Investigating the relationship between drought and clay-shrinkage-induced subsidence damage at the town scale over France

Clay shrink-swell consists in volume changes of clayey, smectite-rich soils as a function of their soil water content. Building foundations can be affected by soil shrinkage during droughts, entailing what is called subsidence damage. This is the second costliest peril covered by the French national natural disaster compensation scheme, the losses amounting to more than 16B€ between 1989 and 2021 (CCR, 2021). As illustrated by the 2022 drought in France, these costs are likely to increase as a result of climate change and of the related amplification of annual soil moisture cycles.

In this context, we investigate the relationship between drought and subsidence damage, using the ISBA land surface model developed by the French meteorological service (Météo-France), geotechnical data from the French geological survey (BRGM) and data from a national claim database operated by the French state-owned national reinsurance company (CCR). We compute several yearly drought indices based on multi-layer soil moisture time series simulated by the ISBA model. Different configurations of the indices are considered, varying in particular the ISBA model settings, and the soil drought definition through a threshold value accounting for a given temporal frequency, for each model soil layer. We assess a large range of configurations by using the Kendall rank correlation of the indices with yearly town-scale insurance claim data from 2000 to 2018, processed using the geotechnical data. The analysis is repeated for five sets of four towns with an important damage history distributed throughout France, in contrasting climate conditions.

Highest rank correlation coefficients are obtained for soil layers deeper than 60 cm, and with temporal frequency threshold values corresponding to intense droughts. Under these conditions, the indices are able to fairly represent the occurrence of damages. The relationship between drought indices and the number of claims is non-linear. This study benefits from the latest improvements in land surface modeling and is a step forwards in climate risk modeling since the indices investigated can be considered as new predictors for subsidence damage. Climate change impact studies will be conducted in a next phase.

[References] CCR: Les Catastrophes naturelles en France, Bilan 1982-2021, 2021.