Comparison of Receptor-Oriented Dispersion Calculations Based On

Home , Schauinsland

Comparison of Receptor-Oriented Dispersion Calculations Based on ECMWF Data and Nested MM5 Simulations for the Schauinsland Monitoring Station Petra Seibert and Paul Skomorowski

Institute of Meteorology, University of Natural Resources & Appl. Life Sci. (BOKU) Peter-Jordan-Straße 82, A-1190 Vienna, Austria European Geosciences Union General Assembly 2007 Met Vienna, Austria, 16-20 April 2007 WorldWideWeb http://www.boku.ac.at/imp/envmet/; E-mail: petra.seibert @ boku.ac.at

1 Motivation • Introducing an option of z-diffusion (horizontal diffusion cannot change the height of a particle above sea level, instead of acting on surfaces parallel to the topography). This feature is important for high-resolution simulations over The German Federal Office for Radiation Protection (Bundesamt fur¨ Strahlenschutz) maintains monitoring sites for air- mountain topography. borne radioactivity measurements. For our work, we are focusing on the mountain station Schauinsland (1200 m) and This new version will soon be made public, at the moment a beta version can be obtained from the author. the station Freiburg im Breisgau (270 m), located in the Rhine Valley (Fig. 1 (left)). Schauinsland is also the only station Setup used: located in Central Europe among the 80 radionuclide monitoring stations being built up for the operational monito- • backward mode ring of the Comprehensive Nuclear-Test-Ban Treaty (CTBT) by its Provisional Technical Secretariate (PTS) in Vienna. • temporal resolution of input meteorological fields 3 h (for ECMWF), 1 h (for MM5) In addition, the Federal Office of Environment (Umweltbundesamt) carries out monitoring of background atmospheric • temporal resolution of output: 3 h ◦ conditions at Schauinsland. • horizontal resolution of output: 0.5 and (nest over Central Europe) approx. 8 km Observations in the past have shown that Schauinsland and Freiburg may sample different air masses depending on the • vertical layer thickness for ouput: 150 m with ECMWF input, 25 with MM5 input meteorological situation. For the investigation of atmospheric trace substances, receptor-oriented dispersion models with • number of particles per receptor day: 150,000 / 180,000 meteorological input data from ECMWF analyses are presently used by the PTS. It is obvious that on this base the differences in the airmass origin are probably not captured. The work presented here was thus guided by the following questions: 3 Selected episodes • Can the origin of air masses at the two stations be simulated better based on a high-resolution, nested meteorological A total of 10 (nomiminally 11) periods with duration from 5 to 10 days have been selected. On this poster, we show model than a based on a global model? results of four periods to illustrate the main findings: • For which weather patterns are differences significant? Could such patterns be diagnosed from available data? • Episode 3 from 3–11 Nov 2004 with strong northwesterly, later northeasterly flow • Is it possible to obtain an improved consideration of orographic effects with simplified methods such as using an • Episode 8 from 31 Jan – 8 Feb 2004 with strong southwesterly flow elevated receptor position for the mountain station? • Episode 5 from 19–24 June 2005 with anticyclonic summer conditions These questions have been answered by simulating several episodes with FLEXPART, based on ECMWF fields and MM5 • Episode 9 from 8–17 Dec 2004 with anticyclonic winter conditions (inversion) calculations.

49˚N 4 Results of meteorological simulations

48˚00'N Freiburg 48˚N Episode 3, northwesterly (later northeasterly) ﬂow Episode 8, southwesterly ﬂow ECMWF MM5(D1) MM5(D5) Freiburg ECMWF MM5(D1) MM5(D5) Schauinsland ECMWF MM5(D1) MM5(D5) Freiburg ECMWF MM5(D1) MM5(D5) Schauinsland Schauinsland 15 10 15 10

10 10 5 5

5 5 0 47˚N 0 Temperatur [°C] Temperatur [°C] Temperatur [°C] Temperatur [°C] −5 0 km 0

0 5 10 ECMWF − Obs. MM5(D1)MM5(D1) − Obs.MM5(D5) MM5(D5)Freiburg − Obs. Obs. ECMWF − Obs. MM5(D1)MM5(D1) − Obs.MM5(D5) MM5(D5)Schauinsland − Obs. Obs. 47˚45'N ECMWF − Obs. MM5(D1)MM5(D1) − Obs.MM5(D5) MM5(D5)Freiburg − Obs. Obs. ECMWF − Obs. MM5(D1)MM5(D1) − Obs.MM5(D5) MM5(D5)Schauinsland − Obs. Obs. 10 7˚45'E 8˚00'E 0 5 5 5 10 10 0 0 −5 0

46˚N 5 5

−5 Differenz [°C] Differenz [°C] Differenz [°C] Differenz [°C] −5 −10

−5 Windgeschw. [m/s] Windgeschw. [m/s] Windgeschw. [m/s] Windgeschw. [m/s] 0 0 0 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 ECMWF − Obs. MM5(D1) − Obs. MM5(D5) − Obs. ECMWF − Obs. MM5(D1) − Obs. MM5(D5) − Obs. ECMWF − Obs. MM5(D1) − Obs. MM5(D5) − Obs. ECMWF − Obs. MM5(D1) − Obs. MM5(D5) − Obs. ECMWF MM5(D1) MM5(D5) Freiburg Obs. ECMWF MM5(D1) MM5(D5) Schauinsland Obs. 03ECMWF NOV 04 NOV 05MM5(D1) NOV 06 NOV 07MM5(D5) NOV 08 NOV Freiburg09 NOV Obs.10 NOV11 NOV 03ECMWF NOV 04 NOV 05MM5(D1) NOV 06 NOV 07MM5(D5) NOV 08 NOV Schauinsland09 NOV 10 Obs. NOV11 NOV 31 JÄN 01 FEB 02 FEB 03 FEB 04 FEB 05 FEB 06 FEB 07 FEB 31 JÄN 01 FEB 02 FEB 03 FEB 04 FEB 05 FEB 06 FEB 07 FEB 360 2004 360 2004 2004 3005 2004 4˚E 5˚E 6˚E 7˚E 8˚E 9˚E 5 5 300 300 300

240 2400 2400 2400 180 180 0 120 120 180 180

Differenz [m/s] Differenz [m/s] −5

Differenz [m/s] Differenz [m/s] −5 Windrichtung [°] Windrichtung [°]

Windrichtung [°] 60 Windrichtung [°] 60 −5 120

[K] [K]

90 5 5 905 905 60 ∆Θ ∆Θ ∆Θ ∆Θ 60 60 60 Differenz [°] Differenz [°]

Differenz [°] Differenz [°] 30 300 0 300 300

0 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 0 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 03 12NOV 0 04 NOV12 0 05 NOV12 0 06 12NOV 0 07 12NOV 0 08 12NOV 0 09 12NOV 0 10 12NOV110 NOV 03 12NOV 0 04 NOV12 0 05 NOV12 0 06 12NOV 0 07 12NOV 0 08 12NOV 0 09 12NOV 0 10 12NOV110 NOV 31 JÄN 01 FEB 02 FEB 03 FEB 04 FEB 05 FEB 06 FEB 07 FEB 31 JÄN 01 FEB 02 FEB 03 FEB 04 FEB 05 FEB 06 FEB 07 FEB 03 NOV 04 NOV 05 NOV 06 NOV 200407 NOV 08 NOV 09 NOV 10 NOV11 NOV 03 NOV 04 NOV 05 NOV 06 NOV 200407 NOV 08 NOV 09 NOV 10 NOV11 NOV 31 JÄN 01 FEB 02 FEB 03 2004FEB 04 FEB 05 FEB 06 FEB 07 FEB 31 JÄN 01 FEB 02 FEB 03 2004FEB 04 FEB 05 FEB 06 FEB 07 FEB 2004 2004 2004 2004 2 Model description Episode 4, anticyclonic conditions in summer Episode 9, anticyclonic conditions in winter with inversion ECMWF MM5(D1) MM5(D5) Freiburg ECMWF MM5(D1) MM5(D5) Schauinsland ECMWF MM5(D1) MM5(D5) Freiburg ECMWF MM5(D1) MM5(D5) Schauinsland 10 25

30 10 20 5 5 25 0 2.1 MM5 meteorological model 15 0 20 −5 Temperatur [°C] Temperatur [°C] Temperatur [°C] Temperatur [°C] 10 15 −5 −10 ECMWF − Obs. MM5(D1)MM5(D1) − Obs.MM5(D5) MM5(D5)Freiburg − Obs. Obs. ECMWF − Obs. MM5(D1)MM5(D1) − Obs.MM5(D5) MM5(D5)Schauinsland − Obs. Obs. ECMWF − Obs. MM5(D1)MM5(D1) − Obs.MM5(D5) MM5(D5)Freiburg − Obs. Obs. ECMWF − Obs. MM5(D1)MM5(D1) − Obs.MM5(D5) MM5(D5)Schauinsland − Obs. Obs. The nohydrostatic meteorological model MM5 (Grell et al., 1994) is used in this study to generate high-resolution 0 10 15 0 0 10 −5 5 5 5 10 meteorolgical ﬁelds. −10 −5 05 −5

Differenz [°C] Differenz [°C] Differenz [°C] Differenz [°C] 5 ◦ ◦ −15 • × Windgeschw. [m/s] Windgeschw. [m/s] Windgeschw. [m/s] −5 Windgeschw. [m/s] −10 initial and boundary condition: 1 1 ECMWF data 0 0 0 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 ECMWF − Obs. MM5(D1) − Obs. MM5(D5) − Obs. ECMWF − Obs. MM5(D1) − Obs. MM5(D5) − Obs. ECMWF − Obs. MM5(D1) − Obs. MM5(D5) − Obs. ECMWF − Obs. MM5(D1) − Obs. MM5(D5) − Obs. ECMWF19 JUN 20 JUNMM5(D1)21 JUN MM5(D5)22 JUN 23 JUNFreiburg Obs.24 JUN 25 JUN ECMWF19 JUN 20 JUNMM5(D1)21 JUN MM5(D5)22 JUN 23 JUNSchauinsland24 JUN Obs. 25 JUN 08 ECMWFDEZ 09 DEZ 10 DEZMM5(D1)11 DEZ 12 DEZ 13MM5(D5) DEZ 14 DEZ 15 DEZFreiburg16 DEZ Obs. 17 DEZ18 DEZ 08 ECMWFDEZ 09 DEZ 10 DEZMM5(D1)11 DEZ 12 DEZ 13MM5(D5) DEZ 14 DEZ 15 DEZSchauinsland16 DEZ Obs.17 DEZ18 DEZ 3605 2005 360 2005 360 2004 2004 • 5 30010 Mother and 4 nests (Fig. 1, right) 300 300 300 0 5 2400 240 240 240 0 0 180 180 180 180 • −5 −5 horizontal resolution (D1–D5): 54, 18, 6, 2 and 0.67 km grid distance 120 120 120−5 120 −5 Differenz [m/s] Differenz [m/s] Differenz [m/s] Differenz [m/s] −10

Windrichtung [°] 60 Windrichtung [°] 60 Windrichtung [°] 60 Windrichtung [°] 60 −10 −10 −15 • 0 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 2-way nesting |ECMWFObs. (Schauinsland − Obs.| − Freiburg)|MM5(D1) − Obs.| MM5−D5 (Schauinsland|MM5(D5) − Obs.|− Freiburg) |ECMWFObs. (Schauinsland − Obs.| − Freiburg)|MM5(D1) − Obs.| MM5−D5 (Schauinsland|MM5(D5) − Obs.|− Freiburg) |ECMWFObs. (Schauinsland − Obs.| − Freiburg)|MM5(D1) − Obs.| MM5−D5 (Schauinsland|MM5(D5) − Obs.|− Freiburg) |ECMWFObs. (Schauinsland − Obs.| − Freiburg)|MM5(D1) − Obs.| MM5−D5 (Schauinsland|MM5(D5) − Obs.|− Freiburg) 19 JUN 20 JUN 21 JUN 22 JUN 23 JUN 24 JUN 25 JUN 19 JUN 20 JUN 21 JUN 22 JUN 23 JUN 24 JUN 25 JUN 08 DEZ 09 DEZ 10 DEZ 11 DEZ 12 DEZ 13 DEZ 14 DEZ 15 DEZ 16 DEZ 17 DEZ18 DEZ 08 DEZ 09 DEZ 10 DEZ 11 DEZ 12 DEZ 13 DEZ 14 DEZ 15 DEZ 16 DEZ 17 DEZ18 DEZ 180 180 180 10 2005 10 2005 2004 2004 • 150 150 15020 15020 120 NOAH land-surface scheme, Kain-Fritsch convection scheme for D1 and D2, Reisner 1 for precipitation, cloud-radiation 120 12015 12015 5 5 [K] [K] [K] [K] 90 90 90 90 10 10 ∆Θ ∆Θ ∆Θ ∆Θ 60 60 60 60

Differenz [°] Differenz [°] Differenz [°] 5 Differenz [°] 5 0 0 scheme with shadowing effects, radiative upper boundary condition for dynamics 30 30 30 30 0 0 ◦ ◦ 0 0 12 0 12 0 12 0 12 0 12 0 12 0 0 0 12 0 12 0 12 0 12 0 12 0 12 0 0 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 0 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 12 0 • × 1912 JUN 0 2012 JUN 0 2112 JUN 0 22 12JUN 0 2312 JUN 0 2412 JUN 250 JUN 1912 JUN 0 2012 JUN 0 2112 JUN 0 22 12JUN 0 2312 JUN 0 2412 JUN 250 JUN 08 12DEZ 0 09 12DEZ 0 10 12DEZ 0 11 12DEZ 0 12 12DEZ 0 13 12DEZ 0 14 12DEZ 0 15 12DEZ 0 16 12DEZ 0 17 12DEZ180 DEZ 08 12DEZ 0 09 12DEZ 0 10 12DEZ 0 11 12DEZ 0 12 12DEZ 0 13 12DEZ 0 14 12DEZ 0 15 12DEZ 0 16 12DEZ 0 17 12DEZ180 DEZ Grid-nudging every 3 h with ECMWF data (1 1 resolution) on the coarse domain 19 JUN 20 JUN 21 JUN 200522 JUN 23 JUN 24 JUN 25 JUN 19 JUN 20 JUN 21 JUN 200522 JUN 23 JUN 24 JUN 25 JUN 08 DEZ 09 DEZ 10 DEZ 11 DEZ 12 DEZ200413 DEZ 14 DEZ 15 DEZ 16 DEZ 17 DEZ18 DEZ 08 DEZ 09 DEZ 10 DEZ 11 DEZ 12 DEZ200413 DEZ 14 DEZ 15 DEZ 16 DEZ 17 DEZ18 DEZ 2005 2005 ◦ 2004 2004 Figure 2: Comparison of observations, ECMWF analysis on 1 , MM5 domain 1 (54 km), and MM5 domain 5 (0.67 km) of temperature and wind in Freiburg (left column) and Schauinsland (right column). 2.2 Lagrangian particle model FLEXPART The last row gives the potential temperature difference as a stability index, it is repeated to provide FLEXPART (Stohl et al., 2005) is a Lagrangian particle dispersion model (LPDM) initially developed at BOKU-Met, the time scale for the right column, too. based to some extent on the older trajectory model FLEXTRA. It simulates the transport, turbulent diffusion, dry and wet deposition, convection and radioactive decay of air pollutants released from any kind of sources. There is a special The comparisons in Figure 2 largely speaks by themselves. In episode 8, MM5-D5 is overestimating the southwesterly option for backward calculations, giving a source-receptor relationship (sensitivity), which is closely related to residence wind at Schauinsland because of local sheltering of the station (at Feldberg [not shown] MM5 still underestimates!) In times, instead of a concentration (see Seibert and Frank (2004) for details). It was used in the latest version 6.2. episode 9, only MM5-D5 can simulate the inversion, though it is dissolved too early. A new MM5 run with z-diffusion for For application with MM5, the old FLEXPART-MM5 version based on FLEXPART v3 was reprogrammed to reflect the temperature hopefully solves the problem. MM5 is generally too cold for daytime maxima under fair weather, the reason latest developments. Two specific improvements have also been introduced: is unknown. In-valley (westerly, during daytime) and out-valley (easterly, nighttime) local winds are simulated well by • Replacing the constant map scale factor for the MM5 grid by exact an transformation in geographical grid MM5-D5 as seen in the first days of episode 5. 5 Results of dispersion simulations indicate nuclear facilities, the isotope factory at Fleurus in Belgium is marked additionally.

Table 1: Comparison of the s–receptor relationship (correlation oeﬃcients) between Schauinsland 20041106 06 (Episode 3) 20040203 06 (Episode 8) (Freiburg) and Fleurus (Belgium) during the studied episode as calculated by diﬀerent models.

FREIBURG MM5 FREIBURG MM5 Only those days when at least one model has non-zero R FRB-0 SIM-0 SIL-0 SIL-1 SIL-2 HYSPLIT CTBTO values are considered (N=22 instead of 44). All values ◦ FRB-0 1.000 0.983 0.613 0.662 0.559 0.243 0.803 based on 1 s–r data. FRB-0 and SIM-0 refer to MM5- SIM-0 — 1.000 0.567 0.625 0.518 0.277 0.791 FLEXPART simulations for Freiburg and Schauinsland, SIL-0 — — 1.000 0.980 0.907 0.615 0.763 SIL refers to ECMWF-FLEXPART simulations (-0 re- SIL-1 — — — 1.000 0.944 0.527 0.783 lease at 5 m agl, -2 release at true height asl, -1 in- SCHAUINSLAND MM5 SCHAUINSLAND MM5 SIL-2 — — — — 1.000 0.502 0.792 between), HYSPLIT is simple evaluation of HYSPLIT HYSPL — — — — — 1.000 0.610 trajectory reseidence times, CTBTO is the ECMWF- CTBTO — — — — — — 1.000 FLEXPART simulation by CTBTO/PTS with release distributed 0-200 m agl.

Table 1 gives an overview of the similarity between diﬀerent simulations for a speciﬁc potential source in the Northwest

SCHAUINSLAND-0 ECMWF SCHAUINSLAND-0 ECMWF of our receptor. The MM5-based simulation do not differ very little for this situation between Freiburg and Schauinsland. Among the ECMWF-based simulations, the one with an intermediate release level is closest to the MM5-based run. The simple HYSPLIT trajectory evaluation compares worst. CTBTO’s ECMWF-based runs have a relatively good correlation with the others, because they have a relatively broad influence region, probably as they release the computational particles initally over a 200-m layer. However, correlation coefficients of quasi-log-normally distributed data a quite sensitive to the few highest values.

Northwesterly flow situation. All simulations are relatively Southwesterly flow situation. MM5-based simulation 6 Conclusions and outlook similar. The isotope factory at Fleurus (Belgium) is a good shows a pronounced weakening of the s–r relationship for Conclusions: candidate for the observed xenon peak on this day, though Schauinsland mountain as compared to Freiburg becau- • The MM5 model is does produce features such as the channelled flow in the Rhine Valley, realistic orographic thermal there are a number of nuclear power plants (NPPs) that se of the high wind velocities on the mountains. The flow circulations and inversions under anticyclonic conditions in the cold season. Despite a cold bias for fair-weather dayti- could also be responsible. MM5-based simulations show channeling in the gap between the Vosges and Jura moun- me temperatures (whose reason could not be clarified) it creates mesoscale wind fields which are clearly better than more structures, and do not show the extension of the tains is nicely visible. ECMWF analyses. This was anticipated. influence region towards central France. • The MM5 model showed that in southwesterly flow conditions a low-level jet passing through the gap between the Jura and Vosges mountains and impinging on the mountains of the southern Black Forest is being produced. Obser- 20050623 06 (Episode 5) 20041214 06 (Episode 9) vations at Feldberg and Schauinsland confirm this result, while the feature is missing from ECMWF fields. This was FREIBURG MM5 FREIBURG MM5 unexpected. • On a scale of several hundreds of kilometres, the rough shape of the influence areas (fields-of-regard in CTBTO ter- minology) produced on ECMWF and MM5 base is relatively similar, though there mayb be significant differences due to orography in detail, which can be important or not, depending on the circumstances. • The least differences are found in fast northwesterly flows. • Due to the mentioned jet, the source-receptor sensitivity for Schauinsland is strongly overestimated with ECMWF-based simulations of southwesterly flows. SCHAUINSLAND MM5 SCHAUINSLAND MM5 • Under strong inversion conditions, ECMWF-based simulations deviate strongly from MM5-based simulations. • Variation of the receptor height only marginally improves ECMWF-based simulations. A receptor height between the true height above sea level of the station and the model topography performs best. CTBTO may wish to consider these results in the interpretation of their measurements, and for the siting of those stations not yet constructed. Outlook SCHAUINSLAND-0 ECMWF SCHAUINSLAND-0 ECMWF There are many ideas for deepening and continuing this work, e.g.: • Study reasons for cold bias in MM5 simulations and other deficiencies and try to improve them • Run FLEXPART-MM5 with a very high-resolution output grid (order of 1 km) to show impact of local circulations more clearly • Full quantitative comparison of source-receptor fields from different simulations as a function of transport distance etc. • Application of the toolbox created to other mountain observatories, especially those in high mountains (the next level Anticyclonic situation in summer. Corresponding to the Anticyclonic situation in winter, strong inversion between of difficulty). Mountains are often preferred sites for background atmospheric monitoring, but from the point of view lower wind velocities and more inhomogeneous flow pat- Schauinsland and Freiburg. ECMWF-based simulations of the modeller these are less-than-ideal sites! terns, a large area is identified as field of regard, with the show an unrealistic advection of the air directly over the strongest influence near the receptors. The MM5 simulati- Alps. MM5-based modelling shows that for Freiburg the Acknowledgments ons concentrate the influence along the Rhine valley. This main influence regions are the low-lying areas north of the holds also for Schauinsland as convection and slope winds Alps, especially Rhine Valley, Swiss Midlands, Lake of Con- This work is financed by the German Federal Office for Radiation Protection (Bundesamt fur¨ Strahlenschutz) Project StSch 4478 (Untersuchung der orographischen Besonderheiten der Probennahmestellen Schauinsland und Freiburg und deren Auswirkung auf die Genauigkeit von adjun- bring the air from the Rhine Valley up to the peaks. Only stance and southeastern France. On Schauinsland, the in- gierten atmosphärischen Ausbreitungsrechnungen ( Einzugsgebiete“)). Access to ECMWF data is granted through the Special Project MOTT. ” the MM5 simulations show a very weak contribution along fluence area goes directly southwards over the crests of The following institutes are greatfully acknowledged for providing measurement data: Umweltbundesamt, Landesamt fur¨ Umwelt, Messungen the Rhone valley, which in many simulations manifests as the Black Forest, the influence of the lower areas is wea- und Naturschutz Baden-Wurttemb¨ erg, Zentralanstalt fur¨ Meteorologie und Geodynamik, Bundesamt fur¨ Strahlenschutz a flow guide for air later reaching southern Germany or ker. Again, the Rhone valley turns out as a weak pathway Austria. with possible influence of areas south of the Alps. These References air masses rather flow around the Alps, not over the Alps as shown by ECMWF-based simulations. Grell, G.,J. Dudhia, D. Stauffer (1994): A Description of the Fifth Generation Penn State/NCAR Mesoscale Model (MM5)., NCAR, TN-398+STR. Seibert, P. and A. Frank (2004): Source-receptor matrix calculation with a Lagrangian particle dispersion model in backward mode, Atmos. Figure 3. Source-receptor relationships for 24-hour recptor days ending at the given date. The rows Chem. Phys., 4, 51-63. are simulations with MM5-FLEXPART for Freiburg and Schauinsland, and ECMWF-FLEXPART for Stohl, A., C. Forster, A. Frank, P. Seibert, G. Wotawa (2005): Technical note: The Lagrangian particle dispersion model FLEXPART version Schauinsland (receptor height 5 m agl). The zoom shows a part of the nested output domain. Crosses 6.2. Atmos. Chem. Phys., 5, 2461-2474.