On the occurrence of rainstorm damage based on home insurance and weather data

Rainstorm damage caused by the malfunction of urban drainage systems and water intrusion due to defects in the building envelope can be considerable. Little research on this topic focused on the collection of damage data, the understanding of damage mechanisms and the deepening of data analysis methods. In this paper, the relative contribution of different failure mechanisms to the occurrence of rainstorm damage is investigated, as well as the extent to which these mechanisms relate to weather variables. For a case study in Rotterdam, the Netherlands, a property level home insurance database of around 3100 water-related damage claims was analysed. The records include comprehensive transcripts of communication between insurer, insured and damage assessment experts, which allowed claims to be classified according to their actual damage cause. The results show that roof and wall leakage is the most frequent failure mechanism causing precipitation-related claims, followed by blocked roof gutters, melting snow and sewer flooding. Claims related to sewer flooding were less present in the data, but are associated with significantly larger claim sizes than claims in the majority class, i.e. roof and wall leakages. Rare events logistic regression analysis revealed that maximum rainfall intensity and rainfall volume are significant predictors for the occurrence probability of precipitation-related claims. Moreover, it was found that claims associated with rainfall intensities smaller than 7–8 mm in a 60-min window are mainly related to failure processes in the private domain, such as roof and wall leakages. For rainfall events that exceed the 7–8 mm h threshold, the failure of systems in the public domain, such as sewer systems, start to contribute considerably to the overall occurrence probability of claims. The communication transcripts, however, lacked information to be conclusive about to which extent sewer-related claims were caused by overloading of sewer systems or failure of system components.


Interactive Comment
Full Screen / Esc

Printer-friendly Version
Interactive Discussion Discussion Paper (section 2.2). To make clear that chapter 2 also covers raw data description we can rename the section 2 heading to "Data and methods". RC2: Conclusion/Recommendations: The authors have related their results to earlier research. They also show how the analysis of insurance data benefit the characterization of damages and the relation with different rainfall events. The recommendations are related to the statistical treatment of the data and not connected with the potential added value of the data for pluvial flood risk management.

AC2:
We would like to discuss the added value for pluvial flood risk management by adding the following paragraph after line 8 on page 5299 (Discussion section): "Results have implications for pluvial flood risk management. The return period of design storms as currently being used to design sewer systems in the Netherlands (see also section 3.2) is largely based on political consensus. Potentially the results presented here can be used to obtain an objective design criterion based on risk assessment. Furthermore, this study provides insights into contributions of urban drainage systems to flood damage at city level. Results will support urban water managers in the evaluation of urban drainage system capacity and decisions about the need for and prioritisation of investment to increase drainage capacity. Further research is needed to explain why damage related drainage capacity occurs below the level of design capacity; this will help water managers to focus efforts on ensuring that their systems reach design capacity." RC3: P5289 L7-13: Rainfall events are presented in different way. In order to have a good comparison, I suggest to present the return period and/or the intensity of the events for both of the cases (if these data are available).

AC3:
Although we agree with reviewer that there is some difficulty in comparing the two examples, we prefer to stick to the statements in the original works. For a good comparison (which was not the primary goal of this introduction), many other aspects have to be taken into account too, such as the spatial extent of the rainfall cell (which is not expressed in the return period) and information about insurance market penetration and insurance coverage for both cases.
RC4: P5295 L5-10: A validation process is mentioned as a reason to discard data corre-sponding to the 3 rain events; however, the validation is not mentioned in previous sections as a part of the modelling process.
AC4: Good point. We have done some basic data checks, which we have not mentioned yet in the paper. Therefore, we would like to add the following at the end of section 2.2 (Insurance data): "The database has been checked on missing and incorrect values, such as duplicated records, inconsistencies in date formats and claim coding.". RC5: P5297 L4-11: The paragraph fits better in section 4 (Discussion). In fact, similar and more explained sentence is written in L25 and following in the same page; therefore, I would suggest to remove L4-11. AC5: There is indeed some repetition here, which can be avoided. We can remove line 6-8 on page 5297 without loosing any information. However, we like to keep the sentences "For rainfall events... occurrence probability." and "Similar conclusions... respectively.", because they describe results that are derived from Fig. 4, which are discussed in this section for the first time.
C2406 RC6: P5299 L26: It is mentioned that the hypothesis of different processes could not be tested using the available database, because there is no explanation of what characteristics the database should have to prove the hypothesis.

AC6:
To explain why the hypothesis could not be tested, we would like to slightly rewrite the paragraph (starting at line 24 on page 5299): "Claims associated with rainfall intensities of 0-5 mm h −1 in Fig. 4 (60-minute window) are possibly generated by a different process than the claims related to rainfall intensities larger than 5 mm h −1 . It maybe the case that more specific damage processes can be distinguished within the existing cause classes. For example, the class "roof leakages" may contain two processes; one related to the presence of latent leaks that are first observed when it it raining and another one related to the exceedance of the "hydraulic capacity" of roofs. The hypothesis could not be tested based on the present database, because it lacked information to distinguish between the sub-processes." C2407