Probabilistic Inverse Method for Source Localization Applied to ETEX and the 2017 Case of Ru-106 including Analyses of Sensitivity to Measurement Data
Round 1
Reviewer 1 Report
General comments:
Several times in the paper ‘large unknown (model) errors’ are mentioned. (e.g. l. 275-276). These ‘errors’ also include uncertainties, not only errors. Therefore, please be more specific in what is meant by ‘(model) errors’; e.g. temporal, spatial, etc. Secondly, the accuracy of the measurements is not taken into account in the study and, moreover, even not mentioned. Please elaborate on the measurement accuracy, preferably in relation to the model error/uncertainty. E.g. is the error in the measurements negligible to the model error?
The used adjoint source-receptor model has not been demonstrated. E.g. by calculating concentrations from a hypothetical release at several locations and times using the forward model, approximating these values using the adjoint model and subsequently doing a quantitative comparison. In addition, the paper would benefit from additional forward simulations for both the ETEX and the Ru-106 case based on the best estimate of the source term. The modelled concentrations can be statistically compared to the datasets.
Regarding the sensitivity to the number of measurements; What happens if a random number of measurements are available? Does the conclusion about non-detections (on lines 481-491) still hold, or do non-detections become relatively more important?
Other comments:
Line 59: consider also mentioning similarities between the datasets; e.g. Ru-106 decays and the ETEX tracer does not, but due to the (long) half-life of approximately 1 year of Ru-106, it can be approximated constant. In addition, in both cases only one chemical/radionuclide is considered.
Line 123: no information ON the time
Line 106: About the ‘lacking measurements’. Please specify if this only occurs at the beginning and/or end of the measurement campaign, or also in the middle.
Line 154: meteorological data goes back to September 25th 00:00 UTC. Some studies suggest an earlier release at a location more to the north. Can you increase the time period and exclude that location? What is the (integrated) correlation if you extend the analysis period?
Line 166: the method we used FOR INVERSE MODELLING relies on…
line 306: Highest posterior density region; refer to [11] where this concept was applied first to inverse atmospheric modelling.
Figures 1 and further: Especially for the maps on European scale the red diamond and blue triangle should be made clearer.
Section 3.1: Consider mentioning the distance between actual and estimated source location. (Same for section 3.2 for Mayak and NIIAR). For ETEX, compare with other studies.
Line 461: a good guess -> the best estimation
Line 486-489: this could already be mentioned in section 3.1
Line 500: that USING ONLY weekly averages…
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
.
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The comments are included in the attached file
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
In general the authors improved the manuscript and generally speaking now it is acceptable for publication.
The last suggestion from my side concerns lines 303-306 'Thus, the main difference is that the method presented here allows for releases of different durations, while the method by Tomas et al. [14] assumes an instantaneous release.'
Here authors mistakenly state that method by Tomas et al [14] assumes insrtantaneous release. Correlation as cost function indeed assumes instantaneous release. But Tomas used time integrated correlation as cost function which, as could be shown, assumes continuous release. Therefore I propose the authors to reformulate statement:
''Thus, the main difference is that the method presented here allows for releases of different durations.'
Author Response
Thank you for your review of our manuscript and for your many valuable suggestions for amendment. Your feedback is much appreciated.
We have incorporated your suggestion with no modifications in the new version of the manuscript.
