Before most volcanic eruptions, restless behaviors, such as gas emission, surface deformation, temperature changes, and earthquakes, can be observed. It is essential to quantify the subsurface processes leading to the volcanic eruptions. Therefore, we will be able to combine monitoring data and numerical models to forecast eruptions like weather forecasts. In this talk, Yan will show a case study that used a thermo-mechanical model to explain the multiple precursory signals observed before the 2006 explosive eruption of Augustine volcano, Alaska. You will also learn many similarities between a volcano and an espresso maker. The models will show how to build high pressure to trigger eruptions or to create crema in espresso.