Download presentation
Presentation is loading. Please wait.
Published byStanley Roberts Modified over 9 years ago
1
CPC Seasonal Forecasts ASO 2005-JAS 2006 Edward O’Lenic Michael Halpert, David Unger NOAA-NWS-CPC 31 st Climate Diagnostics and Prediction Workshop Boulder, Colorado Monday, October 23, 2006
2
Overview During 2005-06 CPC implemented an objective forecast tools consolidation (CON), combining OCN, CCA, SMLR and CFS into a single CON tool. In retrospective forecasts over 1995- 2004 CON improved on official (OFF) temperature (T) and precipitation (P) forecasts by ~18% and ~130%, respectively. We don’t have cross-validated scores yet for this technique and, therefore, don’t know how CON performs on independent data. However, Dave Unger has cross-validated this technique on SST forecasts and found no significant difference in skill between dependent and independent data sets. I will discuss the performance of the 12 ASO2005-JAS2006 ½- month lead T, P forecasts in the context of OFF real-time and CON retrospective forecasts from 1995-2006.
3
Non-ECAll=Non-EC and EC
4
4748 12-month mean skill T Non-EC
5
126 12-month mean skill P Non-EC
6
4748126 12-month mean skill TP
7
4748126 12-month mean skill 1322 3637 1995-2006 standard deviation TP
8
4748126 12-month mean skill 1322 3637 1995-2006 standard deviation 1995-2006 corr. Skill vs % coverage (non-EC) 06-2 TP
9
4748126 12-month mean skill 1322 3637 1995-2006 standard deviation 1995-2006 corr. Skill vs % coverage (non-EC) 06-2 TP 1995-2006 corr. Skill vs % coverage (ALL) 1321 2616
10
4748126 12-month mean skill 1322 3637 1995-2006 standard deviation 1995-2006 corr. Skill vs % coverage (non-EC) 1995-2006 corr. Skill vs % coverage (ALL) 06 1321 -2 2616 TP It was a very good year for both T and P OFF forecasts. OFF followed CON closely for T, mediocre guidance for P. SD is large for T, CON P, implying big swings in skill. There is no relationship between skill and % coverage for non-EC forecasts Area-weighted, all stations skill is more meaningful than non-EC
11
OFFVER CON 01 ASO T 45 31 skill
12
OFFVER CON 01 ASO P -21 32 skill
13
OFFVER CON 01 ASO TP 45 31 skill -21 32
14
OFFVER CON 02 SON T 12 40 skill
15
OFFVER CON 02 SON P skill -23 12
16
OFFVER CON 02 SON TP 12 40 skill -23 12
17
OFFVER CON 03 OND T 55 skill
18
OFFVER CON 03 OND P skill -13 -3
19
OFFVER CON 03 OND TP 55 skill -13 -3
20
OFFVER CON 04 NDJ T 84 85 skill
21
OFFVER CON 04 NDJ P skill 23 -18
22
OFFVER CON 04 NDJ TP 84 85 skill 23 -18
23
OFFVER CON 05 DJF T 77 71 skill
24
OFFVER CON 05 DJF P 77 71 skill -4 -18
25
OFFVER CON 05 DJF TP 77 71 skill -4 -18
26
OFFVER CON 06 JFM T 64 68 skill
27
OFFVER CON 06 JFM P skill 55 8
28
OFFVER CON 06 JFM TP 64 68 skill 55 8
29
OFFVER CON 07 FMA T 31 38 skill
30
OFFVER CON 07 FMA P skill 15 -8
31
OFFVER CON 07 FMA TP 31 38 skill 15 -8
32
OFFVER CON 08 MAM T 38 27 skill
33
OFFVER CON 08 MAM P skill 32 8
34
OFFVER CON 08 MAM TP 38 27 skill 32 8
35
OFFVER CON 09 AMJ T 48 43 skill
36
OFFVER CON 09 AMJ P skill 40 33
37
OFFVER CON 09 AMJ TP 48 43 skill 40 33
38
OFFVER CON 10 MJJ T 39 36 skill
39
OFFVER CON 10 MJJ P skill 23 -6
40
OFFVER CON 10 MJJ TP 39 36 skill 23 -6
41
OFFVER CON 11 JJA T 60 58 skill
42
OFFVER CON 11 JJA P skill -9 6
43
OFFVER CON 11 JJA TP 60 58 skill -9 6
44
OFF VER CON 12 JAS T 12 20 skill
45
OFF VER CON 12 JAS P skill 27 19
46
OFF VER CON 12 JAS TP 12 20 skill 27 19
47
Summary CPC’s objective consolidation (CON) of OCN, CCA, SMLR, CFS improves upon official (OFF), ½-month lead T forecasts (FMA1995-JFM2005) by about 18% and about 130% for P (non-EC). CPC began using CON for T in fall, 2005, and for P in mid- 2006. It was a very good year for both T and P OFF forecasts, with T skill continuously above zero. SD is large for T, CON P, implying big swings in skill. CON was good guidance for T, poor guidance for P for these 12 forecasts. There is no relationship between skill and % coverage for non-EC forecasts. There is a weak skill-%coverage relationship for all stations skill. Subjective deviations from CON P succeeded perhaps because CON predictions cover smaller contiguous areas than OFF. CON predicts much larger area coverage for T than OFF, especially in the eastern U.S. Forecasts appear to change much more than the OBS from lead-to-lead.
48
OFFVER CON 01 ASO TP 45 31 skill -21 32
49
OFFVER CON 02 SON TP 12 40 skill -23 12
50
OFFVER CON 03 OND TP 55 skill -13 -3
51
OFFVER CON 04 NDJ TP 84 85 skill 23 -18
52
OFFVER CON 05 DJF TP 77 71 skill -4 -18
53
OFFVER CON 06 JFM TP 64 68 skill 55 8
54
OFFVER CON 07 FMA TP 31 38 skill 15 -8
55
OFFVER CON 08 MAM TP 38 27 skill 32 8
56
OFFVER CON 09 AMJ TP 48 43 skill 40 33
57
OFFVER CON 10 MJJ TP 39 36 skill 23 -6
58
OFFVER CON 11 JJA TP 60 58 skill -9 6
59
OFF VER CON 12 JAS TP 12 20 skill 27 19
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.