180617 Medium-Range Cloud-Resolving Typhoon Rainfall Ensemble Forecasts for Taiwan through Time-Lagged Approach

Title:
Medium-Range Cloud-Resolving Typhoon Rainfall Ensemble Forecasts for Taiwan through Time-Lagged Approach

Speaker:
陳鑫澔 (Shin-Hau Chen)

Time:

*注意：本週演講往後調整到週日 (US) / 週一 (TW)

06/17 (Sun.) 6 pm PDT, 7 pm MDT, 8 pm CDT, 9 pm EDT
06/18 (Mon.) 9 am Taiwan

Keywords:
Atmospheric science, Typhoon forecast, Precipitation forecast, Typhoon, Quantitative precipitation forecast, Cloud-resolving model

Abstract:

For typhoon hazard reduction in Taiwan, realistic rainfall scenarios associated with different tracks are the most demanded information by the authorities, and can be best provided only by cloud-resolving models (CRMs). However, at high computational expense, such models are deterministic and generally regarded as having inadequate lead time and no probability information to quantify forecast uncertainty, compared to a multi-member ensemble. Thus, this study designs a new forecast strategy to solve the two major drawbacks of CRMs through a time-lagged approach. The ensemble system using the Cloud-Resolving Storm Simulator (CReSS) has a grid size of 2.5 km, a large domain of 1860 x 1360 km2, and an extended range of 8 days, and combines the strengths of high resolution for quantitative precipitation forecast (QPF) and longer lead time for preparation in an innovative way. Its performance is evaluated for six typhoons in 2012-2013.

For the six typhoons, in addition to at short ranges within 3 days, the system produced a high-quality QPF at a longest range up to day 8, 4, 6, 3, 6, and day 7 beforehand, respectively, providing much extended lead time, especially for slow-moving ones that pose higher threats to Taiwan. Moreover, demonstrated in the highlight case of Kong-Rey (2013), since forecast uncertainty (reflected in spread) reduces with lead time, this system can provide a wide range of rainfall scenarios in Taiwan early on at longer lead time (beyond 4-5 days), each highly realistic for its track, for early warning and advanced preparation for the worst case. Then, as the typhoon approaches and the predicted tracks converge at short range, the authority can make adjustments toward the scenario with increasing likelihood. This strategy fits well our conventional wisdom to “hope for the best, but prepare for the worst” when facing natural hazards like typhoons.

Requiring only about 1500 cores to execute four 8-day runs per day, our system compares much favorably in usefulness to a typical 24-member ensemble (5-km grid size, 750 x 900 km2, 3-day forecasts) currently in operation at comparable computational expenses. Thus, it is not only powerful, but also already affordable and feasible.