This paper describes the relationship between surface conditions (temperature and humidity) and subsequent rainfall. The focus is on convective storms that are forced and maintained locally due to conditional instability in the vertical distribution of atmospheric temperature. These storms are described using two probabilistic measures: (1) the probability of occurrence of storms given surface conditions and (2) the average storm rainfall. The surface conditions are described by a single variable: surface wet-bulb temperature. The proposed theoretical relationships are tested using an hourly data set on rainfall and wet-bulb temperature from the Amazon region. These observations confirm that both measures increase linearly with wet-bulb temperature. However, for the occurrence of any storm the wet-bulb temperature has to exceed a threshold of about 22 ¿C. The sensitivity of the frequency of storms to changes in the climatology of surface wet-bulb temperature is larger than the corresponding sensitivity of the average storm rainfall. These general concepts are applied in discussing the potential impact of changes in land cover on rainfall patterns using two specific examples: deforestation in the Amazon region and development of irrigation projects in the Columbia River basin. ¿ American Geophysical Union 1996