A WRF-Chem Realtime Modeling System for Monitoring CO 2 Emissions

P33 A WRF-Chem Realtime Modeling System for Monitoring CO Emissions Aijun Deng*, Thomas Lauvaux, Ken Davis, Natasha Miles, Scott Richardson and David Stauffer Department of Meteorology, The Pennsylvania State University 1. INTRODUCTION Atmospheric inversion methods, used to infer sources and sinks of greenhouse gases, require highly accurate modeling of atmospheric transport and dispersion (AT&D) processes, and this is especially challenging when applied to urban scales and complex terrain. Accuracy of the atmospheric transport and dispersion solely depends on the meteorological conditions represented in numerical weather prediction (NWP) models (Deng et al. 004, Rogers et al. 011). These conditions include parameters critical to AT&D such as wind speed, wind directions and atmosphere stability, etc. Traditionally, AT&D modeling studies were carried out though coupling between an NWP model and an AT&D model using an off-line approach in which the NWP model is run first to produce the various meteorological fields at typically an hour output time frequency, and then a subsequent AT&D model forecast is driven by the NWP solutions (e.g., Deng et al. 004). Commonly-used NWP models for AT&D and air quality studies include the Penn State/NCAR fifth generation mesoscale model (MM5, Grell et al. 1994), and the recently-developed Weather Research and Forecasting model (WRF, Skamarock 008). These NWP models provide gridded meteorological fields that can be used to drive the AT&D/air quality models used to predict the concentrations of air pollutants, given the source information. One example of an off-line AT&D model used for hazard prediction and consequence management is the Second-order Closure Integrated Puff (SCIPUFF) model (Sykes et al. 004). Off-line photochemical air quality models include the Community Multiscale Air Quality (CMAQ) Model (Byun and Schere 006) and the ENVIRON International Corporation s Comprehensive Air Quality Model with Extensions (CAMx) (Kumar and Lurmann 1997), and these are widely used to simulate particulate matter (PM), air toxics, and ozone concentrations. Examples of AT&D/air quality modeling studies using the off-line approach include Deng et al. (008), Lee et al. (009), Byun and Schere (006), Deng et al. (004) and Kumar and Lurmann (1997). Recent development of the WRF model coupled with Chemistry (WRF-Chem, Grell et al. 005) has made in-line coupling possible. WRF-Chem allows users to simulate the interaction between the atmospheric motions and the chemical species including trace gas, aerosol, dust, etc. These calculations are performed in-line so that they are * Corresponding Author: Aijun Deng, Department of Meteorology, 503 Walker Building, Pennsylvania State University, University Park, PA 1680; email: deng@meteo.psu.edu consistent with all conservative transport done by the meteorology model, using the same vertical and horizontal grids (no horizontal and vertical interpolation) and the same physics parameterization for subgrid scale transport. The WRF-Chem modeling system is becoming more widely used to simulate, predict and monitor the emissions and concentrations of the atmospheric trace gases and chemical species (e.g., CO ) that have significant impact in global and local climate changes. Recently, a dense regional deployment of CO surface measurements in the upper Midwest (Miles et al., 01) was used to determine the CO fluxes from agricultural activities at 10-km resolution, using WRF-Chem coupled to a Lagrangian Particle Dispersion Model in backward mode. The inverse fluxes were compared to agricultural inventories and showed the potential of the method to capture the sources and sinks of CO over a region of about 1000x1000km (Lauvaux et al., 01). Several ongoing projects over Indianapolis and Los Angeles are now trying to measure urban emissions using the same inverse methodology. To support a realtime CO monitoring mission during the World Economic Forum (WEF) Annual Meeting 01 over Davos, Switzerland, the Penn State team has developed a time-lagged WRF-Chem realtime modeling system and used it for over two months, including the week of the WEF meeting, with a specified inventory of CO emissions. The WRF- Chem-simulated CO concentrations are compared to the realtime CO concentration observations at two different locations: one site is in the city to evaluate the emissions from Davos, and a second site is at higher altitude measuring the background concentrations. Using a simplified inverse approach, the difference between the observed- and simulated- CO concentrations allowed the evaluation of the reported CO emissions at the daily time scale. This paper describes the realtime WRF-Chem system and presents the main findings on local emissions and discusses the modeling performance in the valley of Davos.. MODEL DESCRIPTION The core of our realtime modeling system used in this research is the WRF model coupled with Chemistry (WRF-Chem, Grell et al. 005). It includes a complete suite of atmospheric physical processes that interact with the model s dynamics and thermodynamics core. These physical processes include: 1) cloud microphysics that predicts the cloud hydrometeor species (e.g., cloud droplets, rain, snow, etc.), ) cumulus parameterization that predicts the thermodynamic effects of the sub-grid scale atmospheric convection, 3) atmospheric radiation that 1

predicts the heating/cooling effects due to atmospheric radiative processes, 4) planetary boundary layer (PBL)/turbulence physics that predicts the atmospheric turbulence transport of heat, moisture and momentum between the earth surface and the lower atmosphere; and 5) land surface processes that predict the exchange of sensible and latent heat between the earth s surface and atmospheric surface layer. With the addition of the AT&D and chemical processes, WRF-Chem can simulate the coupling among trace gas, aerosol, dynamics, radiation and chemistry (Grell et al. 005, Fast et al. 006). WRF- Chem allows users to simulate the interaction between the atmospheric motions and the chemical species including aerosol, dust, etc. in-line so they are consistent with all conservative transport done by the meteorology model. The WRF configuration for the model physics used for the 01 WEF Annual Meeting was based on the modeling experience of many previous numerical modeling studies using WRF (e.g., Gaudet et al. 009, Rogers et al. 011, and Deng et al. 01). The optimal configurations of model physics used in this study include the use of: 1) the single-moment 3-class simple ice scheme microphysical processes, ) the Kain-Fritsch scheme for cumulus parameterization on the coarser outer grids, 3) the Rapid Radiative Transfer Model for longwave atmospheric radiation, and the Dudhia scheme for shortwave atmospheric radiation, 4) the TKE-predicting Mellor-Yamada Level.5 turbulent closure scheme (MYJ PBL) for the boundary layer turbulence parameterization, and 5) the 5-layer thermal diffusion scheme for representation of the interaction between the land surface and the atmospheric surface layer. The WRF modeling system also has fourdimensional data assimilation (FDDA) capabilities to allow the meteorological observations to be continuously assimilated into the model. The FDDA technique used in this study was originally developed for MM5 (Stauffer and Seaman 1994) and recently implemented into WRF (Deng et al. 009). It has several major uses. Firstly, it can be used to create four-dimensional dynamically consistent datasets or dynamic analyses (e.g., Deng et al. 004, Deng and Stauffer 006, Rogers et al. 011). Secondly, it can also be used to create improved lateral boundary conditions for process studies (e.g., Reen et al. 006). Finally, it can be used for dynamic initialization, where the model is relaxed towards observed conditions during a pre-forecast period to improve the initial state and the subsequent short-term forecast (e.g., Deng et al. 01). In this specific realtime application, WMO observations were assimilated into the WRF-Chem system to produce a dynamic analysis, blending the model simulations and the observations to produce the most accurate meteorological conditions possible to simulate the atmospheric CO concentrations in space and time throughout the Davos region. Another challenging issue of the WRF modeling over the Davos region is handling the complex terrain. Terrain slopes greater than 40% caused the model crash due to the instabilities associated with verticallypropagating sound waves. To address this issue, with help from the scientists at the National Center for Atmospheric Research (NCAR), we increased the damping magnitude by increasing the damping factor, time off-centering for vertical sound waves, by a factor of two (i.e., from epssm=0.1 in the default WRF dynamics to epssm=0.). The off-centering is accomplished by using a positive (non-zero) coefficient (i.e., damping factor) in the acoustic timestep of the vertical momentum equation and the geopotential equation. An off-centering coefficient epssm=0.1 is typically used in the ARW applications, independent of time step or grid size. The WRF dynamic solver is designed such that forward-in-time weighting of the vertically-implicit acoustic-time-step terms damps instabilities associated with verticallypropagating sound waves. The forward weighting also damps instabilities associated with sloping model levels and horizontally propagating sound waves. Our sensitivity study using varying damping factors shows that increasing the damping factor by a factor of two eliminated the instability and produced little difference in the model solutions. 3. SYSTEM CONFIGURATION The WRF model grid configuration used for this demonstration is comprised of four grids: 36-km, 1- km, 4-km and 1.33-km (Figure 1), all of which are cocentered at Davos, Switzerland. The 36-km grid, with a mesh of 110x110 grid points, contains the entire continental Europe, and parts of the Atlantic Ocean. The 1-km grid, with a mesh of 151x151 grid points, contains France, Italy, Poland and Germany. The 4- km grid, with a mesh of 175x175 grid points, contains western France, southern Germany, northern Italy, western Austria and all of Switzerland. The 1.33-km grid, with a mesh of 0x0, covers portions of northern Italy, southern Germany, western Austria and western Switzerland, with the grid centered at Davos. The terrain and landuse fields for the 1.33-km grid are shown in Figure. Fifty (50) vertical terrain-following layers are used, with the center point of the lowest model layer located ~1 m above ground level (AGL). The thickness of the layers increases gradually with height, with 7 layers below 850 hpa (~1550 m AGL). Note that WRF s vertical layers are defined based on the dry hydrostatic pressure and the height of the center point of each layer changes with time. The top of the model is set at 100 hpa. A one-way nesting strategy is used so that information from the coarse domains defines the lateral boundaries of the fine domains but no information from the fine domains feeds back to the coarse domains.

13th Annual WRF Users Workshop, 5-9 June 01, Boulder, CO a) Figure 1. WRF-Chem 36/1/4/1.33-km grid configuration used in a realtime CO monitoring mission during the World Economic Forum (WEF) Annual Meeting 01 over Davos, Switzerland. Table 1 lists the FDDA parameters used in this application. As indicated in the table, for this application, 3D analysis nudging and surface analysis nudging were applied on both the 36- and 1-km grids with reduced nudging strength (G) on the 1-km grid, and observation nudging was applied on all grids with the same nudging strength. No mass fields (temperature and moisture) observations are assimilated within the WRF-predicted PBL. The meteorological observation data assimilated into the WRF system, the World Meteorological Organization (WMO) observations distributed by the National Weather Service (NWS), include both 1-hourly upper-air rawinsondes and hourly surface observations. Figure 3 shows the WMO surface observation distributions at 00 UTC, 9 January 01, on the coarsest and the finest grids. b) Figure. Terrain (a) and landuse (b) of the 1.33-km grid, where D at the domain center denotes Davos, Switzerland. Table 1: Multiscale FDDA parameters used in this study. G is the nudging coefficient, RINXY is the radius of influence used in obs nudging, TWINDO is the time window used in obs nudging, and dt is the model time step. 3D Analysis OBS 4km 1.33km 36km 4km 36km 1km 1km 1.33km G (1/sec) Wind field Mass field RINXY (km) 3*10-4 1*10-4 above PBL 180 60 150* 100* above PBL 100* TWINDO (hr) dt (sec) * 0.67 factor for surface,.0 factor at 500 hpa and above, 0.5 factor for surface 3 0 100* 6

a) a single total number for the year 005. In the county of Davos, 79% of the annual direct emissions are due to heating, 18% due to transportation, and 3% due to industry. The inventories account for emissions at the county scale of Davos, which is about 50 km, with only 6 km of urban land cover. Population density maps were used to distribute the county level emissions over the urban areas over the 1.33-km grid, assuming that anthropogenic activities are mainly located within the city limit. The inverse approach is based on a linear interpolation method, applicable in this situation because of the small size of the urban area compared to the grid size and the availability of a background site in the mountains surrounding Davos. First, by subtracting the background site concentrations from those measured at the urban site, we remove signals from distant sources, leaving only the Davos urban signal. The modeled inter-site difference is dependent on the first-guess emissions inventory described above. We apply a multiplicative factor to the initial inventory estimate such that the modeled inter-site CO difference matches that observed, thus determining the daily emissions estimate from this simplified inversion. The inverse system used here does not require the use of an adjoint model, considering that all the emissions from the city are located within the urban tower footprint. 4. PRELIMINARY RESULTS b) Figure 3. WMO observations distribution on the 36-km grid a) and 1.33-km grid b). The gridded meteorological data needed to initialize the WRF-Chem realtime system was the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) analyses (i.e., zerohour forecast) available 6-hourly in realtime. To produce the most accurate initial condition for WRF, the ~0.5-degree GFS analyses are further enhanced by an objective analysis process using a modified Cressman scan approach (OBSGRID, Deng et al. 009). The surface analysis fields used for surface analysis FDDA are generated by OBSGRID at 3- hourly intervals. The objective analysis procedure uses the 1- hourly WMO upper air observations provided by NWS sondes, and the hourly WMO surface observations from the NWS. For the -month period surrounding the WEF meeting, the WRF-Chem realtime system was run twice a day for 4 hours in a time-lagged fashion so that it ends at the current real time. It created the dynamic analyses that represent the most accurate meteorological conditions critical to simulating the atmospheric CO concentrations in space and time over the Davos region. The first guess of the CO emissions from Davos comes from a previous study (Walz et al. 008) that presents emissions inventory data specifically for Davos. This CO emissions initial estimate consists of Figure 4 shows an example of a tracer plume predicted by the PSU WRF-Chem system over the valley of Davos, Switzerland, showing the CO concentration, during the World Economic Forum Annual Meeting 01, (more details may be found at http://cms.met.psu.edu/davos/). Figure 4. An example of a tracer plume predicted by the PSU WRF-Chem system over the valley of Davos, Switzerland, showing the CO concentration, during the World Economic Forum Annual Meeting 01, (more details may be found at http://cms.met.psu.edu/davos/). Figure 5 shows the daily inverse emissions over the two-month period (with 3-day smoothing applied), and observed Heating Degree Days which correspond to the daily difference in temperature to 4

maintain the indoor temperature constant (at 15.5 Cº). As 79% of the annual inventory emissions are due to heating (Walz et al. 008), we expect the daily inverse emissions to be correlated with the observed Heating Degree Days. During the pre-wef period (December 1, 011 - January 4, 01), the emissions are 40% larger than the initial inventory estimate. Emissions estimates larger than the initial (annual) inventory are not surprising, given that lower temperatures in winter necessitate the use of domestic fuel for heating. At the daily time scale, the inverse estimates vary significantly, but are correlated with temperature (r>0.5), with lower values during warmer days. During the WEF meeting (January 5, 01 - January 9, 01), the inverse emissions decrease by 40% compared to the first period. Despite the activity in the city, the emissions show a clear decrease over the week, which can be related to the change of activity during the meeting. Whereas the traffic is substantially larger due to the large number of participants, several public buildings are closed near and around the urban measurement site. Several assumptions remain, such as the possible influence of strong local sources, or the footprint size which could be smaller than the city. The absence of convection near the surface, indicated by low values of observed wind stress, limits also the interpretation of the daily signals. Comparison to PBL Lidar measurements is ongoing and will provide an independent evaluation of the model performances during the period. Finally, the last period of the deployment (January 9, 01 - March 3, 01) includes two singular weather events, with a very cold period of about three weeks corresponding to very large emissions, and a warmer period with lower emissions compared to the first period. These two singular events show the large contribution of house heating related to temperature on the urban carbon budget and confirmed the potential of the system to detect the variability related to weather events over several days. 5. SUMMARY AND CONCLUSIONS The Penn State team has developed a timelagged WRF-Chem realtime modeling system and used it for a realtime CO monitoring mission during the World Economic Forum (WEF) Annual Meeting 01 over Davos, Switzerland. The system uses the multi-scale FDDA capabilities that involve both analysis and observation nudging, and assimilates available meteorological observations during the entire model simulation to produce dynamic analysis meteorological fields coupled in-line with the AT&D processes within WRF-Chem model without the activation of chemical processes. For a two-month period including the WEF meeting, the WRF-Chem realtime system was run twice a day for 4 hours in a time-lagged fashion so that it ends at the current real time. It created the meteorological conditions critical to simulating the atmospheric CO concentrations in space and time over the Davos region. The WRF-Chem-simulated CO concentrations are compared to the observations at two different locations, with a first site in the city used to evaluate the emissions from Davos, and a second at higher altitude measuring the background concentrations. Using a simplified inverse approach, the difference between the observed- and simulated-co concentrations allowed the evaluation of the reported CO emissions at the daily time scale. The inverse emission estimates computed daily were highly correlated with temperature, consistent with the initial inventory study which evaluated the contribution of house heating to 79% of the total carbon budget of Davos. In addition, the emissions increased significantly during the extreme cold wave affecting most of western Europe during February 01, with a 30% increase in emissions compared to the month of January 01. Preliminary results suggest that emissions during the WEF meeting were lower than average, suggesting that changes in the anthropogenic activities or in local sources may have been observed during that period. This first operational inverse system provided daily CO emissions and showed a great potential for larger deployments, despite the complexity of the terrain and the weak vertical mixing near the surface during the campaign. Figure 5: Inverse emissions for the city of Davos, Switzerland, (in blue) during the two-month campaign using the linear interpolation method between WRF- Chem concentrations and observations at the downtown site; Heating Degree Days indicate the number of degrees to maintain constant indoor temperature, and is used as a tracer of energy consumption for house heating (in green). 5 6. ACKNOWLEDGMENTS We acknowledge Jim Dudhia at NCAR for his advice concerning modifying the model s dynamics to reduce the terrain-induced instability over the extremely complex terrain in this study. 7. REFERENCES Byun, D., and K.L. Schere, 006: Review of the Governing Equations, Computational Algorithms, and Other Components of the Models-3 Community Multiscale Air Quality (CMAQ) Modeling System.

Applied Mechanics Reviews, 59, 51-77. Deng, A., N. L. Seaman, G. K. Hunter and D. R. Stauffer, 004: Evaluation of interregional transport using the MM5-SCIPUFF system. J. Appl. Meteor., 43, 1864-1886. Deng, A. and D.R. Stauffer, 006: On improving 4-km mesoscale model simulations. J. Appl. Meteor., 45, 361-381. Deng, A., D.R. Stauffer, I. Sykes, and D. Henn, 008: Comparison of HPAC-SCIPUFF predictions based on MM5 and WRF meteorological inputs, Chemical and Biological Defense Physical Science and Technology Conference, New Orleans, LA, Nov 17-1, 10 pp. Deng, A., D.R. Stauffer, B.J. Gaudet, J. Dudhia, J. Hacker, C. Bruyere, W. Wu, F. Vandenberghe, Y. Liu and A. Bourgeois, 009: Update on WRF-ARW end-to-end multi-scale FDDA system, 10 th Annual WRF Users Workshop, Boulder, CO, June 3, 14 pp. Deng, A., D.R. Stauffer, B.J. Gaudet, and G.K., Hunter, 01: A Rapidly Relocatable High- Resolution WRF System for Military-Defense, Aviation and Wind Energy. 13th Annual WRF Users Workshop, Boulder, CO, June 5-9, 11pp. Gaudet, B., D. Stauffer, N. Seaman, A. Deng, K. Schere, R. Gilliam, J. Pleim, and R. Elleman, 009: Modeling extremely cold stable boundary layers over interior Alaska using a WRF FDDA system. Preprints, 13th Conf. on Mesoscale Processes, Salt Lake City, UT, 17-0 Aug., Amer. Meteor. Soc., 8 pp. Grell, G. A., J. Dudhia, and D. R. Stauffer, 1994: A description of the fifth generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-3981STR, 138 pp. Grell, G.A., S.E. Peckham, R. Schmitz, S.A. McKeen, G. Frost, W.C. Skamarock, B. Eder, 005: Fully coupled online chemistry within the WRF model. Atmos. Environ., 39, 6957-6975. Lauvaux, T., Schuh, A. E., Uliasz, M., Richardson, S., Miles, N., Andrews, A. E., Sweeney, C., Diaz, L. I., Martins, D., Shepson, P. B., and Davis, K. J., 01: Constraining the CO budget of the corn belt: exploring uncertainties from the assumptions in a mesoscale inverse system, Atmos. Chem. Phys., 1, 337-354, doi:10.5194/acp-1-337-01. Lee, J.A., L.J. Peltier, S.E. Haupt, J.C. Wyngaard, D.R. Stauffer, and A. Deng, 009: Improving SCIPUFF dispersion forecasts with NWP ensembles J. Appl Meteor. and Climatol., 48, 305-319. Miles, N. L., S. J. Richardson, K. J. Davis, T. Lauvaux, A. E. Andrews, T. O. West, V. Bandaru, and E. R. Crosson, 01: Large amplitude spatial and temporal gradients in atmospheric boundary layer CO mole fractions detected with a tower-based network in the U.S. upper Midwest, J. Geophys. Res., 117, G01019, doi:10.109/011jg001781. Kumar, N. and F. W. Lurmann, 1997: Peer Review of ENVIRON's Ozone Source Apportionment Technology and the CAMx Air Quality Model. REVISED FINAL REPORT STI-99603-173-RFR. June, 1997. Reen, B. P., D. R. Stauffer, K. J. Davis, and A. R. Desai, 006: A Case Study on the Effects of Heterogeneous Soil Moisture on Mesoscale Boundary-Layer Structure in the Southern Great Plains, U.S.A. Part II: Mesoscale Modelling. Bound.-Layer Meteor., 10, 75-314. DOI 10.1007/s10546-006-9056-6. Rogers, R., A. Deng, D. Stauffer, Y. Jia, S. T. Soong, S. Tanrikulu, S. Beaver, and C. Tran, 011: Fine particulate matter modeling in Central California, Part I: Application of the Weather Research and Forecasting model, 13th Conference on Atmospheric Chemistry, Seattle, WA, Jan. 3 7, 17 pp. Skamarock, W.C., Klemp, J.B., Dudhia, J., Gill, D.O., Barker, D.M., Duda, M.G., Huang, X.-Y., Wang, W., Powers, J.G., 008. A description of the Advanced Research WRF Version 3. NCAR Technical Note NCAR/TN-475+STR. 113 pp. Stauffer, D. R. and N. L. Seaman, 1994: Multiscale Four-Dimensional Data Assimilation. J. Appl. Meteor., 33, 416-434. Sykes, R. I., S. F. Parker, and D. S. Henn, 004: SCIPUFF version.0, technical documentation. A.R.A.P. Tech. Rep. 77, Titan Corporation, Princeton, NJ, 84 pp. Walz, A., G.-P. Calonder, F. Hagedorn, C. Lardelli, C. Lundström and V. Stöckli. 008. Regional CO budget, countermeasures and reduction aims for the Davos region, Switzerland. Energy Policy 36, 811-80. 6