The Risks Digest

Forum on Risks to the Public in Computers and Related Systems

ACM Committee on Computers and Public Policy, Peter G. Neumann, moderator

Volume 5 Issue 50

Tuesday, 27 October 1987

Contents

o Weather
Willis Ware
Geoff Lane
Eugene Miya
o Civil disobedience
David Redell
o Reported Japanese Autopilot Problems
Nancy Leveson
o Amusing bug: Business Week Computer (F)ails
GW Ryan
o Television series "Welcome to my world"
Clive Feather
o Info on RISKS (comp.risks)

Weather (Re: RISKS 5.45)

<willis@rand-unix.ARPA>
Tue, 27 Oct 87 10:46:00 PST
Here's a contribution to the big blow in England.  I was scheduled to
land at Heathrow the morning of the 16th and we sat at Dublin for 6 hours
waiting for the storm to clear. The weather at Dublin was clear and calm.

My comments are based on my participating in an NRC study for NOAA a few
years back; it concerned future needs for computer support to weather
models.  I'm giving you this from memory but I'm sure that the general
facts are correct.

Within the last 10-15 years or so, there were two examples of severe
weather in this country that forecasters missed: the Johntown flood in
western Pennsylvania from unprecedented rains, and the big blizzard in the
Northeast, which interestingly had been forecast by a smallish private
weather service used by some of the trucking companies.  The explanation
from the weather types was that each of them had been a mesoscale
phenomenon -- too small to be included or visible in the global weather
models but too large to fall within the capabilities of local forecasters.

The mesoscale effects seemingly had been overlooked by the WX types and
had just been appreciated at the time we were doing our study.

The large 5-day weather models which typically run once or twice each day
on some huge computer are upper atmosphere models.  They work at a
prescribed pressure profile which it seems to me is 200 millibars,
corresponding roughly to jet altitudes.  As such they predict the large
effects in weather patterns.

The local forecaster uses inputs from satellites, from the large models,
and from local and nearby data sources.  But his ability to predict is
limited to an area perhaps 100 miles or so across.

The mesoscale phenomena are a few-to-several hundred miles in extent and
are lower atmosphere behavior.  The WX types have two problems in getting
models to handle them: first, building the models themselves which must
include things that the upper atmosphere models can ignore (e.g.,
topographic effects, ground-air heat transfer, effects of large lakes
and/or rivers), and getting the data to drive the models.  The models will
have to have a very fine mesh to handle the detail needed, and
consequently the data sources will also have to be fine grained.

As of 7-10 years ago, the work on the models was just starting.  I do not
know the present status.

I suspect that the London and southern England storm was mesoscale
in size.  It definitely was neither a tornado nor a hurricane although both
labels were used by the media.  It clobbered an area a few hundred miles
across but Ireland was untouched.  In addition since the English weather
often comes from the west, there is a lack of weather observations to
drive any models or forecasts.  Aircraft are too high for the observations
needed by mesoscale models, and surface vessels do not normally send up WX
balloons nor take the usual WX measurements.

The damage to London and southern England was stupendous.  Some of the
parks (e.g., Hyde Park) were closed because of the amount of downed
trees, broken branches and trash.  Flooding was extensive and one train
ran into a river when the undermined bridge gave way.  Winds in London
were the highest ever recorded.  The problems continued through the
following week.  Reporting on the telly was confined primarily to
damage; there were only a few comments about the failure to forecast but
no recriminations or blaming.  No mention of mesoscale effects either.

Relevance of all this to RISKS:  you never can be sure how well a model
represents the real world that it purports to describe and mimic.


Weather Prediction in UK

"ZZASSGL" <ZZASSGL@CMS.UMRCC.AC.UK>
Tue, 27 Oct 87 11:29:35 GMT
 1. The UK met office uses a Cyber 205 to prepare its forecasts - It also
    predicted bad weather over north France.  The problem was that the
    weather had not read the forecast and moved north very rapidly after the
    last forecast run of the day.
 2. Without exception the British press has blamed the "computer" for the
    lack of warning of bad weather.  Now we all know that unless there was
    some sort of hardware problem(which there wasn't) it is the software
    that may, and only may, be at fault.  Is this confusion a "risk"?
    It is certainly a misconception and if future funding of the UK Met
    office is cut as a result is will surely be a risk to all the people
    who depend on the very good work that is done there.

  Geoff. Lane.
  University of Manchester Regional Computer Centre


Weather and terrorism (separate)

<eugene@ames-nas.arpa>
27 Oct 87 11:12:48 PST (Tue)
>  Cyber 205 supercomputer as part of its investigation following last week's
>  hurricane-force winds.  The office has been criticised for failing to
>  predict the speed and path of the storm...
>
>My understanding is the main reasons for lack of success in predicting
>the winds were a) lack of data (the storm came from over the sea --
>fairly normal in Britain! -- where there are few weather stations) and
>b) lack of computing power (and lack of good algorithms?).

Please don't completely equate computing power to the problem of these
models.  Also don't blame machine manufacturer (we have both a 205 and
an X-MP/48).  There was a recent Scientific American article on the problems
of weather models (cover was a grid), but a better article appears:

%A Joseph J. Tribbia
%A Richard A. Anthes
%Z NCAR
%T Scientific Basis of Modern Weather Prediction
%J Science
%V 237
%N 4814
%D 31 July 1987
%P 493-499

There is a specific section on why forecasts are inaccurate.  It's not
just algorithms.  The equations of weather are basically good equations.
Read the article, its more detailed than I wish to summarize here.

>Subject: Terrorism (Re: RISKS-5.45)
>
>The computer center at U.C. Santa Barbara was taken over by protesters in
>the spring of '75. Although the computer room was secured behind locked 
>. . .
>One operator inside shut the machine down immediately. Then the protesters
>ushered (all?) the operators out, and took over the entire building - taping
>printouts over all the windows and doors. They threatened to destroy the
>computer if their demands (more money for radical leftist groups) weren't met.
>
>Bill Swan   sigma!bill

Hi Bill--
Yes, Don Davis told us about when he was kicked out.  I did not recall
it was for "radical leftist groups," unless you count the Chicano Studies
program as such.  We have to be careful in distinguishing political judgments.
(Are computer people naturally conservative?)

My suggestion is that if the original requester is interested
in acts of terrorism against computers, contact Computerworld
or maybe Donn Parker (SRI) might have a list.  CW did publish
a list of acts including the bombing (and death) at Wisconsin in
the 1970s, and several bombings in Europe.  Happens ALL the time.

--eugene miya


Civil disobedience

David Redell <redell@src.dec.com>
Tue, 27 Oct 87 11:18:05 PST
Re: Jim Anderson's complaints about the term "Civil Disobedience"

One may admire or despise some or all acts or practitioners of civil
disobedience, but it's simply incorrect to claim that the term stems
from efforts of contemporary newspapers, radio and TV to distort reality
with euphemisms implying polite behavior.  The term originated in the
mid-19th Century and has nothing to do with "civil" as in "civility".
It is based on "civil" as in "civil rights" or "civil servant" -- that is

 civil (adj) 1 a: of or relating to citizens;  b: of or relating to
       the state or its citizenry;... 5: of, relating to, or involving
       the general public...

Dave Redell


Reported Japanese Autopilot Problems

Nancy Leveson <nancy%murphy.uci.edu@ROME.UCI.EDU>
Tue, 27 Oct 87 19:53:14 -0800
Last week somebody told me that Dan Rather had reported that a Japanese
plane, the MU-2, has had 100 crashes which have now been traced to a
computer problem.  Supposedly, the computerized autopilot will, under 
certain conditions, not let the pilot have control back.  I did not hear 
this myself and since nobody has reported it in Risks, I have a suspicion 
that the story is not correct.  Did anyone hear the Dan Rather telecast?

Nancy


Amusing bug: Business Week Computer (F)ails

GW Ryan <gwr@garage.nj.att.com>
24 Oct 87 13:03:31 EDT (Sat)
The following paragraph is taken (without permission) from p. 14 of the
November 2 "Business Week". I thought it was an amusing little bug report.
I tried to imagine how it could happen... you'd think we'd have alphabetical
order down pat by now!

    Some copies of our special bonus issue,
    The Corporate Elite: Chief Executives of
    the Business Week Top 1000, contain
    mistakes in the alphabetical index of
    chief executives on pages 341-350 and in
    the guide to how to read the CEO profiles
    on page 350. Because of a computer problem,
    certain letter combinations with "f" were
    omitted from company names and the guide.

jerry ryan (allegra!cord!gwr)  Bell Labs, Liberty Corner NJ


Television series "Welcome to my world"

Clive Feather <mcvax!root.co.uk!cdwf@uunet.UU.NET>
27 Oct 87 13:25:43 GMT
There is a new series on BBC1 television called "Welcome to my World"
which raises some of the topics we are / ought to be discussing. Anyone
out there seen it and got any comments ?

It's on BBC1, Sundays, at 2305 until 2335.

Clive D.W. Feather     +44 1 606 7799 x 235

Please report problems with the web pages to the maintainer