Diagnosis of Skill Variability of CPC Long- Diagnosis of Skill Variability of CPC Long- Lead Seasonal Forecasts: Implications for Lead Seasonal Forecasts: Implications for Production and Use Production and Use Bob Livezey and Marina Timofeyeva NOAA/NWS/OCWWS/Climate Services Division Climate Diagnostics and Prediction Workshop October 21, 2004 Madison WI
24
Embed
Bob Livezey and Marina Timofeyeva NOAA/NWS/OCWWS/Climate Services Division
Diagnosis of Skill Variability of CPC Long-Lead Seasonal Forecasts: Implications for Production and Use. Bob Livezey and Marina Timofeyeva NOAA/NWS/OCWWS/Climate Services Division. Climate Diagnostics and Prediction Workshop October 21, 2004 Madison WI. Outline. Introduction - PowerPoint PPT Presentation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Diagnosis of Skill Variability of CPC Long-Lead Diagnosis of Skill Variability of CPC Long-Lead Seasonal Forecasts: Implications for Seasonal Forecasts: Implications for
Production and UseProduction and Use
Bob Livezey and Marina TimofeyevaNOAA/NWS/OCWWS/Climate Services Division
Climate Diagnostics and Prediction WorkshopOctober 21, 2004
Madison WI
OutlineOutline
• IntroductionIntroduction• Skill StratificationsSkill Stratifications• ResultsResults• Conclusions and LessonsConclusions and Lessons
IntroductionIntroduction
• Users should only care about the performance of forecasts that can potentially benefit their decision process
• Livezey (1990): There are non-random subsets of seasonal forecasts that were skillful enough to be useful
Seasonal Temperature Forecast Skill Seasonal Temperature Forecast Skill 1960s to 80s1960s to 80s
All Seasons 8.3
Winter 12.6Spring 8.6Summer 9.3Fall 2.8
Seasonal Temperature Forecast Skill Seasonal Temperature Forecast Skill 1960s to 80s1960s to 80s
All Seasons 8.3
Winter 12.6Spring 8.6Summer 9.3Fall 2.8
Introduction (Cont.)Introduction (Cont.)
• This talk will make the point :
– That was made by Livezey (1990)– That there are many non-random subsets of forecasts that do
not have useful skill – That it is critical for this information to be shared with
potential users– That skill analyses with different stratifications are highly
informative– That seasonal forecast production should be dominantly
objective– That long-lead seasonal forecasts of non-trend signal should
only be done on appropriate opportunities, not routinely
Displays and StratificationsDisplays and Stratifications
• CPC Seasonal Forecasts– For 3-equally probable temperature and precipitation classes at 102
Climate Divisions– Made every month from 1995 to present for 0.5-, 1.5-, …, 12.5 month
leads
• Skill Measure: Modified Heidke Skill Score of Categorized Forecasts– 1/3 EC’s scored as ‘hits’
• Displays and Stratifications– Summed over all forecasts for each lead for three overlapping
seasons at a time (DJF to FMA, FMA to AMJ, etc.) – Stratified further by strong ENSO years vs. other years– Mapped for appropriate combinations of leads
– Barely useable national-scale skill entirely confined to Fall/Winter strong ENSO years in short to medium leads
– Otherwise skill is statistically indistinguishable from zero
– Short-lead forecasts overall seem to be no better now than for the 1960s-80s (but at least are now made two-weeks earlier)
PrecipitationPrecipitation
-120 -110 -100 -90 -80 -70
3035
4045
-20 0 20 40
AL
AK
AZAR
CA
CO
CT
DE
FL
GAHI
ID
IL IN
IA
KS
KY
LA
ME
MD
MA
MI
MN
MS
MO
MT
NE
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PARI
SC
SD
TN
TX
UT
VT
VA
WA
WV
WI
WY
Other: OND-DJF (0.5-12.5 lead)
PrecipitationPrecipitation
-120 -110 -100 -90 -80 -70
3035
4045
-10 0 5
AL
AK
AZAR
CA
CO
CT
DE
FL
GAHI
ID
IL IN
IA
KS
KY
LA
ME
MD
MA
MI
MN
MS
MO
MT
NE
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PARI
SC
SD
TN
TX
UT
VT
VA
WA
WV
WI
WY
Other: FMA-AMJ (0.5-12.5 lead)
PrecipitationPrecipitation
-120 -110 -100 -90 -80 -70
3035
4045
-5 5 15 25
AL
AK
AZAR
CA
CO
CT
DE
FL
GAHI
ID
IL IN
IA
KS
KY
LA
ME
MD
MA
MI
MN
MS
MO
MT
NE
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PARI
SC
SD
TN
TX
UT
VT
VA
WA
WV
WI
WY
Other:JJA-ASO (0.5-12.5 lead)
PrecipitationPrecipitation
-120 -110 -100 -90 -80 -70
3035
4045
-20 40 80
AL
AK
AZAR
CA
CO
CT
DE
FL
GAHI
ID
IL IN
IA
KS
KY
LA
ME
MD
MA
MI
MN
MS
MO
MT
NE
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PARI
SC
SD
TN
TX
UT
VT
VA
WA
WV
WI
WY
ENSO: DJF-FMA (0.5-6.5)
ResultsResults
• Seasonal Precipitation Maps:– Justification for long-lead forecasts is
questionable for at least three of six seasonal groups examined
– For certain regions/seasons/situations skill is unambiguously useful even for undisciplined, occasional users
Conclusions and Conclusions and RecommendationsRecommendations
• There are non-random subsets of seasonal forecasts that are skillful enough even for undisciplined, occasional users
• These skills exclusively reflect ENSO and trend signals
• There are many non-random subsets of forecasts that do not have useful skill
• We must share this information with potential users
• Skill analyses with different stratifications are highly informative
• Seasonal forecast production should dominantly rely on objective exploitation of ENSO and trend with subjective modification only after rigorous criteria are met
• Routine production of certain long-lead forecasts should be terminated (can be accommodated by transition to separate trend and high frequency forecasts with latter issued on forecast of opportunity basis only)