Seamless predictionOpportunities and Challenges

Matthew Wheeler1, Hongyan Zhu1, Adam Sobel2, Debra Hudson1 and Griff Young1

The Centre for Australian Weather and Climate ResearchA partnership between CSIRO and the Bureau of Meteorology

1

2 Columbia University, New York, USA

AMOS workshop: The Interface of Weather and Climate, 9 February, 2015

Seamless prediction: What is it?• Seamless prediction is currently a buzz phrase in the

meteorological community.

• Usually it is used to refer to a near continuous time-scale range of prediction products from weather to climate time scales.

"There is no scientific basis to draw artificial boundaries between meso-scale prediction, synoptic scale prediction, seasonal prediction, ENSO prediction, decadal prediction and climate change." Shukla (2009).

• Some argue that a consequence of the above is that predictions across the range of scales should be made with the same models. But others argue against this for practical purposes.

• Finally, an even smaller group have used seamless prediction to refer to a seamless transition between model predictions of atmospheric variables (e.g. temperature/precipitation) and applications (e.g. crop yield forecasts). I don't like this use!

ClimateWeather

In essence, seamless prediction (in my view) is about reducing the barriers between weather and climate in terms of the end-user products and/or the methods and models that are used.

e.g. Recognising that predictions of a tropical cyclone (and other weather-scale phenomena) can be improved by the inclusion of an ocean model component.

e.g. Recognising that atmospheric initial conditions can also be important for seasonal to interannual prediction (and not just weather prediction).

e.g. Providing intermediate-range (e.g. multi-week) prediction products to users.

Another example of a "seamless prediction" approach

Why is this seamless?1. Uses a coupled ocean-atmosphere model that is initialized with both realistic atmospheric and oceanic initial conditions.

2. We look at the performance of the model predictions across a wide range of time scales.

The essence of our approach is:• Compute prediction skill globally for a large range of lead times.

• As we increase the lead time, we also increase the time-averaging window for a seamless transition from weather to climate.

Schematic of window/lead definitions

Data and Method

a. POAMA-2 ensemble prediction system

T47L17 atmosphere; 0.5-2º L25 ocean; and land.

Initialized with realistic atmospheric, land, and ocean initial conditions.

Coupled ocean/atmosphere breeding scheme to produce a burst ensemble of 11

members.

3 versions of the model to provide in total 33 members.

Hindcasts from the 1st, 11th, and 21st of each month (out to 120 days).

b. Observations

GPCP daily precipitation (blended station and satellite).

1º grid converted to POAMA grid.

We use 1996 to 2009 for this work.

c. Measure of prediction skill

We tried different verification measures (ROC score, Brier score, correlation skill).

In the end we chose the simplest: the correlation of the ensemble mean.

We use two versions:

CORt - using total precipitation values

CORa - using anomalies with respect to separate climatologies for the

hindcasts and observations.

CORt is more usual for weather prediction; CORa is more usual for seasonal prediction.

The correlations are computed over time using data from multiple verification times.

Separately for each lead time and each grid point.

Separately for DJF (n=117) and JJA (n=108).

Only CORa is shown in this presentation.

CORa (i.e. removing the influence of the climatological seasonal cycle)

1d1d: Extratropics better than tropics; winter extratropics better than summer.4w4w: ENSO dominates.

Zonally-averaged CORa

The peak in skill at the equator is apparent at all lead times.Extratropical skill drops rapidly from 1d1d to 1w1w and then levels-off.

Skill in tropics (10ºS-10ºN) overtakes skill in extratropics for 4d4d in DJF and 1w1w in JJA.

CORa: plotted as a function of the log(time)

How has this research been applied?• Our seamless verification approach is now being applied by others

as it provides a fairer comparison across time scales.

• The work has given us confidence to provide predictions for a greater range of time scales from our operational coupled ocean-atmosphere model.

Experimenta

l Use O

nly

How "seamless" are current operational products?• At the Bureau there are still separate web pages for weather versus climate

prediction, and they use different models and methods. But progress has been made: "Climate" now includes the monthly outlook, and "weather" now goes out to 7 days.

• ECMWF is perhaps the closest to the seamless paradigm.

All prediction streams use the IFS atmospheric modelling system.

Medium range system is now coupled to an ocean from day 0.

At day 10 the IFS resolution is reduced and continues to day 32 (twice per week) to produce the extended-range products.

However, the long range system is still run separately, using a slightly older version of the IFS.

http://www.ecmwf.int

Seamless Prediction: Opportunities• Scientific cross-fertilization between the weather and climate

communities.• Prediction products covering a wide range of time scales. • Scientifically more satisfying?• Improved prediction skill?

Seamless Prediction: Challenges• Inconsistencies can arise. E.g. will the 10-15 day output from the medium-range system

be consistent with the 10-15 day output from the extended-range system?

• The affordable model resolution for short time scales will always be higher than that for long time scales. This means that different physical processes will need to be parameterized versus resolved, making the model inherently different when run at a different resolution.

• Different processes and phenomena are important for different time scales, and different models are better at different phenomena.

• Display and dissemination to the public.• Compromised prediction skill?

The Centre for Australian Weather and Climate Research

A partnership between CSIRO and the Bureau of Meteorology

THANKS! m.wheeler@bom.gov.au

A buzz phrase in other languages as well?

Extra Slides

Example from David Jones to illustrate the seamless approach. It's the official forecast and overlaid POAMA maximum temperature forecasts for Moree . A basic MOS type calibration of the POAMA output was applied to remove the bias.

Comparison with persistence

An important component of predictability is the prediction skill that can come from persistence. What is its contribution here?

Use the most recent observed precipitation anomalies to predict future anomalies.

1 week average

1 day average

Initial condition

1d1d CORa for persistence (top) model (bottom)

Tropics are generally more persistent than extratropics, but model forecasts convincingly beat persistence almost everywhere.

Persistence of ENSO is obvious, but model still beats persistence in most locations, including around Australia and the US west coast in DJF.However, the model does not get the persistence skill around the sea-ice edges, because it currently uses prescribed climatological sea ice.

4w4w CORa for persistence (top) model (bottom)

Comparison with Potential (or Perfect) Skill:using the assumption that one ensemble member is truth

(All seasons)

Actual skill

Potential skill(perfect model assumption)

Can interpret the difference as either "room for actual improvement" or "too little spread". Here we see the greatest difference for the tropics at the shorter lead times.This points the finger at the tropics (and moist convection) as our biggest handicap.