A soil moisture balance equation over large spatial regions is studied at seasonal and annual time scales for the Arkansas river basin. Interaction and feedback effects between land-surface and atmospheric moisture are studied in the parameterization for this basin. Due to the interaction between the land-surface and atmosphere at large scales, the surface hydrology of large land areas is susceptible to two distinct stable modes in the long-term probability density function: a dry and a wet state. In the soil moisture balance equation, stochastic¯uctuations lead to separate preferred statistical stable states with transitions between these stable states induced by environmental¯uctuations. On the basis of historical data, the soil moisture balance equation is calibrated for the Arkansas river basin. The transition times between the stable modes in the model are studied based on the stochastic representation of the physical processes and the calibrated model parameters. This study has implications for prediction of the transition times between stable modes or residence times, that is, the time the system spends in a given stable mode, since this would be equivalent to predicting the duration of droughts or wet conditions.