the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Spatio-temporal analysis of the role of climate cycles on landslide activity: the case of Majorca (Spain)
- 1Geological and Mining Institute of Spain from the National Research Council (IGME_CSIC), Urb Alcázar del Genil, edf Zulema 4 bajo, 18006 Granada, Spain
- 2Geological and Mining Institute of Spain from the National Research Council (IGME_CSIC), Ríos Rosas 23, 28003 Madrid, Spain
- 1Geological and Mining Institute of Spain from the National Research Council (IGME_CSIC), Urb Alcázar del Genil, edf Zulema 4 bajo, 18006 Granada, Spain
- 2Geological and Mining Institute of Spain from the National Research Council (IGME_CSIC), Ríos Rosas 23, 28003 Madrid, Spain
Abstract. Intense precipitation is one of the main drivers of landslides around the globe. On a global scale, the occurrence of very wet periods is controlled by different, well-known natural cycles: ENSO, NAO, SUNSPOT, etc. In this paper, we present a spatio-temporal analysis of climate cycles on the island of Majorca (Spain) and their correlation with the landslide inventory. Firstly, using spectral analysis techniques, the main climatic cycles that control the rainy periods on the island have been identified. For this purpose, rainfall data from 62 weather stations have been analysed in a comprehensive manner, with time series of more than 30 years. The cycles with the greatest influence on rainfall, from the point of view of statistical confidence, are ENSO (5.6 y and 3.5 y), as well as NAO (7.5 y) and QBO. Then, using geostatistical methods, the distribution of rainfall during dry, average, and wet years was mapped, as was the spatial representation of the statistical significance of the different natural cycles, in order to define the areas of greatest danger from heavy rainfall. The Serra de Tramuntana is not only the rainiest region of the island, but also the area where the highest values of statistical confidence for the set of climatic cycles are concentrated. The 5 largest landslides of the series are very well located in the areas of highest statistical weight, mainly for the NAO (7.5 y), QBO and HALO (22.4 y) cycles, as well as in the wettest sector for the wet type year.
Juan Antonio Luque-Espinar et al.
Status: final response (author comments only)
-
RC1: 'Comment on nhess-2022-50', Stefan Hergarten, 09 May 2022
This paper addresses the occurrence of various climate cycles in precipitation data on the island of Mallorca. As a major finding, the effect of these cycles shows a strong spatial variation. From this finding, conclusions about the importance of the climate cycles on landslide frequency are drawn.
First, I have to state that I am not an expert in climate analysis. While I found the part about the climate cycles interesting, the relation to the landslides does not make much sense to me. Landslide occurrence depends on topography as well as on precipitation. So it is not very surprising that a large amount of the documented landslides took place in a few very wet years. The five biggest landslides happened in regions where relief is -- to my knowledge -- very high. However, I did not find any serious analysis of the spatial distribution of the smaller landslides.
So what can we conclude if we find that one of the considered climate cycles has a quite strong effect on precipitation in the region with the highest number of large landslides, which is rather wet and steep anyway? I think we cannot draw any conclusions about the relation between landslides and climate going beyond what we already know. So the paper does not keep the promise made in the title.
Based on the vague relation to landslides, I cannot recommend publication of the present paper. I might have missed the key point, in which case I would have to apologize.
Best regards,
Stefan Hergarten-
AC1: 'Reply on RC1', Juan Antonio Luque Espinar, 11 May 2022
Dear Dr Hergarten
Firstly, thank you very much for taking the time to read the submitted paper, and thank you for your comments. They will certainly help us to improve the article.
We undoubtedly agree that factors such as relief, geology, slope orientation, etc. condition the occurrence of landslides. It is also widely known that heavy rainfall is the main trigger of landslides, and intense rains are often orographically controlled. There is extensive literature on this subject. Likewise, it is also known that certain natural climatic cycles, such as ENSO, NAO or sunspot, have a clear relationship with landslides and flood records, an aspect that has also been analysed in different publications, including the authors of the present paper. However, the really novel aspect of this work, and the main reason for the publication, is the spatial (and not only temporal) behaviour of climate cycles. As we know, they do not have a homogeneous spatial distribution. In this sense, once the rainfall cycles have been analysed and the ones most related to landslides or floods are identified, the mapping of these influences becomes a (powerful?) prevention tool.
In this work, the use of the inventory of rockfalls in the Tramuntana Range is a first step to validate the methodology here shown, as it can be applied in any region, regardless of its location and climatology. What is relevant is the temporal/spatial analysis of the natural cycles. However, it seems appropriate and a relevant advice from you to validate the climatic results with all the rockfall records, taking into account their volume; not only with the five largest ones.
In summary, we would like to stress that what is important in this work is the methodology, by applying statistical tools. The landslide inventory serves only to validate the results. It is not a work on landslides, as the methodology could be applied to any other rhydroclimatic risk.
Many thanks and kind regards
-
AC1: 'Reply on RC1', Juan Antonio Luque Espinar, 11 May 2022
-
RC2: 'Comment on nhess-2022-50', Omar F. Althuwaynee, 10 May 2022
The article entitled “Spatio-temporal analysis of the role of climate cycles on landslide activity: the case of Majorca (Spain)” discusses the possibility of using spatio-temporal analysis of climate cycles climate cycle and its role as landslide occurrence possible reference phenomenon. And concluded that the 5 largest landslides of the series are very well located in the areas of highest statistical weight.
1- However, the problem statement that need to be solved here is well know hypothesis. Also, the research carried in this article dint come up with interesting facts. More or less the outcome was highly expected.
2- I wish to see a clear sentence that stress on the validity of using the applied method, or compare to current methods that might need to be improved using the presented approach.
3- Some prosaic sentences were mentioned: “the island suffered persistent and abnormal precipitations”, “In this work, ordinary kriging (OK) has been chosen as it is better adapted to the problem under study.”
4- fig. 4 with table 1. you mentioned the wet years, like 2003 but the amount of landslide compare to following years (average to wet) clearly saying different story “dry to wet”
5- fig. 5 However, when there are a significant number of stations where the cycle has not been detected, a dichotomous transformation has been chosen, i.e. observed (1), not observed (0), as in the case of ENSO (6.4y) and Sunspot (11.2y)..!!!!!!
6- What about the false negative and false positive of landslide occurrence within the climate cycle??
I prefer to stop here, as i tried to get along with article searching for significant findings. But, I think due to the poor structure of the manuscript and presented figures, the idea behind it was difficult to follow.
Unfortunately, my decision is against the publication and I recommend to reject the article. In meantime, I ask the authors to have more time to develop the current approach by consider more robust problem statement and methodology.
All the best,
Omar AlThuwaynee
-
AC2: 'Reply on RC2', Juan Antonio Luque Espinar, 23 Jun 2022
The article entitled “Spatio-temporal analysis of the role of climate cycles on landslide activity: the case of Majorca (Spain)” discusses the possibility of using spatio-temporal analysis of climate cycles climate cycle and its role as landslide occurrence possible reference phenomenon. And concluded that the 5 largest landslides of the series are very well located in the areas of highest statistical weight.
We thank all the interesting and useful comments provided by you. Please, find attached a point-by-point response, including all relevant changes that we will made in the manuscript to solve your considerations.
- However, the problem statement that need to be solved here is well known hypothesis. In addition, the research carried in this article dint come up with interesting facts. More or less the outcome was highly expected.
The originality of this work lies in the methodology of the spatial and temporal analysis of the climatic cycles on the island of Mallorca. In this sense, the words "ENSO, NAO, landslides" were only included in three publications, one of them is the preprint of this review, and there are only 4 papers with the words "ENSO, NAO, Kriging" (Web of Science). The authors would like to highlight that the hypothesis put forward in this article is not stated in the previous bibliography. In the present work, authors attempt to solve the lack of methods of spatial analysis of climatic cycles, which would be of great interest to study the occurrence of geohazards (i.e. landslides and floods). The analysis has allowed verifying that the cycles behave as spatial continuous variables, but with a percentage of unstructured behavior due to the uneven influence of these cycles in the rain gauges. The island of Mallorca presents a special topography with a mountainous area in the northern part of the island (Tramuntana range), the rest being of lower altitude and gentle relief. This method provides similar assessments in areas with a complex topography, where the results of the rainfall spatial/temporal distribution will also be more uncertain and complex to determine.
- I wish to see a clear sentence that stress on the validity of using the applied method, or compare to current methods that might need to be improved using the presented approach.
Undoubtedly we will rewrite this part of the article. The literature shows that the climatic analysis carried out has not been proposed so far. This approach shows two well-known robust methods: spectral analysis to estimate the relevance of each natural climatic cycle and geostatistics to determine their spatial behaviour. We will improve the text to emphasise these aspects.
- Some prosaic sentences were mentioned: “the island suffered persistent and abnormal precipitations”, “In this work, ordinary kriging (OK) has been chosen as it is better adapted to the problem under study.”
We will rewrite the text to clarify both sentences. The first one can be deleted, as certainly, it is “prosaic”. In the second one, we wanted to emphasise the analysis of spatial variables carried out by the Theory of Regionalized Variables. This theory was formulated by Matheron (1963, 1965) and applied in different fields of science by authors such as Chilès, Delhomme, Journel or Goovaerts.
- 4 with table 1. you mentioned the wet years, like 2003 but the amount of landslide compare to following years (average to wet) clearly saying different story “dry to wet”
We appreciate the suggestion of the reviewer. We think that Figure 4 and Table 1 are not comparable. Figure 4 only reports the statistical confidence and number of cycles detected, providing different information from Table 1.
You are right on this observation. From 2001 (Mateos, 2001), the inventory is more complete, as it includes all the events detected by Civil Protection authorities and the Road Maintenance Service of Mallorca. Most of them are of small volume. This is one of the reasons we decided to use only the 5 major landslides. Nevertheless, we will carry out a new analysis including the complete rockfall inventory as well as the flooding records on the island. A robust and complete validation will be done.
- 5 However, when there are a significant number of stations where the cycle has not been detected, a dichotomous transformation has been chosen, i.e. observed (1), not observed (0), as in the case of ENSO (6.4y) and Sunspot (11.2y).
Indicator kriging (IK) is a non-parametric estimation method commonly used to solve problems related to the risk of presence of a given dichotomous value (presence-absence). For example, it is very common in pollution problems.
In this case, the method allow estimating the risk of presence of a climatic cycle because it does not reach 90% statistical confidence in some rainfall gauge.
- What about the false negative and false positive of landslide occurrence within the climate cycle??
Thanks for your comment. We will improve the statistical analysis adding the complete landslide database, and also floods recorded in the study area. To quantify the spatial matching (or mismatching) between spatial distribution of climate cycles and hazards density maps.
I prefer to stop here, as i tried to get along with article searching for significant findings. But, I think due to the poor structure of the manuscript and presented figures, the idea behind it was difficult to follow. Unfortunately, my decision is against the publication and I recommend to reject the article. In meantime, I ask the authors to have more time to develop the current approach by consider more robust problem statement and methodology.
We would like to express our sincere gratitude for your in-depth revision. We would like to improve the article considerably and to have another chance. We hope these improvements will lead you to reconsider your decision. Many thanks and kind regards.
-
AC2: 'Reply on RC2', Juan Antonio Luque Espinar, 23 Jun 2022
-
RC3: 'Comment on nhess-2022-50', Matthias Schlögl, 10 May 2022
The manuscript presents a spatio-temporal analysis of climate cycles on the island of Majorca and seeks to explore correlations with evidence of historic landslides, i.e. a landslide inventory featuring 423 events.
Albeit the general idea (especially the topic of identifying main climatic cycles) is interesting, I am afraid that there are several major issues with this manuscript in its present form:
- I am missing a clear red thread throughout the article, thus rendering it slightly difficult to read. It is not ultimately clear to me why these particular methods were chosen to tackle the problem under consideration and what the main findings are that the authors want to convey.
- Moreover, findings are not put into context, and a discussion section is missing completely.
- Sections are not clearly separated. Some parts of section 3.4 on rainfall series data seem more like results than "Materials and Methods" to me.
- The joint consideration of different process categories, which are simply summarized as "landslides", would need better justification. After all, these different processes are most likely characterized by different trigger conditions (e.g. rockfall - earth slides - debris flows). The authors could explore results per process category (at least for those categories where enough events are available).
- Some sections are not written in a balanced way. Especially section 3 does not really present the methods applied in a stringent and reproducible manner. For instance, Section 3.3 provides a rather general introduction of geostatistics, but ends quite abruptly with the last paragraph somehow falling short of actually explaining what is done here and why. Why was OK chosen specifically? I assume that there is a trend due to the topography? Also, it is quite common to perform cross-validation on kriging results, but it is unclear at this point if any validation was performed and if yes, how?
- I am under the impression that some important details were omitted or are at least hidden in the manuscript. For instance, satistical confidence values estimated at each rainfall station have been reclassified from 0 (not detected) to 4 (more than 99% statistical confidence). How were these thresholds chosen, and where are they listed? I assume this information is hidden in l. 136?
- I think that the connex between the climate cycles and landslide events should be motivated in a better way. Currently, I fail to see this connection. Maybe a more detailed exploratory analysis of the landslide inventory against the identified climate cycles might provide interesting insights?
- On a more general note, all recorded landslides seem to have occurred in the north-western part, i.e. in the Serra de Tramuntana. This is statet in l.68, and indicated by the white dots in the maps (Figs. 5, 7). I do not really understand why a geostatistical estimation of the whole island does relate to landslides that only occurred in a quite specific sub-region near to a mountain range?
- Only some large landslides are prominently mentioned. Consequently, the main connection seems to be event magnitude, not event frequency. An exploratory analysis of event frequency could provide interesting insights as well.
Overall, I do not think that the manuscript is suitable for publication in its current form due to these limitations. However, I do encourage the authors to work out the proposed connection between climate cycles and landslides in a more elaborate manner.
Best regards,
Matthias Schlögl
-
AC3: 'Reply on RC3', Juan Antonio Luque Espinar, 23 Jun 2022
The manuscript presents a spatio-temporal analysis of climate cycles on the island of Majorca and seeks to explore correlations with evidence of historic landslides, i.e. a landslide inventory featuring 423 events.
Albeit the general idea (especially the topic of identifying main climatic cycles) is interesting, I am afraid that there are several major issues with this manuscript in its present form:
We would like to express our sincere gratitude for your in-depth revision.
- I am missing a clear red thread throughout the article, thus rendering it slightly difficult to read. It is not ultimately clear to me why these particular methods were chosen to tackle the problem under consideration and what the main findings are that the authors want to convey.
Different authors linked landslides/floods to meteorological processes such as ENSO and NAO. However, there is a lack in the literature regarding the relationship between hydrometeorological disasters and the spatio-temporal behavior of the influence of natural climatic cycles. We will proceed to a better drafting and restructuring of the article to make it easier to understand.
- Moreover, findings are not put into context, and a discussion section is missing completely.
We will rewrite and restructure this section and a “Discussion section” will be included.
- Sections are not clearly separated. Some parts of section 3.4 on rainfall series data seem more like results than "Materials and Methods" to me.
We will rewrite and restructure the entire manuscript for a better understanding.
- The joint consideration of different process categories, which are simply summarized as "landslides", would need better justification. After all, these different processes are most likely characterized by different trigger conditions (e.g. rockfall - earth slides - debris flows). The authors could explore results per process category (at least for those categories where enough events are available).
We use the term “landslides” as generic. The landslide inventory reveals that most of them are rockfalls. Earth slides and debris flows are negligible in the Tramuntana range. Nevertheless, we will carry out a new analysis including the complete rockfall inventory as well as the flooding records on the island. A robust and complete validation will be done.
- Some sections are not written in a balanced way. Especially section 3 does not really present the methods applied in a stringent and reproducible manner. For instance, Section 3.3 provides a rather general introduction of geostatistics, but ends quite abruptly with the last paragraph somehow falling short of actually explaining what is done here and why. Why was OK chosen specifically? I assume that there is a trend due to the topography? Also, it is quite common to perform cross-validation on kriging results, but it is unclear at this point if any validation was performed and if yes, how?
A balanced approach has been sought between the description of the methods, and the extension of the article. Section 3.3 describes the main concepts and the key equations. In addition, some new references have been added to support the concepts used. Finally, it will be revised to complete the description given in the previous version of the article.
The OK was used because provides the least restrictive assumptions and do not also exhibit drift in calculation of experimental variogram. In some cases, points of the experimental variogram, separated by large distances, are distributed around the variance value. This is due to the differences among the precipitation values. The validation of the results has been carried out by matching the values of the estimation variance obtained for each of fitted variogram models. Many author accepts this validation process.
- I am under the impression that some important details were omitted or are at least hidden in the manuscript. For instance, statistical confidence values estimated at each rainfall station have been reclassified from 0 (not detected) to 4 (more than 99% statistical confidence). How were these thresholds chosen, and where are they listed? I assume this information is hidden in l. 136?
We will considerably improve this section.
The methodology is described in the article, although we have rewritten this section to clarify it. When the power spectrum is calculated (Pardo-Igúzquiza and Rodrı́guez-Tovar, 2004), four statistical confidence values <90%, 90%, 95% and 99% are obtained. Value 0 is assigned when a climatic cycle is not estimated (not detected). Therefore, five categories from 0 to 4 are obtained, which will be used to analyze the spatial behavior of each climate cycle (variogram) and subsequently estimate it (kriging).
- I think that the connex between the climate cycles and landslide events should be motivated in a better way. Currently, I fail to see this connection. Maybe a more detailed exploratory analysis of the landslide inventory against the identified climate cycles might provide interesting insights?
We understand your advice. We will carry out a new analysis including the complete rockfall inventory as well as the flooding records on the island. A robust and complete validation/connexion will be done.
- On a more general note, all recorded landslides seem to have occurred in the north-western part, i.e. in the Serra de Tramuntana. This is statet in l.68, and indicated by the white dots in the maps (Figs. 5, 7). I do not really understand why a geostatistical estimation of the whole island does relate to landslides that only occurred in a quite specific sub-region near to a mountain range?
We have only a rockfall inventory in the Tramuntana range. The rest of the island has a gentle relief and slope movements are not a serious problem. In this sense, to complete the entire island, we will include the flooding records.
- Only some large landslides are prominently mentioned. Consequently, the main connection seems to be event magnitude, not event frequency. An exploratory analysis of event frequency could provide interesting insights as well.
We agree. All the rockfall inventory will be included to take into account not only magnitude but also frequency. In this sense, we would like note that the rockfall inventory is more complete since 2001 (Mateos, 2001). Most of the events before this date are lost, except for those of the largest magnitude that left a “footprint” in the media. That is the main reason we used the 5 largest ones.
Overall, I do not think that the manuscript is suitable for publication in its current form due to these limitations. However, I do encourage the authors to work out the proposed connection between climate cycles and landslides in a more elaborate manner.
We will improve the article and in depth modifications will be done taking into account the reviewer's comments.
Many thanks and kind regards,
Juan Antonio Luque-Espinar et al.
Juan Antonio Luque-Espinar et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 279 | 76 | 16 | 371 | 6 | 7 |
- HTML: 279
- PDF: 76
- XML: 16
- Total: 371
- BibTeX: 6
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1