Seasonal Predictability of Tropical Cyclone Activity: Evaluation of Upper Limit and Unrealized Potential Using GFDL's FLOR Prediction System
Abstract
This study explores the seasonal predictability of tropical cyclone (TC) activity and pathways to improve the existing dynamic prediction systems. Using a bootstrap analysis of ensemble hindcasts by GFDL's FLOR prediction system, we show that coastal TC activity in Atlantic and Pacific basins is sensitive to uncertainties of initial conditions. In comparison, the impacts of initial conditions are much weaker for offshore TC activity, especially at lower latitudes. The sensitivity difference suggests that coastal TC activity is inherently less predictable on the seasonal scale. Nonetheless, the predictability of regional and basin-wide TC activity is likely higher than the skill that has been realized by pre-existing FLOR prediction systems. The gap can be mitigated by correcting SST biases and initializing predictions with land-atmosphere conditions. Despite model biases and unforced variability, these remedies help the simulation of the large-scale environment in the tropics and/or the extratropics. Consistently, the skill gains of seasonal TC prediction are statistically significant. The findings highlight the potential of improving seasonal TC prediction and will help guide the design of GFDL's next-generation prediction system.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.A54F..07Z
- Keywords:
-
- 3315 Data assimilation;
- ATMOSPHERIC PROCESSESDE: 3360 Remote sensing;
- ATMOSPHERIC PROCESSESDE: 3372 Tropical cyclones;
- ATMOSPHERIC PROCESSESDE: 4313 Extreme events;
- NATURAL HAZARDS