the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
INSYDE-content: a synthetic, multi-variable flood damage model to household contents
Abstract. This paper introduces INSYDE-content, a novel, probabilistic multi-variable synthetic flood damage model designed to analyze physical damage to household contents on a component-by-component basis. The model addresses a critical gap in current modeling tools, which often overlook the significance of household contents in overall damage assessments. Developed through an expert-based approach and grounded in the scientific and technical literature, INSYDE-content leverages desk-based data to characterize model features, including uncertainty treatment arising from incomplete input data. A sensitivity analysis and a benchmark test against observed losses demonstrate the robust performance of the model and highlight the contribution of different features to damage mechanisms affecting house contents. While in this study INSYDE-content is tailored for illustrative purposes to the hazard, vulnerability and exposure characteristics of Northern Italy, the model is highly adaptable, allowing for its application to different regional contexts through appropriate customization.
- Preprint
(1010 KB) - Metadata XML
-
Supplement
(2602 KB) - BibTeX
- EndNote
Status: open (until 23 Jun 2025)
-
RC1: 'Comment on egusphere-2025-1413', Anonymous Referee #1, 03 Jun 2025
reply
This paper provides a detailed description of the development of an expert based flood content damage model called INSYDE. It seems to be a follow up paper on the structure damage version of INSYDE, a model that seems to have found quite widespread use in the literature. The paper is well written and describes the development process well. The methods are solid but not very innovative and have been around in the grey literature for a long time (e.g. US Army Corps of Engineers). This paper goes in quite some detail describing the methods and adds much needed validation and is therefore definitely a useful addition to the scientific literature. That being said I have concerns about the validation results and more importantly the analysis of the results.
Figure 4 shows that for detached and semi-detached houses the variation in observed damages is much larger than the variation in predicted damages. My first impression is that the model always roughly predicts the same damage regardless of the circumstances (the blue dots are a nearly horizontal line). I think it may not be so bad because the log-log scale masks some of the variation. However, more information is required so readers can actually tell the model performance. For example, I currently cannot see if the variation in observed values is just based on some large outliers or whether there is some more fundamental problem whereby the observed losses have much more variation than the modelled losses. Also is there even any correlation between modelled and observed losses? I understand that there is unexplained uncertainty in the model predictions as indicated by the uncertainty ranges in figure 4. However, if the model typically predicts more or less the same mean how do I know such a complicated model adds any value compared to a simple mean value as prediction?
Also very common error metrics are missing such as Mean Absolute Error, correlation coefficient or R2, so it's nearly impossible to assess how the model is doing from the information presented in the paper. Not all these metrics are needed but at least more information. Table 4 only gives an aggregated comparison, so basically gives a bias value. In one region there seems to be some bias but the authors do not really explain where this bias might be coming from. Lastly, I would expect an in depth analysis and discussion of the model performance in the paper based on the validation. That analysis is missing, making the validation not very useful in its current form.
Some of the input variables for the model validation seem sampled whereas others seem observed and the current text is very unclear about what is sampled and what is observed. This makes it even more difficult to interpreted the validation results.
The word “to” in the title doesn’t read well, maybe you can replace it with “for”? Or another solution..
Citation: https://doi.org/10.5194/egusphere-2025-1413-RC1
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
125 | 28 | 6 | 159 | 13 | 2 | 3 |
- HTML: 125
- PDF: 28
- XML: 6
- Total: 159
- Supplement: 13
- BibTeX: 2
- EndNote: 3
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1