A Multi-resolution Assessment (cmaq) Printer-friendly Version Interactive Discussion Geoscientific Model Development Discussions a Multi-resolution Assessment of the Community Multiscale Air Quality (cmaq) Model V4.7 Wet Deposition Estimates for 2002–2006 Gmdd a Multi-resolution Assessment (cmaq) Pr

This discussion paper is/has been under review for the journal Geoscientific Model Development (GMD). Please refer to the corresponding final paper in GMD if available. Abstract This paper examines the operational performance of the Community Multiscale Air Quality (CMAQ) model simulations for 2002–2006 using both 36-km and 12-km horizontal grid spacing with a primary focus on the performance of the CMAQ model in predicting wet deposition of sulfate (SO = 4), ammonium (NH + 4) and nitrate (NO − 3). Per-5 formance of the wet deposition species is determined by comparing CMAQ predicted concentrations to concentrations measured by the National Acid Deposition Program (NADP), specifically the National Trends Network (NTN). For SO = 4 wet deposition, the CMAQ model estimates were generally comparable between the 36-km and 12-km simulations for the eastern US, with the 12-km simulation giving slightly higher esti-10 mates of SO = 4 wet deposition than the 36-km simulation on average. The normalized mean bias (NMB) was slightly higher for the 12-km simulation, however, both simulations had annual biases that were less than ±15% for each of the five years. The model estimated SO = 4 wet deposition values improved when they were adjusted to account for biases in the model estimated precipitation. The CMAQ model underestimates NH + 4 15 wet deposition over the eastern US using both the 36-km and 12-km horizontal grid spacing, with a slightly larger underestimation in the 36-km simulation. The largest un-derestimations occur during the winter and spring periods, while the summer and fall have slightly smaller underestimations of NH + 4 wet deposition. Annually, the NMB generally ranges between −10% and −16% for the 12-km simulation and −12% to −18% 20 for the 36-km simulation over the five-year period for the eastern US. The underestimation in NH + 4 wet deposition is likely due, in part, to the poor temporal and spatial representation of ammonia (NH 3) emissions, particularly those emissions associated with fertilizer applications and NH 3 bi-directional exchange. The model performance for estimates of NO − 3 wet deposition are mixed throughout the year, with the model 25 largely underestimating NO − 3 wet deposition in the spring and summer in the eastern US, while the model has a relatively small bias in the fall and winter. Model estimates of NO − 3 wet deposition tend to be slightly lower for the 36-km simulation as compared 2316 to …


Introduction
Atmospheric deposition of sulfur and nitrogen cause deleterious impacts on terrestrial and aquatic ecosystems due to acidification and excess nutrients (Lovett and Tear, 2008;Driscoll et al., 2001Driscoll et al., , 2003;;Fenn et al., 2003).Sulfur deposition from SO 2 and SO = 4 emissions contributes to acidification and nitrogen deposition from nitrogen oxide (NO x ) and ammonia (NH 3 ) emissions contribute to acidification and excess nitrogen nutrients.Estimates of wet and dry deposition of nitrogen and sulfur are needed for sensitive ecosystems, as total deposition estimates are used to assess whether current or projected pollutant levels exceed a point where significant harmful effects on sensitive elements of the environment are likely to occur (Geiser et al., 2010).Monitoring of wet deposition is relatively sparse and monitoring of dry deposition is extremely sparse, contributing to significant interpolation errors when these data are used to estimate deposition in unmonitored areas.Thus, a regional air quality model, like the Community Multiscale Air Quality (CMAQ; Byun and Schere, 2006) model, can be used to provide a more spatially complete estimate of total deposition to the sensitive ecosystems.However, the model estimates must first be evaluated to establish the credibility of the model in replicating the observed wet deposition.Introduction

Conclusions References
Tables Figures

Back Close
Full Evaluating the ability of the air quality model to replicate observed net (wet + dry) deposition is difficult.The National Atmospheric Deposition Program (NADP) monitoring sites provide the most complete spatial coverage of observed wet deposition across the US on a temporal scale suitable for air quality model evaluations.Evaluation of dry deposition is even more challenging because monitoring network (e.g.Clean Air Status and Trends Network) dry deposition levels are based on modelled values of deposition velocity and, hence, are not a true measure of dry deposition.Therefore, this work focuses on wet deposition to provide a test of the ability of the model to mix, transport, transform and scavenge the pollutant emissions at the regional scale.Many sensitive ecosystems are in complex terrain where orographic effects influence the precipitation patterns and consequently wet deposition.Thus, quantifying precipitation biases as part of the wet deposition evaluation is critical.
This paper examines the performance of the CMAQ model sulfate (SO = 4 ), nitrate (NO − 3 ) and ammonium (NH + 4 ) wet deposition estimates for the 2002-2006 period over the continental United States (CONUS) using two model grid-spacing options, namely 12-km and 36-km grid spacing.The performance of the CMAQ model estimates is examined temporally using various averaging periods (i.e., monthly, seasonal, annual and multi-annual) and spatially across different regions, as the model performance can vary significantly in space.In cases where deficiencies in model performance are identified, model improvements, such as the production of NO x from lightning and the inclusion of bi-directional flux of NH 3 , are tested and their impacts on model performance assessed.Together, these analyses provide insight into the strengths and weaknesses of the CMAQ model in estimating wet deposition of sulfur and nitrogen to sensitive ecosystems.Figures

Back Close
Full 2 Input data and model configuration

Meteorology
The CMAQ model requires gridded meteorological data to provide estimates of various meteorological parameters such as temperature, wind speed and direction, relative humidity and planetary boundary layer (PBL) height.The 5th generation Mesoscale Model (MM5; Grell et al., 1994) is a Eulerian meteorological model that provides estimates of the meteorological parameters required by the CMAQ model and has been used and tested extensively with the CMAQ model over the past 15 years.For this work, the MM5 version 3.7.4 was used for both the 36-km and 12-km simulations.The 36-km MM5 domain consists of 165 by 129 grid cells covering the entire CONU, and includes portions of Canada and Mexico.The 12-km domain consists of 290 by 251 grid cells covering the eastern two-thirds of the US, southern Canada and northern Mexico.

Emissions
The 2002 National Emissions Inventory (NEI) version 3 was used as the primary basis for the 2002-2006 emissions inputs.Version 3 of the 2002 NEI is documented at http://www.epa.gov/ttn/chief/net/2002inventory.html#documentation.For the major point sources, namely electric generating units (EGUs), year specific continuous emission monitoring systems (CEMS) data were used.Year specific updates to mobile emissions were done using the MOBILE6 model and daily estimates of fire emissions based on satellite detection of fires were included as well.NH 3 emissions from agricultural cropping practices in CMAQ are provided by a separate model based on the Carnegie Mellon University (CMU) ammonia emission model (Goebes et al., 2003), which are then combined with the NEI.Monthly NH 3 emissions from livestock were adjusted according to the inverse-modelling recommendations of Gilliland et al. (2006).
For inventories outside of the US, which include Canada, Mexico and offshore emissions, the latest available base year inventories were used.The CMAQ model-ready emissions were created using the Sparse Matrix Operator Kernel Emissions (SMOKE) modelling system (Houyoux et at., 2000).

CMAQ model configuration
The CMAQ simulations were performed at the 36-km horizontal grid spacing for the CONUS, while for the eastern two-thirds of the US a CMAQ simulation using 12-km horizontal grid spacing was performed.Chemical boundary conditions for the 12-km simulation were provided by the 36-km simulation, while boundary conditions for the 36-km CMAQ simulation were provided by a non-year specific GEOS-Chem (Bey et al., 2001)  The air quality simulations utilized CMAQv4.7 (Foley et al., 2010), the latest version of the model available at that time.The simulations included a 10-day spin-up period for the 36-km simulations, while a 3-day spin-up period was used for the 12-km simulations.The CMAQ simulations were performed using the same horizontal dimensions as their respective meteorology simulation except that the horizontal dimensions were reduced by five grid cells on each of the four lateral boundaries to avoid artifacts that can appear along the domain boundaries in the meteorological simulations.However, unlike the meteorological simulations which utilized 34-vertical layers, the CMAQ simulations used 24-vertical layers.The CMAQ model simulations used the AERO5 aerosol module (Carlton et al., 2010), the Carbon-Bond 05 (CB05) chemical mechanism with chlorine chemistry extensions (Yarwood et al., 2005) and the ACM2 PBL scheme (Pleim, 2007a, b).

Assessing model performance
The The NTN consists of approximately 185 sites in the eastern US (east of 110  Appel et al., 2010), are available for download through the CMAS website.

Precipitation bias adjustment
At least some portion of the error present in the CMAQ estimated wet deposition is due to errors in the precipitation estimates from the meteorological model.Since both the NTN observed and MM5 estimated precipitation data are available for each NTN site, the modelled wet deposition can be adjusted to account for the error present in the model estimated precipitation.This adjustment is accomplished here by linearly adjusting the CMAQ estimated wet deposition by the ratio of the observed to estimated precipitation (see Eq. 1).For example, in the case where the observed precipitation is greater than the model estimated precipitation, the ratio is greater than one and, therefore, the model estimated wet deposition is increased.The precipitation adjustment technique assumes that the observed to modelled precipitation ratio is well correlated with the observed to modelled deposition ratio.In other words, it is not assumed that the wet deposition scales linearly with precipitation, but only that the relationship between the errors in the model precipitation estimates and the error in the CMAQ deposition estimates is linear.Since the bias adjustment was applied over the aggregated seasonal and annual totals, there were no instances in which the observed precipitation was greater than zero while the model estimated precipitation was zero.However, in instances where there is observed precipitation but no model predicted precipitation, the current method of bias adjustment would keep the model estimated wet deposition zero for all species.The impact of the precipitation bias adjustment on model performance will be presented for each of the wet deposition species.

Assessment of CMAQ wet deposition performance
In order to provide a comprehensive assessment of the CMAQ wet deposition estimates, several different types of analyses will be presented.The performance of the model estimates are assessed on several time scales, including monthly, seasonally, annually and finally a multi-annual assessment of model performance.The performance for the 36-km and 12-km CMAQ simulations will be compared to examine how similar or dissimilar the model estimates are for a given time period.Since the 12-km CMAQ domain only covers the eastern two-thirds of the US, comparison to the 36-km results will be limited to the same geographic region (herein referred to as 36-km East).
Results for the western one-third of the US will be limited to estimates from the 36-km CMAQ simulation (herein referred to as 36-km West) only, since no 12-km model data are available for the western US for the current analysis.The model estimates will also be examined spatially to identify regional biases.

Summary of precipitation performance
Simulated precipitation is a critical driver in the performance of the CMAQ-simulated wet deposition estimates, especially since large biases in model estimated precipitation can translate into biases in the CMAQ model estimates.Tables 1 and 2 present seasonal and annual normalized mean bias (NMB) and root mean square error (RMSE) for precipitation for the 12-km, 36-km East and 36-km West domains for the five years simulated.For the eastern US, the precipitation bias and error are lowest in the winter (December, January and February) and spring (March, April and May) seasons, when the majority of the precipitation is on the synoptic scale (i.e.large-scale frontal systems) and can generally be well resolved by the model.In the summer (June, July and August) and early fall (September, October and November), a large amount of the precipitation is sub-grid scale convective rain, which meteorological models tend to have difficultly representing accurately through the various parameterizations, which results in higher precipitation biases in those seasons.See Fig. S1 in the supplemental data for spatial plots of the NTN observed and MM5 estimated annual precipitation (12-km simulation only).
While the precipitation estimates for the 12-km and 36-km East simulations have similar patterns in their bias, the precipitation estimates for the 12-km simulation are consistently higher than those of the 36-km East simulation (indicated by systematically larger NMB values), which results in slightly larger biases in the winter, spring and summer for the 12-km simulation, but a smaller bias in the fall when precipitation is underestimated in both simulations.The bias and error in precipitation tend to be larger for the western US (based on the 36-km West simulation) than for the eastern US.The large bias is especially evident in the summer, when precipitation is grossly overestimated in the 36-km West simulation (summer average NMB = 54.5% for the Figures

Back Close
Full simulations was typically slightly larger than the 12-km East simulation, with annual NMB generally ranging between ±11% for the five year period.

SO
= 4 wet deposition Model estimates from both the 12-km and 36-km simulations capture the seasonal trends in the observed monthly accumulated (accumulated over all sites) SO = 4 wet deposition for the 2002-2006 period, with the estimates from the 12-km CMAQ simulation consistently higher than those from the 36-km East simulation (Fig. 1).The CMAQ model generally overestimates SO = 4 wet deposition in the eastern US, with the 12-km simulation overestimating SO = 4 wet deposition for 50 of the 60 months, while the 36-km East simulation overestimates SO = 4 wet deposition for 33 of the 60 months.However, 88% of the estimates from the 36-km East simulation and 80% of the estimates from the 12-km simulation have a NMB of less than ±15% (Fig. 2).The largest overestimations of SO = 4 wet deposition occur in the late fall and winter, generally between October and March.
Overall, the bias in SO = 4 wet deposition estimates for the eastern US was relatively small for both the 12-km and 36-km East simulations (Table 3).The bias for the 12-km (36-km East) CMAQ simulation is highest in the winter, with the annual NMB ranging from 8.1% (−0.8%) to 30.7% (23.1%) and a five-year average NMB of 17.2% (9.0%).However, SO = 4 wet deposition is relatively small in the winter compared to the other seasons, so RMSE values in the winter are lower than the other seasons (Table 4).The NMB is smallest in the summer, ranging from 1.7% to 14.5% for the 12-km simulation (five-year average NMB = 5.2%) and 0.0% to 9.3% for the 36-km East simulation (fiveyear average NMB = −3.5%).The RMSE is largest in the summer, with annual RMSE values ranging between 1.6-2.1 kg/ha for the two simulations.Bias in the spring and fall periods generally falls between the performance for the summer and winter.The average annual NMB (RMSE) for the five-year period was 7.9% (3.56 kg/ha) for the Introduction

Conclusions References
Tables Figures

Back Close
Full The SO = 4 wet deposition performance for the western US is considerably worse than for the eastern US, with the NMB exceeding 40% in 18 of the 60 (30%) months (Fig. 2).This result is not surprising given the challenging meteorological (recall the large precipitation biases in the western US) and air quality conditions that exist in the western US due to its complex topography.Also note that SO = 4 wet deposition in the western US is an order of magnitude less than that in the eastern US (Fig. 1), which may also contribute to the larger normalized bias.As was the case for the eastern US, the poorest model performance for the western US was in the winter, which had an average NMB of 31.6% (RMSE = 0.28 kg/ha) for the five-year period, while the summer had the best model performance, with a five-year average NMB of just 1.9% (RMSE = 0.25 kg/ha).The model bias was slightly higher in the spring (24.3%) than the fall (13.9%).The average NMB for the entire five-year period was 18.9% (RMSE = 0.82 kg/ha).Given the complexity of the terrain over much of the western US, a simulation utilizing finer grid spacing (e.g.12-km) may result in improved performance, as some of the finer details of the topography would be captured in the modelling system.
Annual SO = 4 wet deposition is highest in the eastern half of the US where the largest SO 2 emissions occur (see Fig. S2 in the supplemental data).The highest amounts of The change in annual SO = 4 wet deposition model bias as a result of applying the precipitation bias adjustment described in Sect.2.5 for the 12-km simulation is shown in Fig. 3, which indicates at least some improvement in model bias for each of the five years by applying the precipitation bias adjustment.However, the improvement varies significantly from year to year, with the largest improvement in model performance for 2002 (annual NMB decreases from 21% to 2%), while for 2003 and 2006 the NMB improves by 3% or less.Spatially, the largest precipitation bias typically occurs in the Northeast and Great Lakes regions (particularly in 2002), and those regions show the largest improvement in bias and error as a result of the adjustment for precipitation bias (see Figs. S3 and S4 in the supplemental data for regional statistics).
A bootstrap sampling technique was used to test the robustness of the precipitation bias adjustment.For each year, the NTN observations were re-sampled with replacement 1000 times.The sample size for each of the 1000 samples matched the number of observations available for that year.The base model SO  minimum in the winter (Fig. 5).Also similar to SO = 4 wet deposition, the NH + 4 wet deposition bias for the eastern US is largest in the summer (Fig. 5).However, unlike the SO = 4 wet deposition, the peak underprediction in NH + 4 wet deposition in the eastern US typically occurs in late spring and early summer (April-June), whereas the underestimation in SO = 4 wet deposition typically peaks in the mid to late summer period.For the western US, NH + 4 wet deposition is more often underestimated than overestimated (Fig. 5), however, there are several months, particularly in the spring and fall seasons, with large NMB (Fig. 6).
The largest bias in NH + 4 wet deposition for the eastern US occurs in the spring, with a five-year annual average NMB of −19.9% (RMSE = 0.38 kg/ha) and −23.6% (RMSE = 0.38 kg/ha) for the 12-km and 36-km East CMAQ simulations, respectively (Tables 5 and 6).Conversely, the spring season has the smallest bias for the western US, with an average NMB of just −3.4% (RMSE = 0.20 kg/ha).The winter has a relatively large bias for both the eastern and western domains, with an average NMB of −13.6% (RMSE = 0.17 kg/ha) and −17.5% (RMSE = 0.15 kg/ha) for the 12-km and 36-km East simulations, respectively, and an average NMB of −37.1% (RMSE = 0.15 kg/ha) for the western US.The NMB for the summer and fall periods is similar for the eastern US and generally ranges between −2.0% to −20.0% across the five years.Overall, for the five-year period NH + 4 , wet deposition is underestimated with the five-year average NMB ranging from −12.8% to −15.7% for the three simulations.
Spatially, the highest observed annual NH + 4 wet deposition occurs in the mid-Atlantic, Great Lakes, Mid-West and portions of Northeast (Fig. S5 in the supplemental data).The CMAQ model estimates the highest annual NH 4 wet deposition generally results in an increase in bias (Fig. 7) and a slight increase in error (Fig. 8) for each of the five years.This suggests that the overestimation in model-estimated precipitation is at least partially compensating for an underestimation in NH + 4 wet deposition.The increase in bias is largest in 2002, where the NMB increases from −3% to −19%, while for the other years the increase in bias is smaller, generally ranging from 3% to 7% (see Fig. S6 in the supplemental data).It is important to note that the NH 3 emissions used in the CMAQ model simulation are constrained using the results of inverse modelling, so some increase in NH + 4 wet deposition bias is expected when the model estimates are adjusted for precipitation bias.
The underestimation in NH + 4 wet deposition may be due, in large part, to the poor temporal and spatial representation of NH 3 emissions, particularly those emissions associated with fertilizer applications and bi-directional exchange of NH 3 from soil and vegetation surfaces.In order to improve the NH 3 emissions, a bi-directional NH 3 exchange mechanism was developed for the CMAQ model which was in turn coupled with an agricultural management tool and a soil nitrogen geochemical cycling model to estimate NH 3 emissions from fertilized croplands (Cooter et al., 2010) bias (Fig. 9).Including the bi-directional exchange which significantly reduces the bias in the precipitation corrected annual NH + 4 wet deposition, with the NMB reduced more than a factor of three (from −19% to −6%).The reduction in the model bias was due to improving the temporal resolution of NH 3 emissions from a monthly profile to hourly, representing grid cell level spatial variability instead of county level and modelling the soil nitrification, de-nitrification, vegetative uptake and soil evasion of NH 3 following fertilizer application rather than using state level fertilizer sales as a surrogate for emissions.It is anticipated that a beta version of the bi-directional NH 3 exchange will be available for the next version of the CMAQ model.

NO − 3 wet deposition
The NO − 3 wet deposition performance is dominated by large underestimations in the summer (Fig. 10), which is consistent with the performance of CMAQ model estimates of aerosol fine particulate NO − 3 (Appel et al., 2008).The CMAQ model estimates of NO − 3 wet deposition for the fall and winter seasons are relatively consistent for the eastern US, with the NMB ranging between ±20% for both the 12-km and 36-km East CMAQ simulations (Fig. 11).In the spring, NO − 3 wet deposition is underestimated in the eastern US, with an average NMB of −14.5% (RMSE = 0.88 kg/ha) and −22.6% (RMSE = 0.95 kg/ha) for the 12-km and 36-km East CMAQ simulations, respectively (Tables 7 and 8).For the western US the NMB is unbiased in the spring.For the summer, the NO − 3 wet deposition is largely underestimated for both the eastern and western US, with a NMB greater than −40% for all three domains, while the RMSE is roughly 1.5 kg/ha for the eastern US and 0.5 kg/ha for the western US.For the entire five-year period, the model underestimates NO − 3 wet deposition with a five-year average NMB of −14.9% (RMSE = 2.54 kg/ha) and −21.4% (RMSE = 2.70 kg/ha) for the 12-km and 36-km East simulations, respectively, and a NMB of −6.9% (RMSE = 1.00 kg/ha) for the 36-km West simulation.Introduction

Conclusions References
Tables Figures

Back Close
Full As was the case with the NH + 4 wet deposition, applying the precipitation bias adjustment to the NO − 3 wet deposition model generally estimates results with an increase in the bias (Fig. 12) and either a slight increase or decrease in error (Fig. 13) for each of the five years (also see Fig. S8 in the supplemental data).One large source of the underestimation of NO − 3 wet deposition is from a lack of lightning generated NO.
Lightning can be a large source of upper tropospheric NO, especially in the summer when lightning activity is high and can contribute significantly to NO − 3 wet deposition (Fang et al., 2010).The lack of NO produced from lightning is less of a problem in the western US, as lightning activity is generally much lower west of the Rocky Mountains as compared to the eastern US.In the base simulations performed here, no lightning generated NO emissions were included in the emissions inventory.In order to estimate the impact of lightning generated NO on NO − 3 wet deposition, this source was added to the CMAQ model simulation using the process described below.
The lightning NO production is calculated using the convective precipitation rate from the meteorological model in order to ensure that the lightning is co-located with clouds, convection and precipitation.A more complete description is available in Allen et al. (2009), but briefly, first the flash frequency is calculated as a function of the convective precipitation rate.Then, for each grid cell, the flash frequency is normalized so that the monthly sum of the modelled flash counts is equal to the monthly sum of Introduction

Conclusions References
Tables Figures

Back Close
Full the flashes observed by the National Lightning Detection Network (NLDN), where the NLDN cloud-to-ground (CG) flash rates are multiplied by Z+1 to account for the contribution of intracloud flashes (IC) to the total flash rate.Z is the climatological IC/CG ratio from Boccippio et al. (2001).This method captures the day-to-day variability in flash rates, while retaining an accurate estimate of the monthly total (Allen et al., 2009).For each flash, it is assumed that 500 moles of NO are produced (DeCaria et al., 2005;Ott et al., 2007), which is a reasonable mid-latitude value.The NO is vertically distributed from the surface to the model layer containing the convective cloud top using climatological vertical flash rate information from the Northern Alabama Lightning Mapping Array (Koshak et al., 2004).
For the summer of 2004, a CMAQ model simulation using 36-km grid spacing was performed for the CONUS that included lightning produced NO as described above.Over the entire summer, NO produced from lightning was equal to 30% of the anthropogenic NO emissions.Because most of the NO produced from lighting is created in the upper troposphere, the impact to surface concentrations is small, as in Kaynak et al. (2008).However, over the eastern US where lightning flash counts are greatest, the impact to NO − 3 wet deposition is substantial.Figure 14 shows the bias in NO − 3 wet deposition at NADP monitoring sites for the CMAQ simulation without lightning NO, including lightning NO, and including lightning NO and the precipitation bias adjustment.For the monitoring locations east of 100 degrees W longitude, the CMAQ simulation with the lightning NO production has a low bias and captures the range of variability shown at the surface monitors.At the monitors west of 100 degrees W longitude, the impact is small and the bias persists, owing to the low lightning flash counts in this region.

Summary
The CMAQ modelling system was used to estimate SO US using a 12-km grid spacing.The resulting wet deposition estimates from the model were compared with surface based observations of wet deposition species available across the US from the NTN for the five-year period.For SO = 4 wet deposition, the operational performance of the CMAQ model estimates were generally comparable for the 36-km and 12-km simulations for the eastern US, with the 12-km simulation on average yielding slightly higher estimates of SO = 4 wet deposition than the 36-km simulation.When compared to observations from the NTN, the NMB for the CMAQ model estimates was slightly higher for the 12-km simulation, however, both simulations had annual NMB that were less than ±15% each year.Bias and error in the model SO = lightning in the upper troposphere, which can be a large source of NO, particularly in the summer in the eastern US when lightning activity is the high.CMAQ model simulations, that include the production of NO from lightning, show a substantial reduction in the NO − 3 wet deposition underestimation in the eastern US in the summer as compared to simulations without lightning NO.There is little impact on bias in the western US when lightning generated NO is included due to the relatively low amount of lightning activity in the western US.
Overall, the performance for the 36-km and 12-km CMAQ model simulations was similar for the eastern US, while for the western US the performance of the 36-km simulation was generally not as good as either eastern US simulation.On an annual basis, the model performance for all three wet deposition species was relatively consistent (NMB <30%), with mostly small variations in normalized bias (standard deviation <3%) over the five-year period for the eastern US.Annual variations in NMB were larger for the western US, with a standard deviation >5.5%.This suggests that the modelling system handles the year-to-year variability relatively well in meteorology and emissions that occur over longer periods of time, particularly for the eastern US.
As annual air quality model simulations become more routine, it is likely that the fiveyear performance assessment presented here could be extended to cover a longer time-period (e.g. a decade).Additionally, expanding the 12-km simulation to include the western US may result in improved model performance over the 36-km simulation given the complexity of the terrain in the western US.Clim., 46, 1396Clim., 46, -1409Clim., 46, , 2007b.Pleim, J. E. and Xiu, A.: Development and testing of a surface flux and planetary boundary layer model for application in mesoscale models, J. Appl.Meteor., 34, 16-32, 1995.Reisner, J., Rasmussen, R. M., and Bruintjes, R. T.: Explicit forecasting of supercooled liquid water in winter storms using the MM5 mesoscale model, Q.J. Roy.Meteor. Soc., 124, 1071-1107, 1998.Xiu, A. and Pleim, J. E.: Development of a land-surface model.Part I: application in a mesoscale meteorological model, J. Appl.Meteor., 40, 192-209, 2001.Yarwood, G., Roa, S., Yocke, M., and Whitten, G.: Updates to the carbon bond chemical Introduction

Conclusions References
Tables Figures

Back Close
Full    Full  Full  Full      Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | simulation.The boundary data for the 36-km CMAQ simulation were created by taking the median value of a 2.0 degree by 2.5 degree (latitude-longitude) 24-vertical layer 2002 GEOS-Chem simulation and averaging the three-hourly data to monthly values.These monthly averages were then used as boundary conditions for all five years of the 36-km CMAQ model simulations.Discussion Paper | Discussion Paper | Discussion Paper | assessment of the CMAQ model's wet deposition estimates is accomplished by comparing the simulated wet deposition estimates to observed wet deposition values available from the National Acid Deposition Program's (NADP; http://nadp.sws.uiuc.edu) National Trends Network (NTN).The NTN measures total weekly wet deposition of several atmospheric pollutants, including SO = all of the SO 2 in rainwater is oxidized to SO = 4 by the time the samples are analysed for the NTN (high prevalence of oxidants), the CMAQ estimates of SO = 4 wet deposition include 150% of the model estimated SO 2 wet deposition to account for the SO 2 captured in the observations.Because in solution the favoured phase of NH 3 is NH Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 12-km simulation and 0.8% (3.10 kg/ha) for the 36-km East simulation, indicating SO = 4 wet deposition is generally overestimated, although only very slightly in the 36-km East simulation.
SO = 4 wet deposition occur in the Ohio Valley and Great Lakes regions and stretching into parts of the Northeast.While these spatial features are well captured by the CMAQ model for all five years, the model tends to overestimate the annual SO = 4 wet deposition in the Ohio Valley region, with some model estimates exceeding 27 kg/ha in areas where observations indicate annual SO = 4 wet deposition of 19-20 kg/ha.The model also underestimates the SO = 4 wet deposition along parts of the coast of the Gulf of Mexico, although to varying degrees throughout the five-year period.Overall, the model captures the spatial variations in annual SO = 4 wet deposition.Discussion Paper | Discussion Paper | Discussion Paper |

=
4 wet deposition estimates and precipitation bias corrected model estimates were matched to these pseudo-sets of observations, and the Root Mean Square Error (RMSE) for each sample was computed.The bootstrap distribution of RMSE values for the base model results and precipitation bias adjusted results is shown in Fig. 4. The largest decrease in RMSE occurs in 2002, 2004 and 2005, while the decrease in RMSE is much smaller in 2003 and 2006, which confirms that the precipitation bias adjustment significantly improves the model performance in 2002, but provides only a minor improvement in 2003 and 2006.The improvement in model performance gained by applying the precipitation bias adjustment is highly dependent on the performance of meteorological model estimates of precipitation, with greater improvement in model performance when the precipitation estimates are poor (as was the case in 2002).
The pattern of NH + 4 wet deposition closely follows the seasonal SO = 4 wet deposition pattern, with a peak in NH Introduction Discussion Paper | Discussion Paper | Discussion Paper |

+
4 wet deposition over the Great Lakes and Mid-West regions, but consistently underestimates the spatial extent of the highest NH + 4 wet deposition in the those regions (Fig.S5).The model does well estimating the localized peak in annual NH + 4 wet deposition in eastern North Carolina, where a large number of confined animal feeding operations contribute to a peak in NH , applying the precipitation adjustment to the CMAQ estimated NH + . The agricultural management tool estimates fertilizer application as a function of crop nutrient demand and the soil geochemical model was used to estimate the nitrification and denitrification processes in the soil column and provided the soil water solution ammonium and hydrogen ion concentrations needed in the bi-directional NH 3 model.Agricultural land use categories and crop profiles were proven by the US Department of Agriculture's 2002 Census of Agriculture (2002 Census of Agriculture, 2004).To evaluate the impact that bi-directional NH 3 exchange has on the CMAQ estimated NH Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | There is a clear downward trend in the NTN observations of NO − 3 wet deposition from 2002-2006, which is also seen in the CMAQ model estimates (Fig. 10).The trend toward lower NO − 3 wet deposition may be due, at least in part, to the implementation of rules under the NO x SIP Call (http://www.epa.gov/ttn/naaqs/ozone/rto/sip/index.html) in mid-2003, which greatly reduced the amount of NO x emissions in 22 states in the eastern US.While the CMAQ model generally does well reproducing the overall observed spatial pattern of NO − 3 wet deposition, the model consistently underestimates the NO − 3 wet deposition in parts of the Northeast and Great Lakes regions, specifically New York, eastern Pennsylvania and Michigan, while overestimating the deposition in western Pennsylvania and West Virginia.
Discussion Paper | Discussion Paper | Discussion Paper | for the years 2002-2006 for the CONUS using a 36-km grid spacing and the eastern Introduction Discussion Paper | Discussion Paper | Discussion Paper |

4
wet deposition estimates were significantly reduced for three of the five years (smaller improvements for the other two years) when the estimates were adjusted to account for biases in the model estimated precipitation.The CMAQ modelling system underestimates NH + 4 wet deposition in the eastern US in both the 36-km and 12-km simulations, with the underestimation tending to be slightly larger in the 36-km simulation.The largest underestimation of NH + 4 wet deposition occurs in the winter and spring periods, while the summer and fall have slightly lower underestimations.The underestimation is likely due, in part, to the poor temporal and spatial representation of NH 3 emissions, particularly those emissions associated with fertilizer applications and bi-directional exchange of NH 3 flux from the soil and vegetation.Implementation of a bi-directional NH 3 flux mechanism in the CMAQ model, along with improvements in the temporal and spatial representation of fertilizer applications, improve the underestimation of NH + 4 wet deposition and these changes will likely be included in the next release of the CMAQ model.The performance for model estimates of NO − 3 wet deposition are mixed throughout the year, with the model largely underestimating NO − 3 wet deposition in the spring and summer in the eastern US, while the bias in the fall and winter is relatively small.Model estimates of NO − 3 wet deposition tend to be slightly lower for the 36-km simulation as compared to the 12-km simulation, particularly in the spring.One large source of the underestimation of NO Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Disclaimer The United States Environmental Protection Agency, through its Office of Research and Development, funded and managed the research described here.It has been subjected to Agency review and approved for publication.Discussion Paper | Discussion Paper | Discussion Paper | Part II: application and evaluation in a mesoscale meteorological model, J. Appl.Meteor.
Discussion Paper | Discussion Paper | Discussion Paper |

Fall
Discussion Paper | Discussion Paper | Discussion Paper |

Fall
Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 4 .Fig. 5 .Fig. 7 .
Fig. 4. Distribution of RMSE based on 1000 bootstrap samples of the modelled and observed SO = 4 wet deposition.Results for model estimates without any adjustment for precipitation bias ("Base Model") are shown in blue and for model estimates adjusted for precipitation errors ("Precip.Adj.") are red.The bold lines indicate the RMSE values from the original dataset.

Fig. 8 .Fig. 9 .
Fig. 8. Distribution of RMSE based on 1000 bootstrap samples of the modelled and observed NH + 4 wet deposition.Results for model estimates without any adjustment for precipitation bias ("Base Model") are shown in blue and for model estimates adjusted for precipitation errors ("Precip.Adj.") are red.The bold lines indicate the RMSE values from the original dataset.

Fig. 12 .
Fig. 12. Box plots of annual modelled -observed NO −3 wet deposition for model wet deposition estimates without any adjustment for precipitation bias ("Base Model"; blue) and for the model estimates adjusted for precipitation errors ("Precip.Adjusted"; red).The black line within the box represents the median bias, shading represents the range of the 25% to 75% quartile and the dashed lines represent the range of the 5% to 95% values.

Fig. 13 .
Fig. 13.Distribution of RMSE based on 1000 bootstrap samples of the modelled and observed NO − 3 wet deposition.Results for model estimates without any adjustment for precipitation bias ("Base Model") are shown in blue and for model estimates adjusted for precipitation errors ("Precip.Adj.") are red.The bold lines indicate the RMSE values from the original dataset.

Table 1 .
Seasonal and annual NMB (%) for precipitation for the 12-km and 36-km CMAQ model simulations.

Table 2 .
Seasonal and annual RMSE (cm)for precipitation for the 12-km and 36-km CMAQ model simulations.

Table 4 .
Seasonal and annual RMSE (kg/ha) for SO =

Table 5 .
Seasonal and annual NMB (%) for NH

Table 8 .
Seasonal and annual RMSE (kg/ha) for NO −