Analysis of the representativeness of backward atmospheric transport modelling at different resolutions at the Takasaki IMS stationD. Arnold [1,4] D. Pino [2,3] A. Vargas [1], P. Seibert [4]

1 Institute of Energy Technologies, Technical University of Catalonia, Barcelona, Spain

2 Applied Physics Department, Technical University of Catalonia, Barcelona, Spain

3 Institute for Space Studies of Catalonia, Barcelona, Spain

4 Institute of Meteorology, University of Natural Resources and Life Sciences, Vienna, AustriaContact:delia.arnold@upc.edu

Introduction & backgroundWithin the Atmospheric Transport Modelling (ATM) topic included in the last International Scientific Studies for the CTBTO, it was recommended to visit ATM-related aspects, as ATM recently proved to be a key technology in the CTBTO verification system. Among them, the determination of the representativeness of large-scale meteorological fields used as input for ATM at different stations and its impacts on the source-receptor sensitivity (SRS) fields was highlighted, as well as possible improvements through to the use of better resolved input data. Previous studies have already shown that this issue deserves attention (Heinrich 2008, Seibert and Skomorowski 2007). In this context, through a voluntary contribution of the State Signatory Spain to the CTBTO Provisional Technical Secretariat a project started last year. In it, the representativeness of the Takasaki RN38 radionuclide station has been investigated.

MM5v3.7 set-up and configuration:

● Input data: ECMWF Op fields 0.5 x 0.5 deg

● MM5 was run in shared memory mode to include orographic shading -> SLOW RUNS

● 5 Nested domains (Fig. 3) with grid sizes from 54 to 0.67 km with a maximum of 100 x 100 grid cells

● 1- way nesting interaction

● 35 vertical levels - more levels caused instabilities in the model

● MRF boundary layer scheme, 5- layer surface scheme (tested also ETA + NOAH)

● Mix-phase moisture scheme

● No snow cover effects: sensitivity tests surprisingly showed worse skill if considered

The station is located on the premises of the Takasaki Radiation Chemistry Research Establishment of the Japan Atomic Energy Research Institute. The city is located in the western part of the Kanto Plain. The Japanese Alps (which here are about 1000 m high) close this plain off in the Northwest, while opens to the Pacific Ocean in the South and East at about 150 km from Takasaki. This topographic situation is conducive to mesoscale meteorological phenomena that may not be resolved in meteorological fields with a resolution of 1 degree as presently used in PTS operations. For this matter, two different episodes, one with westerly advection and one with calm conditions that would allow thermally-induced mesoscale phenomena to develop, have been selected. For each of them, dispersion calculations have been performed using the MM5 meteorological model (with 0.67 km grid spacing in the innermost domain) and the Lagrangian particle dispersion model FLEXPART. The obtained SRS fields will then be compared with the operational CTBTO runs based on ECMWF wind fields with a horizontal grid spacing of 1.0 deg.

MethodologyStep 0: selection of the meteorological episodes to be simulated and that will represent typical conditions occurring in the area of the station

Episodes & meteorological modelling

Modelling vs measurements:

Fig 2: table with the description of the procedure and the modelling needed in the complete study

Fig 3: 5-domain configuration for the MM5 simulations over the Takasaki region

A sample of some of the meteorological modelling results is given in this section but a similar procedure was performed for all the episodes. The first evaluation of parametrizations and set-ups (Fig 4.) shows that the inclusion of the snow cover effects actually worsens the results and was finally not considered in the runs to feed the dispersion model (Fig. 4).

Dispersion calculations

The MM5 modelling results were compared with the ECMWF data used to feed Flexpart v6.2 and with the observations at the RN38 Takasaki, the Maebashi SYNOP and the Karuizwa stations (Fig. 5). In general, MM5 simulations agree better to the measurements than ECMWF data. This is expected since mesoscale phenomena affecting the measurements are better represented in the MM5. The combination of MRF boundary layer scheme plus the simple 5-layer soil scheme gives better skill than the more complex combination ETA BL and NOAH surface scheme.

Fig 4: modelled (solid) versus observed (dashed) 2 m a.g.l. temperature at Maebashi station for episode 9. SN means snow effects considered, NS no snow effects considered

Fig5: comparison of ECMWF (blue), observations (dashed black) and modelled with different configurations (solid coloured) at the Takasaki statio

Fig 1: location of the Takasaki RN38 station

Dispersion calculations were performed with two different versions of FLEXPART: FLEXPART v6.2 fed with the same ECMWF data (FLEX-EC henceforth) used by the CTBTO and FLEXPART v6.2 – MM5v3.7 (FLEX-MM5), which uses the output of the mesoscale model. In both cases convection was switched off to speed up the runs.

SRS for the three episodes

Discussion & conclusions

Fig 6: domains used in the FLEX-EC and FLEX-MM5 simulations, solid red is the limit of the output mother domain, blue the nested one.

Meteorological modelling:

● MM5 in a shared memory mode is slow unless one has a large SM machine (45 days simulated in 11.5 days, in a 2-quad core Intel Xeon(R) CPU 2.27 GHz, 16 GB RAM) but includes orographic shading.

● MM5 is not supported anymore, therefore, its successor, WRF, would be the option.

● Some issues have appeared regarding parametrizations (snow) that should be investigated and that may be specific to the station location

Dispersion calculations:

● Several initial problems have appeared:

● For very small friction velocities, FLEX-MM5 repeatedly crashed and had to be fixed by a setting a minimum friction velocity.

● The nested domain output (not shown) suffers from the low number of particles within a grid cell making it not advisable unless a much larger number of particles can be used (with its computational implications)

● No significant differences appear when changing the release height from a 150 m column to a 10 m release point. However, this statement may not hold in general as there is interference with other model settings.

Ep7

The source receptor sensitivities (SRS) for each of the episodes and set-ups have been compared with the standard products provided by the CTBTO: Some examples are shown below where the differences in strength appear clearly visible. In some cases, the use of high resolution modelling, instead of the current 1degx1deg output, would mean not only a more accurate distinction of the different intensity areas, but also, the identification of possible areas to be considered or not considered as possible source locations. In general, CTBTO SRS strengths are smaller as was also encountered by Seibert and Skomorowski (2007)

Ep10

Ep9

CTBTO SRS (FLEXPART v5.1) FLEX-EC v 6.2 FLEX-MM5 1x1 deg FLEX-MM5 0.5x0.5 deg FLEX-MM5 0.25x0.25 deg

Recommendations & outlook

● The use of a nested output domain with very high resolution (e.g., 1 km) is not recommended as it would require a very large number of particles for obtaining statistically stable results. However, a nested domain with e.g. 0.2 deg grid size will yield additional information that would be significant.

● Using higher resolved meteorological input data together with an increase of output grid resolution provides considerably more realistic SRS fields. Since the current ECMWF products are available at 0.18 deg and even higher resolution is planned for the future, their use should be seriously considered even if there is a price in terms of computing. Then, not having limited-area meteorological models available at IDC would be less limiting.

● If mesoscale meteorological models are to be used for some stations, a model other than MM5 should be considered and checked carefully its implementation in Flexpart so that all the information from the mesoscale model is properly given to the dispersion model. Flexpart-wrf has śtill room for improvement.

● The end of support for the MM5 model and some problems encountered point towards moving from MM5 to another model, e.g. WRF, for possible future nonhydrostatic limited-area modelling tasks.

●A quantitative evaluation of these results and the ones obtained for the RN33 (Schauinsland) station are to be finished in the near future.

CTBTO FLEX-EC FLEX-MM5 1x1 FLEX-MM5 0.5x0.5 FLEX-MM5 0.25x0.25 FLEX-MM5 1x1 10

FLEXPART version 5.1 6.2 6.2 6.2 6.2 6.2

Input data (spatial and temporal resolution)

ECMWF (1 x 1 deg, 3 h)

ECMWF (1 x 1 deg, 3 h) MM5 (0.67 x 0.67 km, 1 h) MM5 (0.67 x 0.67 km, 1 h) MM5 (0.67 x 0.67 km, 1 h) MM5 (0.67 x 0.67 km, 1 h)

Receptor height 0-150 m 0-150 m 0-150 m 0-150 m 0-150 m 10 m

Particles per receptor day

560,000 600,000 (75,000 / 3 h) 600,000 (75,000 / 3 h) 600,000 (75,000 / 3 h) 600,000 (75,000 / 3h) 600,000 (75,000 / 3 h)

Height output layer 150 m 150 m 150 m 150 m 150 m 150 m

Outgrid/nested outgrid Global (1 x 1 deg) Fig 5 (red) 1 x 1 deg Fig 5 (red) 1 x 1 deg / Fig 5 (blue) 0.02 x 0.015 deg

Fig 5 (red) 0.5 x 0.5 deg Fig 5 (red) 0.25 x 0.25 deg Fig 5 (red) 1 x 1 deg / Fig 5 (blue) 0.02 x 0.015 deg

Output interval 3 hours 3 hours 3 hours 3 hours 3 hours 3 hours

Min mixing height ? 10 m 10 m 10 m 10 m 10 m

Length of simulations 15 days 15 days 15 days 15 days 15 days 15 days

Label Meteorological simulation Dispersion calculations Brief description

Ep7 13.03.2007 – 03.04.2007 28.03.2007 – 02.04.2007 Westerly advection of dust from the Gobi region (reaching the Kanto plain on the 01.04.2007)

Ep9 01.02.2007 – 28.02.2007 16.02.2007 – 22.02.2007 High pressure system with an advection of pollutants from China (21.02.2007)

Ep10 10.07.2006 – 12.08.2006 04.08.2006 – 11.08.2006 High pressure system with the developing of mesoscale circulations, including sea breezes reaching all the Kanto plain

Acknowledgement : data and information on the modelling set-up has been provided

by the CTBTO