Uncertainties in weather and climate prediction · weather forecasts and climate projections Future...
Transcript of Uncertainties in weather and climate prediction · weather forecasts and climate projections Future...
Uncertainties in weather and climate prediction
Henk DijkstraInstitute for Marine and Atmospheric research Utrecht
Utrecht University
Weather Forecasts Climate projections
A classical mechanical system: the pendulum (in vacuum)
Given initial position and velocity -> determine future position
θ(0) = θ0
dθ
dt(0) = 0
d2θ
dt2+
g
Lθ = 0
θ(t) = θ0 cos(√
g/L t)
| θ |� 1
θ
mg
:mg sin θ
mLd2θ
dt2= −mg sin θ
Creative aspect 1: Phase space
Geometry of motion!
GibbsBoltzmann
Poincare
d2θ
dt2+
g
Lθ = 0
dx
dt= y
dy
dt= − g
Lx
x = θ ; y =dθ
dt
Nonlinear mechanical (fluid) systems
The Lorenz system
Lorenz model
demo
dx
dt= −c(x − y)
dy
dt= ax − y − xz
dz
dt= b(xy − z)
~ vertical temperature difference
Edward Lorenz (1917-2008)
a
Lorenz attractor
Creative aspect 2: Sensitivity to initial conditions
Trajectory
Lyapunov exponent
0.9056, 0, -14.5723Lorenz system:
λ > 0 → chaotic motion
d(t) = x′(t)− x(t)
λi = limt→∞
1
tln | di(t)
di(0)|
Numerical Weather Prediction Model
Grid: N x M x L
# Observables (temperature, humidity, velocities, etc.): k
Dimension phase space: d = k x N x M x L
Typically: d = 105 − 10
9
Origin of the ‘plume’ in weather forecasts
Numerical weather prediction models: many Lyapunov exponents > 0
Daily mean temperature January 1940 in `grid box’ the Netherlands
Weather prediction
Lorenz (1969): … one flap of a sea-gull’s wing may forever change the
future course of the weather
Leith (1984): … even talking about the weather can
change the weather!
Examples of finite-time error growth on the
Lorenz attractor for three probabilistic predictions
starting from different points on the attractor.
Error growth in the Lorenz attractor
Perron-Frobenius or transfer operator
Lorenz model with ‘noise’
σ : noise amplitude X
Z
= 0.1 density of trajectories
Regime transitions in atmospheric flows
Transition through preferred pathways increases predictive skill
Atmospheric pressure anomalies (hPa)
Mean kinetic energy
Eddy kineticenergy
in press (2015)
Creative aspect 3: Ensemble forecasting
Forecast time
Tem
pera
ture
Complete description of weather prediction in terms of a Probability Density Function (PDF)
Initial condition Forecast
Numbers of observational items assimilated over a 24 hour period on 13 February 2006
Starting a forecast: The initial conditions
Flow dependence of forecast errors
If the forecasts are coherent (small spread) the atmosphere is in a more predictable state than if the forecasts diverge (large spread)
34
30
28
26
24
22
20
18
16
14
12
10
0 1 2 3 4 5 6 7 8 9 10Forecast day
UK
Control Analysis Ensemble
ECMWF ensemble forecast - Air temperatureDate: 26/06/1994 London Lat: 51.5 Long: 0
30
28
26
24
22
20
18
16
14
12
10
8
0 1 2 3 4 5 6 7 8 9 10Forecast day
UK
Control Analysis Ensemble
ECMWF ensemble forecast - Air temperatureDate: 26/06/1995 London Lat: 51.5 Long: 0
26th June 1995 26th June 1994
Example of 66 h probabilistic forecast for 15–16 October 1987.
Slingo J , and Palmer T Phil. Trans. R. Soc. A 2011;369:4751-4767
Processes limiting predictability: formation of H and L pressure systems and their interaction
Surface pressure maps UK, North Sea
Predictability limits
Statistics of ensemble mean forecast error (r.m.s.e.; solid line) and ensemble spread (dotted line) in Northern
Hemisphere systems
Predictability horizon
How about climate projections?
Climate is what you expect, weather is what you get!
Weather: state of the climate system at a given time and place
Climate: statistics of weather conditions over a decade or more
Representative concentration pathways
IPCC, Chapter 11, 2013
Scenario:
Climate models
Community Earth System Model
ocean/sea ice grid box: 10 km
atmosphere/land grid box: 50 km
d ~ 109
Tiled Panel Display Visualization (BBG 611)
Climate projection results
Back to the basics … Pendulum in air
vacuum
air
air
`weather’ of the pendulum: small scale (chaotic) motions induced by interaction with the air
`climate’ of the pendulum: longer time scale (regular) motion controlled by gravity
Also the `climate’ state can display chaotic behavior (but with a very different
Lyapunov exponent than for the weather)
The double pendulum
Projection uncertainties
0 20 40 60 80 1000
0.2
0.4
0.6
0.8
1
Lead time [years from 2000]
Fra
ctio
nal u
ncer
tain
ty
Internal variability
Scenario
Model
Total
Global, decadal mean surface air temperature
SummaryChaos plays a very different role in uncertainties of
weather forecasts and climate projections
Future weather forecasts:- relevant processes are instabilities of the large-scale atmospheric circulation with typical time scales of up to 5 days- limited prediction skill beyond 10 days
Future climate change:- relevant processes are several large-scale feedbacks in the climate system associated with the radiation balance- projection skill is limited by emission scenario
Creative aspect: Phase Space; connecting geometry and motion
Creative aspects: Sensitivity to initial conditions; Ensemble forecasting
Creative aspect: Predictability horizons connected to specific physical processes