Résumé : Kawah Ijen (2386 m) is a stratovolcano located within Ijen Caldera, at the easternmost

part of Java island in Indonesia. Since 2010, the volcano has been equipped with seismometers

and several sensors (temperature and level) have been immersed in its acidic lake waters and in the acidic river seeping on the volcano flanks. While finding instruments capable of resisting to such extreme conditions (pH~0) has been challenging, the coupling of lake monitoring techniques with seismic data improves the knowledge of the volcanic-hydrothermal dynamics. Moreover, the monitoring capabilities have been considerably

enhanced supporting the decision-making of the authorities in case of emergency.

Several methods and processing techniques were used to analyze the seismic data. Much effort has been given to implement the seismic velocities (Moving Window Cross Spectral Analysis (MWCSA)) calculations. At Kawah Ijen, the frequency band that is less affected by the volcanic tremor and the seasonal fluctuations at the source ranges between 0.5-1.0 Hz. Moreover, a stack of 5 days for the current CCF gives reliable results with low errors and allows to detect fluctuations which are missed using a 10-day stack.

The background seismic activity mostly consists in low frequency events and a continuous tremor of low amplitude. Fluctuations of the lake temperature and level result from the recharge of the hydrothermal system during the rainy season. Kawah Ijen lake waters are not perfectly mixed and a shallow stratification occurs during the rainy season, because meteoric waters are less dense than the lake fluids.

Different unrest occurred during our study. Some of them strongly affected the volcanic lake, while others did only weakly. In the first category, a strong unrest commenced in October 2011 with heightened VT (Volcano Tectonic) earthquakes and low frequency events activity, which culminated mid-December 2011. This unrest was correlated with an enhanced heat and hydrothermal fluids discharge to the crater and significant variations of the relative velocities (~1%). This suggests an important build-up of stress into the system. VT earthquakes opened pathways for the fluids to ascend, by increasing the permeability of the system, which latter allowed the initiation of monochromatic tremor (MT) when the steam/gases interacted with the shallow portions of the aquifer. Our calculations evidence a higher contribution of steam in March 2012 that might explain the increase of the MT frequency when bubbles were observed at the lake surface. This period was also characterized by short-lived but strong velocity variations, related to water level

rises containing important amount of bubbles, and important heat and mass discharges

into the lake. On the contrary, the second category of unrest did only slightly affect the

lake system. This could be explained by a dryer hydrothermal system and/or locations of

the seismic sources, which were not directly linked to the lake.

While a magmatic eruption will likely be preceded by a strong seismic activity, the major challenges remain to understand why the unrest we studied did not lead to an eruption and to identify precursory signs of a phreatic eruption. Even a small phreatic eruption would be devastating for the people working everyday in the crater and the ones

who live nearby the voluminous acidic lake.