Review of “Spatio-temporal variation of radionuclide dispersion from nuclear power plant accidents using FLEXPART ensemble modeling”
Impact studies based on a large number of atmospheric transport simulations for hypothetical nuclear accidents in the Middle East with new reactors being built are clearly needed. However, the paper contains numerous scientific deficiencies which make a major revision inevitable.
Major concerns:
1) l. 157: “Modeled concentrations are vertically integrated…” I see no reason what this should be good for (if not for a lack of particles in the surface layer following from too few particles being released). For any impact analysis, it is only layer (mostly surface up to ~ 150 m) concentrations (or doses) that count. Moreover, this approach makes subsections 3.2 and 3.3 incomparable. Model inter-comparison or evaluation will be biased to better outcomes because the vertical cloud or particle positions are no longer important in a total column comparison and upper layer concentrations are less impacted by tricky boundary layer processes. Total columns are mainly used in air quality studies for contrasting model values to satellite observations and this is the first time that I find total column values in a radionuclide dispersion study. The problem culminates in l. 390: “…is of importance in preparedness programs. Figure 9-A shows the frequency of occurrences (FoO) of 131I column densities…” If the authors seriously refer to column loads in the context of nuclear accident preparedness programs, there seems to be a lack of understanding.
2) Meteorological input data: l. 166/167: “…CFSv2 can be used to provide 6-hourly forecast inputs for FLEXPART at the spatial resolution of 0.5 degrees…” Given the scale of the study the use of this data set should be completely avoided. The emitter in UAE and the eastern boarder of the receptor area are just ~1.5° (i.e., three grid boxes) apart. It comes as no surprise that a lot of spatial gradients are lost in Figures 5 and 6. I fear that a lot of differences the authors are discussing between this member and the others are mere artefacts caused by the poor spatial and temporal resolution of the CFSv2 data. The authors even emphasize the problem of low resolution themselves several times in the paper. The CFSv2 data should by no means be used for the purpose of forecasting in the case of an emergency. Additionally, the (re-)analysis ensemble members are correlated (especially ERA5-WRF and FNL-WRF) and not well suited to quantify meteorological uncertainty.
3) The authors often mix integrated with maximum I-131 concentrations or integrated Cs-137 deposition with completion of C-137 deposition. It is only the maximum concentration and the completion of deposition which can be reasonably contrasted to particle release times. Any time-integrated value naturally loses its time stamp information. Terms are correctly introduced by the authors in l. 326-328. But the statement in l. 348-349 already starts to confuse the reader. It culminates in the statement in l. 371/372, l. 523 or in the caption of Figures 7 and 8 that the integrated I-131 [muSv] concentration is converted to maximum hourly dose [muSv/h].
4) The comparison with the results of Maurer et al. (2018) (l. 441/442, 447-450 and l. 569-571) to me demonstrates that the authors have a poor knowledge of atmospheric radionuclide dispersion modelling. Not only that Xe-133 – in contrast to I-131 or Cs-137 – is an inert tracer which undergoes no deposition (thus being easier to model) and that Maurer et al. (2018) employed atmospheric dispersion runs based on NWP (re-)analyses only (no forecast was involved), the scale of this study was completely different. Whereas the scale of the present study covers 200 or 300 km the source and the receptors in Maurer et al. (2018) are mostly several 1000 km (up to 17000 km) apart! Finally, IMS sampling times are not one hour but rather 12 to 24 hours. So, I am sorry to say, this is comparing apples with pears.
The abstract should cover the full paper in a balanced way. E.g., there is no word about the dose calculations. The English is sometimes poor and sentences hard to understand. Tenses are often switched arbitrarily (present tense versus past tense). Some of the minor issues below are mere suggestions, others clearly reflect a lack of care in terms of contents or wording. Sometimes authors are even contradicting themselves within the paper. It is urgently needed to increase coherence and consistency within the paper.
Minor issues:
l. 11 ff.: “intensity of radionuclides” -> “(activity) concentrations of radionuclides”. No proper wording. Occurs numerous times throughout the paper. Remove all occurrences of “intensity” or “intensities” in the very same context in the paper.
l. 12, l. 50, l. 182, l. 184: “a fictitious accident” -> “fictitious accidents”. You investigated in fact 365 scenarios.
l. 23/24: The difference in input PBLH explains well the inter-member variations of simulated radionuclide concentrations. See major concern 2). The PBLH alone will not explain all the differences.
l. 24/25: “Simulated concentrations were found with the same level of consistency as reported for real case studies”. See major concern 4).
l. 38: “…from the Fukushima nuclear power plant accident…”: Somehow a contradiction to what is said above (“…case studies of real accidents of the order of a few days are not suited to examine the impact of seasonal (atmospheric) changes on the radionuclide dispersion.”). This was a real accident and the effect of East Asian northeast monsoon on radionuclide transport was evidently studied.
l. 40: “northern hemisphere” -> “Northern Hemisphere”
l. 48: “…transport and surface concentration and deposition…” -> “transport, surface concentration and deposition”
l. 53: “131 I concentration”: The main reason for the significance of I-131 are thyroid doses, not just the high activity of I-131.
l. 55 and l. 86: Please add “Pisso et al., 2019” to the references.
l. 63: “lack of accuracy”. Please specify. With regard to the internal modelling time step?
l. 68: “perturbations”. Please specify and/or provide a reference.
l. 69: “suite of different meteorological models”: Difference in practice may be limited due to NWP models being similar to each other and thus an ensemble can easily give an incomplete picture of meteorological uncertainty. See for example your ERA5-WFR versus FNL-WRF inputs.
l. 71-73 “…is compared against the (re)analysis members. (Re)analysis-based simulations are expected to be closer to (unavailable in a real-world scenario) actual values than forecast-based ones (Leadbetter et al., 2022).” -> “…is compared against (re)analysis members. (Re)analysis-based simulations (unavailable in a real-world scenario) are expected to be closer to actual values than forecast-based ones (Leadbetter et al., 2022).”
l. 96/97: Please check the suitability of references. Tipka et al. describes the preprocessing of ECMWF fields before being ingested into FLEXPART, the other two papers deal with convection. But likely not removal processes (decay and deposition).
l. 113: “…by (Hanna, 1982).” -> “…by Hanna (1982).”
l. 114: “…method as Maryon (1998) is followed.” -> “…method as in Maryon (1998) is followed.”
l. 115: “In dispersion modeling…” -> “In dispersion modeling of radionuclides…”
l. 118: “…from already calculated the radionuclide…” -> “…from the radionuclide…”
l. 123: “…the Weather Research and Forecasting (WRF)…” -> “…the Weather Research and Forecasting (WRF) model…”
l. 126 + 127: “cloudy pixels” -> “cloudy grid cells”
l. 130: “…any grid cells beneath these grid cells…” Do you mean beneath a value of 80% or beneath in terms of altitude?
l. 131: “…cloud water mixing ratio…” This field is now used to distinguish between below- and in-cloud grid cells. Please state this explicitly.
l. 141: “Eq7” –> “Eq. 7”
l. 148: “…estimate the transport…” -> “…estimate the temporal characteristics of transport…”
l. 150: “…is added to the history output grids that have a horizontal resolution of 10 km…” -> “…is added to the output grid that has a horizontal resolution of 10 km…”
l. 152/153 & caption of Fig. 3: “smooth density estimates”. In how far smooth? Were the distributions smoothed?
l. 153: “normalized difference” -> “maximum normalized difference”
l. 170/171: “…reanalysis data that covers from January 1, 1950, to nearly the present. They are produced at a spatial resolution of about 31 km at hourly time steps.” -> “…reanalysis data that covers January 1, 1950, to nearly the present. They are produced at a spatial resolution of about 0.25° at hourly time steps.
l. 175/176: “A single simulation code is built for each meteorological dataset to be ingested by FLEXPART…” I do not really understand why. In FLEXPART 10.4 there is even only one executable for both, ECMWF and NCEP data. So I wonder why there should be different codes for two NCEP data sets. Finally, authors are contradicting themselves in l. 303 (“same simulation code”).
Table 1 needs to be improved. Use capitals consistently in the header, e.g., “Temporal resolution”. First line with entries: Better remove “x”. Second line with entries: Add “0.25°” and “3-hourly” for FNL. Better remove “x”.
l. 184: “May 2011” - > “March 2011”
l. 193: “This experiment has been performed…” -> “This experiment was performed…”
l. 194/195: “For the diurnal and seasonal stratification of simulations…” –> “For stating particle ages related to simulations…”
Caption of Figure. 1: “Figure 1 A is the study area embracing the B-NPP (red square) and the state of Qatar. The base map and overlaying information are taken from Google Earth. B is the schematic illustration of the LPDM simulation cycle.” -> “Figure 1. A: Study area embracing the B-NPP (red square) and the state of Qatar. The base map and overlaying information are taken from Google Earth. B: Schematic illustration of the LPDM simulation cycle.” I suggest a similar style for all the figures. Use full stops and colons accordingly.
l. 205: “…since it resides mostly in the gaseous phase and has a short half-life of 8 days” This is not true according to my knowledge. The best assumption is a 50:50 partition between gaseous and particulate iodine. Again, the thyroid doses are an important aspect of iodine.
l. 208: “…season and time of day in which…” -> …season and time of day in/at which…”
Figures 2 and S1: “FNL-WRF” -> “Forecast”. Are times in Fig. 2B/S1B UTC-times? Y-axis: “all intensity” -> “all loads”. Caption of Figure 2: “Figure 2 A: the smooth density estimates of air parcel ages corresponding to all intensities of 131I column densities (top row) and of those above the 66th percentile (bottom row). B: the same as A, but for four times of the day.” -> “Figure 2. A: Density estimates of air parcel ages corresponding to all 131I column loads (top row) and of those above the 66th percentile (bottom row). B: The same as A, but for four times of the day.”
l. 210-212: “…differences in the transport characteristics of these radionuclides, such as the wet and dry deposition rate and radioactive decay, are not large enough to cause the abundance of cases where 131I and 137Cs particles are not present in a common grid.” This is even not to be expected, because removal process have no influence on particle positions in LPDM. Just on particle masses. However, the transport for gases and particulates (undergo gravitational settling) will be different to some extent.
Figure S2: “…lines are corresponding kernel density estimates of counts.” Completely obscure to me. Needs (sufficient) explanation in the text.
l. 216: “All ensemble members in all seasons simulated an abrupt increase…” I think this is not true for summer.
l. 225: “In addition to having finer spatial resolution than CFSv2, FNL assimilates observations like ERA5.” Also CFSv2 needs to assimilate observations at some point, i.e., at the analysis time step based on which the forecast is made.
l. 226: “distributed in a wider range” Did you check the significance of this feature?
l. 226/227: “…in FNL…” -> “in FNL-based FLEXPART simulations” or “FNL simulated…” -> “FLEXPART-FNL simulated…” FNL does not yield the output directly. It needs FLEXPART in addition. Numerous analogous formulations throughout the paper. Please adapt them as outlined.
l. 228 + l. 357: “…air parcels reaching further receptors…” -> “…air parcels reaching receptors further away…”
l. 232: “column (mass) densities” (or density). Please note that this is not a proper term. Occurs numerous times in the paper. It has to be “column load”. If you vertically integrate concentrations [Bq/m3] you will end up with column loads [Bq/m2]. However, as stated above, column loads are not used in radionuclide studies.
l. 242/243: “For low intensities of both radionuclides, however, the age distributions are almost the same as that seen in all intensities. The peak of newly arriving air parcels in all intensities occurs earlier in spring than in other seasons.” -> “For low column loads of both radionuclides, however, the age distributions are almost the same as that seen for all column loads. The peak of fast arriving air parcels for all column loads occurs earlier in spring than in other seasons.” I would rather say spring, winter and to some degree also fall behave very similar in terms of fast arriving particles.
l.244: “…in low and all intensity column densities…” -> “…for low and all column loads…”
l. 245: “…led to the more distant transport of radionuclides to northern parts of the study area…” Can this really be concluded that easily? See the options stated at the end of the above paper paragraph.
l. 249: “…the age distribution of the particles that is released…” -> “…the age distribution of the particles that are released…”
l. 250/251: “…the number of long-lived air parcels (including all intensities of concentrations) increased in all members…” -> “…the number of long-lived air parcels (including all levels of column loads) increased for all members…” Anyway, not true for Cs and FLXPART-FNL (Figure S1).
l. 253: “longer Lagrangian particle ages” -> “higher Lagrangian particle ages”
l. 254: “less time to travel by the end of simulation period” -> “less time to travel until the end of simulation period”
l. 257: “…between 20 and 70 hours after release…” I would rather say “…up to 40 or 50 hours after the release…” for moderate and high column loads.”
l. 257/258: “Consequently, the difference in release time of Lagrangian particles is not significantly affected by their age spectrum.” Sentence makes no sense. If, then the other way round.
l. 258/259: “…FNL have simulated a larger number of shorter-aged air parcels…” -> …FLEXPART-FNL simulated a larger number of shorter-lived air parcels…”
l. 261/262: “Like seasonal distribution, the diurnal variations of air parcel ages for high concentrations are very similar in all members.” -> “Like for seasonal distributions, the diurnal variations of air parcel ages for high column loads are very similar for all members.”
l. 276: This equation is quite obscure to me. If j=1 (first simulation time step) "a" can only be equal to one (j=a=1). There cannot be particles older than one hour at this stage of the simulation. For j >1, e.g. j=50, i can only range between 26 and 50 and cannot adopt a value equal to 1 in the nominator given that particles were released over the first 24 hours of the simulation.
l. 277: “…in winter when both dry and wet deposition occur in the study area. We found very similar results in other seasons (Fig. S5).” Something got quite confused here. Comparing Figure S5 with Figure 4 it is easy to see that, e.g. results for FLEXPART-FNL, are not identical for season winter. I guess Figure 4 in fact depicts the full year. I also wonder about similar results in other seasons given that wet deposition will hardly occur in summer in the study area.
l. 277/278: “As expected, the values of the 137Cs deposition increase cumulatively with the time after the accident.” This indeed should be the case. But this is not what can be seen in Figure 4. There is no steady increase in, e.g., the median. Consequently, the statement in the next line, i.e., “deposition (the median of normalized deposition > 0.8) happens within 80 hours after the assumed accident” is misleading to me. According to my understanding Figure 4 probably says that 80% of the deposition is accomplished by particles up to an age of 80 hours integrated over all possible time steps.
Caption of Figure 4: “This plot (S5) shows simulations in winter (other seasons).” Weird sentence in this context.
l. 292: “…the seasonal and diurnal changes…” -> “…the spatio-seasonal distribution…”
Figure 5: Too busy with regard to release time isolines. I think they should be thinned out, local minima and maxima be removed, respectively. Please also state the unit, i.e., muSv.
l. 296: “131I_intg_conc_seas”: Needs to be explained when first introduced.
l. 297: “The model inputs…” -> “Dose coefficients…”?
l. 301: “…in most parts of study area than the other members.” -> “…in most parts of the study area compared to the other members.”
l. 306: “…over emission point in UAE and, with less intensity, within the study area…” -> “…over the emission point in UAE and, with lower heights, within the study area…” What is the reason for this discrepancy? For which time (noon?) is Figure S9 valid?
l.307: “dilution” -> “vertical dilution”
l.308: ”process” -> “processes”
l.311: “cold period” Rather mostly fall instead of winter.
l. 314: “exceptional transport” Why is this feature not present in the both WRF simulations?
l. 315-319: Comparison to the Fukushima accident is too vague. How do you define “adjacent”? If I am not mistaken the paper shows doses over 96 hours for iodine only. But the paper cited evidently provides integrated doses for three months and three nuclides.
l. 324/325: “For example, the simulations of FNL in southern Qatar in fall are more than twice that of FNL-WRF (more details in the subsection 3.3).” -> “For example, the simulations based on FNL in southern Qatar in fall result in more than twice the dose compared to that based on FNL-WRF (more details in the subsection 3.3).”
l. 328-329: “To identify the highest possible level of pollution at each point, regardless of its frequency, local maxima are calculated only from non-zero intensities.” Sentence is completely obscure to me.
l. 330: “is caused” -> “are caused”
Figure 6 and 7B: Needs to be redone with units converted to kBq/m2, critical thresholds (40 and 90 kBq/m2, see text) clearly distinguishable and probably replacing the whitish part of the color scale.
l. 335-343: As a result of the deficiencies of Figure 6, the discussion is not really traceable looking at the figure. Specifically, 40 and 90 kBq/m2 cannot be distinguished (both thresholds fall within the withish part of the color scale).
l. 346: “…the higher CFSv2 137Cstot_depos_seas in winter…” It is not really higher, but rather affects a broader area.
l. 362 + caption of Figures 5 & 7: “…particle ages simultaneous…” -> “…particle ages occurring simultaneously…”
Figure 8: Convert Bq/m2 to kBq/m2.
l. 375: “…extremely high levels of inhalation doses (higher than 200 μSv) and of 137Cstot_depos_seas (higher than 100 kBqm−2)…” Please provide a reference for these – according to your opinion - “extremely high” thresholds.
l. 379: “…131Iintg_conc_seas inhalation doses…” -> “…131Iintg_conc_seas related inhalation doses…”
l. 379/380: “…there are no pronounced seasonal and inter-member differences in 137Cstot_depos_seas at different population levels.” This is hard to believe looking at Figure 6.
l. 391/392: “…north of Qatar…” -> “…in northern Qatar…”
l. 394: “…15% of extreme events have reached receptors in Qatar.” -> “…15% of extreme events occur in Qatar.”
l. 398: “has caused”-> “causes”
l. 400/401: “…occurs mainly in the late winter-early spring period simultaneous with the southward movement of westerlies and the eastward movement of the Saudi Arabian subtropical high pressure.” -> “…occurs mainly in the late winter-early spring period simultaneously with the southward movement of westerlies and the eastward movement of the Saudi Arabian subtropical high pressure system.”
l. 410: “As shown in Figure 10-A, the median of simulated column densities…” -> As shown in Figure 10-A for FLEXPART-ERA5-WRF, the median of simulated column loads…”
l. 411: “There is no comparable change in other seasons that is reflected in a minute increase in the full-year distribution of…” -> "There is no comparable change in other seasons which is reflected in a minute increase in the full-year distribution of I-131 load for…”
l. 413-414: “The upper quartile of 137Cs deposition with STM in winter is around 25% higher than those simulations with GTM in the same period.” -> “The upper quartile of 137Cs deposition related to STM in winter is around 25% higher than that related to GTM in the same period.
l. 415: “The implementation of STM in FNL-WRF…” -> “The implementation of STM in FLEXPART coupled with FNL-WRF…”
l. 417: “mainly in fall”: What is the synoptic pattern related to the feature being pronounced in fall?
Figure 10: Please change the y-axis units to kBq/m2 (you can probably stay with Bq/m3 for concentrations if you remove the vertical integration) and introduce scientific notation.
l. 434: “…from the six International Monitoring System (IMS) stations…” -> …from six International Monitoring System (IMS) stations…” There are more than 70 IMS radionculide stations in total!
l. 436: “…normalized by the sum of the two means…” -> “…normalized by the sum of the simulation and measurement means…”
Figure 11: I wonder about the spurious beams in the regression analysis. Especially about the blue ones in the regression analysis for the full year (yellow data points). I would expect simple lines. Moreover, the seasonal subplots can be skipped in my opinion. In the present layout they are too busy with the red points for fall covering much of the remaining data points. Additionally, they are hardly discussed at all in the text (just l. 443 and 444)
l. 446 ff.: “FB5” -> “F5” Occurs numerous times in the subsequent paragraphs.
l.452: “…the normalized RMSE which is less sensitive to extreme values…” Sorry, I do not understand. Large deviations will be even enforced due to the quadratic term involved. Also, the NMSE is not equal to the normalized RMSE. Apart from normalization the square root is not applied for NMSE.
l. 453: “(2)” -> “(2.6)”
l. 454/454: “Feeding downscaled FNL inputs into FLEXPART-WRF (FNL-WRF) increased the correlation of ERA5-WRF and FNL-WRF to 0.7.” - > “Feeding downscaled FNL and ERA5 inputs into FLEXPART-WRF (FNL-WRF and ERA5-WRF) increased the correlation to 0.7.
l. 456/457: “…indicating that the downscaling of inputs in FNL-WRF did not have much effect on the association of their simulations.” Comparing this statement with Figure 5 (where there is a considerable difference between FNL and FLN-WRF based simulations) the inconsistency introduced by using column loads and surface concentrations (doses) alternately becomes very evident.
l. 458-460: “This suggests that the downscaling of similar (FNL) and different (FNL and ERA-5) meteorological datasets increased and decreased the absolute differences between resulting simulations, respectively.” This sentence makes no sense to me. FNL is similar to what? Overall it is evident from the statistics and does not come as a surprise that simulations based on ERA5-WRF and FNL-WRF are most closely related.
l. 460/461 and caption of Figure 11: “…relative distribution of simulations…” -> “…relative column load distribution of simulations…”
l. 461/462: “All distributions here depicted the higher frequency of low 131I column densities in spring than in other seasons (as also seen in Figure 2-A).” Figure 2-A does not display the frequency of column loads but rather that of particle ages! What is reason for this exceptional feature? Why does it not occur in summer as well?
l. 463: “While the RMSEs have increased tangibly…” I would suggest checking this result. I would have expected it rather the other way round because simulated I-131 levels will overall be higher compared to Cs-137 levels due to the larger source term. Thus, absolute differences between modelled values are prone to be larger for I-131.
l. 474: “The other statistics…” -> “The statistics…”
l. 496: “We quantified meteorological uncertainties by producing an ensemble model…” You can hardly quantify the meteorological uncertainties based on three re-analysis members which are correlated (FNL – FNL-WRF, but above all ERA5-WRF – FNL-WRF). The ECMWF, e.g., uses 50 (!) non-redundant ensemble members for uncertainty quantification. Running FLEXPART based on these (or at least a fraction of these) members would deserve the term “uncertainty quantification”. See major concern 2).
l. 502: “…were also examined concerning the population density…” -> “…were also examined in relation to the population density…”
l. 513 “in spring” I would say in fall too.
l. 527: “stronger” -> “larger”
l. 535: “We found…” -> “We suspect…”
l. 541: “in winter” It looks like (Figure 6) this pattern is even more pronounced in fall.
l. 550: “…in which south/southeast winds transport…” -> “in which south/southeast winds in the cold season transport…”
l. 556/557: “…decreased the median of simulated 131I concentrations…” Only in fall.
l. 561: “In general, CFSv2 simulations of 137Cs and 131I column density are most highly correlated with FNL…” -> “CFSv2 simulations of 137Cs and 131I column loads are most highly correlated with FNL…” Mind that FNL – FNL-WRF correlation is highest.
l. 563: “FB = 0.02” -> “FB = -0.02”
l. 560-575: Please avoid stating numerous statistical scores in the conclusion sections. There can be two or three of them but the conclusion section should contain take-away-messages rather than multiple numbers.
l. 573: “RMSE=14.6 x 10^8 and 2946.3 x 10^7”? Exponents?
l. 576: “…simulations from FNL and FNL-WRF with identical meteorological inputs.” FNL and FNL-WRF are not “identical”, but of course correlated to some degree. The authors should decide at some point in the text whether their ensemble input quantifies meteorological uncertainties (see l. 496) or is identical which makes the use of an ensemble obsolete. |
