Analyzing Projected Changes to the Seasonal Cycle and Daily Extremes Using the STAR Framework

We use the Seasonal Trends and Analysis of Residuals (STAR) technique to analyze how the seasonal cycle and distribution of daily near-surface temperature is projected to change under higher (RCP8.5) simulations of future climate. The STAR methodology (Hayhoe, Scott-Fleming & Stoner, in prep.) decomposes time series data in the frequency-domain into a long-term annual trend, a smooth seasonal cycle which slowly changes in amplitude and phase over time, and a distribution of daily residuals that characterize the model’s changing variability (or “weather”) at daily to sub-seasonal time-scales. This enables us to qualitatively and separately characterize projected changes in both the seasonal cycle and daily extremes.

Consistent with evidence from satellite-derived observed trends and detection & attribution studies (e.g. Santer et al., 2018), we find a zonal-mean pattern of mid-latitude amplification and polar dampening of the seasonal cycle, and an increase in the occurrence of daily maximum temperature extremes over most land regions. This technique is one module of the STAR framework which also includes a novel empirical-statistical downscaling component (STAR-ESDM) that downscales each of these components individually.