1C.5 Skillful Empirical Prediction of High-Impact Temperature Deviations

Monday, 13 January 2020: 9:45 AM
151A (Boston Convention and Exhibition Center)
Patrick T. Brown, San Jose State Univ., San Jose, CA; and M. Evans, A. Mahesh, H. Gupta, and K. caldeira

The sectors of the economy most sensitive to temperature variation include agricultural productivity, outdoor labor productivity, electricity demand, and human mortality. Much of this sensitivity is not the direct result of shifts in mean temperature but rather fluctuations in the proportion of time spent beyond particular thresholds. For example, decreases in crop yields and outdoor labor productivity, as well as increases in electricity demand and human mortality all accelerate as the frequency of daily maximum temperatures above 30°C increases. Thus, foreknowledge of deviations in the shape of the probability distribution of daily high and low temperatures should be of particular value for decision makers. Variation in the relative probabilities of temperatures are driven mostly by internal modes of variability on monthly to decadal timescales and external forcings on longer timescales but both the driving factors and temperature responses are a complex function of geography, season and time-of-day. These complexities, as well as a lack of straight-forward physical constraints in all cases, suggests that data-driven methods can serve as a valuable complement to physics-based forecasts. Here we present such an empirical methodology which forecasts deviations in the distributions of land temperatures on seasonal timescales. The method relates global-scale antecedent sea surface temperature anomalies to subsequent daily temperature distributions (both daily high and low) as a function of location and season. We amalgamate these hindcasts of temperature distributions with the geographic extent of population and agricultural activity to produce seasonal hindcasts of impact-relevant indices like human and crop exposure to daily high temperatures above 30°C. Our out-of-sample hindcasts suggest predictive skill that originates from global warming as well as from the state of internal modes of variability like the El Niño Southern Oscillation. This approach should be useful for the anticipation and mitigation of high-impact temperature deviations across several sectors of the global economy.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner