Envronment Protecton Engneerng Vol. 34 2008 No. 4 PRZEMYSŁAW SZCZYGŁOWSKI*, MARIAN MAZUR* APPLICATION OF BOOT STATISTICAL PACKAGE IN CALCULATING POLLUTANT SPREADING IN AIR The ssue of modellng the pollutant spreadng n ar requres a number of decsons to be taken n order to ensure a proper selecton of the model opton n terms of meteorology or dsperson coeffcents. The dfference n the results of the calculatons of the polluton condtons s observed after each change of opton. Negatve and postve components of Fractonal Bas (FB or Geometrc Mean Bas (MG may prove to be especally useful, relatvely new statstcal ndcators. They allow the overestmaton or underestmaton of the results n the case of calculatng the polluton dstrbuton wthn the specfed area to be evaluated. 1. BASIC STATISTICAL INDICATORS The procedure used to compare the ndvdual optons of models s based on statstcal ndcators developed for the models of pollutant spreadng. These ndcators were gathered n the BOOT Statstcal Model Evaluaton Software Package, Verson 2.0 [1]. The basc statstcal ndcators used n the comparsons were recommended by EPA (Envronmental Protecton Agency, and then collected n the BOOT package [2], [3]. The basc statstcal ndcators are: FB Fractonal Bas, MG Geometrc Mean Bas, NMSE Normalzed Mean Square Error, VG Geometrc Varance, R Correlaton Coeffcent, FAC2 ndcator charactersng model qualty. Asde from the sx basc ndcators defned above, the followng addtonal ndcators were determned as well: FB FN negatve component FB (false-negatve,, FB FP postve component FB (false-postve, negatve (MG FN and postve (MG FP Geometrc Mean Bas MG component. * Department of Management and Envronmental Protecton, AGH Unversty of Scence and Technology, al. Mckewcza 30, 30-059 Kraków, Poland.
152 P. SZCZYGŁOWSKI, M. MAZUR FB 0.5 ( C + ] + FBFN FB FP. (1 FB FN, negatve component FB (false-negatve, that s only these pars (C O, C P are consdered, n whch C P < C O : FB FN + ] ]. (2 FB FP, postve component FB (false-postve, that s only these pars (C O, C P are consdered, n whch C P > C O : FB FP + ], (3 where: C P the concentraton determned by the model, C O the actual concentraton. We may determne negatve (MG FN and postve (MG FP geometrc mean bas components n the same way: 1 MG FN exp [ lnc lnc + (lnc lnc ], (4 2N 1 MG FP exp [ lnc lnc + (lnc lnc ]. (5 2N 2. CALCULATION OPTIONS The wnd feld s determned usng the Calmet meteorologcal pre-processor that transports pollutant puffs over the area surface. For calculatng dsperson of ar pollutants the Calpuff model was used (one of the Gaussan puff models. Informaton on
Applcaton of BOOT statstcal package 153 the topographc profle and land use was obtaned from U.S. Geologcal Survey [4]. Calculatons were performed n the 40 km 40 km network wth a resoluton of 1 km 1 km. Two types of meteorologcal data were used for the calculatons: those from ground meteorologcal statons Katowce (12560, Kraków (12566, Belsko-Bała (11111 and those from 4 aerologcal statons Legonów (12374, Wrocław (12425, Prague (11520 and Poprad (11952. The emsson source was a smokestack n a heat and power staton located n Belsko-Bała, at the coordnates (360.0,5518.0, elevaton of the base of 420 m above sea level. The SO 2 emsson was 100 g/s. Smokestack heght 100 m, and outlet dameter 5.0 m. The outlet velocty of exhaust gas was 15.0 m/s at a gas temperature of 400 K. Calculatons have been performed based on the meteorologcal data for a perod of 744 hours n March 2005 from whch one-hour maxmum concentraton was chosen [5]. The followng optons correspondng to the meteorologcal data used to calculate the dstrbuton of the pollutant concentratons were selected: 1 3 ground statons (Belsko-Bała, Kraków, Katowce, 4 aerologcal statons (Wrocław, Legonów, Prague, Poprad 4up3surf, 2 2 ground statons (Kraków, Katowce, 4 aerologcal statons (Wrocław, Legonów, Prague, Poprad 4up2surf, 3 1 ground staton (Belsko-Bała, 4 aerologcal statons (Wrocław, Legonów, Prague, Poprad 4up1surf, 4 3 ground statons (Belsko-Bała, Kraków, Katowce, 2 aerologcal statons (Wrocław, Poprad 2pow3surf, 5 3 ground statons (Belsko-Bała, Kraków, Katowce, 2 aerologcal statons (Legonów, Prague 2prl3surf, 6 3 ground statons (Belsko-Bała, Kraków, Katowce, 1 aerologcal staton (Legonów leg3surf, 7 3 ground statons (Belsko-Bała, Kraków, Katowce, 1 aerologcal staton (Wrocław wroc3surf, 8 3 ground statons (Belsko-Bała, Kraków, Katowce, 1 aerologcal staton (Prague pra3surf, 9 3 ground statons (Belsko-Bała, Kraków, Katowce, 1 aerologcal staton (Poprad pop3surf. As a real dstrbuton of the pollutant concentratons, the calculatons for opton number 1 were selected (3 ground and 4 aerologcal statons. 3. COMPUTATION RESULTS Maps showng the dstrbuton of one-hour maxmum concentratons were made for each of the aforementoned optons. Fgure 1 shows an example of a polluton dstrbuton map.
154 P. SZCZYGŁOWSKI, M. MAZUR 5540 5535 100 5530 80 5525 40 5520 20 5515 10 5510 5 5505 345 350 355 360 365 370 375 Dstrbuton of one-hour maxmum concentratons (meteorologcal data 4up 3surf, spot emtter heght of 100 m The table presents a lst of statstcal ndcators determned on the grounds of the dstrbuton of one-hour SO 2 pollutant concentratons n the functon of the volume of avalable meteorologcal data. The hghest complance wth the standard accepted for the statstcal error was observed for 4up1surf opton (FB 0.01, and t was addtonally confrmed by the low values of ts negatve and postve components (FB FN 0.01, FB FP 0.02. A hgh FB ndcator value was also observed for 4up2surf opton (FB 0.24, FB FN 0.32, FB FP 0.08. In the case of the other optons, a relatvely low FB ndcator was not confrmed by the negatve and postve components (e.g. for leg3surf opton FB 0.11, but FB FN 0.58, FB FP 0.47. As for the correlaton coeffcent R, the values hgher than 0.7 were reached only by the followng optons: 4up1surf (R 1.00, 2pow3surf (R 0.73. Also for these optons the FAC2 ndcator was hgher than 0.5, that s for 4up1surf FAC2 and for 4up2surf FAC2 t reached the value of 1.00 and 0.85, respectvely. Statstcal analyss of the pollutant concentraton dstrbutons proved that usng the hghest possble volume of meteorologcal data s very mportant to determne the propermmssble level at a gven area. At the same tme, lack of meteorologcal data from ground statons s stll acceptable, although t changes the dstrbuton of the pollutant concentraton, whereas mssng data concernng the vertcal atmosphere profle may have a consderable effect on the results of the calculatons. Ths happened n the case of
Applcaton of BOOT statstcal package 155 optons 6 to 9, whch used nformaton on the vertcal profle of atmosphere condton obtaned only from 1 aerologcal staton. Both the form of the polluton dstrbuton and the statstcal ndcators were completely dfferent than the standard accepted. Lst of statstcal ndcators determned from dstrbuton of one-hour SO 2 pollutant concentratons, dependng on volume of meteorologcal data (spot emtter heght h 100 m Table Computatonal optons selected dependng on type of meteorologcal data used Indcator for computng ar polluton condtons 4up 3surf 4up 2surf 4up 1surf 2pow 3surf 2prl 3surf leg 3surf wroc 3surf pra 3surf pop 3surf FB 0 0.24 0.01 0.46 0.18 0.11 0.69 0.09 0.55 MG 1 1.22 1.01 3.87 0.51 0.62 0.35 0.43 3.50 NMSE 0 1.66 0.02 3.38 5.94 4.85 12.35 3.70 6.25 VG 1 1.34 1.01 2.43E+04 4.24E+05 2.93E+04 106.00 2.70E+04 3.36E+04 R 1 0.73 1.00 0.61 0.18 0.20 0.28 0.16 0.42 FAC2 1 0.85 1.00 0.43 0.27 0.33 0.27 0.33 0.36 Mean 11.5 9.0 11.5 18.2 13.7 10.3 23.6 10.4 20.2 St. dev. 18.7 14.5 19.3 32.3 27.7 19.1 58.8 13.0 40.8 Max 178 147 187 382 300 295 784 145 528 FB FN 0 0.32 0.01 0.14 0.47 0.58 0.23 0.58 0.16 FB FP 0 0.08 0.02 0.60 0.65 0.47 0.92 0.49 0.71 MOE FN 1 0.72 0.99 0.82 0.49 0.45 0.66 0.45 0.78 MOE FP 1 0.91 0.98 0.52 0.41 0.50 0.32 0.49 0.44 MG FN 1 1.29 1.02 5.55 2.61 2.40 1.40 2.11 5.58 MG FP 1 1.05 1.01 1.43 5.14 3.87 4.01 4.93 1.60 As a result of the completed calculatons and analyses t can be stated that the use of only basc statstcal ndcators does not gve a complete mage of the modellng results. For example, a low value of the FB systematc error may result n a wrong evaluaton by the researcher, but the selected model gves a good approxmaton of the real ar qualty condton. However, completely dfferent conclusons may be drawn f ts negatve and postve components are determned as well. It may turn out that some results are much overrated (large value of negatve component FB or underrated (hgh value of addtonal component FB wthn the selected area. ACKNOWLEDGEMENT The work was completed n the scope of AGH-UST statutory research, No. 11.11.150.171. REFERENCES [1] CHANG J.C., HANNA S.R., Techncal Descrptons and User s Gude for the BOOT Statstcal Model Evaluaton Software Package, Verson 2.0, 2005.
156 P. SZCZYGŁOWSKI, M. MAZUR [2] ASTM, 2000, Standard gude for statstcal evaluaton of atmospherc dsperson model performance, Amercan Socety for Testng and Materals, Desgnaton D6589-00. ASTM, 100 Barr Harbor Drve, West Conshohocken, PA 19428-2959. [3] COX W.M., TIKVART J.A., A statstcal procedure for determnng the best performng ar qualty smulaton model, Atmospherc Envronment, 1990, Vol. 24A, No. 9, 2387 2395. [4] http://edc.usgs.gov/ [5] SZCZYGŁOWSKI P., Assessment of usablty of selected Gaussan models n computng of ar polluton condton, Doctoral Dssertaton, Faculty of Mnng Surveyng and Envronmental Engneerng AGH- UST, Kraków, 2007. ZASTOSOWANIE PAKIETU STATYSTYCZNEGO BOOT W OBLICZENIACH ROZPRZESTRZENIANIA SIĘ ZANIECZYSZCZEŃ W POWIETRZU Zagadnene modelowana rozprzestrzenana sę zaneczyszczeń wymaga podjęca welu decyzj mających na celu właścwy dobór opcj modelu zwązanych z meteorologą czy współczynnkam dyspersj. Każda ch zmana powoduje różncę w wynkach oblczeń stanu zaneczyszczeń. Zastosowane w pracy nowe wskaźnk statystyczne pozwalają dokładnej ocenć wpływ zmany poszczególnych parametrów na wynk modelowana. Szczególne przydatnym, stosunkowo nowym wskaźnkam statystycznym, mogą okazać sę ujemne dodatne komponenty błędu systematycznego (FB bądź geometrycznego błędu średnego (MG, które umożlwają ocenę przeszacowana lub nedoszacowana wynków oblczeń w przypadku oblczeń rozkładu zaneczyszczeń na określonym obszarze.