Transcript Seasonal Degree Day Outlooks - Home

Seasonal Degree Day Outlooks
David A. Unger
Climate Prediction Center
Camp Springs, Maryland
Definitions
_
HDD = G65 – t
_
_
t < 65 F
_
CDD = G t – 65 t > 65 F
HD
CD = CDD/N
_ = HDD/N
T = 65+CD-HD
_
_
_CD = T –65 +HD
t = daily mean temperature, T=Monthly or Seasonal Mean
N = Number of days in month or season
CPC Outlook
CPC POE Outlooks
Tools
Overview
Forecaster Input
Skill: Heidke .10
RPS .02
Temperature Fcst
Prob. Anom.For Tercile
(Above, Near, Below)
Model Skills, climatology
Temperature POE
Skill: CRPS .03
Downscaling (Regression Relationships)
Temperature POE
Downscaled
Skill: CRPS .02
Temperature to Degree Day
(Climatological Relationships)
CRPS Skill: CDD .05
Degree Days
HDD .02
HDD CDD POE
Accumulation Algorithms
Degree Days
CRPS Skill: CDD .06
Flexible Region, Seasons
HDD .02
Temperature to Degree Days
Rescaling
Downscaling
FD Seasonal
CD Seasonal
Disaggregation
FD Monthly
CD Monthly
Downscaling
• Regression
• CD = a FD +b
Equation’s coefficients are “inflated”
(CD variance = climatological variance)
Disaggregation - Seasonal to Monthly
• Tm = a Ts + b
Regression, inflated coefficients
• Average 3 estimates
M = M JFM + M FMA + M MAM
3
Verification note
• Continuous Ranked Probability Score
- Mean Absolute Error with provisions
for uncertainty
• Skill Score = 1. –
CRPS
CRPS Climo
- Percent Improvement over climatology
Continuous Ranked Probability Score
CRPS Skill Scores:
.051
.045
.027
.029
.041
.034
.026
.023
.020
.021
.024
.024
.040
.036
.026
.030
.094
.103
.074
.090
.055
.059
.055
.058
Temperature
.013
.016
.027
.026
.044
.038
.050
.047
.002
.001
.011
.004
-.009
.002
-.006
-.008
Skill
.035
.030
.012
.015
High
.10
Moderate
.065
.055
.042
.035
.05
Low
FD
1-Month Lead, All initial times
.01
CD
None
.031 .023
3-Mo
.028 .019
1-Mo
CRPS Skill Scores: Heating and Cooling Degree Days
.040
.071
.114
.085
.036
.073
.019
.028
.047
.102
.023
.048
.101
.121
.014
.076
.088
.115
.079
.111
.090
.090
.029
.035
.058
.043
.021
-.011
.035
.014
.045
-.003
.009
.022
.000
-0.16
-.004
.036
-.026
-.016
Skill
.033
.051
.005
.003
High
.10
Moderate
.044
.024
.046
.030
.05
Low
1-Mo 12-Mo
.02
None
.049 .057
Cooling
.018 .016
Heating
Degree Day Forecast (Accumulations)
Reliability
Reliability
Conclusions
• Downscaled forecasts nearly as skillful as
original temperature outlook
• Skill better in Summer than Winter
• Better understanding of season to season
dependence will lead to improved forecasts
for periods greater than 3-months.
Testing
• 50 – years of “perfect OCN”
Forecast = decadal mean and standard deviation
•
Target year is included to assure skill.
• Seasonal Forecasts on Forecast Divisions supplied
How does the skill of the rescaled forecasts
compare to the original
CRPS Skill Scores – Downscaled and disaggregated
.098
.081
.088
.092
.061
.042
.086
.083
.063
.039
.061
.059
.109
.109
.058
.055
.106
.019
.067
.077
.198
.233
.106
.135
.110
.086
.066
.066
.074
.070
.052
.037
.088
.085
.061
.055
.108
.105
.061
.060
.138
.140
.086
.067
Skill
High
.110
.087
Moderate
.074
.044
Low
FD
CD
None
.104 .109
Seasonal
.066 .057
Monthly
.10
.05
.01
CRPS Skill Scores Temperature to Degree Days
.098
-.027
.088
-.006
.098
.082
.088
.070
.109
.038
.109
.090
.106
.081
.106
.085
.198
.197
.198
.151
.110
.092
.110
.060
.086
.090
.086
.053
.074
.078
.074
.049
.110
.076
.110
.109
.088
.093
.088
.085
.108
.097
.108
.066
.138
.140
.138
.102
Skill
High
.10
Moderate
.05
Low
T
DD
.01
None
.104 .095
Cooling
.104 .074
Heating
Accumulation Algorithm
DDA+B = DDA
+ DD
B
Independent (I)
FA+B = F2A + F2B
Dependent (D)
FA+B = F A +F B
From Climatology
F(I)A+B < FA+B < F(D)A+B
FA+B F(I)
K =
F(D) F (I)
FA+B =F(D) + K(F(D)+F(I))