Meteorological evaluation of a weather-chemistry forecasting model using observations from the TEXAS AQS 2000 field experiment

19 pages
of 19
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Meteorological evaluation of a weather-chemistry forecasting model using observations from the TEXAS AQS 2000 field experiment
  Meteorological evaluation of a weather-chemistry forecasting modelusing observations from the TEXAS AQS 2000 field experiment J.-W. Bao and S. A. Michelson 1 Environmental Technology Laboratory, NOAA, Boulder, Colorado, USA S. A. McKeen 1 Aeronomy Laboratory, NOAA, Boulder, Colorado, USA G. A. Grell 1 Forecast Systems Laboratory, NOAA, Boulder, Colorado, USAReceived 14 May 2004; revised 21 June 2005; accepted 25 August 2005; published 8 November 2005. [ 1 ]  Meteorological forecasts for the period of 25–30 August 2000 from a coupledweather-chemistry model are evaluated both qualitatively and quantitatively using theobservations from different instruments that were deployed in metropolitan Houstonduring the Texas Air-Quality Study 2000 field experiment. The qualitative comparison iscarried out with respect to the meteorological processes associated with the influence of the large-scale flow on the sea breeze that are essential to the development of the surfaceozone exceedances over Houston, while the quantitative comparison is focused on theerrors and uncertainties of the forecasts. The qualitative comparison is performed withrespect to a conceptual model for the influence of the large-scale flow on the sea breeze.The comparison shows that although the overall forecasted influence of the large-scaleflow on the sea breeze compares qualitatively well to the observations, quantitativedifferences do exist between the forecasted and observed wind speed and direction, as wellas with temperature and moisture. It is found that the forecasted low-level winds have asystematic easterly bias and the forecasted low-level temperature has a cold bias. Theerrors in the forecasted low-level moisture appear relatively small, but with a cold biasthey lead to higher relative humidity in the forecast than in reality. There is great sensitivity of the model forecasted low-level winds to different initial conditions. Thequantitative comparison also indicates that the model’s effective horizontal resolutioncorresponding to 1.67-km grid spacing is actually about 10 km. Citation:  Bao, J.-W., S. A. Michelson, S. A. McKeen, and G. A. Grell (2005), Meteorological evaluation of a weather-chemistryforecasting model using observations from the TEXAS AQS 2000 field experiment,  J. Geophys. Res. ,  110 , D21105,doi:10.1029/2004JD005024. 1. Introduction [ 2 ] The 2000 Texas Air Quality Study (hereafter referredto as Texas AQS 2000; see ceer/texaqs/index.html for the details) was conducted duringAugust and September 2000 in order to improve theunderstanding of the processes that control the formationand transport of air pollutants along the southeastern Texascoast. This experiment was the largest air quality study ever carried out in Texas and involved a team of researchers fromfederal and Texas state agencies and universities. Measure-ments of meteorological conditions, gaseous, particulate,and hazardous air pollutants were made by an observationalnetwork to make it possible to study the formation and dailycycles of ozone and its precursors, as well as how these pollutants are controlled or affected by the weather.[ 3 ] To provide real-time weather and air-quality forecastsfor daily operational planning of the experiment, a coupledweather-chemistry forecasting model (MM5-Chem) [ Grell et al. , 2000] was run jointly by the Forecast SystemsLaboratory, Environmental Technology Laboratory, andAeronomy Laboratory of the National Oceanic and Atmo-spheric Administration (NOAA) during the Texas AQS2000 experiment [ Grell et al. , 2002]. A unique feature of this model is that oxidant formation and pollution transport are intricately coupled with the calculation of atmosphericdynamics and physical state. Although the coupled weather-chemistry model provided valuable forecast guidance onwhere to deploy aircraft to take meteorological and chem-ical measurements, evaluation and assessment of the fore-casts were not performed during the field experiment  because the observational data were not ready real-time. JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 110, D21105, doi:10.1029/2004JD005024, 2005 1 Also at Cooperative Institute for Research in Environmental Sciences(CIRES), University of Colorado, Boulder, Colorado, USA.Copyright 2005 by the American Geophysical Union.0148-0227/05/2004JD005024$09.00 D21105  1 of 19  Instead, post-experiment evaluation and assessment of theforecasts have been carried out to provide insights into the performance of the coupled weather-chemistry model and toidentify major errors in the forecasts, so that the skill of thecoupled weather-chemistry model in urban air-quality fore-casts can be improved in the future.[ 4 ] In this work, meteorological observations taken dur-ing the Texas AQS 2000 are used to evaluate and assess themeteorological forecasts of the coupled weather-chemistrymodel. Preliminary analyses of the meteorological observa-tions from the experiment have indicated [  Nielsen-Gammonand Breitenbach , 2002;  McKeen et al. , 2002;  Jiang and  Fast  , 2004;  Banta et al. , 2002;  Darby et al. , 2002;  Senff et al. , 2000] that high ozone pollution episodes are closelylinked to the re-circulation and convergence of ozone and its precursors by the sea breeze. It has long been recognizedthat high surface ozone episodes in the Houston area areassociated with the meteorological conditions in which the pollutants released from various sources are confined to theHouston area by the sea breeze [  Hsu , 1970;  Atkinson , 1981,chapter 5] developed under the influence of weak large-scale flow in the lower troposphere that is associated withthe subtropical high [see, e.g.,  Banta et al. , 2002]. Under such conditions, before the onset of the sea breeze, the pollutants released and accumulated are not transported toofar away from the sources and are enhanced by newlyemitted pollution in the morning. With the onset of thesea breeze, a convergence zone is formed over the Houstonarea by the advance of the sea breeze front. This conver-gence zone is a key meteorological element in producinghigh surface ozone exceedances in the late afternoon be-cause its formation re-circulates the pollutants over thesources, and hence increases the surface ozone concentra-tions. Houston is not the only geographical location wherelocal circulations (such as land/sea breezes) developed inweak large-scale flow play a major role in making highly polluted events [  National Research Council  , 1992,chapter 4]. Previous studies in other areas, such as theLos Angeles area [  Angell et al. , 1972] and along the shoresof Lake Michigan [e.g.,  Lyons et al. , 1995;  Shafran et al. ,2000], have shown that the re-circulation and convergenceof ozone and its precursors by the sea breeze/lake breezecirculation can enhance surface ozone concentrations.[ 5 ] This work focuses on the evaluation and assessment of the meteorological forecasts by comparing the modelresults with aircraft, wind profiler, and rawinsonde obser-vations from the Texas AQS 2000 field experiment for thehigh surface ozone episode that occurred during the time period of 25–30 August 2000. The reason for choosing this period is that it is a part of a 9-day period of surface ozoneexceedances over the Houston metropolitan area and theweather conditions of this period are typical of what wereobserved during the entire period of the Texas AQS 2000field experiment (with high temperatures and not much precipitation). First, a qualitative comparison will be madeto see how well the model output compares against theseobservations with respect to the primary meteorological process involved in the high surface ozone episodes: theinfluence of the large-scale flow on the development of thesea breeze. This qualitative comparison is important be-cause during this period the change in the low-level windsdue to the influence of slowly varying large-scale flow onthe sea breeze is directly responsible for the variation in dailyexceedances of surface ozone. Then, a quantitative compar-ison of the forecasts with the observations will be carried out to reveal the errors and uncertainties in the forecasts.[ 6 ] The paper is arranged as follows. Section 2 provides a brief description of the coupled weather-chemistry model.In section 3, the meteorological conditions over Houstonduring the 25–30 August time period are overviewed.Results of the comparison between the model forecastsand the observations, both qualitative and quantitative, are presented in section 4. Conclusions are presented insection 5. 2. Model Description [ 7 ] The coupled weather-chemistry forecasting model,MM5-Chem, combines a modified version of The Pennsyl-vania State University/National Center for AtmosphericResearch (NCAR) Mesoscale Model (MM5) [ Grell et al. ,1994] and the improved Regional Acid Deposition Model[ Stockwell et al. , 1990]. Details about the coupled modelcan be found in  Grell et al.  [2000]. In the coupled model,the transport of chemical species (grid-scale and sub-gridscale) is determined simultaneously together with meteorol-ogy. The calculations of photolysis, biogenic emissions, anddeposition are also carried out ‘‘online.’’ The approach of directly coupling a meteorological model and a chemicalmodel has two major advantages over the off-line couplingapproach in which chemistry-transport-models are run usingoutput from a meteorological model as input. First, thechemical model can take the entire meteorological informa-tion as input at every time step of the meteorological model,including subgrid-scale dynamical motions like turbulenceand moist convection. Second, the online approach allowsone to consider chemical-meteorological feedback (such asthat of aerosols on incoming solar radiation) that is not  possible in the off-line approach. The validation and eval-uation study conducted by  Langmann  [2000] indicates that the on-line approach is able to better reproduce measurednear surface concentrations of pollutants than the off-lineapproach does. This is attributed to a better representation of transport by convective clouds with the high temporalresolution afforded by the online treatment of pollutiontransport. The mixing efficiency between the ABL and thefree troposphere is believed to have a significant impact on photo-oxidant concentrations in the lower troposphere.[ 8 ] During Texas AQS 2000, the coupled model was runon multiple 1-way nested meshes of 60 km, 15 km, 5 km,and 1.67 km horizontal grid spacings (Figure 1) with 25vertical layers (9 levels within the lowest 2 km), to providereal-time forecasts for daily operational planning. The initialconditions for the 60-km mesh were specified at 0000 UTCusing the NOAA Forecast Systems Laboratory/Rapid Up-date Cycle (RUC) analyses of 40-km resolution that use the National Centers for Environmental Prediction’s (NCEP)ETA model output as the first guess. The lateral boundaryconditions for the 60-km domain are specified using the6-hourly ETA forecasts. The SST in the RUC analyses isspecified by interpolating the National Environmental Sat-ellite, Data and Information Service (NESDIS) multi-chan-nel SST analysis on a 50 km grid [  McClain et al. , 1985] tothe RUC grid. The land use is specified using the U.S. D21105  BAO ET AL.: EVALUATION OF A WEATHER-CHEMISTRY MODEL2 of 19 D21105  Geological Survey’s 24 surface categories (including 1category for urban). The sensitivity of the model forecast to different initialization datasets but the same land usespecification is reported later in section 4.2.3. A 24-h fore-cast was carried out daily, beginning at 0000 UTC. Thechemical fields were initialized with the previous forecast totake into account the effect of accumulation [  Jiang and  Fast  , 2004;  Grell et al. , 2002;  Bao et al. , 2002]. Theemission inventory is based upon the compiled databasesfrom the Environmental Protection Agency (EPA) NET-96version 3 inventory with modifications for southeast Texasaccording to the Texas Commission on EnvironmentalQuality [see  McKeen et al. , 2002;  Jiang and Fast  , 2004].[ 9 ] The configuration of MM5 physics (see  Grell et al. [2000] for more details) includes the mixed-phase cloud physics, the revised Grell scheme (only for 60 km, 15 km,and 5 km meshes), a 1.5 order ABL scheme, the RUC land-surface parameterization, and the MM5 simple short-waveand long-wave radiation parameterization schemes. Boththe revised Grell scheme and the RUC land-surface param-eterization are not available in the standard MM5. Also, theversion of the coupled model used for the real-time forecast during Texas AQS 2000 does not include chemical feed- back to meteorology. 3. Case Overview [ 10 ] As mentioned in the introduction, during the periodof 25-30 August 2000, observations indicate that there wasa variation in the exceedances of surface ozone over theHouston area associated with the variation in the dailydevelopment of the sea breeze circulation due to large scaleforcing. In fact, the maximum hourly averaged surfaceozone concentrations during this period were 185 ppb onAugust 25, 119 ppb on 26 August, 80 ppb on 27 August,78 ppb on 28 August, 146 ppb on 29 August, and 199 ppbon 30 August (see monops/psi_rpt). The location of the maximum hourlyaveraged surface ozone concentrations also varies. Sincesurface ozone concentration is directly related to the ozoneconcentration in the ABL [see, e.g.,  Ryerson et al. , 2003;  Jiang and Fast  , 2004], the connection between the windregime change and the corresponding variation in thedistribution of surface ozone can be further illustrated byexamining the observed winds and ozone distribution abovethe surface layer. Figures 2a, 2c, and 2e show the observedwinds and ozone concentrations within the layer between550 m and 750 m above ground level (AGL) measured by NCAR’s aircraft (Electra) on 25, 27, and 30 August 2000.As can be seen in these figures, there is a variation of theobserved ozone distribution associated with the change inthe observed winds. On 25 August, the winds are south-easterly with the maximum ozone concentrations over metropolitan Houston. The observed winds on 27 August  become significantly stronger and have more of a southerlycomponent than on 25 August, while the maximum ozoneconcentrations are located to the northwest of metropolitanHouston. In contrast, the observed winds on 30 August arevery weak, and the maximum ozone concentrations are tothe southeast of metropolitan Houston.[ 11 ] The large-scale flow pattern for this time period wascharacterized by a ridge over Texas that was associated with Figure 1.  The location of the four nested MM5 domains. D21105  BAO ET AL.: EVALUATION OF A WEATHER-CHEMISTRY MODEL3 of 19 D21105  a sub-tropical high in the mid-troposphere, which slowlymoved east (not shown). Correspondingly, at lower levelsthe location of the high and the ridge axis over southeast Texas varied slightly during the time period. Although thelow-level winds over Texas remained fairly light (5– 7.5 ms  1 at 850 hPa), this variation in the location of thehigh induced a change in wind direction between south-easterly and southerly over southeast Texas. The location of the surface ridge axis varied during the time period in such away that the surface wind direction over Houston changedfrom south-southeasterly to having a westerly component towards the end of the six days. With the weak large-scale pressure gradient that was in place, the surface winds over Texas were weak to moderate. The surface air over south-east Texas was warm and humid over the time period, withthe maximum surface temperatures between the middle 30 0 s  C to 40 0 s   C. The weak-to-moderate large-scale surfaceflow, along with the relatively cloud-free skies (except for 25 August) due to large-scale subsidence, were not onlyconducive to the production and accumulation of ozone and Figure 2.  The observed ozone and wind vectors versus the forecasted counterparts within the layer  between 550 m and 750 m AGL along the flight track of NCAR Electra. (a), (c), and (e) are theobservations on 25, 27, and 30 August 2000, respectively, while (b), (d), and (f) are the correspondingforecasts which are initialized at 0000 UTC on each particular day. The scale of wind vectors and thecolor scale for ozone concentrations are shown in (a) and (c), respectively. D21105  BAO ET AL.: EVALUATION OF A WEATHER-CHEMISTRY MODEL4 of 19 D21105  its precursors, but also provided favorable large-scale flowfor the development of the sea breeze circulation. As will bediscussed later in this section, because of the differences inthe large-scale background flow before the onset of the sea breeze (that are associated with the change of the large-scalesurface-ridge alignment and the land-sea temperature con-trast), the development of the sea breeze varied during thesix-day period, leading to the difference in the low-levelozone concentration and distribution.[ 12 ] It has long been recognized that variation in thelarge-scale flow (i.e., the background flow) influences thedevelopment of sea breezes. The dynamics of the influenceare well revealed by  Estoque  [1962] using a series of 2-Dnumerical model experiments. Estoque’s simulations pro-vide a conceptual model for understanding the impact of thelarge-scale flow on the development of the sea breezeconvergence zone. The essence of this conceptual modelis that a strong sea breeze convergence zone is most likelyto develop where the large-scale low-level pressure gradient is such that the large-scale flow moves from land to sea withweak to moderate speed (  5 ms  1 ). Although Estoque’ssimulations are idealized and, thus, are not representative of extreme conditions, they provide insight into the effect of the large-scale pressure gradient and flow on the strengthand location of the sea-breeze convergence zone in the lateafternoon.[ 13 ] The conditions in the real atmosphere under whichsea breezes develop often lie between  Estoque ’s [1962]idealized and extreme ones, depending on the magnitude of the large-scale pressure gradient and the geographical loca-tion [ Yoshikado , 1981;  Banta et al. , 1993]. Realistic irreg-ular coastlines (such as Galveston Bay in this case study)distort the straight sea-breeze convergence zone shown inEstoque’s idealized simulations [  McPherson , 1970] andcause it to follow the coastline. Nevertheless, Estoque’ssimulations indicate that the large-scale low-level pressuregradient and corresponding flow are important parametersin determining the onset of sea breezes, how intense the sea breeze convergence zones are, and how far they advanceinland. They provide an essential dynamic framework for understanding the influence of the large-scale flow on thesea breeze.[ 14 ] The evolution of a sea-breeze convergence zone iscritical to the air-quality over Houston, given its geograph-ical location near the coast and the 30   N latitude. 30   N is acritical latitude in the dependence of the evolution of theland-sea breeze cycle on latitude. Given the same land-seathermal contrast, the resonance of the diurnal heating cyclewith the earth’s rotation frequency maximizes the amplitudeof the land-sea breeze circulation at 30   N [  Rotunno , 1983; Yan and Anthes , 1987]. Idealized numerical modeling of sea breeze by  Anthes  [1978] also indicates that the sea breezemay recirculate pollutants lofted by upward motion at thesea-breeze front on the previous day. Pollutants offshore, if caught in the subsiding branch of the sea breeze circulation,can be returned to the inflow layer. Therefore, the Houstonarea is influenced more than other places at different latitudes by the land-sea breeze-related local accumulation,spread, and recirculation of ozone and its precursors.Moreover, the intensity of the sea-breeze front varies withthe large-scale flow. Over the Houston area, onshore large-scale flow tends to blow the ozone precursors away fromtheir sources and result in a weak sea breeze front, and thustends to reduce the severity of high ozone levels. On theother hand, offshore large-scale flow tends to create astronger sea breeze front that favors local accumulation of  pollutants. 4. Results [ 15 ] To better compare the model forecasted influence of the large-scale flow on the sea breeze with observations for the 6-day period of interest in this work, we divide the period into three low-level flow regimes. These regimes arecategorized according to three types of distribution of theambient sea-level pressure over Galveston Bay at 1200 UTC: (1) northeast-to-southwest pressure gradient with a high to the northeast of Galveston Bay (25, 26August), (2) almost east-to-west pressure gradient with ahigh to the east of Texas (27, 28 August), and (3) almost north-to-south gradient with a high to the north or northwest of Galveston Bay (29, 30 August). Figures 2b, 2d, and 2f show the forecasted winds and ozone concentrations abovethe surface layer along the transects made by NCAR’saircraft (Electra) corresponding to the three wind regimes(on 25, 27, and 30 August 2000, respectively). As with theobservations, there is a correspondence between the windchanges and the corresponding variation in the ozoneconcentrations in the forecast. As will be shown, theforecasted winds have an easterly bias compared with theobservations, and thus the forecasted maximum ozoneconcentrations are located to the west or northwest of theobserved maxima. The change in the forecasted winds maynot be the only factor affecting the forecasted variation inozone concentrations on the three days. Air temperature,cloud cover and the emission inventory also affect themodel forecasted ozone concentrations. Nevertheless, theseresults indicate that the change in the flow regime on thelarge scale affects the distributions and severity of the ozoneexceedances.[ 16 ] The forecasts from the MM5-coupled weather chem-istry model are compared to the observations taken by wind profilers, aircraft, and rawinsondes during the Texas AQS2000 experiment in order to investigate the overall meteo-rological performance of the model during the 25– 30 August 2000 time period. First, a qualitative comparisonof the model forecasts with the observations will be made tosee how well the model performs in simulating the influenceof the large-scale flow on the sea breeze. Then, a quantita-tive comparison of the model forecasts with the observa-tions will be carried out to reveal the errors and uncertaintiesin the model’s meteorological forecasts. 4.1. Qualitative Comparisons of the Influence of theLarge-Scale Flow on the See Breeze [ 17 ] The formation and transport of surface ozone and its precursors are consequences of anthropogenic emissionsand the influence of large-scale meteorological conditionson locally driven processes (such as the sea breeze and thedevelopment of the ABL). Since the model and its initialand boundary conditions are not perfect, definite quantita-tive disagreement between the model forecasts and obser-vations is inevitable. Thus, the first important step in theevaluation of the model’s performance is to compare the D21105  BAO ET AL.: EVALUATION OF A WEATHER-CHEMISTRY MODEL5 of 19 D21105
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks