the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Nov 2023
60-years analysis of drought in the western Po River basin
Abstract. Since the start of the 21st century, greater focus has been put on drought and its wide range of environmental and socioeconomic effects, particularly in the context of climate change. This is especially true for the North-western region of Italy, comprising the Piedmont and Aosta valley, which have been affected in recent years by droughts that have had acute effects on water resources and water security in all sectors, including agriculture, energy and domestic use. The region also belongs to the Mediterranean hot-spot, characterized by faster than global average warming rates and higher vulnerability to their effects. Therefore, characterizing the observed changes and trends in drought conditions is of particular significance. To this end, 60 years of precipitation and temperature data from the North West Italy Optimum Interpolation data set are used to calculate the drought indices SPI (Standardized Precipitation Index) and SPEI (Standardized Precipitation Evapotranspiration Index) at a shorter (3-month) and at a longer (12-month) time scale. First, trend analysis on precipitation and temperature is performed, finding limited areas with significant precipitation decrease and, conversely, a general temperature increase over the region, with higher values found in the higher elevation areas. Changes in meteorological drought are then evaluated, both in terms of drought indices trends and in terms of changes in the characteristics of drought periods, on both a local and regional scale. A relation between the altitude of the area and the observed changes is highlighted, with significant differences between the plain and mountainous portion of the region. The differences are mainly related to the observed trends, with the low-altitude part of the region displaying a tendency towards dryer conditions not in common with the mountainous area. Significantly, no trend is found at a region-wide level but is instead found when considering homogeneous areas defined by terrain ruggedness. Furthermore, changes in the number of drought episodes and in their severity, duration and intensity are found to be correlated with terrain ruggedness at all time scales.
- Preprint
(31711 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on hess-2023-218', Anonymous Referee #1, 12 Dec 2023
The authors explore trends in meteorological droughts, annual and seasonal precipitation and temperature in the northwest of Italy (western Po river basin) from 1957 to 2023. In this case study, the authors also investigate if trends in lower altitudes are similar to those in the mountains. Trends are analyzed with the Sen’s slope (for rainfall and temperature), with a t-test (for SPI and SPEI drought indices), and compared with elevation indices (elevation mean and spatial variance). Conclusions show that the trends are different for the plains and mountainous regions. Drought characteristics increased in the plains due to precipitation and temperature changes. Droughts did not change much in the mountains due to non-significant precipitation change, despite significant warming.
The methods are mostly appropriate and the study region is interesting due to the diverse landscape. However, the manuscript does not present substantial conclusions or methodological advances. The findings are not contrasted with other studies and the novelty for hydrology and earth system sciences is unclear. Trend detection has been largely explored in Europe (as cited in the Introduction section and references therein) and trend attribution is not investigated or discussed. Results and Figures are not discussed clearly and concisely.
Specific comments are provided below.
- The first paragraph of the Introduction is too general and not directly associated with the aim of the study. It would read better if presented more concisely.
- Line 36 needs citations for “changing patterns of meteorological droughts”; and also for “an increase in drought occurrence in the area has been detected”.
- Lines 36-43 are unclear. When and where are such changes detected? How do the “reported changes differ significantly”? How consistent are the following results: “studies considering both precipitation and temperature (Hanel et al., 2018; Falzoi et al., 2019; Arpa Piemonte and Regione Piemonte, 2020b; Vogel et al., 2021; Baronetti et al., 2022) have found more consistent results”? Why is “the rise in evaporation as a main factor in drought increase”?
- Lines 48-49: What are the results conflicting and what is the consensus?
- Lines 51-54: When and where did these studies investigate? Does it also cover the study area of the current manuscript?
- Lines 58-60: Why did the authors choose the present study area? For example, does it have longer or more extensive observed data sets to analyze the effect of elevation on meteorological trends? Please specify that only meteorological droughts are investigated. It is also proposed to contrast the drought conditions in northern Italy to those in northwestern Italy, but the manuscript only presents for northwestern Italy.
- Figure 1. What are the elements shown in the larger map (colors, lines, names)? Is it elevation, land cover, rivers, roads? Every element should be either described or removed from the map. The river network should also be included, as many readers may not know the extent of the Po river. Roads and region names that are not relevant to the study should be removed. Also, the same projection should used in all maps, so that the shape of the study area (and the latitude-to-longitude ratio) on the larger and smaller maps are the same. Similar reasoning applies to the other figures.
- Line 71-72. Do you mean “bordered by France on the west and south-west”? And “two other Italian regions (…) on the east and south-east?
- The latitudes described in lines 76-77 and 96 are not needed and can be inferred from Fig. 1.
- A description of the data projection in Line 96 is not needed. The total number of grids described in lines 96 and 106 are also not needed because the spatial resolution of the data is already provided.
- Please provide a reference to the data set (Line 95) and the interpolation method (Line 97).
- Section 2.2. Why are these data explored and how do the results change if other data sets are analyzed? Do the data explored have a dense gauging network? How many gauges does it include? It is important to provide specific data details due to the small study area and the high result sensitivity to different data sources (cited in the Introduction).
- Figure 2. Is “Terrain roughness” the same as “Elevation standard deviation”? Also, is it “terrain roughness” or “terrain ruggedness”? It would be nice to standardize throughout the text. Please clarify the units of “Terrain roughness”. What does F and F(x) refer to?
- Lines 113-116. What is the definition of terrain ruggedness (its concept) and how is it calculated? Please provide precise details in the main text. Also, why did the authors choose the present metric and how does it compare with other metrics (e.g., terrain slope, amplitude, other terrain ruggedness indices)? Line 113: “height”. Do you mean “elevation”?
- Lines 116-118 are a repetition of the previous section. Either make it more concise or remove the sentence.
- Lines 120-123. A description of the results is not needed here. Can remove these sentences.
- Lines 124-127. I could not find the results compared with the K3 Mountain classification. How did the authors find that the classification is “quite satisfactory”? How satisfactory is it?
- Table 1. These thresholds were not used or discussed in the manuscript. Thus, the table could be removed.
- Lines 157-159. These sentences are unclear.
- Line 164. Is it potential evaporation?
- Line 184. How were the series deseasonalized? With “seasonal precipitation series”, do you mean the series for each season, or the series with the seasonal time scale?
- Figure 3. I could not find DDr and DSr in the figure.
- Lines 200-202. It would be nice to clarify the differences between Fig. 3a and 3b. Also, please define precisely when the drought ends in each case. How sensitive are the results for different thresholds in defining when the drought ends?
- Line 213. “discarded”. Do you mean “analyzed”?
- Sections 2.4.4 and 2.4.5. I find it hard to distinguish between “drought run”, “drought event”, “drought episode” in the Methodology and the Results sections. It might be clearer to refer to droughts in a single cell as “local droughts” and in multiple cells as “regional droughts”, or some other term related to the spatial differences. Please standardize throughout the manuscript.
- Figure 4. (a) Should also be presented and discussed in relative terms (units of % per year relative to the long-term precipitation). The negative values in the colorbar of (b) and (c) should have a smooth transition from zero, symmetric to the positive values. The colorbar in (a) is ok. Please label the numbers in the x and y axis (◦N and ◦E) and use the same projection as Figure 1.
- Line 251-252. Figure 4 does not present seasonal trends.
- Line 255-257. Do the authors mean that precipitation trends are not the same in the entire area?
- Correlation values should always specify their associated p-values and which correlation method is used (Pearson, Spearman). If Pearson correlation is used, are its assumptions met? These should be clarified in Lines 260, 265, and throughout the manuscript.
- Line 279-281. Why are these associated with soil moisture and groundwater? Please provide citations or the results of an analysis.
- Lines 283-285. These sentences are unclear, please revise them. What are “worse conditions”?
- Figure 5. (a) Is precipitation units in mm per month? This figure is not discussed in the manuscript. Either discuss it or remove it.
- Lines 290-293. How relevant is the magnitude of the SPI and SPEI trends?
- Figure 6. (e) and (f) should avoid differentiating the variables by blue and red colors because it creates some confusion with (a) – (d). Also, what exactly is the unit Δindex? Why are the trends presented in month units here but in year units in Fig. 8?
- Table 2. Are “Number of runs” the number of drought events? Is C the correlation coefficient? Why is it that only one variable has units?
- Figure 8. This figure is hard to understand and is also not much discussed in the manuscript. Red and blue colors are used in the other figures to differentiate between increasing or decreasing trends, but here denote different variables, creating some confusion.
- Figure 9. (a) Do the negative and positive y values represent different variables? If so, this should be clarified by using two different y axis and by describing in the figure caption.
- Line 422. What does “worse” refer to here?
- Section 4 – Discussion and conclusion. I could not find any Discussion in this section, only an overview of the results. The manuscript should have a discussion section showing how the current results contrast with other studies, what is its novelty, how it is relevant for hydrology or earth system sciences, and how it could be extrapolated to other study areas.
Citation: https://doi.org/10.5194/hess-2023-218-RC1 -
AC1: 'Reply on RC1', Emanuele Mombrini, 19 Dec 2023
Authors’ reply to Reviewer #1
We thank Reviewer #1 for her/his thorough and in-depth analysis of our work, and for all the comments and suggestions regarding the conciseness, readability and clarity of our paper. We think that addressing all these specific concerns in the revised version of our manuscript will greatly increase its quality. We feel that it is important to address the concerns raised about the novelty of our analysis by including a broader discussion of our results in the revised manuscript. Still, we would like to offer a response here to what we think are the main critical points of Reviewer #1’s comments:- Overall novelty of the study
We think that our paper, through the detailed analysis of the meterological variables in north-western Italy, advances our understanding of hydrological systems in general, and thus falls within the scope of the HESS journal. In particular, this study shows evidence of complex interactions between terrain characteristics and wetting/drying trends in the area. Namely, the ruggedness of the terrain is better correlated than elevation to the observed trends/changes (which has not been shown in other studies, as far as we could find). This emerges thanks to the characteristics of our domain—which includes a low-elevation hilly area and allows us to distinguish the trends in pre-alpine and hilly areas located at similar elevations. Besides this point, our study also highlights the spatial features of the temporal evolutions of drought indexes, coupling the spatial and temporal analyses to offer a less than usual approach to trend analysis (e.g. In Figure 12). Given these points, we feel that the methodological approach and results originality may justify the interest of the scientific community. It will be our goal, during the review phase, to better emphasize these points in the revised manuscript, given the suggested lack of clarity about them.- Choice of domain
This analysis is performed in a region that corresponds to the headwater of the largest catchment in the Southern European Alps, where high mountains surround densely populated areas and fertile agricultural plains. To our knowledge, the region hasn’t been the subject of similar studies before. While the region has been included in the domain of numerous drought trends studies, as noted by Reviewer #1, this has generally been done either without sufficient spatial resolution or without a specific focus on the distribution of such trends on a smaller scale. The choice of the domain is also determined by the authors’ familiarity and interest in the region, the aforementioned terrain characteristics, and by the occurrence of a particularly severe drought in the 2021-2022 period, which sparked widespread interest in the phenomena.- Choice of dataset
Our choice of dataset is functional for the purposes of our study, because of its spatial resolution and its observation-based nature. As detailed in Appendix A, the interpolation method used for the dataset doesn’t add or remove elevation trends and is thus suitable for studying the relations between meteorological variables and terrain characteristics. Furthermore, the dataset has a much higher number of gauging stations in the region compared to other datasets, with hundreds of stations in the area (see https://doi.org/10.5194/nhess-13-1457-2013).- Contrast with previous studies
Studies regarding the relation between drought trends and elevation have been performed in other parts of the globe (namely China https://doi.org/10.1038/s41598-020-71295-1, Iran https://doi.org/10.1007/s00704-020-03386-y, India https://doi.org/10.1016/j.atmosres.2023.106824 and the Canary Islands https://doi.org/10.1038/s41612-023-00358-7), but have found contrasting results regarding the distribution of wetting/drying trends in high/low elevation areas; thus, we feel that further study in this field is warranted, as no general conclusions can be drawn given our current understanding of the involved processes. As stated, our study also shows that areas with particular topographical features may show not only elevation-dependent effects, but also more complex topography-related effects.Finally, we respond to other more specific concerns raised in Reviewer #1’s comment.
- Specifying that only meteorological droughts are investigated:
We agree that the title could be misleading. Although the use of the SPI/SPEI indices at different time scales to monitor other types of drought is established in the literature (https://library.wmo.int/viewer/39629?medianame=wmo_1090_en_#page=1&viewer=picture&o=bookmarks&n=0&q=, https://doi.org/10.1175/2012EI000434.1) this is not the ultimate goal of our study, and we feel that the choice of studying meteorological data should be made clearer. Apart from making the necessary changes to the manuscript, we propose to change its title to reflect that a drought analysis of meteorological data is performed.
- Choice of terrain ruggedness as a variable:
The choice of terrain ruggedness as a variable is motivated by a need to differentiate areas with distinct terrain characteristics (plains, hills and mountains) which would have been grouped together if classified through elevation bands. We acknowledge that other variables could have been used, such as those proposed by Reviewer #1, but the impact of a similar but different classification would not change the results. The comparison with the K3 mountain classification (which we will make more explicit in the revised manuscript) was in fact used to ensure that the chosen variable could distinguish correctly the different types of relief present in the domain.Citation: https://doi.org/10.5194/hess-2023-218-AC1
-
RC2: 'Comment on hess-2023-218', Anonymous Referee #2, 19 Dec 2023
The authors analyse droughts in North-West Italy. They aim to investigate the difference in drought trends across different levels of terrain ruggedness. Overall, it seems like an interesting study that could contribute some interesting findings. However, the manuscript needs some improvements. In general, the main aim of the paper and how this contributes to the existing literature could be made clearer. In addition, if the aim is indeed to investigate the correlation between drought trends and elevation/ruggedness this should be better reflected in the methods and results, since as it is now only a small part of the results takes into account the elevation/ruggedness and the rest just focuses on the trends in droughts.
Some more specific comments:
- The first part of the introduction is very general and not very relevant.
- In lines 36 to 38 the authors state that the fact that drought occurrence increases is contradictory to the finding that recent droughts are not exceptional. However, this does not necessarily contradict each other, droughts can occur more frequently, even if the individual droughts are not more exceptional than previous droughts.
- In general, the introduction describes a lot of research on drought and its relation to orography that has already been done in Italy. It is not very clear to me which research gap the authors aim to address with this study and how this will contribute to an improved understanding of drought.
- In section 2.2 the authors describe that the data set they use is a gridded dataset, based on the interpolation of station data. How are these results affected by the interpolation method used? Why not analyse the station data, instead of the interpolated data?
- In section 2.3, according to the section title, the authors describe how they divide the areas based on elevation. However, from the text I understand that the division is actually based on ruggedness and not on elevation. Also, what is the difference between terrain roughness and ruggedness? Or are they the same? They seem to be used interchangeably. Please explain the difference and clearly state which one is used, or if they are the same, make sure to be consistent throughout the manuscript. Also, the authors state that they investigate orography, meaning the combination of elevation and ruggedness, but in the end they define groups based only on ruggedness, and not elevation. This should be corrected in the rest of the manuscript, where it is sometimes stated that the correlation between drought and orography is investigated.
- The caption of figure 2 mentions ruggedness, while in the figure roughness is used. Also, in the caption “(d) correlation …” should be (f) and the caption states that it is the correlation between elevation and elevation standard deviation, while the axes in the figure describe mean elevation and terrain roughness. Although I understand that this is how the roughness is defined, it is better to be consistent and use the same term.
- From section 2.3 it seems that you are actually also investigating the differences in trends for different elevation groups and comparing that to the ruggedness groups. It would be good to make this more clear throughout the manuscript (e.g. also in the introduction). In addition, you could consider also showing the classification based on elevation in figure 2.
- To calculate the parameters of the SPI, the authors use the maximum likelihood method and for the SPEI, they use probability weighted moments. Why not use the same method (if there is a good reason, please explain) and could this affect the results (e.g. the difference in the trends between SPI and SPEI)?
- The section on trend analysis does not describe the method that is used for trend analysis, please add this. In addition, how are seasonal precipitation series defined and how are temperature series deseasonalised?
- The difference between drought runs and drought events is not very clear. Are drought runs based on one pixel and drought events based on multiple pixels? For the drought events, is the same method used as for the drought runs, but with the additional condition that 25% of the domain needs to be in drought? In addition, why did you choose 25% as threshold and what does “domain” mean? Is this the total case study area or the area within the different ruggedness areas? If the latter, could this introduce some bias in your results? Since the areas with low terrain ruggedness are very close together and the higher terrain ruggedness are more spread out, so less likely to be all in drought conditions at the same time?
- The authors use a t-test to calculate the difference between the means of the two periods. Why was this method used for trend detection and not another method? What are the underlying assumptions of this method? Do they hold and what are the potential implications for the results? In addition, why not compare the number of drought events between the two periods (in addition to severity, duration)?
- Figure 5 shows time series for a representative point in the domain, where is this point? And how can one point be representative if four different areas are investigated?
- From section 3.3, it seems that linear regression was used to calculate trends? This is not mentioned in the methods.
- When analysing the trends in drought runs and events, this seems to not be separated by area. Why not, since the main aim of the paper is to show the effect of ruggedness on drought trends?
- There is no discussion of the results, only a summary. Are the results similar to the findings of the studies discussed in the introduction? And if not, why not? How are the results affected by the choice of methods?
Citation: https://doi.org/10.5194/hess-2023-218-RC2 -
AC2: 'Reply on RC2', Emanuele Mombrini, 29 Dec 2023
We thank Reviewer #2 for the detailed analysis and the precise comments about our methodology and presentation/discussion of the results. While we will address all the concerns raised in a revised version of our manuscript, we would like to offer responses to two relevant points here.
1) Reviewer #2 raises concerns about the novelties of the work. We refer to our comment to Revier #1 for the concerns regarding the novelty of our study, both in terms of research gaps addressed, comparison with previous studies and need for a more in-depth discussion of our results. In general, we feel that the high resolution of the studied data and the topographical characteristics of our domain show evidence of a complex terrain dependency between drought trends and terrain characteristics (and not only elevation) which is, as far as we know, not addressed in other studies. Also, the space-time identification of flood events and their trend analysis is not usually performed in drought analyses and constitutes a contribution of this paper. Still, we will give greater attention to these aspects in the revised version of our manuscript.2) One point that we wish to address in more detail is the use of the maximum likelihood method for the estimation of gamma distribution parameters in the calculation of the SPI and of the PWM for the estimation of the log-logistic distribution in the calculation of the SPEI. Our choice of parameter calculation was based on the available literature (for the SPI, see https://doi.org/10.1007/s12145-014-0178-y, https://doi.org/10.1007/s11269-012-0026-0, both citing the formulas proposed in https://doi.org/10.1175/1520-0493(1958)086%3C0117:ANOTGD%3E2.0.CO;2, for the SPEI, see https://doi.org/10.1002/joc.3887). In Beguería et al., (2014), the authors discuss briefly the effects of choosing the PWM method over the maximum likelihood method for the estimation of the log-logistic paramerters, stating that “[t]he SPEI series based on maximum likelihood were very similar to those based on the unbiased PWM method […]. Given that calculation of the maximum likelihood estimation was about two-fold more time consuming, we conclude that the unbiased PWM method should be preferred for computation of SPEI series”. The negligible lack of impact, apart from computation performance, of the choice of parameter estimation is also cited by Laimighofer et Laaha (2022, https://doi.org/10.1016/j.jhydrol.2022.128385). We will however check whether the results are affected by the choice of the parameter estimation method in the revised version of the manuscript.
Citation: https://doi.org/10.5194/hess-2023-218-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 468 | 171 | 35 | 674 | 32 | 30 |
- HTML: 468
- PDF: 171
- XML: 35
- Total: 674
- BibTeX: 32
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1