National soil class mapping for Switzerland

Soil classification is one of the two essential foundations of seismic building codes. It simplifies the representation of local site amplification in elastic response spectra by grouping subsurface structures that behave analogously with respect to soil transfer functions. In Switzerland, soil classification has been implemented within the SIA 261 building code, adapted from Eurocode 8 (EC8, 2004). Until now, several cantons have produced soil class maps according to different SIA 261 standards (2003, 2014, 2020), often relying mainly on geological descriptions. 

The need for a national initiative is underscored by the recent approval of the second-generation EC8. The updated standard expands soil categories from five (A–E) to six (A–F) and mandates a geophysical definition of classes based on two parameters: the average shear-wave velocity of the sediment cover (Vs,H) and the depth to engineering bedrock (H800). Where H800 is not known, the fundamental resonance frequency (f0) may be used. Local assignment of soil class based solely on geological or geotechnical proxies is no longer permitted, making the development of geophysically defined national maps both timely and essential.

This project was launched to produce a comprehensive seismic soil class map of Switzerland. The mapping approach will prioritize geophysical parameters, while integrating geotechnical and geological datasets to ensure complete spatial coverage. Standardized resources such as the Swisstopo “Geocover V2” geological atlas and national soil databases will be complemented with geophysical datasets including shear-wave velocity profiles, ambient noise H/V spectral ratios, and amplification functions from seismic stations. Geotechnical input such as Standard Penetration Tests and Cone Penetration Tests will further constrain subsurface properties.
The project will proceed in phases: (i) collection, harmonization, and quality control of geophysical, geotechnical, and geological datasets; (ii) development of a robust methodology to translate these datasets into soil class maps and parameter layers, explicitly accounting for uncertainties; (iii) implementation of a logic-tree framework with weighted data integration; and (iv) spatial prediction using machine-learning techniques such as random forest algorithms.

The outputs will include national maps of Vs30, Vs,H, and H800, each accompanied by uncertainty layers. These parameter maps will be combined into soil class maps at 1:25,000 scale, produced according to both the current Swiss norm (SIA 261) and the new EC8.