the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Climate change impacts on regional fire weather in heterogeneous landscapes of Central Europe
Julia Miller
Andrea Böhnisch
Ralf Ludwig
Manuela I. Brunner
Abstract. Wildfires have reached an unprecedented scale in the Northern Hemisphere. The summers of 2021 and 2022 demonstrated the destructive power of wildfires especially in Northern America and Southern Europe. Global warming indicates that fire seasons will become more extreme and will extend to more temperate regions in northern latitudes in the future. Multiple studies claim that natural variability hides the trend of increasing fire danger in climate model simulations for future potentially fire-prone areas. Single Model Initial-Condition Large Ensembles (SMILEs) help scientists to distinguish climate trends from natural variability. So far, the SMILE framework has only been applied for fire danger estimation on a global scale. In this study, we use a regional SMILE of the Canadian regional climate model version 5 (CRCM5-LE) over Central Europe under the RCP 8.5 scenario from 1980 to 2099, to analyze fire danger trends in a currently not fire-prone area. We use the meteorological Canadian Fire Weather Index (FWI) as a fire danger indicator. The study area covers four heterogeneous landscapes, namely the Alps, the Alpine Foreland, the lowlands of the Southern German Escarpment and the Eastern Mountain Ranges of the Bavarian Forest. We demonstrate that the CRCM5-LE is a suitable dataset to disentangle climate trends from natural variability in a multivariate fire danger metric. Results show the strongest increases in the median (50th) and extreme (90th) percentile of the FWI in the northern parts of the study area in the summer months July and August, where high fire danger becomes the median condition and extremes occur earlier in the fire season. The southern parts of the study region are affected less strongly, but due to weaker variability in these regions, time of emergence (TOE) is reached there in the early 2040’s. In the northern parts, the climate change trend exceeds natural variability in the late 2040’s. We find that today’s threshold for a 100-year FWI event, will occur every 30 years by 2050 and every 10 years by 2099. Our results highlight Central Europe’s potential for severe fire events from a meteorological perspective and the need for fire management in the near future even in temperate regions.
Julia Miller et al.
Status: open (until 25 Jun 2023)
-
RC1: 'Comment on nhess-2023-51', Anonymous Referee #1, 09 May 2023
reply
The manuscript “Climate change impacts on regional fire weather in heterogeneous landscapes of Central Europe” by Miller et al. presents a study on climate change impacts on the fire danger index FWI in hydrological Bavaria in central Europe, using a single model initial-condition large ensemble (SMILE) data set and the RCP 8.5 emission scenario. Changes in FWI are evaluated in terms of fire danger levels over the whole region, as well as on a sub-regional scale. The study provides new and interesting knowledge of projected changes in FWI in the studied region, and the topic is suitable for the journal. Strengths of the manuscript include the multiple approaches applied to investigate the hypotheses, and the clearly communicating figures. The manuscript has potential, but is not at the required level for a scientific paper in its current state. Parts of the analyses are wrong, methods are not clearly presented, the discussion is not clearly presented, and references do not always reflect work supporting the claims. I would recommend all authors to carefully revise the manuscript, and correct and clarify where necessary. General and specific comments are provided below.
General comments
- The method to derive the return period is wrong. Thus, the analysis using return periods must be omitted or corrected and clarified. The applied temporal resolution is not stated, but the method is wrong regardless of the applied resolution. If all daily (or monthly) data in each year is used to extract the 99th percentile value, this value represent the 100 days (or 100 months) return period and not 100 years as stated in the manuscript. See e.g. Camuffo et al. (2020): https://doi.org/10.1007/s11600-020-00452-x. If instead the maximum value in each year was extracted, the period (30 years, i.e. 30 values) is too short to extract the 99th
- Several references do not represent the original work reflecting you statements. Examples are line 46, 126, 157, 181 and 212. Please make sure the original references are used throughout, or in cases where this is not possible, add “e.g.” before the references to avoid the reader to believe the reference is the original work.
- The discussion comprise multiple detailed comments on different aspects of the analysis or results, with too general subtitles. Introduced topics (e.g. uncertainties related to the chosen climate model on line 302) and summaries of results are not always followed up. Overall, the current state of the discussion chapter makes it hard for the reader to know what to expect and to follow the line of arguments of the authors. Please revise and clarify the discussion chapter, avoid mentioning topics without commenting on them in relation or your study, and lift part of the discussion to a more general level.
- The text is in several places informal with the use of unnecessary introductions (e.g. “another aspect which has to be discussed” on line 283 or “needs to be critically reflected upon” on line 318) or subjective words, inconsistent used of concepts, and imprecise descriptions of methods and results. Further, most of the manuscript is written in present tense. Papers are usually written in past tense when presenting analysis and results in abstract, data and methods, results and conclusions. Thus, I would recommend changing to past tense. In general, please carefully revise the whole manuscript and clean and clarify the text.
- The manuscript refer to similar analysis in France and UK, but does not refer to results over the same region (HydBav) by others, e.g. from global or regional studies that cover the region. Please include other studies that cover your region (for example https://doi.org/10.1029/2018GL080959 and https://doi.org/10.1007/s10584-016-1661-x). In addition, please comment on potential differences between France, UK and your region, when you refer to results over these regions.
Specific comments:
- You state in the title that your study area represents “heterogeneous landscapes”. However, in line 110-111 you “assume that the water availability, climatology and landscape of different river systems reflect the fire regime of an area”. A more in-depth reflection on the heterogeneity of the regions, and even subregions of your study, in relation to your spatial aggregations and findings would be beneficial to better reflect the title of your study.
- «Central Europe» is a concept that refers to a considerably larger region than the study area “Hydrological Bavaria”. Please clarify the region to avoid exaggerating your study domain, in particular in the Abstract (e.g. line 7 changing “over Central Europe” to “in a region in central Europe” or similar), the conclusion (e.g. line 395 changing “for Central Europe” to “in central Europe” or similar), and the title (“of Central Europe” to “in central Europe”). Specifically, change to lower case ‘c’ (“central Europe”) throughout.
- Titles: please use NHESS house rules (sentence-style capitalization: https://www.natural-hazards-and-earth-system-sciences.net/submission.html#manuscriptcomposition)
- Line 27 and line 396-397: You state that Central Europe has not been exposed to wildfires before recent years (line 27) or to date (line 396-397). However, central Europe has been exposed to multiple wildfires at least the past three decades. Rephrase to correct statements and provide reference(s) that have fire records underlying your statements.
- Line 38-40: The claim that fire indices represent a statistical correlation between fire events and meteorological conditions is wrong. Please correct.
- Line 40-41: Please provide reference that “They have been proven to produce reliable ratings of fire danger in short and long term weather predictions on a global scale”.
- Line 42: please rephrase “do not guarantee”, as this is an unclear statement. Fire indices have nothing to do with ignition at all.
- Line 45 and other: Various concepts are used for the fire indices; please be consistent and potentially introduce relevant relations between concepts such as “fire risk”, “fire weather”, “fire danger”, “fire indices”, “likelihood of fire” and “probability of fire”.
- Line 49, 57 and 59: Central Europe (line 49) and temperate climate (line 58 and 59) refer only to studies of England and France here. As these concepts are used for the HydBav later, please clarify the use and links between geographical regions and climate regions. Are results for England and France directly transferrable to HydBav?
- Line 49-50: Please provide reference that trends are not distinguishable from natural variability in Central Europe. Further, this sentence relates to the weak trend signal relative to natural variability independent of how models represent the natural variability, and thus the link to the next sentence is wrong or not clear.
- Line 52-53: Unclear sentence “In France, …” Please rephrase.
- Line 56: clarify the meaning of “natural variability of changes”.
- Line 59-63: The claim here is that climate model ensembles using multiple models (but fewer simulations per model compared to SMILE) underrepresent natural variability, whereas SMILE does not. Please clarify how large ensembles using a single model (SMILE) better represent natural variability as compared to large ensembles from different models, and add references that support this claim.
- Line 70: The reference period (1980-2009) is not one of the established reference periods. Please explain the choice of the period. Why not use the almost identical period 1981-2010, which is a widely used reference period?
- Line 71: Please change “increases” to “changes”, because your analyses were also able to detect if there were any decreases.
- Line 74: Please clarify which TOE you refer to (TOE of what?)
- Line 82: Please clarify whether your mean two domains in Europe, or two domains of which one is in Europe.
- Line 88: clarify “independent” (in which regards?). Fifty members based on the same model are far from independent as such.
- Line 91: please clarify “at this time”
- Line 93: please clarify the link between your study choice and the provided references.
- Line 93: Please comment on this assumption (internal variability = natural variability) in the discussion or here. Potential limitations or lack thereof?
- Line 95: How does smaller (temperature) and equal (precipitation) member spread in your SMILE compared to EURO-CORDEX relate to the previous claim that SMILE overcome the limitation of multi-model ensembles related to under-representation of climate variability. The results of Von Trentini (2019) imply that multi-model ensembles represent a larger variability as compared to CRCM5-LE.
- Line 98: state which observational data were used for the bias correction.
- Line 100: please rephrase and clarify. Better represented when evaluating against what (wouldn’t that be against climate data, which you state is what should be bias-adjusted in the first place)? “climate data” is very general, do you mean data from climate models?
- Line 100: Has there been any studies evaluating the data you use against meteorological variables from observations (independent of the bias adjustment) or reanalysis over the region?
- Line 103: Please insert the stated rivers in Figure 1 in order to inform a reader, who is not familiar in the region, how the named rivers relate to the study area.
- Line 108: Does ‘s’ refer to ‘see’? Please write out. Also, use capital F in figure names.
- Line 110: ‘water’ and ‘water availability’ are imprecise. Clarify what water you mean and add a supporting reference. If you mean soil moisture or precipitation, these (and thus also the fire regimes) are likely highly heterogeneous within each subregion (in particular in mountainous areas).
- Line 109-111: Please clarify and justify your assumption. The subregions are not defined according to the river systems, i.e. river catchments (as seen from Fig. 1). As you state earlier (line 102 and title) and later (line 120-123), hydrology, climatology and landscape are highly variable in the study area, and is likely highly variable in particular in the mountainous areas, within a subregion, with consequences for fire characteristics. What is mean by “an area”?
- Line 112-120: Which period and data underlie the numbers presented here? Could a figure (temperature and precipitation spatial patterns) be added (e.g. in appendix) to make the information more intuitive to the reader?
- Line 129: As written, ‘noon’ refers only to wind. Rephrase so it refers to all variables (even 24h precipitation is measured at noon).
- Line 132: what is meant by ‘bookkeeping’? This concept is linked to financial transactions. Can it be replaced by a more commonly used concept within natural sciences to be more intuitive to the reader?
- Figure 2 caption: suggest to replace ‘vegetation’ with ‘organic matter’, ‘fuel layers’ or similar (as used in the main text) for clarity.
- Line 143: please rephrase or clarify “without memory of past conditions”. Because the fuel moisture codes have memory of past conditions, BUI and ISI have too.
- Line 156: High-altitude parts of the region will likely have snow in the beginning of the defined fire season. Please state if/how you have accounted for snow in the evaluation here. E.g. see last section in https://www.nwcg.gov/publications/pms437/cffdrs/fire-weather-index-system. If you are neglecting the effect of snow on fire danger, it is worth reflecting on it in the discussion.
- Line 157: Vitolo et al (2019) does not apply any fire season, and should be replaced by reference(s) using or arguing for using April-September.
- Line 158: “annually calculated FWI values” is unclear (may refer to annual values, which I assume is not the case). Suggest to delete “of annually calculated FWI values” for clarity.
- Figure 3: Please add a legend (ref. NHESS figure composition: https://www.natural-hazards-and-earth-system-sciences.net/submission.html#manuscriptcomposition).
- Line 170: As you state yourself, the results of the evaluation you have performed does not affect the climate change impact assessment of your study. Why do you not evaluate your data using measures that can actually reflect your data’s ability to assess climate change impacts? For example its ability to represent historical changes.
- Figure 4: The applied FWI colour scale is almost identical to the FWI colour scale provided in Table 1, but they reflect different FWI intervals. Please change for clarity.
- Section 2.5.1: The description is unclear in terms of when the different aggregations were applied (both in space and time), and when the continuous analysis vs the data split are applied. Consider reorganise the section to better fit each part of the analysis. Please also clarify how the ‘extreme condition’ (90th percentile) is computed (is it over the analysed time period, region or models, and in which order is it calculated). Clarifications in this section are necessary for reproducibility.
- Line 174: Please provide which trend method was applied.
- Line 177: You state that you are “summarizing FWI values over a fire season on daily, monthly or annual basis”. Summarizing would provide very different ranges of FWI on the different temporal scales, and it does not look like they are summarized in e.g. Fig. 5 (looks like average or median over each month). Please correct or clarify.
- Line 178 and line 182: Suggest to replace ‘increasing’ with ‘changes in’ for clarity. The analyses allow for changes in both directions.
- Line 216: Your results are scenario-specific. Please specify the scenario, e.g. “according to RCP 8.5”.
- Line 219-222: Please rephrase to clarify your reasoning.
- Line 223: Why is not the Southgerman Escarpment mentioned here (regions of strongest rises in extreme FWI in July to Sep)?
- Line 226, 228 and 229 and potentially other places: ‘median case’ and ‘extreme case’ are unclear concepts. Do you refer to ‘median FWI’ and ‘extreme FWI’? Please be consistent with concepts or clarify newly introduced ones.
- Line 133: Why do you state ‘mean conditions (median)’ and not ‘median conditions’ (what is the difference)? Please clarify in the text. Similarly, clarify similar statement in line 349: ‘On average (median)’
- Line 236: Please consider using the phrasing ‘mid 21st century’ instead of ‘middle of the 21st century’
- Line 240-241: Please clarify how your findings indicate that the distribution of the FWI extremes resembles the distribution of the FWI median? Figure 7 clearly shows that the distributions are different both in terms of mean and standard deviation.
- Line 242: consider replacing ‘changes’ to ‘increases’ for clarification.
- Line 244: please specify what you mean by ‘strongly’
- Line 244-245: Please clarify what you mean and which parts of the results you refer to. Your statement here seem opposite compare to the preceding sentence (median FWI increase strongly vs median FWI shows hardly any changes).
- Line 245: Why state the EFFIS reference here, when every classification of fire danger level in the (also previously mentioned) results is based on it?
- Line 247: clarify which average you are referring to.
- Line 247-248: the interval signs in parenthesis are wrong.
- Figure 5: The levels referred to in the result section would be more easily recognisable if a colour scale using discrete colours was used. Discrete colours would provide a more clear message to the reader, in particular when these levels are the main message of these results, and not the small varieties in between. Please consider changing to discrete colours.
- Figure 5 caption: do you mean “at least” two levels (indicated by thick black dot), or are there never more than two levels?
- Line 259-260: please provide numbers or proportions in parenthesis.
- Line 262: Please provide in what ways they are similar. EMR is not described in other terms than relative to Alpine Foreland.
- Figure 8: Please consider adding proportions on the right y-axis, as proportions are used in the text.
- Figure 8 caption: Please clarify by specifying what is meant by frequency (e.g. “number of days within a fire season”)
- Line 278: As in line 215, clarify the scenario dependence of your results (in line with your statement in line 306-307). The way it is phrased now imply more certainty about the future than we can state.
- Line 279: Why move away from the defined classes? How is hazardous defined?
- Line 285-286: please state the variable (FWI) that is compared.
- Line 289-291: Please state in what relevant ways the formulas have been adjusted (i.e. relevant implications). Is this a more likely reason for the differences as compared to the fundamental differences in how the underlying meteorological data are produced?
- Line 293: please rephrase sentence to be more to the point. It is unclear how the tiling patterns referred to in the text ‘has to be discussed’ and not the ones seen e.g. in September (Fig. 5 [2]f) or at smaller scale in the Alps in July-Sep (Fig 5 [1]def and [2]def).
- Line 296: please change ‘correlates’ with a more appropriate word, or provide correlation results.
- Line 302: you mention the uncertainty related to the chosen climate model. Please elaborate on this point in relation to the specific model you applied.
- Line 316: please remove “potential of”. FWI describes the fire weather, not the potential of fire weather.
- Line 318-327: Please consider deleting this paragraph, and alternatively reduce the main message to a single sentence in the methods chapter arguing for your use of danger levels.
- Line 328-331: Please elaborate briefly on the flammability of the surface in your study region.
- Line 333-339: Please reflect/explain results rather than summarise them.
- Line 340: increases in variability (line 337) is not the same as high variability in general (line 340). Please elaborate what you mean by your findings (increasing variability over time in mountainous regions) corroborate the findings by Wastl et al (2012; higher variability in mountainous regions than other regions).
- Line 345: Unclear whether ‘extreme FWI conditions’ represent the 90th percentile or the classes (FWI>50). In case of the former, do you mean elevated conditions compared to former months or compared future to present. In case of the latter, is that not seen directly from the figure and not ‘implied’ from your findings? Please clarify the meaning of this sentence.
- Line 348: ‘tremendous’ is subjective, please clarify. See also ‘dramatic’ in line 362 and ‘strikingly’ in line 364-365.
- Line 349: by ‘seasonal’, do you mean ‘monthly’? In which ways are they hotspots, in terms of general conditions/increases/other?
- Line 358: The use of vegetation in Figure 2 caption implies also litter and organic matter on the ground. In this context, vegetation is necessary for fire development because it comprise the fuel. Is it the same use of vegetation here? I assume vegetation is highly present during winter also, although parts are covered by snow, and deciduous trees lack their green leaves. Please clarify the text.
- Line 359: ‘half year’ typically refers to six months. Consider changing to ‘period’ or similar, as you refer to December-April.
- Line 358-360: would FWI be suitable for the winter season? The reasoning provided here include lack of vegetation, whereas this is not accounted for in FWI. And what about snowfall and snow cover? Further, would you assume the temperature thresholds included in FWI calculation be exceeded in the Alps in winter? Please reflect on the considerations needed for such assessments.
- Line 366: states ‘exists currently no fire danger’, however you have fire danger everywhere (as fire danger is defined as the estimates from the index, regardless of values). Please clarify.
- Line 372-373: you mention overestimation of natural variability. How does this relate to line 59? What about potential underestimation when using SMILE? If a model has a limitation (e.g. in representing natural variability), all realisations from that model suffer from the same limitation. If you or other have validated the ability of SMILE to represent natural variability, please state this in the text and refer to relevant evidence. Applies also for line 375.
- Line 373: ‘slight overestimation of the CRCM5-LE’. Clarify, what does it overestimate?
- Section 4.4: the title and content of the section does not match (impacts [title] vs conditions influencing flammability, emergency in other regions. Further, the content is not coherent. Please revise and clarify the message.
- France and UK: Several places in the manuscript, results of France and UK is used for guiding and comparing the results of the present study, and to make final recommendations for fire emergency. However, you do not reflect on potential relevant differences between the regions (e.g. hydroclimotology and vegetation). Please consider commenting on such aspects.
- Line 397 (and line 406): You state that the study area is not affected by high fire danger to date, but high fire danger is present in relatively large areas in current climate (Fig. 5[2]def, where the dots indicate a change from a currently high level).
- Line 397-398: Please clarify ‘by accounting for natural variability’.
- Line 398: Please clarify the difference between “strongest increase” and “most hazardous developments”
- Line 400: please clarify in what terms, and in which results the statement “less strongly affected” applies. For example, in fig 5[2], Alps is the only region with dots in April and May, and the two regions you mention increase multiple fir danger levels as seen e.g. in Fig. 6[2] august. As mentioned earlier in the manuscript (line 324-326), increases in classes may provide a better approach to assess increases due to non-linearity, and thus and linear comparison (e.g. Fig 7) may not the best way to conclude the strongest trends.
- Line 401: the statement that FWI has a stronger variability for Alps and Eastern Mountain Ranges contradicts the findings in Fig. 7, where the standard deviation is smaller for these regions compared to the other subregions. Please clarify.
- Line 404: please consider repeating the hypothesis, and structure the conclusions by these.
- Line 407: please clarify what ‘also’ refer to.
- Line 410: What about the data of the subregions and land cover (Fig. 1 and A1)?
- Figure C1: Why do you use 95th percentile and not 90th percentile as done in the remaining analysis?
- Why number the Figures A1, B1 and C1 instead of A1, A2 and A3 as is normally done?
Citation: https://doi.org/10.5194/nhess-2023-51-RC1
Julia Miller et al.
Julia Miller et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
325 | 77 | 10 | 412 | 4 | 5 |
- HTML: 325
- PDF: 77
- XML: 10
- Total: 412
- BibTeX: 4
- EndNote: 5
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1