How Climate Change Affects River and Lake Water Temperature in Central-West Poland—A Case Study of the Warta River Catchment

Reviewer 1 comments

Responses to the Reviewers

Major comments

In my opinion it would be clearer to analyze measured temperature only and discuss the trend calculated from that. Most time series still would cover a period of 30 years, which is equivalent to the time periods used in climate analyses and thus long enough.

Thank you for the comment. Thank you very much for your insight into the results of our analysis. Indeed, as indicated in the paper we also used data that were reconstructed. The method of data reconstruction in rivers is described in detail in the paper presented by Sojka and Ptak [49]. Due to the fact that Poland stopped measuring water temperature in rivers after 2014. At the same time, this period is one of the most interesting from the point of view of climate change. Therefore, this paper is not limited to the analysis of the period 1984-2015 for which direct data on river water temperature were available. As demonstrated earlier in the paper Sojka and Ptak [49], it is possible to reconstruct these data using artificial neural networks. Using air temperature and lake water temperature data, such a reconstruction can be performed and the results obtained are reliable. This was confirmed by their independent verification on the basis of direct measurement data. The objective of the paper is the assessment of the direction and extent of changes in water temperatures in rivers and lakes in the context of changes in air temperature.

In the current form, with a mix of measured and simulated data, it needs to be made clear from the beginning, that the analysis is (partly) based on modelled data, i.e. in the abstract, maybe in the introduction

Thank you for the comment. Thus, measured data and data reconstructed using artificial neural networks were used in the analysis of river water temperature changes. Therefore, the following information in the abstract section has been completed: The analysis of river water temperatures used both measured data and data reconstructed using artificial neural networks.

Moreover, the following information has been added to the conclusions section: “Furthermore, it should be emphasized that the analysis of changes in river water temperatures was carried out on the basis of both measured and reconstructed data. In order to clearly assess the response of rivers to climate change, direct measurements of river water temperature should be restarted as soon as possible.”

p11, l 365 – 368: The authors discuss that they showed that changes in air temperature are the main factor for changes in water temperature. This is not surprising when water temperature is calculated from air temperature. I think based on the method they used, they cannot discuss this relationship or discuss the influence of local factors on river water temperature, when it has been calculated also based on lake water temperatures measured up to 100 km and more away. Can the authors please explain?

Thank you for the comment. The reconstruction of river water temperatures was based on water temperature data from one lake and air temperatures from three meteorological stations. The distances from the river station to the hydrological station during the reconstruction ranged from 2.1 to 109.9 km. The average distance at the meteorological station from the hydrological station is 51.5km (Table 1 - below).

Three stations were taken into account because the river water temperature at the hydrological station is the resultant of the climatic conditions in the catchment, and in this case it is required to have data not from one station but several in this study of three meteorological stations. Finally, each artificial neural network used to reconstruct the data was validated, using the procedure described in the methodology section. To assess the quality the MLP the following statistical parameters were used: coefficient of determination, root mean square error, normalised root mean square error, Nash–Sutcliffe model efficiency coefficient, and mean absolute percentage error.

Minor comments

Could the authors please specify how they chose the time period for the analyses, why they choose the years 1984-2020

The period 1984-2020 was chosen because, at the time of beginning working on this paper (calculations and statistical analyses), this was the longest verified data set that had been provided by the Institute of Meteorology and Water Management - National Research Institute (IMGW-PIB). Now the verified dataset is available for the period 1984-2021.

L32: “rivers waters” should be “rivers”. “lotic systems” could also work

Thank you for the comment. Revised as suggested.

L35: cross out “An increase in inland waters is greater than for seas and oceans”, because it is a repetition of the previous sentence. If this sentence is supposed to refer to a global level, then a reference is needed

Thank you for the comment. The indicated paragraph has been removed. As pointed out, this was a repetition.

L44: “lowland rivers are the most susceptible to climate change”: could the authors please provide an (additional) reference from an international journal (English written) for this general statement?

Thank you for the comment. Revised as suggested. The pointed paragraph has been made more precise. Graf and Wrzesiński (2020) performed the analysis on the example of Polish rivers. Therefore, it is only appropriate that “In Poland, the lowland rivers are the most susceptible to climate change”.

L55: I don’t understand why an increase in temperature results in a higher exposure to cryophilic organisms, that is organisms, that is organisms that grow at low temperatures. Rephrasing the argument or being more specific may help here.

Thank you for the comment. The indicated paragraph has been removed.

L66/67 “ecosystems of lotic waters are more susceptible to climate change…” more susceptible than what? reference not clear

Thank you for the comment. Revised as suggested. “In Europe, river waters are more affected by climate change than marine and ocean waters [14, 15]. Rivers are therefore more vulnerable to biodiversity loss than seas and oceans.

L68 “it is forecasted… over the upcoming years…” This sounds like a relatively certain change in the near future. It may be better to use the word “project/projection” and specify the time range, e.g. “until the end of the century”, “over the next decade”…

Thank you for the comment. Revised as suggested. “Moreover, it is projected that in the near future the greatest losses will occur in the case of common species of ichthyofauna and invertebrate macrofauna.”.

L97: “by forest” instead of “by forest complexes”

Thank you for the comment. Revised as suggested.

Table 1: How were the lake stations chosen and how was is chosen which lake refers to which river station?

·   For example, for river station d it looks like lake II is closer than lake III

·   station h is very far away from lake III

·   how are the lakes characterized? Are they shallow/deep, large/small, what is their residence time? Were the lakes similar to each other or do they differ? All this will affect their water temperature and may bias the estimate of the river water temperature

Thank you for the comment. As indicated in section 2.3 (2.3 Data reconstruction), the methodology developed by Sojka and Ptak [49] was applied to the data reconstruction. That research showed the best results of river water temperature reconstruction were obtained based on data from three meteorological stations and one hydrological station. Sojka and Ptak [49] considered stations located at the closest distance. In this study, they also selected meteorological stations and lake hydrological stations that were closest to the river hydrological station.

·        The river station d is located at a distance of 71.2 km and 71.5 km from lake stations III and II respectively. Therefore, the nearest station was used for the reconstruction following the adopted methodology. For river station d, is lake station III.

·        Lake station III is located at a distance of 185 km from river station h. According to adopted methodology, MLP learning was performed in the first stage to find the optimal architecture employing the automated network design method. To assess the quality the MLP the following statistical parameters were used: coefficient of determination - 0.995, root mean square error - 0.412 ^oC, normalised root mean square error - 0.021, Nash–Sutcliffe model efficiency coefficient - 0.9988, and mean absolute percentage error - 5.64%. These results show that there is no reason not to adopt this MLP architecture (using data from Lake Station III) for data reconstruction at station h.

·        The lakes taken from the analysis have the following morphometric parameters. Water temperature studies are carried out by the largest lakes. The influence of morphometric parameters of the lakes and residence time on the results of the river water temperature reconstruction was not investigated in this study. This is an interesting scientific problem that could be addressed in subsequent. analyses.

Table 2. Morphometric data of the lakes

L134: “analogical parts” what is meant by this? two parts of the same size?

Thank you for the comment. The sentence has been revised to "Data were divided into two parts. The first includes 70% of the data used during the artificial neural network learning stage, and the second includes 30% of the data used during the testing and validation stage.”.

L179-183: was the autocorrelation coefficient significant in this study? and was the correction factor calculated in this study?

Thank you for the comment. The aim of the study was not to present autocorrelation results. In each case, autocorrelation was calculated. Only when the autocorrelation coefficient rkR was significant at a level of 0.05, according to Hamed and Rao [60], the correction factor n/n* was calculated, followed by corrected variance Var*(S). Then, the value of the corrected variance Var*(S) is considered in the calculation of the test statistic Z. Such a method of data analysis was applied by Abdul Aziz and Burn [61], Khattak et al. [62], Kumar et al. [63]. The analysis of the direction and extent of changes by means of MK and S tests employed the modifiedmk package developed by Patakamuri and O’Brien [65]. The statistical analysis by means of a Mann-Kendall test was conducted with the assumption of the standard levels of significance of 0.05 and 0.01.

L195: variables of the equation need to be explained

Thank you for the comment. Revised as suggested.

L210: “at meteorological station” (not “in”)

Thank you for the comment. Revised as suggested.

Figures 3 & 4: colours need to be explained; vertical lines in the right plot need to be explained

Thank you for the comment. The legend and description of the vertical lines in Figures 3 & 4 were completed.

Figures 3 & 4: the horizontal bars are not well legible. Either they can be combined or the figure needs to be larger, e.g. higher

Thank you for the comment. The vertical size of the figures has been increased.

L250: station name/letter missing in both brackets

Thank you for the comment. Based on the comments made by Reviewer 2, Section 3.2 River water temperature has been rewritten. The indicated paragraph has been deleted.

L252/253: repetition to previous sentences. Also, could the authors quantify the variability? It looks like the range is larger in April

Thank you for the comment. Revised as suggested. The indicated paragraph has been edited. “The highest range of temperature changes between river stations also occurred in July.”

L266 “mid 1990’s”

Thank you for the comment. Revised as suggested.

L268: The authors write that organic soils may affect water temperature, can they please specify the process?

Thank you for the comment. Revised as suggested.

“Moreover, the Ner River within the range of the hydrological station flows through a flat valley filled with peats. The above situation may be of great importance, in terms of water temperature changes, in the context of groundwater alimentation [9].”

L277: “were recorded” recorded or calculated? That is, measured or simulated data?

Thank you for the comment. In this study, some measured water temperature data and some reconstructed data were used during the trend analysis. A better expression would be to use “were indicated” instead of “were recorded”.

L302: Are lake temperatures surface water temperatures?

The following information has been completed in the methodology section. The water temperature measurements were conducted at a depth of 0.4 m under the water surface at 6.00 UTC. The data were compiled for the hydrological year beginning on 1 November and ending on 31 October of the following year.

Figure 5: Why does lake III show such a large variability?

Thank you for the comment. The changes for Lake III are different from others. Looking at the other lakes I and II also clearly indicates their individuality. The reasons for this variability in each lake have not been investigated in this paper. It is likely to be influenced by both climatic and non-climatic local factors. A detailed study should be undertaken to answer this question.

L405: Here the authors conclude that e.g. groundwater and forest play a role in differences between river water temperature and air temperature. Although I truly believe this statement, I think this conclusion needs to be backed up by data.

Thank you for the comment. Revised as suggested. Information on the effect of forests on river water temperature has been removed. We agree that this was not the scope of this study.

River - Station

Meteorological station

Distance (km)

Warta - Gorzów Wielkopolski (a)

Gorzów Wielkopolski (1)

2.1

Piła (2)

109.9

Słubice (3)

60.4

Obra - Zbąszyń (b)

Leszno (4)

62.3

Wielichowo (5)

33.3

Zielona Góra (6)

44.9

Gwda – Piła (c)

Chrząstkowo (7)

56.4

Gorzów Wielkopolski (1)

108.3

Piła (2)

2.4

Warta – Śrem (d)

Leszno (4)

43.8

Poznań – Ławica (8)

37.9

Wielichowo (5)

45.2

Prosna – Bogusław (e)

Kalisz (9)

15.7

Koło (10)

58.9

Poznań – Ławica (8)

95.8

Ner – Dąbie (f)

Kalisz (9)

61.2

Koło (10)

17.0

Łódź – Lublinek (11)

56.2

Widawka – Szczerców (g)

Łódź – Lublinek (11)

47.3

Sulejów (12)

52.5

Wieluń (13)

41.0

Warta – Bobry (h)

Lgota Górna (14)

48.6

Sulejów (12)

48.7

Wieluń (13)

62.8

Mała Noteć – Gębice (i)

Koło (10)

61.7

Kołuda Wielka (15)

16.9

Toruń (16)

62.0

Noteć – Nowe Drezdenko (j)

Gorzów Wielkopolski (1)

39.7

Piła (2)

68.5

Poznań – Ławica (8)

83.0

Reviewer 2 comments

Responses to the Reviewers

In this manuscript, some minor English errors and expressions need to be seriously checked and revised. For instance, in line 208 and 238, “multiannual” should be “multi-annual”.

Thank you for the comment. Revised as suggested.

Line 362-363, In the study of Ptak and Seidhar, the impact of watershed forests on river water temperature can be described in more detail.

Thank you for your comment. Completed as suggested. Ptak (2017) showed, significant differences in the thermal regime of two rivers located 17 km apart in the period from May to October. The reason for the differences in the thermal conditions of the waters of the two rivers, with relatively homogeneous climatic and hydrological conditions, was the differences in the catchment use structure - mainly forest cover. The share of forests in the area of these catchments was 71.6% and 39.3%. The length of the main river in direct contact with the forested areas represents 68.3% and 21.6% respectively. Water temperatures in the non-forested catchment between May and October were even several degrees higher, moreover, they were characterised by higher variability. Modelling studies by Sridhar at al. (2004) showed that the greatest reduction in river water temperature was predicted for mature forest (high LAI) areas that are located by the river (i.e. within 10 m of the stream bank) and that are about 30 m wide.

The scale in (a) and (b) of Figure 2 should be used uniformly

Thank you for your comment. Figure 2 a and b present different information (a) rate of changes in ^oC per decade and (b) the significance level of the changes given that the legend cannot be normalised. In comparison, the scale of the figures is the same.

3.2. River water temperature is too wordy. It can be simplified.

Thank you for the comment. Section 3.2. River water temperature was revised. Less important information has been deleted.