1. Why do we need atmospheric models?  to get comprehensive source attribution information...we don’t just want to know how much is depositing at any given location, we also want to know where it came from:  different source regions (local, regional, national, global)  different jurisdictions (different states and provinces)  anthropogenic vs. natural emissions  different source types (power plants, waste incin., smelters…)  to estimate deposition over large regions …because deposition fields are highly spatially variable, and one can’t measure everywhere all the time…  to estimate dry deposition... presently, dry deposition can only be estimated via models  to evaluate potential consequences of future emissions scenarios

2. Models are not perfect “…Everyone believes monitoring results except for the person making the measurements… and nobody believes modeling results except for the person doing the modeling…” How not perfect are they? Results are encouraging, but difficult to evaluate models due to lack of contemporaneous monitoring and emissions inventory data More certain info at a few locations (monitoring) vs. less certain info region-wide (modeling) Models are a test of our knowledge… If they don’t work, fundamental things about our understanding of atmospheric mercury that are wrong or incomplete…

3. OctNovDecJanFebMarAprMayJunJulAugSep 20072008 …. Recent Reactive Gaseous Mercury concentrations at the Grand Bay NERR, MS Then down for ~2 months due to hurricanes Air Toxic Phenomena Can be Very Episodic … sporadic measurements can miss events

4. What Do We Need to Know Regarding Atmospheric Mercury? Type of InformationMonitoringModeling Atmospheric deposition* Can give us “exact” answers at a few locations Can give us approximate answers throughout the domain* Source-attribution for deposition For monitoring site only -- using receptor-based techniques & enhanced monitoring Can give us approx. information with suitably designed methodology Deposition for historical periods -- Possible if historical emissions inventories can be estimated Deposition for alternative future scenarios -- “Easy” as long as emissions scenarios are specified * consistent with the needs of subsequent analyses (e.g., ecosystem modeling) with respect to spatial, temporal, and “species” resolution (e.g., Hg(0) vs. RGM vs. Hg(p))

5. atmospheric chemistry phase partitioning Atmospheric Mercury Model wet and dry deposition Wet and dry deposition of different Hg forms to sensitive ecosystems Source attribution information for deposition Model Outputs Speciated ambient concentration data Wet deposition data Model Evaluation Model Inter-comparison Model Visualization meteorology Inputs to Model emissions land use For model evaluation, model inputs must be for the same time period as measurement data

6. 6 We are organizing the initial collaborative work around specific episodes of high concentration of one or more mercury forms

7. Can we learn what is needed about atmospheric mercury deposition by making atmospheric measurements alone? NO…

8. Monitoring needed to provide solid deposition estimates at a given location and for model development and evaluation Modeling needed to help interpret and extend measurements and to estimate source- receptor relationships To get the answers we need, we need to use both monitoring and modeling -- together

9. Atmospheric mercury concentrations and deposition varies with time and location throughout the Gulf of Mexico region  We don’t really know very much about the variation, as there are only a few measurement locations and models have not generally been run in the region to estimate these variations.  Variations will be effected by proximity to sources, meteorological conditions, including but not limited to precipitation, and surface exchange phenomena

10. 2. Dry deposition can be important for mercury, but its very difficult to measure directly (maybe impossible with current methods?)… So, most estimates of dry deposition are developed from models anyway 1. You can’t measure everywhere all the time, and since there can be significant spatial and temporal variations in deposition, its difficult to even estimate the magnitude of deposition from a set of measurements 3. Its essentially impossible to develop information on source-receptor relationships from measurements alone… The simplest modeling approaches (e.g., Receptor-based techniques such as back-trajectory analysis) have substantial limitations – and even if they work perfectly, you only get information for that particular monitoring site… Comprehensive atmospheric fate and transport models have the potential to give us the answers we seek 4. But, there are many uncertainties in modeling, and measurements must be used to “ground truth” the results… …So, need both modeling and monitoring – alone, each is incomplete

11. Measurements can tell us concentrations and deposition at a given location... But, measurements can’t tell us everything we want to know  Concentrations & deposition in the surrounding region -- there might be large spatial gradients -- want information for an entire ecosystem Atmospheric mercury fate and transport models are needed to obtain these types of information  Source attribution and other explanatory information -- where is the mercury coming from? -- why are we seeing what we are seeing?  Impacts of potential future emissions scenarios -- due to alternative domestic regulatory actions -- due to possible international developments We also need:

12. Atmospheric mercury concentrations and deposition varies with time and location throughout the Gulf of Mexico region  Not much known about the spatial and temporal variations –  only a few measurement locations, and very few long-term sites  models not generally run to show these variations.  Variations caused by  changing emissions  proximity to sources  precipitation  wind speed and direction  other meteorological factors  surface exchange phenomena  Monitoring atmospheric mercury is expensive and there are necessarily a limited number of sites…  Impossible to capture the variations by measurements alone.  Thus, impossible to determine the deposition to the Gulf of Mexico ecosystem or sub-ecosystems by measurements alone…

14.  We are generally not actually interested in the concentration or deposition at a single monitoring site…  We are just using the few monitoring sites that we might have to give us a clue as to what the total impact might be (and where it comes from)…  We are interested in the deposition to an entire water body, or to a particular ecosystem If the deposition was constant throughout the ecosystem domain, all we would need is one monitoring site … but there may be large spatial variations in deposition… If you’re far enough away from any local or regional sources that the deposition is relatively constant (in space)… …then you are likely too far away to measure any of the potentially significant impacts of local and regional emissions sources

15. 959697989900010203040506070809101112131415 1995 Canada Inventory 15 speciated atmospheric Hg measurements at site x speciated atmospheric Hg measurements at site y speciated atmospheric Hg measurements at site z 2002 US Inventory 1999 US Inventory 2000 Global Inventory 2000 Canada Inventory New York inventory ? Ontario inventory ? Hypothetical – just for illustration purposes For model evaluation, inventory must be accurate and for same period as measurements (a big challenge!)

16. What do atmospheric mercury fate and transport models need?  Emissions inventories -- speciated -- geographically & temporally resolved -- accurate and complete -- local, regional, national, global -- anthropogenic and natural TRENDS? Emissions inventories are needed – with all these characteristics – to interpret trends in measurements  Meteorological data -- based on measurements and models -- good spatial and temporal resolution needed  Scientific understanding of atmospheric mercury -- atmospheric phase-partitioning and chemistry? -- “real” speciation, i.e., what is “RGM”? -- wet and dry deposition, surface exchange processes -- thus, additional “process” research needed UNCERTAINTIES? There are still a lot of uncertainties in our understanding of atmospheric mercury, but, models have not been able to adequately evaluated to date, so we don’t know how far off we are...  Ambient measurements for “ground truthing” -- speciated concentration measurements are most useful -- wet deposition useful but not as helpful for model evaluation -- also need measurements “aloft”, e.g., tall towers, aircraft UNDERSTANDING MEASUREMENTS? Models can help explain why we see what we see TRENDS? Emissions inventories are needed – with all these characteristics – to interpret trends in measurements UNCERTAINTIES? There are still a lot of uncertainties in our understanding of atmospheric mercury, but, models have not been able to adequately evaluated to date, so we don’t know how far off we are...

17. Atmospheric Models Emissions Inventories To evaluate and improve atmospheric models, emissions inventories must be: Accurate for each individual source (especially for large sources), including variations For the same time periods as measurements used for evaluation For all forms of mercury Mercury Deposition Network (MDN) (wet deposition only) Air Concentration Network Atmospheric Monitoring To evaluate and improve atmospheric models, atmospheric monitoring must be: For air concentrations (not just wet deposition) For all forms of mercury For sites impacted by sources (not just background sites) At elevations in the atmosphere (not just at ground level) ? Understanding and Decisions ? ? ? Source-attribution; Scenarios Understanding Trends Inputs for Ecosystem Models Deposition For Entire Region Incin Manuf Fuel (not coal electric) Hg Dep to Lake Michigan (g/km2-yr) Coal-electric Coal Scenarios

18. Hg from more distant sources emissions of Hg(0), Hg(II), Hg(p) atmospheric deposition to the water surface atmospheric deposition to the watershed SOURCE-ATTRIBUTION How much of the deposition arises from natural vs. anthropogenic? from global vs. national? from local vs. regional? from different source types? how do answers change with time? how do answers change with location? Hg(0) Hg(II) Hg(p) atmospheric chemistry inter-converts mercury forms 18

19. What is an atmospheric model? a computer simulation of the fate and transport of emitted pollutants two different types of models –Eulerian –Lagrangian 19

20. Why do we need atmospheric mercury models?  to get comprehensive source attribution information --- we don’t just want to know how much is depositing at any given location, we also want to know where it came from…  to estimate deposition over large regions, …because deposition fields are highly spatially variable, and one can’t measure everywhere all the time…  to estimate dry deposition  to evaluate potential consequences of alternative future emissions scenarios 20

21.  For areas without large emissions sources  the deposition may be relatively low,  but what deposition there is may largely come from natural and global sources  For areas with large emissions sources  the deposition will be higher  and be more strongly influenced by these large emissions sources... Source-apportionment answers depend a lot on where you are 21

22. some challenges facing mercury modeling emissions inventories need all sources accurately divided into different Hg forms U.S. 1996, 1999, 2003 / CAN 1995, 2000, 2005 temporal variations (e.g. shut downs) meteorological data precipitation not well characterized scientific understanding what is RGM? what is Hg(p)? accurate info for known reactions? do we know all significant reactions? natural emissions, re-emissions? ambient data for model evaluation Mercury Deposition Network (MDN) is great, but: also need RGM, Hg(p), and Hg(0) concentrations also need data above the surface (e.g., from aircraft) also need source-impacted sites (not just background) 22

23. Objectives and Rationale of Atmospheric Modeling in Conjunction with Great Lake Multi-Compartment Mercury Modeling Project  Estimate deposition amount of different mercury species and/or forms to different regions of Lake Ontario lake surface and watershed, for use in ecological assessment and modeling  dry deposition generally estimated with models  modeling can help fill in spatial gaps between measurement sites  modeling can help estimate deposition for other times past future (for different emissions scenarios)  Estimate source attribution for deposition of different mercury species and/or forms to different regions of Lake Ontario lake surface and watershed, including estimation of the relative importance of:  different source regions (local, regional, national, continental, global)  different jurisdictions (different states and provinces)  anthropogenic vs. natural emissions  different anthropogenic source types (power plants, waste incin., etc) 23

24. Atmospheric models can potentially provide valuable deposition and source-attribution information. But… models have not been adequately evaluated, so we don’t really know very well how good or bad they are… … air pollution model or error pollution model? Challenges / critical data needs for model evaluation: Ambient Monitoring Data  speciated ambient concentrations (need RGM and Hg(p), not just total gaseous mercury)  wet deposition Emissions inventories  complete  “accurate”  speciated  up-to-date (or at least for the same period as measurements)  temporal resolution better than annual (e.g., shut-downs, etc) Atmospheric models can potentially provide valuable deposition and source-attribution information. But… models have not been adequately evaluated, so we don’t really know very well how good or bad they are… … air pollution model or error pollution model?

25.  policy development requires:  source-attribution (source-receptor info)  estimated impacts of alternative future scenarios  estimation of source-attribution & future impacts requires atmospheric models  atmospheric models require:  knowledge of atmospheric chemistry & fate  emissions data  ambient data for “ground-truthing”

26. Variations on time scales of minutes to hours  CEM’s needed – and not just on coal-fired power plants  CEM’s must be speciated or of little use in developing critical source-receptor information  Clean Air Mercury Rule only requires ~weekly total-Hg measurements, for purposes of trading We don’t have information about major events  e.g., maintenance or permanent closures, installation of new pollution control devices, process changes  Therefore, difficult to interpret trends in ambient data Temporal Problems with Emissions Inventories Long delay before inventories released  2002 inventory is being released this year in U.S.; till now, the latest available inventory was for 1999  How can we use new measurement data?

27. 27 Hg from other sources: local, regional & more distant emissions of Hg(0), Hg(II), Hg(p) atmospheric deposition to the water surface atmospheric deposition to the watershed

28.  policy development requires:  source-attribution (source-receptor info)  estimated impacts of alternative future scenarios  estimation of source-attribution & future impacts requires atmospheric models  atmospheric models require:  knowledge of atmospheric chemistry & fate  emissions data  ambient data for “ground-truthing”

29. 29 Hg from other sources: local, regional & more distant emissions of Hg(0), Hg(II), Hg(p) atmospheric deposition to the water surface Hg(0) Hg(II) Hg(p) atmospheric chemistry inter-converts mercury forms atmospheric deposition to the watershed Measurement of ambient air concentrations Measurement of wet deposition WET DEPOSITION  complex – hard to diagnose  weekly – many events  background – also need near-field AMBIENT AIR CONCENTRATIONS  more fundamental – easier to diagnose  need continuous – episodic source impacts  need speciation – at least RGM, Hg(p), Hg(0)  need data at surface and above

30. speciated ambient concentration data is scarce  few measurement sites at ground level  very few measurements aloft collaboration between measurement and modeling community is key  measurers need modelers to help interpret data  modelers need measurements to evaluate models therefore, atmospheric mercury models have not really been comprehensively evaluated yet  we don’t really know how good or bad they are

31. Emissions Atmospheric Chemistry Meteorology Transport and Dispersion Wet and Dry Deposition Atmospheric Mercury Model Measurements at specific locations Ambient concentrations and deposition Model evaluation Model results Source attribution

32. Some Current Emissions Inventory Challenges  Re-emissions of previously deposited anthropogenic Hg  Emissions speciation [at least among Hg(0), Hg(II), Hg(p); more specific species if possible]  Reporting and harmonization of source categories  Mobile source emissions?  Enough temporal resolution to know when emissions for individual point sources change significantly Note: Hg continuous emissions monitors now commercially available 32

33. Some Current Atmospheric Chemistry Challenges  Plume chemistry, e.g., rapid reduction of RGM to elemental mercury?  Boundary conditions for regional models?  Oxidation of elemental mercury by O 3 and OH may be over-represented, leading to overestimation of the contribution of global sources to regional deposition Calvert, J., and S. Lindberg (2005). Mechanisms of mercury removal by O3 and OH in the atmosphere. Atmospheric Environment 39: 3355-3367.  Atmospheric methyl-mercury: significance? sources? transport? chemistry? deposition? e.g., Hall et al. (2005). Methyl and total mercury in precipitation in the Great Lakes region. Atmospheric Environment 39: 7557-7569.  Source-Receptor answers influenced by above factors 33

34. Some Model Evaluation Issues Data availability Simple vs. Complex Measurements

35. Data availability A major impediment to evaluating and improving atmospheric Hg models has been the lack of speciated Hg air concentration data There have been very few measurements to date, and these data are rarely made available in a practical way (timely, complete, etc.) Situation may be getting better, largely because of meetings like this!

36. wet dep monitor Simple vs. Complex Measurements: 1. Wet deposition is a very complicated phenomena...  many ways to get the “wrong” answer – incorrect emissions, incorrect transport, incorrect chemistry, incorrect 3-D precipitation, incorrect wet-deposition algorithms, etc.. ambient air monitor  models need ambient air concentrations first, and then if they can get those right, they can try to do wet deposition... ? ? ?

37. monitor at ground level Simple vs. Complex Measurements: 2. Potential complication with ground-level monitors... (“fumigation”, “filtration”, etc.)... monitor above the canopy  atmospheric phenomena are complex and not well understood;  models need “simple” measurements for diagnostic evaluations;  ground-level data for rapidly depositing substances (e.g., RGM) hard to interpret  elevated platforms might be more useful (at present level of understanding) ?

38. Simple vs. Complex measurements - 3. Urban areas: a.Emissions inventory poorly known b.Meteorology very complex (flow around buildings) c.So, measurements in urban areas not particularly useful for current large-scale model evaluations

39. Sampling near intense sources? Must get the fine-scale met “perfect” Ok, if one wants to develop hypotheses regarding whether or not this is actually a source of the pollutant (and you can’t do a stack test for some reason!). Sampling site? Simple vs. Complex Measurements – 4: extreme near-field measurements

40. Complex vs. Simple Measurements – 5: Need some source impacted measurements Major questions regarding plume chemistry and near-field impacts (are there “hot spots”?) Most monitoring sites are designed to be “regional background” sites (e.g., most Mercury Deposition Network sites). We need some source-impacted sites as well to help resolve near-field questions But not too close – maybe 20-30 km is ideal (?)

41. Some Current Model Evaluation Challenges  Lack of speciated atmospheric concentration measurements, at ground level and aloft  Emissions inventory uncertainties, including speciation and temporal resolution – i.e., is the model wrong or is the inventory wrong?  It has not really been possible to adequately evaluate current atmospheric mercury models 41

42. 9. A model does not have to be perfect in order to be useful  Often, most decisions just require qualitatively reasonable results  And realistically, most if not all data and information used in decision-making has uncertainties (e.g., public health impacts, economic impacts)  So, we shouldn’t demand perfection of models

43. 11. MDN is GREAT!…but there are some big gaps in atmospheric monitoring – making it very difficult to evaluate and improve models  We desperately need national MDN-like network to measure ambient air concentrations of Hg0, Hg(p), and RGM, with readily available data  What is RGM? What is Hg(p)?  Both “background/regional” and near-source measurements needed…  Measurements at different heights in the atmosphere

44.  Dry deposition is important, and difficult – if not impossible – to measure reliably with current techniques…  Essentially all dry deposition estimates made currently are made by applying models  National ambient network of speciated ambient measurements will help to evaluate and improve models of dry deposition Dry Deposition?

45. Conclusions Impacts are episodic & depend on form of mercury emitted Source-attribution information is important Modeling needed to get source-attribution information (more!) Monitoring for model evaluation & refinement Many uncertainties but useful model results are emerging Models don’t have to be perfect to give useful information Many opportunities exist for improvements in modeling/monitoring integrated approaches to develop source-attribution information