Abstract. Operative geologists who are involved in emergency management have often to deal with the consequences of assuming critical and strongly impacting decisions in uncertain conditions.

Geohazards induced by active landslides are one of the civil protection situations requiring such decisions.

Nowadays, the monitoring of active landslides is almost always supported by numerical early warning systems, based on instrumental geotechnical and topographic networks. These networks provide numerical early warning thresholds, which are set up in order to activate alert conditions at various levels of criticality in an objective way.

Despite these progresses the issue related to the possibility to dispatch false alerts has not yet effectively solved and that’s the reason why the critical stages of the decisional processes are frequently relying not only on quantitative thresholds but also on the subjective experience of the emergency managers.

Therefore it is not so uncommon to read landslide-monitoring procedures that combine the quantitative information provided by the monitoring systems with the qualitative decisional elements coming from their professional experience in order to assume the most correct decision.

It's therefore evident that such an approach weakens the objectiveness provided by instrumental monitoring systems but, at the same time, collecting geological empirical and qualitative data can strengthen an hypothesis like the one that an active landslide could finally collapse.

Bayesian methods are frequently used in clinical decision making, another field of the human activity where critical decisions have to be made in a short time, combining objective values such as those provided by medical tests with diagnostic qualitative markers.

Based on the methods of clinical diagnosis, the has author has elaborated a reliable and objective Bayesian Decision Support System (or DSS), developed to support the decision makers in assuming the most correct decisions based on all the elements, both quantitative and qualitative, that are available at a certain step of the decision process.

Thanks to the Bayesian approach, the DSS allows also to assess the predictivity of any single decisional step, which is the probability that a monitored landslide actually collapses when particular diagnostic evidences are detected, either instrumental or observational.

Hence the decision makers who are able to issue a civil protection alert when a given degree of confidence about the chance that a monitored landslide will collapse is reached. The degree of confidence associated to the civil protection alert can be declared in the alert bulletin (e.g.: 80 % or 93 %). The decisional process can be tracked and replied by everyone in complete transparency.

It's therefore evident that such a DSS allows the civil protection authorities to increase the reliability of the alerts, reducing at the same time the so-called “cry wolf” effect and the discomfort related to evacuations and to other civil protection measures. As a matter of fact, the decisional process becomes clearer and the people’s trust in the civil protection systems is being strengthened by a more transparent emergency communication.

The DSS here described is an evolution and a statistical improvement of the method adopted in 2013 and 2014 during the emergency of the Mont the la Saxe landslide, and is now being successfully applied to two other hazardous situations in the Aosta Valley Alps: the Brenva Site (Mont Blanc Massif) and the Berlachu site in the municipality of Lillianes (Lower Lys Valley).