughfall 4/14/201425 Total N DepositionTotal S Deposition YEARRegionLatitudeLongitudeTDEPTFALLTDEPTFALL 2007WA46.739-121.8162.6260.4701.545 2007WA47.836-123.9813.7490.9402.582 2008C2CCRLAC42.902-122.2796.8730.5001.469 2008CRG45.816-121.9465.9621.2442.600 2008MBS47.539-121.5734.4522.7623.165 2008Sierra38.730-120.5506.8113.0491.017 2008Sierra39.316-120.8497.0997.1761.065 2008Sierra37.530-119.6706.5813.2711.197 2008WIL44.213-122.2532.9850.8971.602 2008WIL44.282-122.0342.6920.6261.446 2011Yosemite37.819-119.7816.8832.7651.4250.982 2011Yosemite37.526-119.6666.3613.0321.2441.153 2011Yosemite37.759-119.8147.3224.6571.4131.691 2011Yosemite37.802-119.8636.9673.0051.4161.345 2011Yosemite37.855-119.6597.0933.0031.5150.693 2011Yosemite37.712-119.7037.0733.8281.3791.143 2011Yosemite37.734-119.6096.3532.8111.2770.944

26. 4/14/201426 Comparison With Other Methods 2 * NTN Wet Deposition TDEP Total N TDEP Total S 2*NTN Total S2*NTN Total N

27. Future Steps for Consideration (Page 1) Characterizing uncertainty Extending estimates before 2000 Incorporating new CMAQ bidi runs –2002-2011 –12 km for all years –NH3 improved –How to fuse with AMoN? Incorporating 1-in-3 networks (IMPROVE, CSN) –Bias correction –Weekly deposition Adjustment for autocorrelation 4/14/201427

28. Future Steps for Consideration – Page 2 Use model for particle size distribution instead of 80/20 rule –pSO4 assumed to be all accumulation mode –pNO3 is assumed to be 80% accumulation mode and 20% coarse mode –pNH4 is assumed to be 80% associated with pSO4 and 20% with pNO3 Adjust deposition to preserve mass balance Fusion on CMAQ and NTN wet deposition –building Dennis and Foley work that uses NTN data to precip adjust and bias correct CMAQ wet deposition Further model evaluation/intercomparison work Development of web page? –CASTNET or NADP Other? 284/14/2014

29. Longer term research suggestions from Fall 2012 NADP meeting 4/14/201429 Project“Volunteers”Status Testing the range of influence with CMAQ and passives Robin, John??? Use 12 km CMAQ vd’s for all years Move to finer scale of deposition Donna, John, RichCMAQ runs will be done this summer Investigate NO 3 VdJohn, Donna, Mike??? Better characterization of organic N Research on other forms of Organic N. Donna, Mike, JohnWork underway Intercomparison of CMAQ/CAMx/AURAMS for 2009 (ROMANS II, more AMoN sites) Donna, Mike, Bret, Eladio, Krish Initial discussions; will use 2011 Inclusion of NH 3 bidirectional fluxDonna, Robin, John, Gary CMAQ data will be available; need to look at data fusion One to Five year variation studyGeorge, Gary Donna???