e-Communications Sep 2019

The 8th NCEP Ensemble User Workshop S2S Prediction and Applications Session

The 8th NCEP Ensemble User Workshop was held at NOAA Center for Weather and Climate Prediction in College Park, Maryland on August 27-29, 2019. The central theme was to better convey forecast uncertainties to users to optimize the utility of NWS forecast information. The workshop had a session on subsesonal-to-seasonal (S2S) climate prediction and applications on day 3. Climate Prediction Center (CPC) Director David DeWitt gave an overview, which broached the subject. The presentations by ensemble developer, forecast producer and external partner representatives followed to share their valuable experiences in evaluation and utility of the products. Following is the main content of the overview.

Requirements and challenges to improve S2S climate prediction

The subseasonal to seasonal predictions are characterized by small signals and large noise, hence they are inherently probabilistic. The key tool for informing forecasts is using ensemble of dynamical models. The S2S threat vector describing the forecast uncertainty narrows as time to event decreases. But it never becomes deterministic. CPC hazard forecasts present the opportunity when we anticipate a period of increased predictability.

CPC S2S prediction products use ensembles with focus on the first month. The week-2 probabilistic hazards outlook is designed to forecast probability of extreme events. We use the Global Ensemble Forecast System (GEFS) based Week-2 Global Probabilistic Extremes Forecast Tool to inform the development of probabilistic hazard, which is available online. Our standard temperature and precipitation outlooks include week-2 and week 3-4 forecasts, which are issued in tercile classes (above, near, or below normal) daily and in two classes (above or below normal) once per week on Friday, respectively. CPC also issues GEFS-based week-2 forecast for the probability of exceeding 50 mm rainfall.

CPC technical requirements for the day 6 to ~ day 35 ensemble forecast system are (1) consistent reforecasts with each real-time upgrade of length at least 20 years, and 5 members, (2) equilibrated boundary conditions in reforecasts and real-time, i.e. no discontinuities or non-physical trends in soil moisture, upper ocean temperature, and sea ice etc.,

Fig. 1 Standard deviation of January precipitation from one month lead forecasts by 3 state of the art S2S models (top 3 panels) and the observation (bottom panel). It demonstrates that models have errors of 100% or more in predicting mean statistics of tropical precipitation. Result is even worse if large ENSO events being removed.

(3) metrics for evaluation of upgrade, such as i) temperature and precipitation skills over the same domain as that of CPC outlooks, ii) global skill for relevant fields for Global Tropics Hazards Outlook and International Desk products, and iii) fidelity of simulation of major modes of S2S variability and associated teleconnections from El Niño-Southern Oscillation (ENSO), Madden-Julian Oscillation (MJO), Arctic Oscillation (AO), North Atlantic Oscillation (NAO), Pacific/North Amerian pattern (PNA), and Sudden Stratospheric Warming (SSW).

CPC's wishlist to improve the day 6 to ~ Day 35 ensemble forecast system contains (1) reliable predictions of periods of enhanced versus suppressed predictability, (2) reduction of persistent systematic errors in coupled general circulation models (CGCMs), and (3) improved precipitation forecasts on all timescales.

There has been dramatically increased skill in predicting MJO as a result of tremendous efforts to improve physics made by Environmental Modeling Center. To derive the benefit for stakeholders, we still need to drastically improve teleconnections from MJO, which involve other important processes, like nonlinear interactions with the jet stream in the mid-latitude as one of them.

More scientific challenges that limit CPC S2S forecast skill include (1) El Niño false alarms, a major systematic error that current generation coupled models have in timing and amplitude of equatorial Pacific SST anomalies forecast, (2) systematic errors in forecasting magnitude and distribution of average precipitation in the tropics (even at first month lead) (ex. Fig. 1), (3) inability to predict onset of flash drought as occurred over Northern Plain in 2017, which highlights challenges on predictability of precipitation beyond week-2 for the spring and summer seasons when variability is controlled by convective as opposed to advective processes, and accuracy of land surface model simulation of enhanced evaporation, (4) development of erroneous double Inter Tropical Convergence Zone (ITCZ), a common model problem across time scales from S2S to long-term climate change.

The Program for the Oral and Poster Sessions are available online at NOAA/CPC website.

Online registration to the Workshop is open and the details of registration, hotel and other information can be found at the website of the Meeting's local host organizer, Department of Earth and Ocean Sciences at the Nicholas School of the Environment, Duke University.

Societal demand for S2S forecasts continues to grow. S2S models are just starting to be run at resolutions where transient eddies are resolved. Hopefully, this leads to improvements in representation of their statistics from a forecasting perspective. Persistent systematic errors in S2S (and climate change) models likely limit forecast skill. Reducing these errors should be a priority for the global S2S enterprise in order to meet societal demand for improved S2S forecast skill.

(Collated by Jiayu Zhou)