An Analysis of CPC’s Operational 0.5-Month Lead Seasonal Outlooks

Abstract

n analysis and verification of 15 years of Climate Prediction Center (CPC) operational seasonal surface temperature and precipitation climate outlooks over the United States is presented for the shortest and most commonly used lead time of 0.5 months. The analysis is intended to inform users of the characteristics and skill of the outlooks, and inform the forecast producers of specific biases or weaknesses to help guide development of improved forecast tools and procedures. The forecast assessments include both categorical and probabilistic verification diagnostics and their seasonalities, and encompass both temporal and spatial variations in forecast skill. A reliability analysis assesses the correspondence between the forecast probabilities and their corresponding observed relative frequencies. Attribution of skill to specific physical sources is discussed. ENSO and long-term trends are shown to be the two dominant sources of seasonal forecast skill. Higher average skill is found for temperature than for precipitation, largely because temperature benefits from trends to a much greater extent than precipitation, whose skill is more exclusively ENSO based. Skill over the United States is substantially dependent on season and location. The warming trend is shown to have been reproduced, but considerably underestimated, in the forecasts. Aside from this underestimation, and slight overconfidence in precipitation forecast probabilities, a fairly good correspondence between forecast probabilities and subsequent observed relative frequencies is found. This confirms that the usually weak forecast probability anomalies, while disappointing to some users, are justified by normally modest signal-to-noise ratios.