Climatological contingency analysis and seasonal forecast skill in the United States

Tercile contingency tables were computed using 75 years (1895–1969) of state‐weighted monthly temperature and precipitation data. The data were grouped into 3‐month seasons and ranked for each state into three equally probable terciles. By comparing each season with the antecedent season, the magnitude of interseasonal persistence (or reversal) which exceeded chance was measured and mapped for the United States. The patterns of interseasonal tendencies are quite varied and demonstrate that there are geographical regions with substantial persistence of climatic anomalies between seasons. There were also regions which reverse tendencies (for example, a wet spring tending to follow a dry winter). The average excess tendency over chance for all contingencies is approximately 9 per cent. Seasonal forecasts were then generated (for an independent data period of 1970–1978) using only these contingency relationships. Standard skill scores were calculated and it was determined that these climatological interseasonal tendencies could not alone produce seasonal forecasts exceeding chance.