Towards a quantified and global landslide hazard assessment for New-Caledonia (South Pacific) in a regulatory mapping context

Requiring spatial and temporal quantified information on landslide hazard over a large area is a prerequisite to forecast them. However, in many cases, the quantification remains partial, because of a lack of information on the phenomena, on predisposing and triggering factors or because the scientific approaches used in research domain are complex to apply in a regulatory framework. Thus, in this context, for many sites and end-users, the documents produced by empirical methods are used, without quantification of hazards.

In 2019, a collaboration between the DIMENC Geological Survey Service of New-Caledonia (South-Pacific) and the BRGM planed the development of a global methodology of landslide hazard assessment at the 1:25,000 scale of work according to the recommendations of the JTC-1. Indeed, landslide hazard in New Caledonia is insufficiently assessed and few taken into account in land-use planning. However, this large mountainous island is regularly affected by different type of instabilities (i.e. rock-falls; rock-slides; slides; debris-flows) due to intense rainfalls. The consequences can be material and human, as in 2016 for the municipality of Houaïlou, where debris-flows occurred, inducing 5 deaths, 3 missing persons, 8 injuries along with large material damages. Few heuristic landslide hazard maps based on expert opinion are available, but the methodology is not homogeneous and harmonized. Therefore, even if these maps constitute a solid base of knowledge, their valorization for land use planning remains difficult.

To overcome these shortcomings, the methodology chosen is quantitative, taking into account the susceptibility of the territory (i.e. spatial probability of phenomena occurrence with discrimination of initiation and run-out), the temporal probability of occurrence (i.e. from diachronic analyses) and the phenomena intensity (i.e. through the considered velocity of runout and the potential of induced damages). The methodology is declined by type of phenomena and is based on a comprehensive inventory. Six main steps are defined with:

• An inventory of the events by visual remote sensing and field observations;
• Discriminated mapping of bedrock and surficial formations (i.e. regolith: weathered formations and gravitational deposits);
• Computation of each landslide initiation susceptibility by a bivariate method;
• Integration of the temporal occurrence probability;
• Computation of the phenomena runout by a numerical approach taking into account the reach angle;
• Integration of the intensity of the phenomena according to the estimated volumes and/or velocity to quantify landslide hazard.

The classes of spatial and temporal probabilities are based on the JTC-1 agreement and allow obtaining quantified hazard maps. The validation of the results is performed by a field validation, by phenomena not used for the computations, and by statistical tests. The method is tested in the municipality of Mont-Dore (643 km²), which was heavily impacted in 1988 by cyclone 'Anne'. Beyond the fact that the methodology will be applied throughout the territory in an operational framework and will allow the adaptation of local planning, the project allows the improvement of:

• Knowledge of the different kind of landslides in a volcano-sedimentary and metamorphic context strongly weathered;
• Knowledge of the regolith, which newly integrated for this type of analysis for the island’s municipalities.