Evaluating a new low-cost radiosonde system
for use in adaptive sounding networks
(and its implications)
Michael Douglas, National Severe Storms Laboratory
Norman, OK
John Mejia, Desert Research Institute, Reno, Nevada
(formerly at OU)
Original Motivation
Much of my research has involved work with developing
countries. These Weather Services usually cannot
support dense radiosonde networks...
How can these countries/regions carry out effective
Numerical Weather Prediction activities?
Very Recent History
NSSL Director’s Discretionary Funds allowed for the
purchase of a new, low-cost (~$12K) radiosonde system
this past year, along with some testing costs.
Tests have been carried out in Albuquerque and Norman.
InterMet 3050 overall system layout
InterMet 3050 radiosonde system
Antenna assembly…
simple vertical, not a parabolic
tracking antenna, which is much
more expensive
iMet-1 radiosonde…
4 “AA” batteries
On-off-frequency
Switch
(info from Company brochure - claims to be “verified”)
Radiosondes ~ $175-$200 each in low quantity
Initial results
from SippicanIMET
comparisons at
Albuquerque
QuickTime™ and a
decompressor
are needed to see this picture.
example of a dual
sounding
on same balloon
R.H. difference
T difference
Average of 7 soundings - Sippican RH is ~10% less; T
difference less than .1˚C. Winds (GPS) identical.
Get pdf of preliminary report from PACS-SONET web site
(Google “PACS SONET”)
Wind agreement ... perfect...sort of...
1-sec positions
Note: too little attention
to wind details...
U-component
What are some implications of low-cost
radiosonde systems?
Can we design a cost-effective adaptive sounding
network?
“Cost effective” implies benefits exceed costs! (Wx
Forecasting is fundamentally an economic activity)
Routine (permanent) sounding site costs
historically have been high...
• Ground station (~$100K)
• Inflation shelter for 300-600 gm balloons (~$100K+)
• Hydrogen generator (mostly foreign sites) ~$80K
• Staffing for 2x daily soundings, high reliability, 7 days/week, 365
days/year (~$50K/year??)
Thus, at least $300K to get started, then significant staff effort to keep
running.
Initial start-up costs limit experimenting with
“sounding network design”
Requirements for inexpensive
adaptive sounding network
• Ground stations must be inexpensive ($10K vrs $100K)
• Network must be adaptive (say ~120 obs’s/year instead
of 730 obs’s/yr)
• Labor must be local, part-time (pay-by-observation),
even at $100 per ob’ this is only ~$15K/year.
There is an upper limit to useful density of
radiosonde observations...
Balloon ascent rate (~5 m/s) requires ~30-60 minutes to
profile troposphere - inadequate for storm-scale
forecasting.
Time changes and practical limits of synchronizing
launches may limit spatial density, perhaps 100 km
spatial separation is a practical lower limit.
Point of diminishing return rapidly approaches as
network density increases - depends on incremental
benefits of add’l fcst skill (rarely quantified).
Concept of adaptive sounding network...
Frequency of adaptive observations
• This is motivated by perceived cost-benefit of
additional forecast skill... How many days justify
additional radiosonde obs’s?
– Must quantify ($$$) impact of incremental forecast
skill... (meteorology is a science, wx forecasting is
an economic activity!)
– Must be a fraction of current 2-daily observations
to justify “adaptiveness” - like 25% or less of “full
time obs’s”
Cost per observation
Radiosonde cost (function of quantity purchased)
~ $150 - $200 in low quantities depending on vendor
Balloons (size dependent ~$25)
Inflation gas (Helium is 3 times H2 cost but safer...)
~ $10 - $30 per balloon
Labor cost is controllable...to a point. Pay by observation is
most economical. Lets assume $100 per observation for
argument.
Can use ~$300 per observation as plausible (USA) cost...
Perhaps most obvious example of an adaptive sounding network:
for improving hurricane track forecasting
Red= routine obs’s, black = adaptive
Cost of this “small” adaptive network…
23 sites across Caribbean Sea region for hurricane
track/intensity forecasting
Set-up cost ~$500K
Annual operation for 60 obs’s/year per site
@$300/ob ~ $500K
How does this compare with other possible hurricane
monitoring and forecast activities? (HRD in 2008 used
~1200 dropsondes and 39 research/operational P-3 flights;
these total ~ $2M+)
... for forecasting over Central and Eastern US
~15 sites
60 obs’s/yr
~
$300K/yr
Possible to “easily” reconfigure network for regional focus or
evaluation of adaptive strategies (research potential)
Scaling things up a bit: Imagine adaptive network consisting of
100 additional sites across western North America (~4x current density)
Set-up cost ~ $2 M, annual cost for 120 obs’s/year
per site @ $30K/site: $3 M How does this cost
(~60 sites shown)
compare with other efforts to improve 12-72 hr fcsts
over the central and eastern US?
Problems with adaptive approach
• Deciding when to make observations
• Maintaining observer proficiency with infrequent obs’s
• Reliable 2-way communications is required
• Initial purchase of more equipment
• Additional gas logistics issues with more sites
• Determining whether extra observations have impact
on objective and subjective forecasts and the value
($$$) of these impacts...
Pluses of adaptive sounding network…
• Focus on critical weather forecasts downstream
• Flexibility in deciding which stations operate on a daily
basis
• Should be great motivator for NWS forecasters - control
over special sounding network (“Democratic” decisionmaking for go-days?)
• Technology straightforward, little R&D required
• Could be implemented quickly…
How to evaluate potential impact of extra data?
When to make observations?
• Must decide on what stations “to activate”. Decision
based on:
– objective guidance ~ where add’l data will have
greatest positive impact on downstream forecast
skill... (must identify “priority regions”)
– Subjective input - which sectors of economy likely to
benefit most from a better forecast...
• Blend of objective and subjective tools likely required
for go/no-go decisions. Perhaps a blend of SOO input
and objective guidance? This area needs work...
Why not just carry out OSSE’s?
OSSE’s do cost $$$ (computer and research time) and are model
and procedure dependent.
OSSE’s cannot evaluate any subjective use of additional soundings
and improved analyses at the WFO’s.
Why has the current NWS raob network had the same density for
the past ~ 50 years? Has it been OSSE-optimized? That is, is the
current radiosonde network adjusted to just reach the point of
diminishing returns? Not likely...
It may actually be less expensive, more convincing, and take
considerably less time to to carry out operational trials than to carry
out an OSSE.
We are not saying...
... That the current radiosonde network should be
replaced by an adaptive network... Daily obs’s are
needed for many purposes...
Rather,
Any adaptive sounding network should complement the
current network and be focused on high impact weather
events (however those are defined!)
This makes it much easier to justify the operational costs
Who should sponsor/organize such an effort?
Ought to be NWS...but.... could easily take a decade to
program funds...
Universities could do it...but which ones...and why should
they?
Many economic sectors might benefit and could easily
fund it... but... why should they if someone else will?
NSF-type basic research activity lacks operational
component to engage NWS...
Summary: this kind of sounding network is made feasible by
introduction of low-cost radiosonde ground stations…
UAV-domain?
Adaptive sounding
domain?
AMDAR-domain?
SUMMARY OF MAIN POINTS
Low initial cost is key to everything!
– With small initial investment there is less pressure to make
routine, unsustainable, radiosonde observations
– little reluctance to establish more stations, since no cost penalty
– Pay-by-observation means low annual labor costs
– Logistics are relatively economical (fewer gas transport issues)
– Wx prediction focus - smaller balloons - troposphere primarily
– But need to quantify the value ($$$) of incremental forecast skill
improvement...
This may be most cost-effective means to increase short range forecast
skill (to 72 hr or thereabouts) in many parts of world - including the USA.