Teleconnections & Extended-Range Prediction Steven Feldstein The Pennsylvania State University March 3, 2016

Teleconnection Patterns Definition: Teleconnection patterns are recurring and persistent spatial patterns that connect weather (wind, temperature, precipitation, pressure) over great distances across the globe.

The dominant Northern Hemisphere teleconnection patterns Climate Prediction Center

Impact of teleconnections on the jet stream

Earliest NAO Observations Norse (Viking) settlers arrived in Greenland in CE 985. The Norse, who appeared to be very interested observers of the weather, also seemed to be aware of teleconnection patterns in the North Atlantic basin. There was an anonymous Norwegian book (approx. CE 1230), entitled the `King's Mirror'. This book, in the form of a discussion between father and son, wrote that severe weather in Greenland coincides with warmer weather at distant locations, and vice versa.

Wallace and Gutzler (1982) WP NAO PNA A modern study which unified previous teleconnection research and systematically Identified the key patterns WP NAO PNA

Two types of teleconnections Northern Annular Mode Southern Annular Mode Annular (dipole) Modes (Thompson and Wallace (2000)) Wave train patterns PNA Pacific South American (PSA) pattern

What is the timescale of teleconnections? Teleconnections last for about 2 weeks For most of 20th century, scientists thought the teleconnections persisted for several months and were probably forced by SST anomalies Using a climate model, Lau (1981) showed that teleconnections remain in model solution if the SST field is fixed Feldstein (2000) showed with daily observational that all teleconnections except the Southern Oscillation have an e-folding time scale of about 10 days

How are the PNA and PSA excited(a modeling study)? Wavetrain Model’s tropical convection Horel and Wallace (1981) PNA and PSA are excited by tropical convection Wallace and Gutzler (1981)

How are the NAO, NAM, and SAM excited How are the NAO, NAM, and SAM excited? (The NAO is generated by Rossby wave breaking) Black contours show two different potential temperature contours on the tropopause. Wave breaking is identified by warm air being located poleward of cold air. Benedict et al. (2004)

The PNA is amplified by breaking Rossby waves (after tropical convection) As with the NAO, the PNA is also associated with wave breaking. However, wave breaking plays a smaller role for the PNA.

8-14 day medium-range NOAA CPC forecasts

Gap in forecasting between 10 and 30 days NWP forecasts from 1-10 days have been steadily improving. These forecasts are limited by chaos theory. Probabilistic forecast (for lead times > 30 days) are now being performed. The physical basis for these forecasts is that anomalies in SST change slowly over many months. Up until the past few years, because of this gap, 10-30 forecasts have not been performed. Using our knowledge of the properties of teleconnections, and the physical processes that drive them, we are optimistic that 10-30 day probabilistic weather forecast will soon be taking place.

Edward Lorenz (1917-2008) Pioneered chaos theory, which placed the 7-10 day limit on forecast predictability under a firm theoretical foundation (Lorenz 1963). Lorenz showed that two atmospheric states, with an imperceptible difference, can evolve into very different states. Since there are always errors in observational measures of the atmosphere, this implies that there is a limit to skillful weather forecasts. This is the so-called Butterfly Effect.

Why is SST linked to long-range forecasting? SST anomalies slowly evolve over many months.

Why do SST anomalies last for several months? Tropical SST anomalies are associated with ocean waves that propagate very slowly and persist for many months.

Madden Julian Oscillation A 30-60 day tropical oscillation in convection and the wind Eastward propagating convection The MJO is the most prominent mode of intraseasonal variability in the tropics. There are many theories that offer explanations for the occurrence of the MJO.

A link between the NAO and the MJO The relationship between the MJO phase and the anomalous frequency of occurrence as a function of lead time for 4 teleconnection patterns The MJO has a big influence on both NAO phases Cassou (2008) A physical explanation for the link between the NAO and MJO is unclear

PNA occurrence depends upon where tropical convection is When Convection is over Indonesia (Phase 4), +PNA occurs more frequently 12-18 days later small pattern numbers represent +PNA large pattern numbers represent -PNA When Convection is over Indonesia (Phase 4), --PNA occurs less frequently 15-21 days later Johnson and Feldstein (2010) Physical mechanism linking the MJO to the PNA is Rossby wave propagation

MJO convection excites Rossby waves that propagate from the tropics to midlatitudes

Development of a Probabilistic Extended-Range Forecast Model To test if probabilistic 2-4 weeks forecast based upon the phase of the MJO, ENSO, and long term-trend are feasible As we will see, our project was a success. The results also show the importance of using basic science to advance weather forecasting.

Wheeler and Hendon diagram Source: http://cawcr.gov.au/staff/mwheeler/maproom/RMM/ RMM1 and RMM2 define the phase and amplitude of the MJO

Extended-range prediction: MJO and ENSO Extended-range (10-30 day) prediction is based on two key ideas: MJO and ENSO tropical convection persists for more than 10-30 days. The MJO has a period of 30-60 days and remains in a single phase for about 5-6 days. The ENSO SST anomalies persist for many months. It takes tropical convection about 7 days to excite midlatitude teleconnection patterns, and once established these patterns typically last for about 2 weeks.

The method for obtaining probabilistic forecasts. Generate two probability density function: all days (left panel), days for a specific MJO and ENSO phase (right panel). Comparing the solid and dashed curves in the right panel gives the probabilistic forecast.

Heidke Skill Score HSS = (H - E)/(T - E) X 100 H = number of days correctly forecasted E = expected fraction of correct forecast by chance (E=1/3 total number of days) T = number of forecasted days Use leave-one-year-out forecast approach HSS is used to evaluate the forecast skill. HSS = 0 corresponds to a random forecast

Comparing HSS values for observational and GFSv2 model data HSS value for all days with observational data (statistical model) HSS values for those days when the MJO and/or ENSO are active (statistical model) HSS values for GFSv2 numerical model (state-of-the-art US model) Statistical model better beyond 3 weeks

At many locations HSS values are high Forecast lead time The MJO and ENSO influence different locations

HSS values by MJO phase: Some MJO phases yield better forecasts

Examples of a 2-week forecast Teleconnection pattern HSS Forecast Forecast HSS

Examples of 3 and 4 week forecasts HSS

Further improvement of forecasts Our research suggests that accounting for the following features has the potential to further improve probablistic forecasts The initial flow in midlatitudes The state of the stratospheric polar vortex Arctic sea ice Interference between stationary and planetary waves in midlatitudes

Examining the initial flows that leads to a large/small amplitude extratropical response to the MJO The 300-hPa geopotential height field that is followed 7-10 days later by a large (left column) or a small (right column) amplitude wave field in midlatitudes. Include a new parameter in the forecast based on the similarity between the observed flow and the above patterns

The NAM in the stratosphere and troposphere When the NAM is active in the stratosphere, the NAM and NAO in the troposphere sometimes persist for up to two months. This has potential Implications for extended and long range forecasting. (Baldwin and Dunkerton (1999, 2001)

Summary of impact of sea ice Lee and Feldstein (2014)

Interference in the upper troposphere Constructive Interference: Positive SWI Destructive Interference: Negative SWI Constructive and destructive interference between the climatological stationary eddies and transient eddies for all seasons (Goss et al. 2015)