344
Andy Delcloo and P. De Meutter
One of the possible ways to deal with this issue is to simulate the contribution
of civilian sources to the IMS station detections explicitly. To ascertain the level of
agreement that can be made between such atmospheric transport modelling and IMS
observations, two international challenges have been set up, of which the results of
the latest challenge have been published recently [1]. During the “blind phase” of this
second challenge, participants were asked to simulate the Xe-133 contribution from
the nuclear facility Ansto (near Sidney, Australia) to six IMS stations in the southern
hemisphere. Real emission data were not made available then and participants were
asked to perform simulations with unit releases. In the current “open phase” of the
challenge, participants are encouraged to perform additional in-depth research while
having available the emission data and IMS observations for research purposes.
In this paper, we redo this second international ATM challenge using the new
ERA5 dataset from ECMWF. ERA5 is unique compared to previous reanalyses
since it consists of 10 members. The spread between the members represents the
meteorological uncertainty.
54.2 Method
We have used detailed time-resolved emission data from the nuclear facility Ansto.
Flexpart [3] has been run multiple times (using each of the ERA5 ensemble members)
in forward mode to simulate Xe-133 activity concentrations at six IMS stations in the
southern hemisphere for which observations were available for the period 11 May
2013 until 10 June 2013.
The (dis)agreement between the observed and simulated Xe-133 activity concentration is quantified as in Maurer et al. [1], which is the following rank score
(Eq. (54.1)):
Rank = R
2
+ (1 − |FB|/2) + F5 + ACC
(54.1)
with R the correlation coefficient, FB the fractional bias, F5 the factor of simulated
activity concentrations within a factor 5 of the observations and ACC the accuracy
with the associated threshold being the sample-specific minimum detectable concentration (roughly the accuracy of discriminating detections from non-detection)
(Fig. 54.1).
54.3 Results
See Fig. 54.2.
Andy Delcloo and P. De Meutter
One of the possible ways to deal with this issue is to simulate the contribution
of civilian sources to the IMS station detections explicitly. To ascertain the level of
agreement that can be made between such atmospheric transport modelling and IMS
observations, two international challenges have been set up, of which the results of
the latest challenge have been published recently [1]. During the “blind phase” of this
second challenge, participants were asked to simulate the Xe-133 contribution from
the nuclear facility Ansto (near Sidney, Australia) to six IMS stations in the southern
hemisphere. Real emission data were not made available then and participants were
asked to perform simulations with unit releases. In the current “open phase” of the
challenge, participants are encouraged to perform additional in-depth research while
having available the emission data and IMS observations for research purposes.
In this paper, we redo this second international ATM challenge using the new
ERA5 dataset from ECMWF. ERA5 is unique compared to previous reanalyses
since it consists of 10 members. The spread between the members represents the
meteorological uncertainty.
54.2 Method
We have used detailed time-resolved emission data from the nuclear facility Ansto.
Flexpart [3] has been run multiple times (using each of the ERA5 ensemble members)
in forward mode to simulate Xe-133 activity concentrations at six IMS stations in the
southern hemisphere for which observations were available for the period 11 May
2013 until 10 June 2013.
The (dis)agreement between the observed and simulated Xe-133 activity concentration is quantified as in Maurer et al. [1], which is the following rank score
(Eq. (54.1)):
Rank = R
2
+ (1 − |FB|/2) + F5 + ACC
(54.1)
with R the correlation coefficient, FB the fractional bias, F5 the factor of simulated
activity concentrations within a factor 5 of the observations and ACC the accuracy
with the associated threshold being the sample-specific minimum detectable concentration (roughly the accuracy of discriminating detections from non-detection)
(Fig. 54.1).
54.3 Results
See Fig. 54.2.