The presence of a volcanic lake increases the probability of certain types of volcanic hazard. So far, this topic has only been approached in a purely deterministic way, e.g.: phreatic-phreatomagmatic eruption behavior, dynamics of Nyos-type gas bursts and gas accumulation and dispersion, models on lahars and their triggers, dispersion of acidic brines upon lake seepage, gas hazard from degassing lakes, etc. BET_UNREST, the Bayesian Event Tree structure to better assess the state of volcanic unrest, is an advanced code based on the previous BET_EF, the Bayesian Event Tree for Eruption Forecasting. BET_UNREST represents a flexible tool to provide probability distributions of any specific event, linked to volcanic unrest, which we are interested in (magmatic or non, eruptive or non), by merging all the relevant available information such as theoretical models, a priori beliefs, monitoring measures, and any kind of past data. The BET_UNREST structure permits to introduce such "non-magmatic lake scenarios", given an appropriate selection of "monitoring parameters" at each node of the event tree. Exactly these non-magmatic scenarios are those helpful to track volcanic unrest for volcanoes hosting active magmatic-hydrothermal systems and volcanic lakes. An important paradigm is that, for volcanic lakes, hazard does not necessarily has to be induced by an increase in magmatic activity, as often is the case for purely magmatic systems and erupting volcanoes.
At Node 1, we need to evaluate the probability of a "lake system" being in state of unrest or not; at Node 2, given unrest, if it is in a state of magmatic unrest or not; at Node 3 both the magmatic unrest and non-magmatic unrest scenario can lead to an eruption or not; at Node 4 the hazardous feature should be specified to better assess the hazard, as also non-magmatic and non-eruption unrest can pose volcanic hazard.
Scanning some lake situations: (1) Nyos-type lakes are possibly hazardous without renewal of magmatic input; monitoring parameters should reflect CO2 input fluxes, thresholds and eventually the rate of artificial degassing, and cloud dispersion models (e.g., Lake Nyos and Monoun, Albano Lake, Lake Kivu), (2) phreatic eruptions at highly active crater lakes (e.g., Poás, Yugama, Aso, White Island) are often anticipated by long-term heating and sometimes by short-term cooling episodes, by floating sulphur spherules or by changes in the evaporative regime, (3) phreatic eruption activity itself can become precursory for later phreatomagmatic eruptions, due to pressure drop in the system (e.g., Ruapehu), (4) prolonged seepage of acidic lake brines and hydrothermal alteration can mechanically weaken the volcano flank, eventually leading the flank collapses (e.g., Rincón de la Vieja, Irazú), (5) lake throw out or overflow, or simply anomalous rain events can trigger lahars on the (weakened) volcano flank (e.g., Ruapehu), (6) persistent evaporative degassing from a lake can harm crops, cattle and human activity on surrounding (farm) lands, and can become more intense when the lake eventually disappears (e.g., Poás, Aso, Copahue), (7) dispersion of acidic brines originating from lakes into the hydrologic network can threaten the health of surrounding populations (e.g., Kawah Ijen).
This deterministic knowledge should now be translated properly into adequate monitoring parameters with specific thresholds at the various Nodes within the BET_UNREST structure, in order to provide probability density functions (PDFs) for specific hazardous events, for each specific lake. The numerical outcome of the application of BET_UNREST will provide (1) a guideline for future deterministic research in search for the "missing link", (2) a less subjective, less "emotion or philosophy based" frame for volcanologists deciding to descend active craters or not, increasing the operator's safety, (3) a structured, science-based bridge between volcanologists, the population and decision-maker (e.g., civil protection), during periods of unrest, (4) a tool to organize and protect future land-use, tourism and urbanization near a lake-hosting volcano, and, last but not least, (5) a method to quantify probabilities of possible occurrences of lake-related hazards, useful in volcanic surveillance.

1. crater lakes
2. Probabilistic hazard assessment