Crop-yield forecasts provide useful information to growers, marketers, government agencies and other users. Yields for several crops in California are currently forecast based on field surveys and farmer interviews, although official forecasts do not exist for many crops. Because broad-scale crop yields depend largely on the weather, measurements from existing meteorological stations have the potential to provide reliable, timely and cost-effective predictions. We developed weather-based models of statewide yields for 12 major California crops and tested their accuracy using cross-validation from 1980 to 2003. Many of the weather-based forecasts were highly accurate, as judged by the percentage of yield variation explained by the forecast, the number of yields with correctly predicted direction of yield change, or the number of yields with correctly predicted extreme yields. The most successfully modeled crop was almonds, with 81% of yield variance captured by the forecast. Predictions for most crops relied on weather measurements well before harvest time, in many cases allowing longer lead times than existing procedures.