The 10-Year Return Levels of Maximum Wind Speeds under Frozen and Unfrozen Soil Forest Conditions in Finland

Round 1

Reviewer 1 Report

In this manuscript, the authors determined the 10-year return levels of maximum wind speed between frozen and unfrozen soil conditions in Finland. They authors use weather reanalysis data to obtain wind speeds and then down-scaled the wind data. Overall, the study presented in the manuscript is novel and deserves publication following moderate revisions. Below are my specific comments.

Abstract, line 26 -"relieved"?

Introduction, lines 32-39 - The authors should quantify some of these impacts and economic losses to provide an understanding for the reader of why the study could be important.

Line 43 - put space between -0.32 m/s

Line 44 - "stilling"?

Line 62 - why do the authors mention "in leaf" here?

Lines 65-70 - Under what conditions were the soil frost models developed? Where? How accurate are they?

Line 83-84 - what is the accuracy of the GIS-produced multipliers?

Lines 108-110 - what are the authors referring to when they saw "lower soil" for the last few variables? Is lower soil referencing deeper soil in the profile? If so, what depth exactly?

Line 113 - what is "it"?

Line 138-139 - why?

Line 139-141 - is there data to support this claim?

Line 159 - what is "shielding"? Why exactly was it not considered?

Line 169-170 - more detail is needed about the development of the modifiers.

Line 206-207 - I don't understand this sentence

Fig 1b - needs a scale bar

Line 212 - One sentence is not sufficient here. I'm assuming this is support to be a validation site? Much more detail is needed. Why was this location selected? What was measured at the site to validate? Was ground truthing conducted?

Line 250-253 - The authors need to display data regarding the wind direction here.

Line 260-264 - this section is unclear

Line 282 - Most readers will not know specifically where Lapland is, and it is the first time the authors use this geographic region. Previously in the manuscript they had only used "norther Finland", etc...I would stick to that terminology because it is unambiguous.

Line 285 - effect on what?

Line 287-293 - Instead of describing if the data were more negative or positive, it would be more helpful if the authors provide an interpretation to the reader...say the winds were stronger in the frozen soil season, etc...

Fig 4 - the text on this figure is too small to read

Line 318-326 - what example area??? this paragraph is disjointed from the rest of the results. See comment about line 212.

Fig 7 - when I look at this figure, the ERA v. OBS trend line looks like it has the same slope but is slightly offset, indicating that the ERA values are merely an underestimation of the OBS values. In contrast, the WM vs. OBS trend line has a different slope. So why is the ERA v. OBS a better estimator? Also given the discussion that the weather stations for the OBS data may be biased anyway, what interpretation can the authors give here?

Line 340 - improvement from what to what?

Line 340-343 - The authors should run statistics to determine whether there truly are differences or not.

Line 355-364 - There are no data in this manuscript that supports this text. Furthermore, Fig 10 is mentioned but absent. Fig 9 is not even mentioned.

Discussion - the authors need to do a better job throughout the section integrating the results of their study with what is in the literature and put their results in context

Line 406-407 - "especially as we restricted our work to current climate" ?

Line 420 - "over one in only 15%" ?

Line 426 - the authors need to elaborate on how their findings can be implemented into comprehensive wind damage risk assessment

Line 431-432 - it would be easy enough to run some statistics to determine if the seasons need to be separated

Line 434 - why would the authors assume there is no support from better tree anchorage in frozen soil? I thought this concept was introduced at the beginning of the manuscript and used as justification for why they authors were evaluation frozen and unfrozen soil conditions?

Conclusions - the authors have missed a clear opportunity here to place the findings of their study in a much broader context of its implications. in this study, the authors have demonstrated how these techniques can be applied to determine wind assessment in Finland, and while the results are restricted to Finland, the applications are not!

Author Response

Comments and Suggestions for Authors

In this manuscript, the authors determined the 10-year return levels of maximum wind speed between frozen and unfrozen soil conditions in Finland. They authors use weather reanalysis data to obtain wind speeds and then down-scaled the wind data. Overall, the study presented in the manuscript is novel and deserves publication following moderate revisions. Below are my specific comments.

------------------------------------------------------

Abstract, line 26 -"relieved"?

Typo: relieved -> revealed

Introduction, lines 32-39 - The authors should quantify some of these impacts and economic losses to provide an understanding for the reader of why the study could be important.

We added some examples about the damages caused by recent storms

Line 43 - put space between -0.32 m/s

Fixed

Line 44 - "stilling"?

Stilling is commonly used in the context of weakening wind speeds, like in e.g. [17]. However, to make it more straightforward, we changed “stilling” to “weakening”.  

 

Line 62 - why do the authors mention "in leaf" here?

”In leaf” was mentioned to emphasize that birch is vulnerable to uprooting when in leaf, but not when there is no leaves in birch. We modified this part to more well-defined.

Lines 65-70 - Under what conditions were the soil frost models developed? Where? How accurate are they?

We modified this paragraph slightly to emphasize that frost models were developed in Finnish conditions. Development of soil frost models were also covered in the lines 111-113 of the original manuscript, paragraph which is also further modified in the revised version.

111-113: “By using soil temperature observations from several stations across Finland, model parameters were further optimized for three different soil types: clay or silt soil, sandy soil, and peatlands [31]. Based on it, the revised model explained roughly 90–99% of the observed soil temperature variability.”

Line 83-84 - what is the accuracy of the GIS-produced multipliers?

With this sentence we indicate that technically it is easy to use softwares such as ArcGIS, QGIS and R to calculate the multipliers following the theory introduced in the publications [39-41]. We edited slightly this sentence.

How accurately wind multiplier downscaled wind speeds resemble observed wind speeds is covered in the section 3.4 of the manuscript.

Lines 108-110 - what are the authors referring to when they saw "lower soil" for the last few variables? Is lower soil referencing deeper soil in the profile? If so, what depth exactly?

The original equation presented by [30] described only heat flow from the surface and [42] extended the equation to take into account heat flow from below the soil layer of consideration (below 20 cm depth in our case). The latter part of the equation consists the model extension. In naming the variables, we follow [42], i.e. while, e.g., soil thermal conductivity refers to the thermal conductivity of soil between the surface and layer of consideration (20 cm depth in our case), lower soil thermal conductivity refers to the thermal conductivity of soil between the layer of consideration (20 cm depth in our case) and Z_l. Following to [31], Z_l was set to 6.8 m.

We edited this paragraph accordingly to be more specific.

Line 113 - what is "it"?

This section (lines 111-113 of original manuscript) was edited based on prior comment about lines 65-70 to be more specific.

Line 138-139 - why?

We added a sentence about the 10-year return level period being relatively short compared to our data period of 35 years.

Line 139-141 - is there data to support this claim?

We added a reference to soil frost calculations of [31] and stated with “(not shown)” that detailed results considering this is not covered here.

Line 159 - what is "shielding"? Why exactly was it not considered?

Shielding multiplier takes into account the effect of upwind (tall) buildings that provide wind cover for the point of interest. In [40] it is only considered in urban areas. Therefore, it was not considered relevant for our application and we decided to simplify our wind multiplier calculations accordingly.

We added some explanation about the shielding factor. 

Line 169-170 - more detail is needed about the development of the modifiers.

As development of wind multipliers used in this study is described in detail in our previous work [41], we ended up to keep this section rather simplified. We modified this part of the manuscript slightly and referred more specifically to sections of [41] where development of multipliers is covered.

Line 206-207 - I don't understand this sentence

We ended up to remove the whole sentence as it was more or less unnecessary.

Fig 1b - needs a scale bar

Scale bar added.

Line 212 - One sentence is not sufficient here. I'm assuming this is support to be a validation site? Much more detail is needed. Why was this location selected? What was measured at the site to validate? Was ground truthing conducted?

Area was not an actual validation site, at least in the sense of measurements to be validated. It was more like an area chosen for a more detailed “zoom in” into large scale results presented with maps of Finland, i.e. this was area chosen for more close inspection of the influences of our downscaling approach. Main reasoning to choose this area was simply the interesting results with large differences mainly driven by area’s topography.

We edited this part of the text to make the purpose of used example area clearer.

Line 250-253 - The authors need to display data regarding the wind direction here.

We added “(not shown)” to the end of last sentence to make it clear we are not presenting findings mentioned in the text with figure.

Line 260-264 - this section is unclear

We tried to simplify this section to make the point clearer, i.e. that there are quite often years when maximum wind speed is observed multiple times, and sometimes even during both soil frost seasons.

Line 282 - Most readers will not know specifically where Lapland is, and it is the first time the authors use this geographic region. Previously in the manuscript they had only used "norther Finland", etc...I would stick to that terminology because it is unambiguous.

We removed Lapland and just talk about the most northwestern part of Finland.

Line 285 - effect on what?

We added “on distribution of differences” into this sentence.

Line 287-293 - Instead of describing if the data were more negative or positive, it would be more helpful if the authors provide an interpretation to the reader...say the winds were stronger in the frozen soil season, etc...

We changed all the mentions about positive and negative differences to expressions about winds being stronger during frozen or unfrozen soil season.

Fig 4 - the text on this figure is too small to read

Figure edited.

Line 318-326 - what example area??? this paragraph is disjointed from the rest of the results. See comment about line 212.

This paragraph is continuation from previous paragraph, where we bring up the small scale features visible in the large scale maps of Finland and introduce the idea of example area where we examine these features in more detail. Besides additions to section 2.5, where this example area is first introduced, we edited text a bit to emphasize the meaning of this area.

Fig 7 - when I look at this figure, the ERA v. OBS trend line looks like it has the same slope but is slightly offset, indicating that the ERA values are merely an underestimation of the OBS values. In contrast, the WM vs. OBS trend line has a different slope. So why is the ERA v. OBS a better estimator? Also given the discussion that the weather stations for the OBS data may be biased anyway, what interpretation can the authors give here?

Yes, the slope of OBS vs WM itself is a bit alarming compared to slope of OBS vs ERA. As can be seen from Figure 7, slope of OBS and WM is clearly affected by values over 15 m/s that are overestimated due to wind multipliers. However, like we point out in the text, these overestimations are from a few specific weather stations with more or less unique characteristics. Still, used statistics point towards the overall improvement in our return level estimates when wind multipliers are used. 

Line 340 - improvement from what to what?

This sentence was corrected to be more definitive.

Line 340-343 - The authors should run statistics to determine whether there truly are differences or not.

We added 95 % confidence intervals of both R²’s to show that those are not overlapping. For differences we run Mann-Whitney U test, resulting p < 2.2e-16). Considering how we have used K-S test D statistic in this study to describe similarity of ERA and WM to OBS, I don’t exactly understand how these could be tested.

Line 355-364 - There are no data in this manuscript that supports this text. Furthermore, Fig 10 is mentioned but absent. Fig 9 is not even mentioned.

We added “(not shown)” as we think showing results e.g. as a table for each of the 40 stations and comparison statistics is not that meaningful.

We corrected the figure numbering of older version of manuscript that was still mistakenly in place in this section.

Discussion - the authors need to do a better job throughout the section integrating the results of their study with what is in the literature and put their results in context

Section was slightly edited to describe how our results could be used in wind risk assessments, e.g. when probabilities of wind speeds over CWSs are considered, and the role of soil frost in it.

Line 406-407 - "especially as we restricted our work to current climate" ?

We clarified this sentence.

Line 420 - "over one in only 15%" ?

We added some explanation about the wind multiplier values being over one (1.0).

Line 426 - the authors need to elaborate on how their findings can be implemented into comprehensive wind damage risk assessment

We modified this part of the text and attempted to provide more comprehensive example.

Line 431-432 - it would be easy enough to run some statistics to determine if the seasons need to be separated

We added some statistical testing when comparing differences in seasonal maximum wind speeds between soil frost seasons. We also emphasized the small differences compared to confidence intervals of return level estimates.

Line 434 - why would the authors assume there is no support from better tree anchorage in frozen soil? I thought this concept was introduced at the beginning of the manuscript and used as justification for why they authors were evaluation frozen and unfrozen soil conditions?

This part was edited extensively. Even if differences are small, soil frost provide support by excluding strong winds from the return level calculations of unfrozen soil season. As occurrence of strong winds varies so much, soil frost season reduces the overall risk of wind damage.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments to the author

In this manuscript, Laapas et al. present findings of a modeling study examining differences in maximum wind speed between frozen and unfrozen periods for three soil types in Finland. The authors find small differences, and highlight underlying causes of wind speed variability and challenges in modeling it. The question is novel and impactful, since wind speed is not a well examined aspect of climate and has significant consequences for ecosystem function and tree survival. The general approach is also robust, and the manuscript is well written, especially in the introduction and discussion. That said, I have two major concerns regarding the manuscript.

First, it seems to be a missed opportunity that more rigorous statistical comparisons were not conducted. The authors acknowledge that the comparisons presented in the manuscript are largely qualitative, but the small effects observed warrant a more robust statistical approach. Had differences been more obvious, perhaps this would not be so, but as they are, I find the qualitative comparisons unfulfilling. Could some summary metrics for wind speed not be analyzed as a function or region, soil type, or other environmental factors?

Second, it would be helpful if more attention was given to validating the model and overall approach. It makes sense that frost days will vary based on soil conditions, but the model output suggests extreme differences in frost days between soil types over small spatial scales. Especially in northern Finland, the difference in the minimum number of frost days between spruce and peat soils can be over 100 days. It is not intuitive that soil type would make that large of a difference in frost days for soil types that could be found in the same location.

Specific comments

89. Is it not possible to combine this with data from previous studies to estimate damage risk?

112. So the model estimates soil temperature (frost days) independently for each soil type across all of Finland?

122. what evidence supports this approach? It is intuitive that frost varies by soil type, but are the strong differences between soil types in the same local area projected by this model likely?

269. Could this (and all other comparisons made between regions) not be tested statistically?

Fig. 4 Text is far too small to read, lines should be thicker as well. Same for Fig. 7

Author Response

Comments to the author

In this manuscript, Laapas et al. present findings of a modeling study examining differences in maximum wind speed between frozen and unfrozen periods for three soil types in Finland. The authors find small differences, and highlight underlying causes of wind speed variability and challenges in modeling it. The question is novel and impactful, since wind speed is not a well examined aspect of climate and has significant consequences for ecosystem function and tree survival. The general approach is also robust, and the manuscript is well written, especially in the introduction and discussion. That said, I have two major concerns regarding the manuscript.

First, it seems to be a missed opportunity that more rigorous statistical comparisons were not conducted. The authors acknowledge that the comparisons presented in the manuscript are largely qualitative, but the small effects observed warrant a more robust statistical approach. Had differences been more obvious, perhaps this would not be so, but as they are, I find the qualitative comparisons unfulfilling. Could some summary metrics for wind speed not be analyzed as a function or region, soil type, or other environmental factors?

We tried to add statistical robustness to our approach by using some additional statistical tests when comparing station level differences in seasonal maximum wind speeds between soil frost seasons and when comparing observed return levels to ones derived straight from reanalysis and to ones downscaled with wind multipliers. However, we did not test differences in return levels, which already are a product of statistical analysis and have associated uncertainties. These uncertainties were considered in detail in 40 stations, where we found the confidence intervals to be on average +/- 1 m/s. 

Second, it would be helpful if more attention was given to validating the model and overall approach. It makes sense that frost days will vary based on soil conditions, but the model output suggests extreme differences in frost days between soil types over small spatial scales. Especially in northern Finland, the difference in the minimum number of frost days between spruce and peat soils can be over 100 days. It is not intuitive that soil type would make that large of a difference in frost days for soil types that could be found in the same location.

Large observed variability in soil frost depending on soil type was one of the main reasons why soil frost in [31], also used in this study, was modeled for different soil types. For example, attached here are two figures about the soil frost (and snow depth) measured by the Finnish Environment Institute at the same location (Kittilä) in the northern Finland. As can be seen, there are huge differences in soil frost depending on soil type. Figure with soil type “fine gravel”, similar to sandy soil used in this study, have soil frost depths beyond 20 cm already since December, and maximum depth of over one meter. Conversely for “mire” (peat), soil frost penetrates barely into 20 cm even in the coldest periods of winter. Figures can be found here (10.04.2019): http://wwwi3.ymparisto.fi/i3/tilanne/ENG/frost/LAP.htm

We added two sentences into Introduction about the large variability depending on soil type and the main reasoning for the variability.

  

Specific comments

89. Is it not possible to combine this with data from previous studies to estimate damage risk?

Yes, soil frost data from previous study [31] was used in this study to determine the soil frost seasons. We modified this paragraph to make it clearer that calculations by [31] was used in this study.

112. So the model estimates soil temperature (frost days) independently for each soil type across all of Finland?

Yes, this was the case. We added one sentence to clarify this.

122. what evidence supports this approach? It is intuitive that frost varies by soil type, but are the strong differences between soil types in the same local area projected by this model likely?

For the first part of comment, we added references and pointed out to typical rooting depth of main boreal tree species. The second part is same as second major concern and addressed above.

269. Could this (and all other comparisons made between regions) not be tested statistically?

We added results from Kolmogorov-Smirnov test where we determined if there was statistically significant difference between soil frost seasons, individually for all three forest-soil types and station groups from southern, central and northern Finland.

Fig. 4 Text is far too small to read, lines should be thicker as well. Same for Fig. 7

Figures edited.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Overall, the authors' responses to my previous comments were helpful and I find this version of the manuscript much improved. I recommend minor edits. 

39. It is confusing to declare that winter storms are the most damaging but then emphasize damages in autumn and summer. I presume this is provided to contextualize the magnitude of damages between seasons (e.g. 100 v 25 milj.m3). Reword to clarify.

62. “Already” is confusing. Reword.

68. comma after “spring”

77. provide an example of magnitude to contextualize “a lot”

135. What about Betula? Considering differences in angiosperm and gymnosperm rooting depths, your results would not apply to Betula necessarily. However, it is mentioned as an important species earlier. It might be worth mentioning at some point why you did not consider Betula since it is mentioned earlier as an important species.  

Fig 2. Subset titles are too thin and appear inconsistent on page. Panel titles could still be larger.

280. be specific where possible. Roughly half is how many?

289. This is a helpful inclusion, but please specify the direction of difference/effect based on each KS test.

Fig 4. Legend fonts should be larger. Axis on inset panels are not readable.

456. These last two paragraphs are very helpful.

Author Response

39. It is confusing to declare that winter storms are the most damaging but then emphasize damages in autumn and summer. I presume this is provided to contextualize the magnitude of damages between seasons (e.g. 100 v 25 milj.m3). Reword to clarify.

We changed “Also in Finland” to “In Finland” for not to confuse Western and Central Europe, where most damages have occurred in winter storms, to the situation in Finland.

62. “Already” is confusing. Reword.

Changed “Already” to “Even”.

68. comma after “spring”

Comma added.

77. provide an example of magnitude to contextualize “a lot”

Added “even up to few months in mean duration”.

135. What about Betula? Considering differences in angiosperm and gymnosperm rooting depths, your results would not apply to Betula necessarily. However, it is mentioned as an important species earlier. It might be worth mentioning at some point why you did not consider Betula since it is mentioned earlier as an important species.

We added a mention about the exclusion of birch to the section where we mentioned birch is not vulnerable to wind damage from late autumn to early spring. Also, instead of “the most common combination of forest and soil types”, we changed it to “some of the most common…”

Fig 2. Subset titles are too thin and appear inconsistent on page. Panel titles could still be larger.

We ended up to remove subset titles and simplify panel titles.

280. be specific where possible. Roughly half is how many?

We added specific percentages for each soil type.

289. This is a helpful inclusion, but please specify the direction of difference/effect based on each KS test.

As we were comparing distributions, which were heavily overlapping, we thought quantifying the direction of difference was not that meaningful. It would have demanded to simplify distributions to e.g. mean value. Further, in our context the whole section is emphasizing the similarities between seasons, so we ended up to keep our approach of two-sided KS test.

Fig 4. Legend fonts should be larger. Axis on inset panels are not readable.

We enhanced parts of this figure.

456. These last two paragraphs are very helpful.