Title: Making Climate Products Digestible
1Making Climate Products Digestible Kelly
Redmond Western Regional Climate Center Desert
Research Institute
31st Climate Diagnostics and Prediction
Workshop Earth System Research Laboratory Boulder
CO, 2006 October 23-27
2Climate questions span a large range of
complexity and purposes (orders of magnitude),
and span a range from past to future. In most
cases, some kind of decision is
involved. Decisions take place in the future,
and involve future conditions. Actual forecasts
pertinent to a specific question would be best,
but most often such forecasts are not available,
or would not be credible. Many requests for
past data or information pertain to an
upcoming decision. Thus past behavior is often
treated as a de facto forecast. Past is
prologue. The fact that something has happened
once conveys considerable credibility.
3A programmers perspective If its easy to
use, it was hard for the programmer. If its
hard to use, it was easy for the
programmer. (from an NPS software
consultant) The engineers who design cars
never seem to have to fix them parental
observation ))_at_!_)!_)!)((
supplemental observation
4Products are much better if users, or user
perspectives, are integrated into design. Better
yet if designer has experience as both a user and
a provider of information. Benevolent Dictator
? vs. Committee ?
5Learning by doing. Start with a best
guess. This can often be pretty good.
Dictated by experience. Listen. Iterate. T
his approach can be slow and halting, but it has
a rich track record.
6(No Transcript)
7Allowing for growth First time users
handholding and help Experienced users
shortcuts, and elaboration and sophistication We
ll designed web pages have both levels available
at once. Fewer web pages to manage. Short
Attention Span Syndrome (Faster, James
Gleick) Even industrious people can be lazy
about certain things Saving labor (minimal
hand movement, clicking and scrolling) Ease and
efficiency Compactness
8CDC Composites page With thanks to Cathy
Smith Many other examples as well
9WRCC Divisional Data plot page
10 2006 Oct 23 1100 PDT
11 think like a mountain (Aldo Leopold, A
Sand County Almanac, 1949) With climate
products think like a user How to do
this??? Know the user. Firsthand is
better, not always practical on an industrial
scale. Secondhand can work, but needs
intermediaries. Provider push vs user
pull Provider has (actually or potentially)
useful information for user User wants certain
information (perhaps needs, but does not
recognize) Role of climate services
provider Help the user to structure their
thinking
12Suppose your own personal resources were on the
line not just an abstraction! What would you
do ? How would you decide ? How confident would
you want, or need, to be ? Is this a voluntary
decision, or a mandatory (imposed) decision
? What makes you trust information ? Track
record is probably most frequently cited as
desired information. How do I know I am not
getting a pig in a poke? (recent
question) What kind of track record information
is most convincing? What kind of track record
information is most relevant (objectively (?)
)? According to a climate scientist
? According to an average user ? Skill measured
how? Overall patterns, events, specific
locations, specific circumstances, new or
emergent phenomena How well can humans evaluate
pattern similarities by eye ? (OLenic talk)
13Precip (Fcst / Obs) Jul-Aug-Sep 2006
Temp (Fcst / Obs) A real track record based on
real forecasts, of which this is just one. But,
it is the most recent.
14Precip (Fcst / Obs) Nov-Dec-Jan 2005-06
Temp (Fcst / Obs) A real track record based
on real forecasts, of which this is just one.
This one is from the previous October.
15Monthly Oct init 700 mb Height Skill
1982-2003 Seasonal
16All areas CFS monthly 2006 Oct 22
With skill mask
17All areas CFS seasonal 2006 Oct 22
With skill mask
18Monthly Oct init Precip Skill 1982-2003
Seasonal
19Monthly Oct init Temp Skill 1982-2003
Seasonal
20The track record of real forecasts rather than
simulated forecasts. Here, CFS. Disadvantage
short record. Many other measures could also
be utilized Correlations Skill scores Percent
right Percent of application-dependent events
captured
21(No Transcript)
22(No Transcript)
23A Noble Experiment The Drought Outlook Most
of the forecast consists of expansion,
contraction, weakening or strengthening of
existing areas, seldom new areas of
formation. This is partly the nature of drought,
and partly the nature of drought forecasters
24An new source of concern What reference period
should be used ?
25(No Transcript)
26(No Transcript)
27Annual Mean Temperatures, 2000-2005. Departures
from 1895-2000 Mean.
Non-standardized. Units Degrees F.
Normalized (standard deviations).
The West dominates recent U.S. warming.
28Summer (Jun-Jul-Aug) Mean Temperatures,
2000-2006. Departures from 1895-2000 Mean.
Non-standardized. Units Degrees F.
Normalized (standard deviations).
The West dominates recent U.S. warming.
29Autumn (Sep-Oct-Nov) Mean Temperatures,
2000-2005. Departures from 1895-2000 Mean.
Non-standardized. Units Degrees F.
Normalized (standard deviations).
The West dominates recent U.S. warming.
30Winter (Dec-Jan-Feb) Mean Temperatures,
1999/2000-2005/2006. Departures from 1895-2000
Mean.
Non-standardized. Units Degrees F.
Normalized (standard deviations).
The West dominates recent U.S. warming.
31Spring (Mar-Apr-May) Mean Temperatures,
2000-2006. Departures from 1895-2000 Mean.
Non-standardized. Units Degrees F.
Normalized (standard deviations).
The West dominates recent U.S. warming.
32(No Transcript)
33(No Transcript)
34Climate events unfold slooooooowwwwwwwly The
relation between verification time, and
preparation time, is different than for daily
weather forecasts. Weather Three work shifts
per day, each 8 hrs long. Four 12-hour forecasts
updated every 6 hrs (4x per day). Approximately
0.75 work shift per forecast update. Climate 90
days is 270 work shifts, or 90 daytime work
shifts. One daytime work shift is 1/30.033
forecast period (monthly), and 1/90.011 daytime
work shift per seasonal forecast. Thus, lots
of time to watch the forecast verify. Or, not.
Perhaps, more temptation to update. Also,
time enough to forget forecast. Easier to
intermingle observations and forecasts, the
longer the duration and lead time of the
phenomena being forecast. Aspects of the
forecast system can evolve during as the
verifying period is unfolding. Forecasts can be
nudged, modified, or otherwise re-interpreted as
they play out. This is more likely to occur the
longer the interval covered monthly, seasonal,
annual, decadal ENSO discussion Forecasts
less tentative when ENSO event appears to have
started. Are these really forecasts? More like
observacasts
35What matters to users can be, and often is, very
application-specific.
36A small difference in circumstances or context,
and it can be a whole new problem.
Prof. Julien Clinton Sprott, UW Madison Physics
37Provider and user relationship. A lock and key
situation. Like biochemistry.
38Steroid and nuclear hormones Human estrogen
receptorligand binding domain hER-alpha-LBD Tane
nbaum et al PNAS, 1998 Feb 18 Stereo Image Fix
on estradiol carbon black oxygen red
39Evaluation Not of forecasts themselves, but of
their use, and of system effectiveness Done
properly, this takes lots of attention, and takes
skill Attention and skill imply need for
resources, perhaps significant Process has
often expended most of its energy and budget by
this point This part is often viewed as
extra rather than integral Resource
decision How to allocate resources to the
forecast system Basic understanding (need for
better forecasts) Better packaging and
presentation of forecast information Better
understanding of the use and decision
environment How effectively is this knowledge
plowed back into the forecast system?
40Local applicability. User orientation. My back
yard.
41Whale Point (600 ft) and Highlands Peak (2500
ft), Big Sur. 2 miles apart.
Whale Point 600 ft
Highlands Peak 2500 ft
42Whale Point, Big Sur, 600 ft, 10-min Temperature,
July 2006 Heat Wave.
43Highlands Pk, Big Sur, 2470 ft, 10-min
Temperature, July 2006 Heat Wave.
44Audiences for climate and climate forecast
products General - casual users, minor economic
decisions Interested - modest economic
decisions (eg, whitewater guide
company) Dependent - major decisions (eg,
utilities, ag commodities, derivatives)
45Water Year Temperature Departure 700 mb (10,000
ft) 1 Oct 2005 Through 31 Mar 2006
46Warm Season Temperature Departure 700 mb
(10,000 ft) 1 Apr 2006 Through 30 Sep 2006
47Water Year Temperature Departure 700 mb (10,000
ft) 1 Oct 2005 Through 30 Sep 2006
48Climate Test Bed offers many opportunities Hydro
logic forecasts could potentially benefit,
perhaps considerably NWS and NRCS, and many
water managers. Other resource
managers Providing model output that can link,
or pipe directly, into applications In an
atmosphere of limited total resources, the
ultimate importance of developing a supportive
constituency for maintaining the research and
operational infrastructure should not be
underestimated. Along with substance (assumed to
be a given (!?!) ), the associated
marketing salesmanship presentation all take
unexpectedly large amounts of time and
resources. A willingness to commit to this can
mean the difference between success and failure.
49What makes for a better dining experience? Taste
or presentation? What makes for a better
forecast use experience? Accuracy or
presentation?
50Whats for supper?