Valley fever (coccidioidomycosis) is a disease endemic to arid regions within the Western Hemisphere, and is caused by a soil-dwelling fungus, Coccidioides immitis. Incidence data for Pima County, reported to the Arizona Department of Health Services as new cases of valley fever, were used to conduct exploratory analyses and develop monthly multivariate models of relationships between valley fever incidence and climate conditions and variability in Pima County, Arizona, USA. Bivariate and compositing analyses conducted during the exploratory portion of the study revealed that antecedent temperature and precipitation in different seasons are important predictors of incidence. These results were used in the selection of candidate variables for multivariate predictive modeling, which was designed to predict deviation from mean incidence on the basis of past, current, and forecast climate conditions. The models were specified using a backward stepwise procedure, and were most sensitive to key predictor variables in the winter season and variables that were time-lagged 1 year or more prior to the month being predicted. Model accuracy was generally moderate ( r(2) values for the monthly models, tested on independent data, ranged from 0.15 to 0.50), and months with high incidence can be predicted more accurately than months with low incidence.