Conclusions:

1

TRENDS REVISITED
C P C
Huug van den Dool Climate Prediction
Center NCEP/NWS/NOAA CDPW Reno October, 22,
2003
2
(Trends not a straight line, LF ups and
downs.) Trends Diagnostics OR rather How to
deal with trends in a real time forecast
setting.? How to improve Trend forecast
tools? How to physically explain Trends?
3
Intro IWhere does 2003 stand over the US
trendwise???Is it another warm year??
4
(No Transcript)
5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
Sofar, DJF thru JAS 2003B N A at 102 US
locations23 37 41
15
(No Transcript)
16
Intro II The Great Performance Measure (PM)
17
The PM (blue line)Retro-active OCN (pink line)
18
What is OCN? (Optimal Climate Normals).
Essentially a forecast in which one persists the
average of the anomalies observed in the same
named season over the last K years.Example of
OCN for JFM 2004 The average anomaly for JFM
over 1994-2003 (K10 T no space averaging)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
(No Transcript)
34
What might explain the skill of such simple
forecasts?
35
Table 1. Weights (X100) of the constructed
analogue on global SST with data thru Feb 2001.
An example.Yr(j) Wt(aj) Yr Wt Yr Wt Yr Wt 56
5 67 -8 78 -1 89 8 57 2 68 -5 79 -3 90 13 58 -
4 69 -3 80 -4 91 7 59 -7 70 -5 81 -8 92 11 60 -
3 71 -2 82 1 93 -6 61 1 72 6 83 0 94
2 62 -1 73 1 84 -1 95 7 63 -1 74 1 85 3 96
2 64 -3 75 2 86 12 97 14 65 -8 76 5 87 5 98
2 66 -5 77 1 88 0 99 26sum -24
sum -7 sum 4 sum
86----------------------------------------------
------------------------------------------
CA-SST(s) 3 aj SST(s,j), where aj is given as
in the Table. j
36
Table 1. Weights (X100) of the constructed
analogue on global SST with data thru Feb 2001.
An example.Yr(j) Wt(aj) Yr Wt Yr Wt Yr Wt 56
5 67 -8 78 -1 89 8 57 2 68 -5 79 -3 90 13 58 -
4 69 -3 80 -4 91 7 59 -7 70 -5 81 -8 92 11 60 -
3 71 -2 82 1 93 -6 61 1 72 6 83 0 94
2 62 -1 73 1 84 -1 95 7 63 -1 74 1 85 3 96
2 64 -3 75 2 86 12 97 14 65 -8 76 5 87 5 98
2 66 -5 77 1 88 0 99 26sum -24
sum -7 sum 4 sum
86----------------------------------------------
------------------------------------------

• CA-SST(s) 3 aj SST(s,j), where aj is given as
  in the Table.
• j
• OCN-SST(s) 3 aj SST(s,j), where aj0 (1/K)
  for older(recent) j.
• j

37
Trends in lower boundary conditions? global SST
38
EOFs for JAS global SST 1948-2003
39
(No Transcript)
40
Trends in lower boundary conditions? global Soil
Moisture
41
(No Transcript)
42
(No Transcript)
43
Is the inter-decadal component of climate
variation accurately known ???Probably not.
Nature provides just one realization.
44
Evidence 1) 70 of skill of OCN over US can be
obtained by replacing the K year average of
T(s,m) by the annual mean spatial mean value,
i.e. we can ignore some, if not most, of the
spatial and seasonal dependence.
45
2) We can try to fight noise by a) determining
optimal K in EOF space ( Peitao Peng), i.e. build
a smooth spatial dependence b) We could generate
more data with a credible model
46
Courtesy Marty Hoerling
47
(No Transcript)
48
DJF US Nationwide (NCDC)
49
JJA US Nationwide (NCDC)
50
East Anglia Climate Unit
51
(No Transcript)
52
Closing comments

• LF (inter)decadal variability (trends) are
  important for seasonal forecasts, even at short
  leads.
• Are there any situations we can identify a-priori
  where trend tools should be played down ?
• Trends over the US appear related to trends in
  the NH, even worldwide

53
Do trends have spatial patterns, and seasonality
? (probably yes)Can we extract such patterns
(and seasonality) from limited observations ?
(probably, barely)So either we fight noise by
EOF or other spatial smoothing, ORWe generate
artificial data by running a trustworthy GCM
54
Explaining trends may require understanding
global changeShall we start forecasts for K
year averages ?? (Regress from the global mean ?)