An Empirical Prediction Model for Summer Temperatures in Washington State

An empirical model has been developed and tested for forecasting seasonal temperature, with the initial application focusing on summer mean temperatures for Washington state as a whole. This model is based on three predictors: sea surface temperatures (SSTs) off the coast of the Pacific Northwest during the month of May; forecasts of sea surface temperatures in the tropical Pacific Ocean (represented by the NINO3.4 index) for June through August; and predictions from NOAA's Climate Forecast System (CFS) model of the mean 500 hPa geopotential height (Z) for the months of June through August for an area encompassing the states of Oregon and Washington.

The empirical forecast model relates the three predictors to Washington summer temperature using a generalized additive model, which can capture complex (i.e., not just a linear fit) relationships between predictor and predictand. The model has a training period of 1949 through 1992 and a validation period of 1993 through 2017. The results indicate that the forecast model is skillful, and that much of the variability in summer temperatures can be accounted for by the values of the regional SST, NINO3.4 index, and predicted 500 hPa Z.

Our focus on forecasts of summer temperatures emerged from discussions with managers familiar with planning decisions related to wildfires and agriculture. Use of our forecast could be instructive for learning more about climate change adaptation options in the region. The actions that managers contemplate pursuing in response to a warmer-than-average forecast for the summer are likely to resemble the options in preparation for the impacts associated with an upward trend in summer temperatures.