A fracturing state map for Chiavenna Valley: preliminary geostatistical analysis and optimal spatial sampling design

The interplay between tectonic processes and post-glacial debuttressing reflects in the actual fracturing state of the Chiavenna Valley (SO), which is the focus of this study. Fractures in rock masses condition water flow and stability, therefore knowledge of their pattern and properties is crucial to understand and predict the evolution of these processes. The overarching goal of the study is the creation of a fracturing map, at the slope scale, of the study area. Specific objectives are: i) to explore the interaction between meso-scale rock mass properties and understand the geomechanical and hydrogeological relationships between them; ii) to evaluate the available rock mass properties database and highlight its weakness according to the goal of the study; iii) to define a strategy to improve the database; and quantify its effectiveness. For each specific objective, the methods applied and the main results obtained are presented below. I) The geomechanical database available for the study area includes 132 survey points, with different levels of information regarding both primary variables (e.g. Jv, JRC, Aperture, Joint Compressive Strength, Weathering grade) and inferred variables (e.g. RMR, GSI and permeability). Correlation matrices between primary properties were calculated and a Principal Component Analysis performed. The Joint Volumetric Count (Jv), the index that summarizes better the general fracturing state of the rock mass, did not show any strong correlation with other primary variables. II) A tentative map of Jv was derived by means of Ordinary Kriging (following the approach of Ferrari et. al, 2014) following standard verifications of data non-clustering, normality, and casualty. The best result was obtained with a Gaussian model with a Lag of 1000 m and an anisotropy in the direction 35°-215°. However, the cross validation showed a high prediction standard error. This is mainly attributable to the distribution of the geomechanical field surveys locations, which are highly concentrated along roads (i.e. accessible areas). The result suggested the necessity of additional, spatially distributed field surveys. III) Thirty new sampling locations were selected with the Spatial Simulated Annealing model-based sampling algorithm (SSA) implemented in the R platform (Brus, 2018). The algorithm was specifically modified and adapted for the present study. The additional 30 geomechanical field surveys (both classical and with UAV) in the selected locations are now ongoing. The evaluation of the resulting variogram and Jv map improvement will follow, eventually exploring different geostatistical techniques.