Water is an extremely limiting resource in arid regions and wildlife managers need accurate inventories of water sources to better manage natural resources. Many of the water sources in the Sonoran Desert are tinajas, solid rock-bottom pools of varying sizes. These and other isolated and ephemeral water resources are essential for desert wildlife. We developed an approach to predict the location of unidentified ephemeral waters in the Sonoran Desert of Arizona, USA. We used Mahalanobis distance based on topographic wetness and slope to indicate groups of pixels in GIS that are the most similar in these aspects to locations of currently known waters. We tested this model in southwestern Arizona at the U.S. Air Force's Barry M. Goldwater Range - East by comparing polygons of predicted waters with random polygons. Seventy-four percent of standing surface water features found were attributed to the predicted polygons derived by our model. The model found a significantly larger water capacity in predicted polygons than in random polygons. This modeling technique could provide a new tool for researchers and land managers to better estimate potential water resources for wildlife conservation objectives in arid landscapes.