The Distribution and Activity of the Invasive Raccoon Dog in Lithuania as Found with Country-Wide Camera Trapping

Round 1

Reviewer 1 Report

The manuscript submitted by the authors on the distribution and activity patterns of the raccoon dog in Lithuania is of interest. In my opinion, however, it needs major revisions before it can be accepted for publication in Forests. Statistical analyses in particular must be revised. Standard t-test and ANOVA, which assume normal distributions and homogeneity of variances, are inappropriate in the present case. In addition, the authors seem to forget that their samples are sometimes “paired”, the same sites being involved across samples. Below, I suggest the use of simple non-parametric tests. Another possibility, smarter but probably more time-consuming, would be to retain t-tests and ANOVAs (taking into account that the same sites are sometimes involved) but to compute their P-values, performing random permutations of “RSF” values between samples. Other referees will possibly propose other alternatives, such as data transformation by log(x+1), the use of generalized linear models, or time-series analysis. The authors will then have to choose between the different proposals.

Specific comments:

Line 13. How can 57 sites out of 101 give a percentage of 63.4%?

Line 13. I suggest to replace “relative shooting frequency” by “shooting rate” throughout the manuscript (figures included), and thus “RSF” by “SR”. “Relative frequency” typically refers to a proportion or a percentage. Lines 208-219, the authors do compare relative shooting frequencies.

Line 138. I think that the authors mean “was calculated for each site”.

Line 140. I suggest: “number of days the camera was working”.

Lines 141-142. In the present case, standard Student t-test is inappropriate for comparing regions: the distribution of “RSF” values is clearly far from normal (it includes many 0 values) and within-group variance increases with group mean (as can be deduced from the results given by the authors lines 176-180 and the number of sites per region). I suggest to use Mann-Whitney test (= Wilcoxon test for independent samples) instead of Student t-test. The authors could also compare the three regions simultaneously, using Kruskal-Wallis test.

Lines 142-143. In the present case, standard t-test and ANOVA are inappropriate for comparing periods (years, seasons or months). As outlined above, the distribution of “RSF” values is far from normal, and within-group variance likely varies with group mean. Furthermore, the same sites are monitored throughout the study. As a consequence, samples are not independent, but paired. Finally, an additional difficulty is that there is probably a seasonal pattern (due to raccoon dogs’ reduced activity in winter) and that only two years (2020 and 2021) were entirely sampled. I suggest to use Wilcoxon paired-sample test to compare the 101 “RSF” values of winter and the 101 values of each other season (alternatively, the 101 values of winter and the 101 values computed for the remainder of the year). The four seasons might also be compared simultaneously using Friedman test. Then, I suggest to compare only years 2020 and 2021. This can be done season by season (4 tests) or month by month (12 tests), using Wilcoxon paired-sample test. Alternatively, this can be done, comparing the 12 monthly means of 2020 and the 12 monthly means of 2022 with Wilcoxon test paired over the months (1 test).

Line 144: Seasons are not defined. Are they delimited by equinoxes and solstices? Are they periods of three consecutive calendar months, such as April-May-June for spring?

Lines 152-156. In order to avoid confusion between “sr” and “SR” (see second comment for line 13), ‘sr” can be replaced by “r”, the single letter ordinarily used by Manly for the selection ratio (“selection ratio” rather than “selection rate”).

Line 164. See first comment for line 13.

Lines 164-167. Something is necessarily wrong: 38+14+6+6 ≠ 57.

Lines 172-175. A unique G-test (or Pearson’s chi-square test) comparing the three regions would be better than two G-tests.

Lines 172-175. G-test (and Pearson’s chi-square test) must be given with associated degree of freedom: df = 1 for a comparison between two independent frequencies, df = 2 for a comparison between three independent frequencies.

Lines 178-182. See comment for lines 141-142.

Figs 2a and 2b. Boxplots of the “RSF” values would be more interesting than plot of means and 95%CI. This would show that distributions are not normal.

Line 192. Standard Student t-test is also inappropriate in the present case. As for regions, a possibility for comparing the sites with and without large predators is to use Mann-Whitney test (= Wilcoxon test for independent samples). A better option would be to compute an independence chi-square (or G-test) on the 2x2 contingency table (presence/absence of raccoon dogs x presence/absence of large predators). Indeed, detection of raccoon dogs and detection of large predators are both random variables. However, this test is probably less powerful than Mann-Whitney test.

Lines 195-196. A test is lacking (see also comment for lines 142-143), and the sentence should be rewritten.

Lines 198-202. See comment for lines 142-143.

Fig. 3. This figure is of interest. However, the authors should display a broken line linking monthly means (as that linking hourly means in Fig. 4) rather than a curve obtained with a non-specified method.

Lines 205-206, first sentence. I suggest instead: “Overall, 69.5% (CI = 66.5–72.4%) of the photographs were recorded at night, xx.x% (CI = xx.x–xx.x%) during daytime, and xx.x% (CI = xx.x–xx.x%) during twilight hours.”

Lines 212-214 and 218-219. See second comment for lines 172-175.

Lines 223-230. After reading the Discussion, I guess that the authors use Manly’s selection ratio because other authors studying raccoon dogs’ activity used it. I must here point out that the null hypothesis of G-test (or Pearson’s chi-square) is not necessarily that relative frequencies are equal: the null hypothesis could be that the relative shooting frequencies recorded for two periods are proportional to the length of the considered periods.

Fig. 4. Day length varies throughout each season. So, black, grey and white areas cannot be simply night-time, twilight and daytime. As already outlined, the seasons must be defined (the four graphs suggest that they are not delimited by equinoxes and solstices).

Fig. 5. The authors should use a broken line (as in Fig. 4) rather than a curve obtained with a non-specified method.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors conducted a nationwide census of racoon dogs activity in Lithuania in the period 2019-2022 to understand distribution and seasonal activity patterns of this invasive species.

The study is generally well written with only few sentences to edit. The methods and stats require some major revisions. Furthermore, the discussion is only a collection of info from other studies, rather than a thorough discussion of the results of the present study and therefore is in need a major rewriting. See my comments below.

Line 15: correct with “sites where these predators were absent”

Line 16: not sure about the policy of this journal, but usually results are not reported in the abstract, I would delete this

Line 48: please define “passive days”

Line 137: please correct the title

Line 140: I assume there is a “day” missing in “number of cameras were working”

Line 141: difference between what? Sites, cameras, time of day?

Line 142: influence of years and month on what?

Line 164: here and throughout the result sections, I do not understand the need to report the mean±SE (or it is SD?) if the CI are also reported. This should be 4.30±1.42. Especially if your aim is to statistically compare the shooting frequencies among sites, seasons or else then only CIs are needed.

Lines 165-168: are these values CIs or min-max range?

Lines 197-202: ANOVA and t-test are parametric tests that can be used when variables are normally distributed. From the methods it is not clear if you checked for the normal distribution of the variables, but I doubt this was the case as you have integer numbers (i.e. number of photos per camera). Therefore the distribution should likely be poisson, rather than Gaussian, which means you should use a non-parametric Kruskal-Wallis test and not ANOVA.

Line 231-235: I understand you used the temperature value printed in the recorded videos, however in my experience I found that wildlife photo-traps very rarely collect reliable temperature data. Did you check the reliability of the temperature data before deployment in the field? Otherwise it will be much more reliable to use temperature data from local weather stations.

Lines 239-253: all this info should go in the introduction in a condensed form

Lines 262-285: again, this is info that should go in the introduction in a much shortened form. Furthermore, your study did not focus on feeding habits of the racoon dog, therefore there is no need at all to mention their diet habits. Additionally, you are only listing what is known from a series of studies and are completely shifting away from the main point of discussing your results in the wider context.

 Line 286-303: this is all useless info at this point. How does this fit within your study? Where are you discussing your results?

Lines 304-308: so far these are the only sentences that truly belong to the discussion…

Line 315: which neighbourhoods? You mean that racoon dogs share territories with wolf and lynx?

Line 317: change “registrations” with “recordings” here and throughout

Line 321: I am not sure that “pupping period” is the right term, rather just “pup period”

Line 322: change “contract” with “contrast”

Lines 338-343: see my previous comment about the reliability of temperature recordings by camera traps. It would be worth to double check you temperature data with data collected from weather stations

Lines 344-348: how do they find amphibians during winter? I would expect that they use other food sources as there are no amphibians

The English language requires only minor editing

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Thank you for addressing all my previous comments. I still have some comments though, see below.

Lines 137-141: thanks for addressing the issue I raised regarding the temperature recorded by the cameras. However, you should report quantitatively the difference you measured, i.e. mean±CI (or SE, SD) of the cameras and thermometer measurements. Perhaps more important is the consistency in temperature measurements of the cameras over time, given that you used the cameras for long periods. So, even if the cameras’ temperature recording is significantly different from the thermometer, you can still report that such difference is consistent over time and therefore make a proper recalibration.

Lines 150-152: I assume this is a formatting error rather than the title of the subsection.

Lines 157-159: it is not clear how you calculated the difference between years: mean values across 2019-2022, only for 2020-2021, mean value per month (i.e. 12 values per year), mean per seasons, all the pairwise comparisons between years? The Wilcoxon Signed-rank test is a non-parametric version of the paired sample t-test, so I am not sure why you used this, unless you are considering photos from the same camera across years as a repeated (paired) sample. But in this case a better way to analyse this would be with a GLMM. If you do not have repeated photos from the same cameras across years I would suggest to use a Kruskal-Wallis test as a non-parametric version of the ANOVA and test for difference of the mean RSF across the 4 years, or a Mann-Whitney if you are only analysing 2020-2021.

Lines 228-232: I don’t understand if this is only one of the pairwise comparisons or rather the difference between 2020-2021 was actually the only test you performed. In this case you need a Mann-Whitney test if samples are independent or a Wilcoxon if samples are paired (see my previous comment).

Lines 299-302: I would specify here that this is the temperature recorded by the cameras, not by the thermometer. Unless you clearly show in the methods that the two are equivalent.

Line 318-326: A discussion should always start by presenting in few sentences the most important results of the study and then elaborate on these. All the info presented here still belong to the introduction or, if you really want to present it here, should not be presented at the very beginning. It just does not read well and I struggle to understand how this can help a reader understand your results without some context.

Author Response

Please see the attachment

Author Response File: Author Response.docx