forecast quality and predictability of severe european cyclones
DESCRIPTION
Forecast quality and predictability of severe European cyclones. Jenny Owen Peter Knippertz , Tomasz Trzeciak . University of Leeds, School of Earth and Environment, Leeds, UK. Motivation. Xynthia. Damaging weather Important for Europe Cause fatalities and economic losses. Daria. - PowerPoint PPT PresentationTRANSCRIPT
Forecast quality and predictability of severe European cyclonesJenny OwenPeter Knippertz, Tomasz Trzeciak.
University of Leeds, School of Earth and Environment, Leeds, UK
• Damaging weather• Important for Europe• Cause fatalities and
economic losses
Xynthia
LotharDaria Friedhelm
Motivation
• Select historic storms – Storm Severity Index: – Measures ‘unusualness’ of wind
speed – Cubed ~ power of the wind ~
damage• Track storms automatically
– Minima in mean sea level pressure (MSLP)
– Connect together at consecutive timesteps
Method IHow well are severe European windstorms forecast?
What factors affect forecast quality?
Daria
Selected Storms
• Daria• Nana• Vivian• Wiebke• Udine• Verena• Agnes• Urania• Silke• Lara
• Anatol• Franz• Lothar• Martin• Kerstin• Rebekka• Elke• Lukas• Pawel• Jennifer
• Frieda• Jeanette• Gero• Cyrus• Hanno• Kyrill• Emma• Klaus• Quinten• Xynthia
Method II
• Categorise storms:
1. Jet stream shape, relative to the track of the storm
2. Processes that govern deepening, by pressure tendency equation
Categorising Storms: Jet
• Based on jet stream (wind speed at 300hPa).• Meridional sections that move with the storm track.• Similar plots for θe showed no clear groupings.
KlausKyrill
Xynthia
Split Jet
Cross Early Cross Late
Jeanette
Edge
Categorising Storms: Jet
Split
Cross EarlyEdge
Cross Late
• Klaus• Vivian• Wiebke• Kyrill• Lothar• Martin• Emma• Jeanette• Daria• Agnes
• Anatol• Udine• Rebekka• Lara• Xynthia• Jennifer• Gero• Hanno• Silke• Elke
• Urania• Nana• Quinten• Verena• Kerstin• Pawel• Cyrus• Lukas• Franz• Frieda
Pressure Tendency Equation
• Fink et al. (2012, GRL) applied the Pressure Tendency Equation to mid-latitude cyclones
• 3o x 3o column • From surface to 100hPa • Box moves along storm track
and compares properties from one time step to the next
• Identify processes that add or remove mass from column and affect core pressure
Pressure Tendency Equation
horiz vert diab
Density tendency
Precip
Stratosphere
Categorising Storms: PTE
storm dphidt ep res horiz vert diabEmma 0.00 1.30 0.00 76.98 0.00 21.71Kyrill 1.44 1.45 0.33 66.83 0.00 29.82Daria 2.70 1.55 0.00 64.41 0.00 31.20Martin 0.26 2.16 0.00 59.53 0.00 38.05Jennifer 4.19 2.29 0.00 56.88 0.00 36.64Vivian 0.00 1.52 1.14 53.76 0.00 43.58Klaus 2.52 3.36 0.00 43.30 0.00 50.57Wiebke 18.11 1.92 0.22 41.66 0.00 38.07Xynthia 1.81 3.99 0.00 33.11 0.00 61.09Lothar 6.77 3.71 0.00 31.40 0.00 57.91
PTE terms’ contribution to deepening for ten of the storms
Baroclinicity Diabatic Processes
Categorising Storms: PTE
• Klaus• Vivian• Wiebke• Kyrill• Lothar• Martin• Emma• Jeanette• Daria• Agnes
• Anatol• Udine• Rebekka• Lara• Xynthia• Jennifer• Gero• Hanno• Silke• Elke
• Urania• Nana• Quinten• Verena• Kerstin• Pawel• Cyrus• Lukas• Franz• Frieda
HorizDiab
Linking Categories
Horiz DiabCross Early 7 1Edge 9 1Cross Late 6 3Split 2 2
• Storms that spend longer on the north side of the jet tend to be more baroclinic – stronger temperature gradients.
• Diabatic storms tend to spend more time on the south side of the jet – warmer and wetter, more potential for latent heat release.
Method III
• Run automatic tracker on ECMWF Ensemble Control Forecast– Spatial and temporal resolution– Initialisation time
• Match forecast tracks to analysis tracks– Quantify best match based on proximity of analysis
and forecast tracks at similar time– Quality control: reject if tracks > 20 degrees apart at
any matched point
Matched Tracks: Location
KlausKyrill
Xynthia
Split Jet
Cross Early
Edge
Cross Late
Jeanette
• Some storms are better forecast than others• Some tracks are not a good match
Matched Tracks: Pressure
KlausKyrill
XynthiaJeanette
• Storms not always weaker in forecast – but difficult to see big picture
Method IV
• Assess how forecast quality varies with lead time
• Correlations– Correlation coefficient, R– Test for significance of correlation, T
• Future Work: Perform more rigorous statistical tests
Results: Latitude & Longitude
0 20 40 60 80 100 120 140 160 180
-12
-8
-4
0
4
8
12
16
f(x) = 0.0105853701735742 x − 0.170403825717322R² = 0.0679506436132739
f(x) = 0.0259708200495926 x − 0.627258235919236R² = 0.0751527330139643
LongitudeLinear (Longitude)LatitudeLinear (Latitude)
Lead Time
Anal
ysis
- Fo
reca
st
• Storms move more slowly W-E in forecasts, than in analysis• Storms slightly further south in forecasts
r t Sig?Latitude 0.261 2.30 Longitude 0.274 2.53
Results: Core Pressure
0 20 40 60 80 100 120 140 160 180
-50
-40
-30
-20
-10
0
10
20
f(x) = − 0.109502733955367 x − 1.20664273232731R² = 0.175392067572964
Pressure
Lead time
Anal
ysis
- Fo
reca
st
• Storms have higher core pressure in forecast => storm less intense in forecast• Agrees with previous work e.g. Froude et al.
r t Sig?Pressure -0.419 4.10
Jet Stream Type: Pressure
0 20 40 60 80 100 120 140 160 180
-50
-40
-30
-20
-10
0
10
f(x) = − 0.055654958645443 x + 2.28705098314606R² = 0.0720957782387813
f(x) = − 0.084335804218684 x − 4.26868775664918R² = 0.101438967935236
f(x) = NaN x + NaNR² = 0f(x) = NaN x + NaNR² = 0
CELinear (CE)EdgeLinear (Edge)CLLinear (CL)
Anal
ysis
- Fo
reca
st
r n t Sig?Cross early -0.63 20 3.45 Edge -0.42 20 1.99 Cross late -0.32 27 1.68 (0.1)Split -0.27 13 0.92
• Core pressure underprediction stronger in some jet stream types than others
PTE Type: Core Pressure
0 20 40 60 80 100 120 140 160 180
-50
-40
-30
-20
-10
0
10
f(x) = − 0.102317916264443 x + 1.98644315454146R² = 0.179841782079172f(x) = − 0.108323254944458 x − 3.239246408431R² = 0.175891664489393
HorizLinear (Horiz)Diab
r n t Sig?Horiz -0.419 54 3.33 Diab -0.424 26 2.29
• Indication that core pressure underprediction stronger in storms where baroclinic processes dominate deepening, than in those where diabatic processes dominate.
• Needs further statistical testing
Resolution: Core Pressure
• Operational forecast, so forecast system upgraded regularly (dynamics and resolution).
• Some evidence of relationship between forecast quality and system evolution.
100 150 200 250 300 350 400
-35
-30
-25
-20
-15
-10
-5
0
5
10
f(x) = 0.0523009605263158 x − 23.1280243026316R² = 0.337452399082539
Forecast at 36 hours lead time
Resolution (TL)
Fore
cast
- An
alys
is
Summary I
• Selected 30 European windstorms.• Categorised by:
– Jet stream – Processes that dominate deepening (PTE)
• Assessed forecast quality:– Longitude & latitude– Core pressure (intensity)
Summary II
• Storms in forecast too slow.• Core pressure generally underforecast:
– Strength of relationship with lead time depends on jet stream type.
– Baroclinic storms may be more underforecast than diabatic ones.
• Tendency for improvements of forecast system to affect forecast quality.