This paper develops a model which is able to forecast exchange rate turmoil. Our starting point relies on the empirical evidence that exchange rate volatility is not constant. In fact, the modeling strategy adopted refers to the vast literature of the GARCH class of models, where the variance process is explicitly modeled. Further empirical evidence shows that it is possible to distinguish between two different regimes: "ordinary" versus "turbulence". Low exchange rate changes are associated with low volatility (ordinary regime) and high exchange rate devaluations go together with high volatility. This calls for a regime switching approach. In our model we also allow the transition probabilities to vary over time as functions of economic and financial indicators. We find that real effective exchange rate, money supply relative to reserves, stock index returns and bank stock index returns and volatility are the major indicators.