Archive AGMODELS-L: file log9402, part 1/1, size 152859 bytes:

------------------------------ Cut here ------------------------------

Hello: Getting the specifices out of the way:

Robert P. Eastburn, CPSC
Eastburn JOAT (Jack of All Trades <BG>)
1436 Clayton-Delaney Road
Clayton, DE 19938
302-653-0769 (tel/fax)

Areas of interest include soil genesis/classification and soil and
water. Did some water irrigation scheduling on corn, soybeans, and
vegetables. Also, a small amount on modeling concerning broiler
production.

OB-? What is the current state of seasonal high water table
modeling and can such be used to address the question of soil
saturation regarding jurisdictional wetland determination ?

[Robert P. Eastburn (himself) My opinions are my company's -- ]
[Clayton, DE, USA I own it. That means I'm ]
[Eastburn @ delphi.com.us last out, last paid and sweep up!]

I am developing and using models for the assessment of
exposure to pesticides via water supplies and other
environmental media. The area of most interest and promise
for me right now is an attempt to link such models with very
easy-to-use GIS systems such as ArcView2 and the new Excel
mapping tools.

David Gustafson voice: 314/537-7194
Monsanto Ag Group -- BB5F fax: 314/537-6134
700 Chesterfield Vill Pkwy N
St. Louis, MO 63198 digust@ccmail.monsanto.com

I am a graduate student at NCSU co-majoring in Crop Science and
Biomathematics. My research deals with modeling the half-life of
herbicides, in order to take into account the variation that the half-
life values show over different soil pH's, temps., %OM, etc. This
model will then be put into another model in the form of a
computer program which will predict the potential
of a herbicide/soil combination to be a groundwater
contamination problem.

My interests include the above research, as well as
herbicide/pesticide resistance and soil seedbank dynamics.

Cassie Webster
webster@cropserv1.cropsci.ncsu.edu
(919) 515-5817
fax: (919) 515-7959

TRAINING PROGRAM ON COMPUTER SIMULATION OF CROP GROWTH
AND MANAGEMENT RESPONSES

We think that you and your readers would be interested to know that
the International Training Division at the University of Florida in
conjection with the International Fertilizer Development Center (IFDC),
the International Benchmark Sites Network for Agrotechnology
Transfer (IBSNAT) Project, and the International Consortium for the
Application of Systems Approaches to Agriculture (ICASA), is preparing
to offer a workshop on Computer Simulations of Crop Growth and
Management Responses from Aug. 8-19, 1994 on the campus of
the University of Florida in Gainesville.

This 2-week course, jointly conducted by the University of Florida,
IFDC, IBSNAT and ICASA emphasizes how comprehensive simulation
models of crop growth, soil water and nutrient dynamics are
developed and can be applied to cropping systems in various regions
of the world. Extensive computer time will be devoted to the use
of the new version 3 CERES models (maize, wheat, rice, sorgum and cassava)
and the new CROPGRO models (peanut, soybean and dry bean) for crop growth,
development and response to water and nutrients. Participants will receive
copies of all models as well as the newly released Decision Support System
for Agrotechnology Transfer Version 3.0 software package which supports
the application of the new crop simulation models.

The fee for the course is US $ 1400.
For more information or to register for the course, please contact:

Ms. Lisette Staal PHONE:904-392-3166
University of Florida FAX:904-392-3165
International Training Division, IFAS BITNET:ITD@IFASGNV
P.O. Box 110480 INTERNET:ITD@GNV.IFAS.UFL.EDU
Gainesville, FL 32611-0480 TELEX:568757 CABLE:CENTROP
-----------------------------------------

Thank you very much for your time and support, please contact us if
you would like information on other related training programs at
the University of Florida.

DR. GERRIT HOOGENBOOM | DEPT. OF BIOL. AND AGRIC. ENG.
ASSISTANT PROFESSOR | GEORGIA STATION
INTERNET : GHOOGEN @ | THE UNIVERSITY OF GEORGIA
GAES.GRIFFIN.PEACHNET.EDU | GRIFFIN, GEORGIA 30223-1797 USA
DIALCOM: 157:AGS634 PHONE: (404)-228-7216
FAX: (404) 228-7218

As per Jpier's request, this is me.
I'm Michael Volkmer, a graduate student at the University of
Nebraska-Lincoln. I'm doing my thesis research, funded by the Nebraska
MSEA project, in the area of predicting the effect of tillage practices
on nitrate leaching in irrigated corn production using finite element
modelling. Currently I'm using SWMS_2d for modelling purposes. My one
burning issue? Is there a UNIX based grid generator available? To do
meaningfull modelling of the rootzone, the finite element grid must
consist of up to 1000 (or more) nodes. This fries my DOS based
generator. Any suggestions?

Dear Colleagues,

my subject is radiocarbon and stable isotopes in soil organic matter
studies. In our lab we have done a lot of analyses and sampled a neat
set of data for a soil carbon model.

Recently we received a research grant to engage a PhD student to
work on this model. The following job description might be interesting
for subscribers of AGMODELS-L in Germany or other German speaking
colleagues.

-----------------------------------------------------------------------

UNIVERSITAET HAMBURG
Fachbereich Geowissenschaften
Institut fuer Bodenkunde

Ab 1.3.94 ist die Stelle

einer Wissenschaftlichen Mitarbeiterin/
eines Wissenschaftlichen Mitarbeiters

der Verguetungsgruppe IIa BAT zu besetzen. Das Vertragsverhaeltnis
soll gemaess Par. 24 Abs. 3 Hamburgisches Hochschulgesetz* in Verbindung
mit Par. 57b Hochschulrahmengesetz sowie der Verwaltungsanordnung fuer
die Beschaeftigung von Wissenschaftlichen Mitarbeitern auf 34 Monate
befristet werden.

Aufgabengebiet:

Im Rahmen des BMFT Klimaschwerpunkt-Verbundvorhabens "Spurenstoff-
kreislaeufe" soll ein Simulationsmodell der Dynamik des organischen
Kohlenstoffs des Bodens entwickelt werden. Das Modell soll insbeson-
dere nach Tiefenschichten differenzieren und delta-13C-Wert und Radiokoh-
lenstoffalter beruecksichtigen. Produktion, Emission und Verbrauch
der Spurengase CH4 und CO2 in Abhaengigkeit der Bodeneigenschaften,
der klimatischen Bedingungen und der Landnutzung sollen dargestellt
werden koennen. Das Modell soll als Modul moeglichst rechnerunabhaengig
programmiert werden. Es soll mit Schnittstellen, ausser fuer benutzer-
freundliche manuelle Dateneingabe sowie den Import aus anderen
Datenbanken, vor allem fuer ein globales terrestrisches Biosphaeren-
modell und daran gekoppelte Klimamodelle versehen werden.

Ausserhalb der Dienstaufgaben besteht Gelegenheit zur wissenschaftli-
chen Weiterbildung, insbesondere zur Anfertigung einer Dissertation.

Einstellungsvoraussetzungen:

Abschluss eines den Aufgaben entsprechenden Hochschulstudiums.

Schwerbehinderte haben Vorrang vor gesetzlich nicht bevorrechtigten
Bewerberinnen/Bewerbern gleicher Eignung, Befaehigung und fachlicher
Leistungen.

Die Universitaet strebt eine Erhoehung des Anteils von Frauen am wis-
senschaftlichen Personal an und fordert deshalb qualifizierte Frauen
nachdruecklich auf, sich zu bewerben.

Bewerbungen mit den ueblichen Unterlagen werden bis zum 20.2.94
erbeten an

Universitaet Hamburg
Institut fuer Bodenkunde
Dr. Peter Becker-Heidmann
Allende-Platz 2
D-21406 Hamburg

* "... Wissenschaftliche Mitarbeiter, die die Promotion oder eine vergleichbare
Qualifikation anstreben, werden in der Regel mit der Haelfte der regelmaessigen
Arbeitszeit des oeffentlichen Dienstes beschaeftigt. ..."

-------------------------------------------------------------------------------

Peter Becker-Heidmann PBeckerH@rrz.uni-hamburg.de

: : /YY\ : : : Phone: +49 40 4123 2003
: : (YYYY). : : Fax: +49 40 4123 2024
: : . \YY/. : :
: : .||. : : : Institut fuer Bodenkunde
_:__:__:_||__:__:__:_ Allende-Platz 2
SOILSOILSOILSOILSOILS D-20146 Hamburg, FRG

Hello there.

I came across a watershed scale nonpoint source (NPS) model for water, chemical and sediment routing in my routine browsing in water resources abstract, AGRICOLA, and Compendex. The name of this NPS model is "SWAM" ( small watershed model). The publications cited from the database search were USD-ARS report published in 1990 , and a proceedings on interagency workshop on sedimentation in 1986.a The name of the authors are D. DeCoursey and Carlos Alonso of USDA-ARS.
I would like know if there are other publications, such as user manuals and journal papers that describe the model in detail. I tried to get hold of these report
s thro' interlibrary loan but to no avail so far.

I am a civil eng. graduate student in the Univ. of Connecticut. My area of interest is in NPS modeling in field and watershed scales.

Thank you for your help.

Hock Heng
U-37
Dept of Civil Engineering
Univ. Connecticut
Storrs, CT 06269

hengh@eng2.uconn.edu

On Wed, 2 Feb 1994 19:49:08 -0600 Cassie Webster said:
>
You must be working for Jerry Weber--you have my condolences and tell
him hello for me.
Don Wauchope
> I am a graduate student at NCSU co-majoring in Crop Science and
>Biomathematics. My research deals with modeling the half-life of
>herbicides, in order to take into account the variation that the half-
>life values show over different soil pH's, temps., %OM, etc. This
>model will then be put into another model in the form of a
>computer program which will predict the potential
>of a herbicide/soil combination to be a groundwater
>contamination problem.
>
> My interests include the above research, as well as
>herbicide/pesticide resistance and soil seedbank dynamics.
>
>
> Cassie Webster
> webster@cropserv1.cropsci.ncsu.edu
> (919) 515-5817
> fax: (919) 515-7959
>
>

>From Michael Volkmer:
>
> ... Is there a UNIX based grid generator available? To do
> meaningfull modelling of the rootzone, the finite element grid must
> consist of up to 1000 (or more) nodes. This fries my DOS based
> generator. Any suggestions?
>
Commercial packages including Ideas & Ansys have good mesh generators.
They run on a variety of different machines. Check with the structural
or mechanical engr groups.

--
R.S. Gates, Ph.D. P.E. gates@aen.uky.edu
Associate Professor ph (606) 257-3000 x 208
208 Agricultural Engineering Building fax (606) 257-5671
University of Kentucky
Lexington, KY 40546-0276
USA

>Hi Michael, I use ansys to do finite element modelling here
at University of Monitoba. If you want more details pls. feel
free to communicate on the address below so that we don't have
to enter into hectic discussion just we had on the network regarding
DSS<>KB<>ES.
Vinod
kalband@ccu.Umanitoba.ca
> As per Jpier's request, this is me.
> I'm Michael Volkmer, a graduate student at the University of
> Nebraska-Lincoln. I'm doing my thesis research, funded by the Nebraska
> MSEA project, in the area of predicting the effect of tillage practices
> on nitrate leaching in irrigated corn production using finite element
> modelling. Currently I'm using SWMS_2d for modelling purposes. My one
> burning issue? Is there a UNIX based grid generator available? To do
> meaningfull modelling of the rootzone, the finite element grid must
> consist of up to 1000 (or more) nodes. This fries my DOS based
> generator. Any suggestions?
>

In regards to information on the SWAM model, with authors DeCoursey and Alonso.
The last address/phone listing I have for Alonso and DeCoursey are:

USDA-Agricultural Research Service
Great Plains Systems Research
P.O. Box E
Fort Collins, CO 80522

Phone: (303)490-8300
FAX: (303)490-8310

Good luck.

Dennis Flanagan

My name is Kees Plaizier. I am working at the Department of
Animal and Poultry Science of the University of Guelph, Canada. I
have very recently started a project on the development of a
stochastic model simulating the feed intake, milk
production, reproduction and profitability of dairy herds. The
objective of this project is to relate reproductive performance to
economic and biological efficiencies.
As I am very new in the modelling business, I would much
appreciate suggestions and\or information that you might have.

Regards,
J.C.B. Plaizier
Department of Animal and Poultry Science
University of Guelph
Guelph, Ontario
Canada N1G 2W1
tel: 519-8244415
fax: 519-7670573

> In regards to information on the SWAM model, with authors DeCoursey and Alonso.
> The last address/phone listing I have for Alonso and DeCoursey are:
>
>
> USDA-Agricultural Research Service
> Great Plains Systems Research
> P.O. Box E
> Fort Collins, CO 80522
>
> Phone: (303)490-8300
> FAX: (303)490-8310
>
> Dennis Flanagan
>
My colleague, Mike Volkmer was interested in contacting Carlos
Alonso also and he is no longer at GPSR. He took a job at
Mississippi (I can't tell you State or University) according to
Ken Rojas at GPSR. I suppose you could call GPSR to get the
forwarding address. Don DeCoursey can be found at TERRA at Ft.
Collins, CO.

Jerome Pier
jp@unl.edu

unsubscribe agmodels-mg

>
> > In regards to information on the SWAM model, with authors DeCoursey and Alonso.
> > The last address/phone listing I have for Alonso and DeCoursey are:
> >
> >
> > USDA-Agricultural Research Service
> > Great Plains Systems Research
> > P.O. Box E
> > Fort Collins, CO 80522
> >
> > Phone: (303)490-8300
> > FAX: (303)490-8310
> >
> > Dennis Flanagan
> >
> My colleague, Mike Volkmer was interested in contacting Carlos
> Alonso also and he is no longer at GPSR. He took a job at
> Mississippi (I can't tell you State or University) according to
> Ken Rojas at GPSR. I suppose you could call GPSR to get the
> forwarding address. Don DeCoursey can be found at TERRA at Ft.
> Collins, CO.
>
> Jerome Pier
> jp@unl.edu
>
>

Hello -

My name is Mike Dwyer, and I am a software developer (systems analyst) with
ADC, a contractor to the Soil Conservation Service (located in Fort Collins).
My project is the implementation of the Revised Universal Soil Loss Equation
(RUSLE) model for the SCS field offices.

>From our perspective, we are not developing a model but putting a new front-
end on the model developed by ARS, Purdue, etc. and released by the SWCS.
The project also involves the integration of "our" application with other
datasets in a single field office database (using Informix/SQL). The data
used in our application is perhaps a bit more restricted than what could be
enterable in the SWCS software, in order to keep field office users within
the bounds of agency policy and accepted agronomics.

This gives me an interest in some of the topics discussed in months past:

I can agree with Georg Hoermann, 7 Jan, about user interfaces discouraging
use of some models -

> My interest in user interfaces comes from my teaching experience: I
> think the *real* reason why models are not applied is the interface and
> not the contents. There are nearly thousands of hydrologic models but
> none of them can be used by our biologists. Students can work with a
> graphics package or a spreadsheet within a week, but if it comes to
> modelling they have to deal with really awkward FORTRAN formatting

RUSLE doesn't force use of FORTRAN, but the familiarity and consistency of
a common interface across applications is beneficial to the user. A related
issue is the development of modelling/simulation software such that the
computational engines can run "behind the scenes" as it were to various user
interfaces designed for different environments.

Ken Rojas said (December) -

> I could go on forever on this topic, it really controls model development
> processes and eventual customer satisfaction. One can not just create a
> model with basic features and expect users to be satified, users are
> getting too sophisticated and want the same look and feel as everything
> else they use. Is this a reasonable request to put on the scientist
> developing the model?

There was also the discussion of user involvement in development of the models
themselves. Would Greg McIsaac agree that the users' biggest interest might
be in the interface to a model more than the science behind it?

Perhaps the "hottest" issue for us right now is the sharing of data that
drives several models or evaluation tools. Climate data is one such area.
Crop charateristics is another. At least the climate data seems to break
down in terms of frequency of measurement and normalized vs observed data.
Plant data, however, seems to vary with each model. The attributes that
are significant for one are irrelevant to the other, or each research
project defines the variations a little differently. Furthermore, the
characteristics that identify the differentiation also vary from one crop
to the next (corn by days to maturity, alfalfa by year of rotation or by
interseeding, e.g.).

This latter is part of the data modeler's problem, of course, but it may
be helpful for developers of a _single_ model to consider that their tool
may somewhere or someday be integrated into a multi-tool environment.

Thought this might interest some on soils-l or agmodels-l:
Sorry if you have already seen this...

Jerome Pier
jp@unl.edu

Forwarded message:
> From SUSTAG-L@WSUVM1.CSC.WSU.EDU Fri Feb 11 21:36:02 1994
> Message-Id: <199402120333.AA18500@crcnis1.unl.edu>
> Date: Fri, 11 Feb 1994 19:33:34 -0800
> Reply-To: Discussions about Sustainable Agriculture <SUSTAG-L@WSUVM1.CSC.WSU.EDU>
> Sender: Discussions about Sustainable Agriculture <SUSTAG-L@WSUVM1.CSC.WSU.EDU>
> X-Ph: V3.7@crcnis1
> From: "Tom Hodges (moderated newsgroup)" <sustag@BETA.TRICITY.WSU.EDU>
> Subject: AGRONOMIST POSITION ANNOUNCEMENT (fwd)
> X-To: Principles of Sustainable Agriculture
> <sustag-l@wsuvm1.csc.wsu.edu>
> To: Multiple recipients of list SUSTAG-L <SUSTAG-L@WSUVM1.CSC.WSU.EDU>
>
> ---------- Forwarded message ----------
> Date: Fri, 11 Feb 94 09:40:51 CST
> From: J.M. Moynihan <moyni001@maroon.tc.umn.edu>
> To: sustag@beta.tricity.wsu.edu
> Subject: AGRONOMIST POSITION ANNOUNCEMENT
>
>
> Tom: Please post this on the network. I've sent it to sustag-mg, too,
> so many people may receive it twice, but better that than some deserving
> agronomist goes jobless...
>
> Thanks muchly,
>
> Meg Moynihan
> Research Assistant
> Dept. of Agronomy & Plant Genetics
> Univ. of Minnesota
> moyni001@maroon.tc.umn.edu
> *************************************************************************
>
> Please send all inquiries to Dr. Craig Sheaffer at the address below.
> Thank you.
>
> POSITION DESCRIPTION
>
> Department of Agronomy and Plant Genetics, Southwest Experiment Station,
> and College of Agriculture, University of Minnesota
>
>
> TITLE: Cropping Systems Agronomist
>
> RANK: Assistant Professor
>
> BACKGROUND:
>
> This is a twelve-month, tenure track, faculty position located at the
> Southwest Experiment
> Station (SWES) in Lamberton, MN with tenure held in the Department of
> Agronomy and Plant
> Genetics, Twin Cities Campus, University of Minnesota. The appointee will
> be responsible to
> the Head of the SWES and the Head of the Department of Agronomy and Plant
> Genetics. We
> prefer employment to begin not later than September 1, 1994.
>
> The Southwest Experiment Station is located 150 miles southwest of
> Minneapolis/St. Paul in a
> major corn and soybean producing region. Lamberton is a rural community
> with a population
> of about 1000. The Southwest Experiment Station consists of 668 acres and
> is one of five
> branch experiment stations in the Minnesota Agricultural Experiment
> Station system. Research
> at the Station is primarily concerned with crop production as it pertains
> to the climate and soils
> of southwest Minnesota. The adjacent 160 acre Koch Farm provides a unique
> opportunity for
> research on sustainable cropping systems. The Cropping Systems Agronomist
> will work with
> a soil scientist, a weed scientist, and a support staff of 15 full-time
> personnel located at the
> Southwest Experiment Station. The facilities and location provide
> interesting research
> opportunities for multidisciplinary studies involving several departments
> on the St. Paul
> Campus, other branch stations, and federal and state agencies. The
> Station is also a center of
> extension education activities in southwestern Minnesota.
>
> RESPONSIBILITIES:
>
> RESEARCH: The Cropping Systems Agronomist will be responsible for
> providing project
> leadership and for participating in cooperative applied research.
> Collaborative research with
> other agronomists and scientists from other disciplines will be strongly
> encouraged. Research
> goals must be coordinated with the needs and directions of the Southwest
> Experiment Station
> and the Department of Agronomy and Plant Genetics. Appointment to the
> Graduate School
> faculty and advising of graduate students is expected. Base funding for
> research will be
> provided through the Minnesota Agricultural Experiment Station with an
> active search for
> supplemental funding expected.
>
> Priority is for multidisciplinary research leading to the development of
> sustainable cropping
> systems for corn, soybean, small grain, forages, and other potential
> crops. Research will
> include interdisciplinary efforts at the Koch Farm and at satellite farm
> sites.
>
> OUTREACH AND PUBLIC SERVICE: The appointee will participate in the
> planning,
> development and implementation of cropping system programs on request;
> communicate
> research results through a variety of channels; and respond to public
> requests for
> speaking/teaching opportunities related to the scientist's research
> program. Coordination of
> field days and assistance in planning the crop management program for the
> station is expected.
> The agronomist will also advise the Station Head with respect to the
> overall research program
> at the station, outreach needs, personnel acquisition and evaluation, and
> budget preparation.
> This person will also provide service by participating in activities and
> committees which
> support the Station, Department of Agronomy and Plant Genetics, College of
> Agriculture and
> the Minnesota Agricultural Experiment Station.
>
> QUALIFICATIONS:
>
> Minimum: A Ph.D. in agronomy or a closely related field by the date of
> appointment.
>
> Desired: Experience in cropping systems and crop production research.
> On-farm work or
> research experience. Skills in communication to diverse audiences
> including scientists, crop
> advisors, extension educators, and growers. Disciplinary and
> interdisciplinary collaborative
> research efforts. Understanding and interest in sustainable and
> ecologically sound agricultural
> systems. Coursework in agronomy, statistics, soil science, plant
> breeding, weed science, plant
> pathology, entomology, and plant physiology. Two years of post-graduate
> experience.
>
> SALARY AND BENEFITS:
>
> Salary will be commensurate with education and career experience.
> Benefits will include
> group life and health insurance, dental insurance, faculty retirement
> program, vacation and sick
> leave, and opportunity for study leaves and sabbaticals.
>
> EQUAL OPPORTUNITY AND AFFIRMATIVE ACTION:
>
> The University of Minnesota is committed to the policy that all persons
> shall have equal
> access to its programs, facilities, and employment without regard to race,
> color, creed, religion,
> national origin, sex, age, marital status, disability, public assistance
> status, veterans status, or
> sexual orientation.
>
> APPLICATIONS:
>
> Please send a letter of application, undergraduate transcripts, graduate
> transcripts, curriculum
> vita, and a one-page summary of career goals in the context of this
> position; and have 3 letters
> of reference sent by May 1, 1994 to:
>
> Dr. Craig C. Sheaffer, Chair
> Cropping Systems Agronomist Search Committee
> Department of Agronomy and Plant Genetics
> 411 Borlaug Hall
> University of Minnesota
> St. Paul, MN 55108
> TEL: (612) 625-7224
> FAX: (612) 625-1268
> EMAIL: sheaf001@maroon.tc.umn.edu
>

Hello -

Intro: I am John Bolte, Assistant Prof in Bioresource Engineering at
Oregon State University working in the areas of computational ecology,
and artificial intelligence. Particular areas of interest are in
aquatic systems modeling and decision support, self-organizing systems
representations, object-oriented programming and intelligent
simulation systems.

On Feb 12, Mike Dwyer posted an intersting note on his work with
developing a user interface for the Revised Universal Soil Loss
Equation (RUSLE), in which he wrote:

> There was also the discussion of user involvement in development of
> the models themselves. Would Greg McIsaac agree that the users'
> biggest interest might be in the interface to a model more than
> the science behind it?

I agree that this MIGHT be the case. Whether it is the case might
be confirmed by surveying users. But I disagree that this SHOULD be
the case. Being an old Jeffersonian FORTRAN hacker, I think people
should know what they are doing. When they apply a packaged model,
such as RUSLE, they are applying a set of assumptions and empirical
relations. Where the assumptions or empirical relations do not apply,
the model may not preform well. If the user is not aware of these
limitations, it will increase the liklihood that they will apply
or interpret the predictions from the model in ways that are not
supported by observation. And when the user is an agent of the
government, there is a chance that the best modern science will
be used to treat people unfairly.

Model developers could set limits internally in the model or in the
interface so that the user would be less likely to misuse it. But
model developers can be guilty of exaggerating the applicability of
their work. Consider the UNIVERSAL Soil Loss Equation: In a paper
entitled "The use and misuse of the Universal Soil Loss Equation"
(Journal of Soil and Water Conservation vol 31, No.1) Walt
Wischmeier, one of the chief developers of the USLE, stated that
based upon a 189 plot sample, "About 5% of the [USLE] predicted
losses differed from the measured losses by little more than 4.5 tons
[per ac per year]..." In metric units, that is 1.03 kg/sq m -yr.
In a more recent paper Risse et al. (1993, Soil Science
Society of America Jounal 57:825-833) analyzed erosion from 208 plots
and based on their histogram of the errors it appears to me that 95%
of the measured average annual soil loss rates could be as large
as 5 kg/sq m - yr greater than or less than the USLE prediction.
Furthermore, the 95% confidence intervals for the regression of
prediced vs measured soil loss indicated that a future observed soil
loss may be more than 10 kg/sq. m per year greater than or less than
the USLE predicted value -- about 10 times greater than
Wishmeier's earlier estimates of the confidence limits.

But I don't want to single out Walt Wischmeier for a phenomenon that
appears to be widespread in science and engineering. The
physicist and historian of science Evelyn Fox Keller observed the
following: "To know the history of science is to recognize the
mortality of any claim to universal truth. Every past vision of
scientific truth, every model of natural phenomenon, has proved in
time to be more limited than its adherents claimed. The survival of
productive difference in science requires that we put all
claims for intellectual hegemony in their proper place - that we
understand that such claims are by their very nature political rather
than scientific." (Reflections on Gender and Science, 1985, Yale
University Press.) Note Keller's reference to claims to UNIVERSAL
truth corresponds in an interesting way to the name of the
UNIVERSAL Soil Loss Equation (and the Revised USLE, for that matter).

Having said all that, I wonder whether confidence limits on model
predictions will be a part of the RUSLE package? If so, what will be
the basis of the limits?

But speaking of intellectual hegemony, Mike writes:

> The project also involves the integration of "our" application with
> other datasets in a single field office database (using
> Informix/SQL). The data used in our application is perhaps a bit
> more restricted than what could be enterable in the SWCS software,
> in order to keep field office users within the bounds of agency
> policy and accepted agronomics.

Who determines what the "accepted agronomics" are? Will farmers who
are outside the agronomic mainstream, such as Amish and organic
farmers, be subjected to the same assumptions and empirical relations
that may not apply very well to their farming operations? I know
of some evidence which suggests that infiltration rates may be
greater and soil erosion less on Amish farms as well as land
treated with manure see: Mueller et al., 1984, Soil and Water Losses
as affected by Tillage and Manure" Soil Science Society of America
Journal 48:896-900, and Mary Jackson, 1988, "Amish Agriculture and No-
till: the hazards of applying the USLE to unusual farms" Journal of
Soil and Water Conservation, 43(6):483-486. These studies are very
limited and in my opinion do not provide enough of a basis for
adjusting the RUSLE for manure applications or the use of horses
rather than tractors on farmland. But they do suggest some
limitations of assuming that the "accepted agronomics" apply equally
well to all situations.

I realize that the USDA-SCS has a national charge to impliment
Conservation Compliance, and are working within a fixed budget and
therefore need to make "executive decisions" in the interest of
efficiently accomplishing their nationwide mandate. But in our quest
for efficiency, let us not lose sight of the limitations of our
knowledge and the very real impacts that our assumptions and
empirical baises might have on real people. One possible safeguard
against the misuse of models is to have well informed model users.

Best regards,

Greg McIsaac

> Date sent: Sat, 12 Feb 1994 20:51:06 -0600
> Send reply to: <agmodels-l@unl.edu>
> From: Mike Dwyer <miked@pss5.ftc.scs.ag.gov>
> To: Multiple recipients of list <agmodels-l@unl.edu>
> Subject: Introduction

>
> Hello -
>
> My name is Mike Dwyer, and I am a software developer (systems analyst) with
> ADC, a contractor to the Soil Conservation Service (located in Fort Collins).
> My project is the implementation of the Revised Universal Soil Loss Equation
> (RUSLE) model for the SCS field offices.
>
> >From our perspective, we are not developing a model but putting a new front-
> end on the model developed by ARS, Purdue, etc. and released by the SWCS.
> The project also involves the integration of "our" application with other
> datasets in a single field office database (using Informix/SQL). The data
> used in our application is perhaps a bit more restricted than what could be
> enterable in the SWCS software, in order to keep field office users within
> the bounds of agency policy and accepted agronomics.
>
> This gives me an interest in some of the topics discussed in months past:
>
> I can agree with Georg Hoermann, 7 Jan, about user interfaces discouraging
> use of some models -
>
> > My interest in user interfaces comes from my teaching experience: I
> > think the *real* reason why models are not applied is the interface and
> > not the contents. There are nearly thousands of hydrologic models but
> > none of them can be used by our biologists. Students can work with a
> > graphics package or a spreadsheet within a week, but if it comes to
> > modelling they have to deal with really awkward FORTRAN formatting
>
> RUSLE doesn't force use of FORTRAN, but the familiarity and consistency of
> a common interface across applications is beneficial to the user. A related
> issue is the development of modelling/simulation software such that the
> computational engines can run "behind the scenes" as it were to various user
> interfaces designed for different environments.
>
> Ken Rojas said (December) -
>
> > I could go on forever on this topic, it really controls model development
> > processes and eventual customer satisfaction. One can not just create a
> > model with basic features and expect users to be satified, users are
> > getting too sophisticated and want the same look and feel as everything
> > else they use. Is this a reasonable request to put on the scientist
> > developing the model?
>
> There was also the discussion of user involvement in development of the models
> themselves. Would Greg McIsaac agree that the users' biggest interest might
> be in the interface to a model more than the science behind it?
>
> Perhaps the "hottest" issue for us right now is the sharing of data that
> drives several models or evaluation tools. Climate data is one such area.
> Crop charateristics is another. At least the climate data seems to break
> down in terms of frequency of measurement and normalized vs observed data.
> Plant data, however, seems to vary with each model. The attributes that
> are significant for one are irrelevant to the other, or each research
> project defines the variations a little differently. Furthermore, the
> characteristics that identify the differentiation also vary from one crop
> to the next (corn by days to maturity, alfalfa by year of rotation or by
> interseeding, e.g.).
>
> This latter is part of the data modeler's problem, of course, but it may
> be helpful for developers of a _single_ model to consider that their tool
> may somewhere or someday be integrated into a multi-tool environment.
>
>

I have a contrary opionion to that expressed recently by two
correspondents. Ken Rojas claimed that the front end was important because
users are more sophisticated now, and Georg Hoermann claimed that models
are not used because of similar reasons, and complained that none of the
hydrologic models can be used by their biologists. Part of the latter
problem is that people are beginning to think of computer models of complex
processes like they think of computer games. The biologists, for example,
(not to pick on them) want a hydrologic model that will substutute for
hydrologic knowledge. That is in my opinion part of the problem. The only
models I know with "front ends" that make it easy for a person totally
outside the discipline to run are scientifically specious, and cannot deal
properly with complex problems. I suggest that in fact users who expect
such easy substitution for knowledge are much LESS sophisticated, and are
unrealistically expecting models to substitute for understanding. None of
this is to defend models that are difficult even for the person in the same
discipline to use.

Roger E. Smith
roger@lily.aerc.colostate.edu
303 491 8263

On Mon, 14 Feb 1994 10:25:21 -0600 Greg McIsaac said:
>
>On Feb 12, Mike Dwyer posted an intersting note on his work with
>developing a user interface for the Revised Universal Soil Loss
>Equation (RUSLE), in which he wrote:
>
>> There was also the discussion of user involvement in development of
>> the models themselves. Would Greg McIsaac agree that the users'
>> biggest interest might be in the interface to a model more than
>> the science behind it?
>
>I agree that this MIGHT be the case. Whether it is the case might
>be confirmed by surveying users. But I disagree that this SHOULD be
>the case. Being an old Jeffersonian FORTRAN hacker, I think people
>should know what they are doing. When they apply a packaged model,
>such as RUSLE, they are applying a set of assumptions and empirical
>relations. Where the assumptions or empirical relations do not apply,
>the model may not preform well. If the user is not aware of these
>limitations, it will increase the liklihood that they will apply
>or interpret the predictions from the model in ways that are not
>supported by observation. And when the user is an agent of the
>government, there is a chance that the best modern science will
>be used to treat people unfairly.
>
...
...
Gee, as a "government agent" it's distressing to know that I'm likely to
misuse a model to treat people unfairly. It's more likely that I will
just be ignorant of it's inner workings like anyone else. Or maybe I will
have to make a decision based on less knowledge than would be ideally
available and just don't have the luxury of taking a pure moral stance. You
see, no decision is also usually unfair and if we wait 'till the models are
omniscient...you get the idea.
...
...
>Model developers could set limits internally in the model or in the
>interface so that the user would be less likely to misuse it. But
>model developers can be guilty of exaggerating the applicability of
>their work. Consider the UNIVERSAL Soil Loss Equation: In a paper
>entitled "The use and misuse of the Universal Soil Loss Equation"
>(Journal of Soil and Water Conservation vol 31, No.1) Walt
>Wischmeier, one of the chief developers of the USLE, stated that
>based upon a 189 plot sample, "About 5% of the [USLE] predicted
>losses differed from the measured losses by little more than 4.5 tons
>[per ac per year]..." In metric units, that is 1.03 kg/sq m -yr.
>In a more recent paper Risse et al. (1993, Soil Science
>Society of America Jounal 57:825-833) analyzed erosion from 208 plots
>and based on their histogram of the errors it appears to me that 95%
>of the measured average annual soil loss rates could be as large
>as 5 kg/sq m - yr greater than or less than the USLE prediction.
>Furthermore, the 95% confidence intervals for the regression of
>prediced vs measured soil loss indicated that a future observed soil
>loss may be more than 10 kg/sq. m per year greater than or less than
>the USLE predicted value -- about 10 times greater than
>Wishmeier's earlier estimates of the confidence limits.
>
>But I don't want to single out Walt Wischmeier for a phenomenon that
>appears to be widespread in science and engineering. The
>physicist and historian of science Evelyn Fox Keller observed the
>following: "To know the history of science is to recognize the
>mortality of any claim to universal truth. Every past vision of
>scientific truth, every model of natural phenomenon, has proved in
>time to be more limited than its adherents claimed. The survival of
>productive difference in science requires that we put all
>claims for intellectual hegemony in their proper place - that we
>understand that such claims are by their very nature political rather
>than scientific." (Reflections on Gender and Science, 1985, Yale
>University Press.) Note Keller's reference to claims to UNIVERSAL
>truth corresponds in an interesting way to the name of the
>UNIVERSAL Soil Loss Equation (and the Revised USLE, for that matter).
>
>Having said all that, I wonder whether confidence limits on model
>predictions will be a part of the RUSLE package? If so, what will be
>the basis of the limits?
>
>But speaking of intellectual hegemony, Mike writes:
>
>> The project also involves the integration of "our" application with
>> other datasets in a single field office database (using
>> Informix/SQL). The data used in our application is perhaps a bit
>> more restricted than what could be enterable in the SWCS software,
>> in order to keep field office users within the bounds of agency
>> policy and accepted agronomics.
>
>Who determines what the "accepted agronomics" are? Will farmers who
>are outside the agronomic mainstream, such as Amish and organic
>farmers, be subjected to the same assumptions and empirical relations
>that may not apply very well to their farming operations? I know
>of some evidence which suggests that infiltration rates may be
>greater and soil erosion less on Amish farms as well as land
>treated with manure see: Mueller et al., 1984, Soil and Water Losses
>as affected by Tillage and Manure" Soil Science Society of America
>Journal 48:896-900, and Mary Jackson, 1988, "Amish Agriculture and No-
> till: the hazards of applying the USLE to unusual farms" Journal of
>Soil and Water Conservation, 43(6):483-486. These studies are very
>limited and in my opinion do not provide enough of a basis for
>adjusting the RUSLE for manure applications or the use of horses
>rather than tractors on farmland. But they do suggest some
>limitations of assuming that the "accepted agronomics" apply equally
>well to all situations.
>
>I realize that the USDA-SCS has a national charge to impliment
>Conservation Compliance, and are working within a fixed budget and
>therefore need to make "executive decisions" in the interest of
>efficiently accomplishing their nationwide mandate. But in our quest
>for efficiency, let us not lose sight of the limitations of our
>knowledge and the very real impacts that our assumptions and
>empirical baises might have on real people. One possible safeguard
>against the misuse of models is to have well informed model users.
>
>Best regards,
>
>Greg McIsaac
>
>
>> Date sent: Sat, 12 Feb 1994 20:51:06 -0600
>> Send reply to: <agmodels-l@unl.edu>
>> From: Mike Dwyer <miked@pss5.ftc.scs.ag.gov>
>> To: Multiple recipients of list <agmodels-l@unl.edu>
>> Subject: Introduction
>
>>
>> Hello -
>>
>> My name is Mike Dwyer, and I am a software developer (systems analyst) with
>> ADC, a contractor to the Soil Conservation Service (located in Fort Collins).
>> My project is the implementation of the Revised Universal Soil Loss Equation
>> (RUSLE) model for the SCS field offices.
>>
>> >From our perspective, we are not developing a model but putting a new front-
>> end on the model developed by ARS, Purdue, etc. and released by the SWCS.
>> The project also involves the integration of "our" application with other
>> datasets in a single field office database (using Informix/SQL). The data
>> used in our application is perhaps a bit more restricted than what could be
>> enterable in the SWCS software, in order to keep field office users within
>> the bounds of agency policy and accepted agronomics.
>>
>> This gives me an interest in some of the topics discussed in months past:
>>
>> I can agree with Georg Hoermann, 7 Jan, about user interfaces discouraging
>> use of some models -
>>
>> > My interest in user interfaces comes from my teaching experience: I
>> > think the *real* reason why models are not applied is the interface and
>> > not the contents. There are nearly thousands of hydrologic models but
>> > none of them can be used by our biologists. Students can work with a
>> > graphics package or a spreadsheet within a week, but if it comes to
>> > modelling they have to deal with really awkward FORTRAN formatting
>>
>> RUSLE doesn't force use of FORTRAN, but the familiarity and consistency of
>> a common interface across applications is beneficial to the user. A related
>> issue is the development of modelling/simulation software such that the
>> computational engines can run "behind the scenes" as it were to various user
>> interfaces designed for different environments.
>>
>> Ken Rojas said (December) -
>>
>> > I could go on forever on this topic, it really controls model development
>> > processes and eventual customer satisfaction. One can not just create a
>> > model with basic features and expect users to be satified, users are
>> > getting too sophisticated and want the same look and feel as everything
>> > else they use. Is this a reasonable request to put on the scientist
>> > developing the model?
>>
>> There was also the discussion of user involvement in development of the
>models
>> themselves. Would Greg McIsaac agree that the users' biggest interest might
>> be in the interface to a model more than the science behind it?
>>
>> Perhaps the "hottest" issue for us right now is the sharing of data that
>> drives several models or evaluation tools. Climate data is one such area.
>> Crop charateristics is another. At least the climate data seems to break
>> down in terms of frequency of measurement and normalized vs observed data.
>> Plant data, however, seems to vary with each model. The attributes that
>> are significant for one are irrelevant to the other, or each research
>> project defines the variations a little differently. Furthermore, the
>> characteristics that identify the differentiation also vary from one crop
>> to the next (corn by days to maturity, alfalfa by year of rotation or by
>> interseeding, e.g.).
>>
>> This latter is part of the data modeler's problem, of course, but it may
>> be helpful for developers of a _single_ model to consider that their tool
>> may somewhere or someday be integrated into a multi-tool environment.
>>
>>

Roger,

Very well said. I agree entirely and bow to your eloquence.

Greg McIsaac

> Date sent: Mon, 14 Feb 1994 12:16:48 -0600
> Send reply to: <agmodels-l@unl.edu>
> From: "" <roger@lily.aerc.colostate.edu>
> To: Multiple recipients of list <agmodels-l@unl.edu>
> Subject: Re: Modelling Comments

> I have a contrary opionion to that expressed recently by two
> correspondents. Ken Rojas claimed that the front end was important because
> users are more sophisticated now, and Georg Hoermann claimed that models
> are not used because of similar reasons, and complained that none of the
> hydrologic models can be used by their biologists. Part of the latter
> problem is that people are beginning to think of computer models of complex
> processes like they think of computer games. The biologists, for example,
> (not to pick on them) want a hydrologic model that will substutute for
> hydrologic knowledge. That is in my opinion part of the problem. The only
> models I know with "front ends" that make it easy for a person totally
> outside the discipline to run are scientifically specious, and cannot deal
> properly with complex problems. I suggest that in fact users who expect
> such easy substitution for knowledge are much LESS sophisticated, and are
> unrealistically expecting models to substitute for understanding. None of
> this is to defend models that are difficult even for the person in the same
> discipline to use.
>
> Roger E. Smith
> roger@lily.aerc.colostate.edu
> 303 491 8263
>

Mike Dwyer commented:

>Perhaps the "hottest" issue for us right now is the sharing of data that
>drives several models or evaluation tools. Climate data is one such area.
>Crop charateristics is another. At least the climate data seems to break
>down in terms of frequency of measurement and normalized vs observed data.
>Plant data, however, seems to vary with each model. The attributes that
>are significant for one are irrelevant to the other, or each research
>project defines the variations a little differently. Furthermore, the
>characteristics that identify the differentiation also vary from one crop
>to the next (corn by days to maturity, alfalfa by year of rotation or by
>interseeding, e.g.).

>This latter is part of the data modeler's problem, of course, but it may
>be helpful for developers of a _single_ model to consider that their tool
>may somewhere or someday be integrated into a multi-tool environment.

Object-oriented systems may provide a glimmer of hope here. Not only
sharing data but sharing/reusing functional subsystems as well. We
at OSU (and many others) have been actively involved in this area,
examining architectures which enable expertise embedded within
software objects/agents to communicate with eachother via
well-defined, high level interfaces. Some successes. some failures.
Our conclusion at this point is that we as modelers must pay much more
close attention to system design than we have done in the past.
Doing so opens up possibilities for rapid model development and
deployment reusing preexisting model components. The trends in the
software industry towards object-based systems provides many lessons
for the modeling community, as well as opportunities for systems-level
support for language independant object sharing which have not been
available in the past. Taking advantages of these opportunities will
involve the modeling community as a whole to upgrade our view of
software development from the FORTRAN style "lets do the minimum to
solve a set of equations" view to one that more focuses on software
engineering for reusability and embeddability of submodels.

Our own work at OSU is moving more and more to a view of simulation as
a set of "domain experts" sitting around a table discussing how they
expect a system to respond, giving a particular set of driving forces
and conditions. Details of implmentation of each expert is irrelevant;
that is, they might be expert systems, algorithm databases, neural
network conglomerations, whatever. Doesn't matter, as long as they
all communicate through a well-defined high-level interface. To
conduct a simulation of a fishpond, we might grab a FishBiologist, a
WaterChemist, a Climatologist, etc. out of our "Expert Object"
grabbag, assemble them around a software blackboard, describe the
intial conditions and constraint we are interested in, and have
the experts tell us what they think will happen. As long as we do a
good job of "packaging" our models inside the various domain experts,
it is relatively simple to reuse them. Requires a lot of thoughtful
design up front, though, rather than the traditional somewhat ad hoc
approach to model implementation.

- John Bolte
Oregon State University
Bioresource Engineering
Boltej@ccmail.orst.edu

On Feb 12, Mike Dwyer posted an intersting note on his work with
developing a user interface for the Revised Universal Soil Loss
Equation (RUSLE), in which he wrote:

> There was also the discussion of user involvement in development of
> the models themselves. Would Greg McIsaac agree that the users'
> biggest interest might be in the interface to a model more than
> the science behind it?

I agree that this MIGHT be the case. Whether it is the case might
be confirmed by surveying users. But I disagree that this SHOULD be
the case. Being an old Jeffersonian FORTRAN hacker, I think people
should know what they are doing. When they apply a packaged model,
such as RUSLE, they are applying a set of assumptions and empirical
relations. Where the assumptions or empirical relations do not apply,
the model may not preform well. If the user is not aware of these
limitations, it will increase the liklihood that they will apply
or interpret the predictions from the model in ways that are not
supported by observation. And when the user is an agent of the
government, there is a chance that the best modern science will
be used to treat people unfairly.

These concerns are appropriate regardless of the implementation or
interface. One could easily argue that a well designed interface
(both user and internal) would allow the model assumptions to be more
explicitly presented to the user and enforced by the model. In any
case, Mike's original point regarding model usability seems well
taken. Our experience is that this is the largest factor influence
model usage. Models are typically assumed to be (relatively) correct,
regardless of ease of use. This is a given. The only question now,
in the age of sophisticate user interfaces and programming
environments, is whether this is sufficient. I have concluded it is
not, for several good reasons: 1) Models must be usable to be used, 2)
data inputs can be more clearly delineated and defined in a
"user-friendly" dialog-driven environment, 3) most importantly, model
results can be much more effectively presented in a graphical
environment than as a file full of numbers. We have had several cases
where we have converted FORTRAN style models with limited model
interpretation capabilities to a graphical analysis environment and
identified bugs/poor assumptions not readily discernable in the
original model (this with some pretty widely used models!)

The drawback to all this is that it is a substantially more difficult
programming task to design interfaces, graphical interpreters, and
program architectures which allow all this to happen. It is typically
a task for which we receive little funding and little reward. Whats
the solution? Becoming more efficient at reusing and sharing code is
a step in the right direction.

- John Bolte
Oregon State University
Bioresource Engineering

A 'friendly' User Interface is really only required
if the 'User' needs it and generally most computer users need one.
However, in the past it has been very difficult to write a good
user interface so scientists just could not spend
time doing this. But if your model is one that you
want to go beyond your personal research use or has a
practical application then a good user interface is essential.

We've been putting doing this and retaining our perfectly
good FORTRAN code by writing Visual Basic front end
with dialog boxes, buttons, etc. the whole GUI thing.
The Resulting VB program then calls the FORTRAN
model as a DLL, it all runs under MS Windows.
No modification FORTAN program was required
just modifications of compile and link options.

The improved user interface also allows users
to run iterations of similar models much faster
than our old user interface -- which was just a
flat data file.

>>From our perspective, we are not developing a model but putting a new front-
...stuff deleted
>
>Ken Rojas said (December) -
>
> > I could go on forever on this topic, it really controls model development
> > processes and eventual customer satisfaction. One can not just create a
> > model with basic features and expect users to be satified, users are
> > getting too sophisticated and want the same look and feel as everything
> > else they use. Is this a reasonable request to put on the scientist
> > developing the model?

Bruce Heimbigner
(509)335-2441
FAX (509)335-1082
bruce@heimbig.ansci.wsu.edu

In response to my posting on model users and builders, Don Wauchope
wrote:

> Gee, as a "government agent" it's distressing to know that I'm
> likely to misuse a model to treat people unfairly. It's more
> likely that I will just be ignorant of it's inner workings like
> anyone else. Or maybe I will have to make a decision based on
> less knowledge than would be ideally available and just don't have
> the luxury of taking a pure moral stance. You see, no decision is
> also usually unfair and if we wait 'till the models are
> omniscient...you get the idea. > ...

Don,

Thanks for your comment, but I did not write, as you suggest, that
all goverment agents are "likely to misuse a model to treat people
unfairly."

Rather I wrote that the less people know about the
assumptions and empirical limitations of a model, the
more likely they are to misuse it. If the user happens to be a
government agent, such as the SCS in the case of RUSLE, then "there
is a chance that" the misuse may lead to unfair treatment.

I really don't follow your statements:

> It's more likely that I will just be ignorant of it's inner
> workings like anyone else. Or maybe I will
> have to make a decision based on less knowledge than would be
> ideally available and just don't have the luxury of taking a pure
> moral stance.

Do you have some data which supports your statement that "no decision
is usually unfair?"

Greg McIsaac

By the way, I, too, am a government agent.

>

Unsubscribe

Nachi

************************************************************************
* Nachi Narayanan WORK (703)-231-7470 *
* Research Scientist HOME (703)-552-7332 *
+ Information Systems Lab +
* Virginia Tech FAX (703)-231-9131 *
* 202. Price Hall INTERNET: *
* Blacksburg, VA 24061-0319 nachi@vt.edu *
* nachi@atum.ento.vt.edu *
************************************************************************

I _am_ working with Jerry Weber! I told him 'Hello' but failed to
mention the condolences :) I'm also working with Gail Wilkerson
(she's my major advisor) and Harvey Gold in Biomath.

By the way, does anybody out there know of a source which lists
all the degradation products of most of the herbicides and
insecticides . . . maybe even along with their half-life values under
different soil conditions?

(Thought I'd at least ask.)

Cassie Webster

> You must be working for Jerry Weber--you have my condolences and
tell > him hello for me.
> Don Wauchope

> Cassie Webster
> > webster@cropserv1.cropsci.ncsu.edu
> > (919) 515-5817
> > fax: (919) 515-7959
> >
> >
>

I am seeking names and information on models which estimate long and
short term groundwater quantity as a function of various water
conservation techniques used in urban areas. It is especially desireable
if the model interfaces with a GIS system. Please refer me
to another network if you know of one.

Pat Bartling
mx"pat@gpsrv1.gpsr.colostate.edu"

On Mon, 14 Feb 1994 14:40:18 -0600 Greg McIsaac said:
>Don,
>
>Thanks for your comment, but I did not write, as you suggest, that
>all goverment agents are "likely to misuse a model to treat people
>unfairly."

Greg:
OK, I was being touchy--if your comment was a concern for the possible mis-
application of a model in your own sphere of responsibility that's really
good. I'm touchy because information I've developed has been misused, and
the results were unfair. The unfairness was due to someone not under-
the limitations of the model, not malice. Thus, I said:

>> It's more likely that I will just be ignorant of it's inner
>> workings like anyone else. Or maybe I will
>> have to make a decision based on less knowledge than would be
>> ideally available and just don't have the luxury of taking a pure
>> moral stance.
>
>Do you have some data which supports your statement that "no decision
>is usually unfair?"
>
I was thinking of the situation of registration of pesticides (in which I am
somewhat involved) where EPA is called upon to render a timely decision on a
petition and they will never have all the information needed to make a
unambiguous decision. This is one place a lot of modeling is going to be used.
My point is if they make no decision at all that is also unfair to the
company--the clock is ticking on their patent protection.

Anyway, I take it back.
>Greg McIsaac
>
>By the way, I, too, am a government agent.
>
>>

To the ag. modelers:

On Monday 14 February, Cassie Webster wrote:

> By the way, does anybody out there know of a source which lists
> all the degradation products of most of the herbicides and
> insecticides . . . maybe even along with their half-life values under
> different soil conditions?
>
> (Thought I'd at least ask.)
>
> Cassie Webster

One helpful reference might be

Howard, P.H. 1991. Handbook of Environmental Fate and Exposure
Data for Organic Chemicals. Volume III: Pesticides. Lewis
Publishers, Inc. Chelsea, Michigan. 684pp.

Library of Congress cat. no. TD176.4.H69 1989 ??

This does not have as complete a list of pesticides as you might
want, but it does have quite a few, and gives detailed information
on them. If you cannot find a copy over on campus, we have one
at the Air Resources Library (on Varsity Drive, just beyond
Fraternity Court). You could have trouble with the modeling if you
are looking for exact half-lives of compounds -- this reference
gives pretty wide ranges sometimes. Good luck.

[I know I haven't introduced myself yet to this group, but I'll
do that today.]

-Mike

<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>
| MIKE MUNSTER |
<> EMAP-Agroecosystems phone 919-515-3311 <>
| 1509 Varsity Dr. fax 919-515-3593 |
<> Raleigh, NC 27606 email mike_munster@ncsu.edu <>
| USA |
<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>

Greetings to the AGMODELS-L. We have been silently reading the
postings to this group, but several weeks ago (maybe months),
the comments of Drew Laughland and Jerome Pier's reply inspired
me to join the fray, only I waited until now. More on that
at the bottom.

My name is Mike Munster (B.S. in agronomy, M.S. in plant pathology).
I am currently a research assistant at North Carolina State University in
Raleigh, NC, USA. My colleague Gyanendra 'Gyani' Dhakwha has a
Ph.D. in meteorology, with experience in crop modeling. We both
work for the Agroecosystem Resource Group of an EPA-sponsored
project called the Environmental Monitoring and Assessment Program (EMAP).
The goal of the program is to describe the condition of the natural
resources of the U.S.A. (agriculture, forests, arid lands, surface waters,
estuaries, the Great Lakes, and lancscapes in general). Our group
is concerned with the ecological state of agricultural areas.

Gyani and I work on indicators of crop productivity, and we are
attempting to use of crop growth models in a sort of backward way
to account for known sources of variation (managment, soil,
weather) in crop growth. We hope that (a) the variance is
reduced and (b) subtler aspects of the system health might be
uncovered. Our data come from surveys conducted across the
states of North Carolina (1992) and Nebraska (1993). Models to be
tested are SOYGRO, CERES-maize, CERES-wheat, and EPIC.

Another colleague of ours says that if a model has an acronym, it
is too complicated to be trusted. A simpler idea that we have used, but
that still has an underlying model, is nitrogen use efficiency. This
is just the total nitrogen applied to a field divided by the harvested
product. A simple ratio, but it sort of implies a linear relationship
that passes through the origin. Yet it does say something about the
efficiency of the agricultural system as a whole in its use of N.

Drew Laughland's comments caught my attention because I, too, need simple,
regionally applicable indices. Someone had suggested calculating
nitrogen excess as the

(N applied) - (crop yield)*(concentration of N in the grain).

I have been reassured by at least one person that the N
concentration is consistent enough to make the index useful, at
least for maize, but I am skeptical, especially since I must
consider other grain and nongrain crops, too. Drew's suggestion
of (N applied) - (crop yield)*(minimum N rate needed) sounds
interesting. Has anyone out there done determinations of either
N concentrations in harvested products or N response curves
(seems like the minimum N rate needed might come from there)?

-Mike

<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>
| MIKE MUNSTER |
<> EMAP-Agroecosystems phone 919-515-3311 <>
| 1509 Varsity Dr. fax 919-515-3593 |
<> Raleigh, NC 27606 email mike_munster@ncsu.edu <>
| USA |
<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>

On Feb 14, John Bolte wrote:

> One could easily argue that a well designed interface
> (both user and internal) would allow the model assumptions to
> be more explicitly presented to the user and enforced by
> the model.

Yes, this is a good point. I was assuming that user interfaces would
largely insulate the user from the model assumptions and limitations.
But, I can see where the interface could also be a educational tool.
Glad to hear that you are working in that direction. Sorry to hear
that it is an under-rewarded activity.

There are other limitations to this approach, however. Some
assumptions may be "subliminal" or not entirely obvious to the
model builders, interface creators, or the users. Such
"cultural" assumptions may become obvious as models are used
internationally, across cultures which operate under very different
sets of basic assumptions.

Best regards,

Greg McIsaac

> Date sent: Mon, 14 Feb 1994 14:33:07 -0600
> Send reply to: <agmodels-l@unl.edu>
> From: boltej@ccmail.orst.edu
> To: Multiple recipients of list <agmodels-l@unl.edu>
> Subject: Re: Model users and builders

>
>
> On Feb 12, Mike Dwyer posted an intersting note on his work with
> developing a user interface for the Revised Universal Soil Loss
> Equation (RUSLE), in which he wrote:
>
> > There was also the discussion of user involvement in development of
> > the models themselves. Would Greg McIsaac agree that the users'
> > biggest interest might be in the interface to a model more than
> > the science behind it?
>
> I agree that this MIGHT be the case. Whether it is the case might
> be confirmed by surveying users. But I disagree that this SHOULD be
> the case. Being an old Jeffersonian FORTRAN hacker, I think people
> should know what they are doing. When they apply a packaged model,
> such as RUSLE, they are applying a set of assumptions and empirical
> relations. Where the assumptions or empirical relations do not apply,
> the model may not preform well. If the user is not aware of these
> limitations, it will increase the liklihood that they will apply
> or interpret the predictions from the model in ways that are not
> supported by observation. And when the user is an agent of the
> government, there is a chance that the best modern science will
> be used to treat people unfairly.
>
> These concerns are appropriate regardless of the implementation or
> interface. One could easily argue that a well designed interface
> (both user and internal) would allow the model assumptions to be more
> explicitly presented to the user and enforced by the model. In any
> case, Mike's original point regarding model usability seems well
> taken. Our experience is that this is the largest factor influence
> model usage. Models are typically assumed to be (relatively) correct,
> regardless of ease of use. This is a given. The only question now,
> in the age of sophisticate user interfaces and programming
> environments, is whether this is sufficient. I have concluded it is
> not, for several good reasons: 1) Models must be usable to be used, 2)
> data inputs can be more clearly delineated and defined in a
> "user-friendly" dialog-driven environment, 3) most importantly, model
> results can be much more effectively presented in a graphical
> environment than as a file full of numbers. We have had several cases
> where we have converted FORTRAN style models with limited model
> interpretation capabilities to a graphical analysis environment and
> identified bugs/poor assumptions not readily discernable in the
> original model (this with some pretty widely used models!)
>
> The drawback to all this is that it is a substantially more difficult
> programming task to design interfaces, graphical interpreters, and
> program architectures which allow all this to happen. It is typically
> a task for which we receive little funding and little reward. Whats
> the solution? Becoming more efficient at reusing and sharing code is
> a step in the right direction.
>
> - John Bolte
> Oregon State University
> Bioresource Engineering
>

Greetings,

My name is David Bosch and I work for the USDA-ARS at the
Southeast Watershed Research Laboratory in Tifton, Georgia. My
primary area of interest, and current research, is vadose zone
hydrology and transport. Currently I am working on determining
the dominant transport pathways in various landscapes in the
Coastal Plain area. We have a wide variance of soil types in
this area, and thus a wide variance of transport pathways.

I have several modeling interests which I work on off and on.
One is lateral subsurface flow and transport above shallow (< 3
m) soil restrictions. Another is flow and transport in Riparian
areas in the Southeast. We are in the advanced stages of
developing a model to examine water and chemical movement in near
stream areas in the Coastal Plain.

Recently I have seen a lot of discussion concerning user
interfaces for models. Although I can see the benefits of
creating 'user-friendly' shells around models, I also see a
danger. The danger concerns model application. If a model is
made so easy to use that anyone can, you leave the door open to
improper application and interpretation of model results. While
we are building these protective shells around the models, we
must also incorporate safeguards into the models to prevent such
abuse.

*****************************************************
* David Bosch *
* Research Hydrologist *
* USDA-ARS *
* Southeast Watershed Research Laboratory *
* PO Box 946 *
* Tifton, Georgia *
* *
* e-mail: dbosch@tifton.cpes.peachnet.edu *
* phone: 912-386-3515 *
* fax: 912-386-7294 *
*****************************************************

I am trying to find a program which will draw flow nets using
gridded hydraulic head data. I am using the software package
SURFER to interpolate between observations and establish a
regularly spaced data set. This is then used to draw hydraulic
head contours for the area. I would also like to draw flow lines
between the contours based upon the head contours. Does anyone
have such a program or could tell me where I could obtain it?
Thanks in advance.

*****************************************************
* David Bosch *
* Research Hydrologist *
* USDA-ARS *
* Southeast Watershed Research Laboratory *
* PO Box 946 *
* Tifton, Georgia *
* *
* e-mail: dbosch@tifton.cpes.peachnet.edu *
* phone: 912-386-3515 *
* fax: 912-386-7294 *
*****************************************************

I am working on a program that identifies channels from point elevation
data. Look at The American Society of Agricultural Engineers paper
#892159 to see if this would work for you.
The program is not in a finished state and currently works on about a
200 by 200 matrix. I am working on a version that will handle 1000 x 1000
the first version is limited to 255 points wide. I can use surfer data
with it. Unless you are an accomplished C programer the current version
would not work for you but I would be glad to try to run your data through
the program. The only catch is the edges have to be fixed as higher or
lower than the data on real world data I usualy assume that all sides
are lower than the data. I also have difficulties with large flat areas.
I can handle depressions but the flow path is crooked. If the flat areas
are not too large they don't cause the program to fail but it wanders around
a lot. The program was designed to be used on areas 8x32 feet but we have
used it on 1 mile data with good results except in the flat areas.
Gordon

My name is Tony Starfield and I am a modeler in the Dept of Ecology,
Evolution and Behavior here at the Univ of Minnesota. My main research
interest is the METHODOLOGY of modeling, i.e. how to build useful models
and how they might be used. I have written two books on modeling:
Building Models for Conservation & Wildlife Management (2nd ed, Bellwether
Press, Edina MN, 1991) and How to Model It: Problem-solving for the
Computer Age (McGraw-Hill, 1990). Current research interests include the
uses of modeling and decision analysis in conservation biology and novel
techniques for building "quick and dirty but useful" ecosystem models (see
A I Applications, 7(2&3):1-13).

I have no experience in agricultural models but joined the list-server
because Jonathan Haskett said there was a lot of discussion going on about
the methodlogy of modeling. I must admit that I joined a few months ago
and have been eavesdropping without introducing myself. The recent
discussion about user interfaces has brought me out of the closet, as it
were.

I believe a model is both a tool and a communication device. If it is
misused, then there is something wrong with the way in which it was
designed and/or the system within which it is used. There is a lot to be
said for designing models for use, and the main purport of this note is to
share with you a book that I have found illuminating: The design of
Everyday THings by Donald A Norman, Doubleday 1988. The system within
which a model is used is something I have not seen discussed or written
about -- it is something we need to think about. Modeling is usurping the
role of statistics in some areas, without the appreciation of what it can
or cannot do. In other areas it is being grafted onto decision-making
systems without careful thought. I'd be interested in comments on how to
educate the players and design human systems to make proper use of models.

Tony Starfield

Tony
Wander over to the St. Paul campus and talk to Bruce Wilson in Ag
Engineering. He has a lot of experiance in models and some damn good
ideas.
Gordon

Gordon Couger
Biosystems & Agricultural Engineering
Oklahoma State Universtiy Stillwater OK

February 15, 1994

To combine the discussion on user interfaces with the debate on
the knowledge level of the person using the model. . . .

And to continue Tony Starfield's comments on model design . . .

For a model that is going to be used by researchers who will be
examining the data generated from the model, or using the model for
registration decisions, or legislative decisions, what about
incorporating the assumptions and limitations of the model _into_ the
user interface?

Maybe the majority of models used for these purposes
are already like this, I'll let you tell me since I'm not familiar
with a lot of these models. But in addition to outlining these
assumptions in the "user guide" that comes with the program, why not
include the information in either an optional "Help about the model"
section, or a mandatory "Enter in the number of weeds per acre"
follwed by a "***Note: This model assumes a uniform distribution of
weeds in any given acre, which may effect predicted yield
reduction***".

For decision-aid models EVERY assumption would probably not want
to be outlined, but for researchers and especially with complex
models this might be useful.

I guess my point is this - are we as model builders providing the
user with enough information to make intelligent decisions based on
the model's prediction, or are people
treating models too lightly, no matter the information available.

It seems many people fall in the extreme categories of "well,
it's a model, it's not the REAL world, we really can't use it for
anything" or "the model said that this compound will leach, so it
does."

Cassie Webster

On February 15 Tony Starfield wrote:

> I'd be interested in comments on how to educate the players and
> design human systems to make proper use of models.
>
> Tony Starfield
>

Perhaps I misunderstand again, but "designing human systems
to make proper use of models" seems like a backwards formulation.
Didn't human systems come first and models created by humans as
attempts to enhance a certain aspect of the human system, such as
learning or testing out new ideas and designs? And what is "proper"
to you, may be totally "improper" to me.

The February 4, 1994 issue of Science contains an article titled
"Verification, Validation, and Confirmation of Numerical Models in the
Earth Sciences" by Oreskes, Shrader-Frenchette and Belitz (p 641-646.)
This paper seems to address some of the limitations and purposes of
that Dr. Starfield may be also concerned about. The abstract of
the article reads as follows:

"Validation and Verification of numerical models of natural systems is
impossible. This is because natural systems are never closed and
because model results are always non-unique. Models can be confirmed
by the demonstration of agreement between observation and prediction,
but confirmation is inherently partial. Complete confirmation is
logically precluded by the fallacy of affirming the consequent and by
incomplete access to natural phenomenon. Models can only be evaluated
in relative terms and their predictive value is always open to
question. The primary value of models is heuristic."

A similar article was published in Advances in Water Resources volume
15 (1992) p 75-83 titled "Ground water models cannot be validated" by
Konikow and Bredehoef.

Although this may sound like a radical position, I think the
substantive consequence of these papers is mostly one of terminology.
They argue that terms "validation" and "verification" imply that the
model itself is "true" when it really is an approximation of truth.
Oreskes et al. are particularly concerned about the misuse of these
terms in describing models used in the public policy arena.

There appears to be a philosophically deeper but less substantive
implication of these papers: if models are not and can never be proven
to be fully "true" but can only be demonstrated to correspond to some
degree to the available measured observations, then models are
partially works of art. Oreskes et al. argue:

"A model, like a novel, may resonate with nature, but it is not the
'real' thing. Like a novel, the model may be convincing - it may
'ring true' if it is consistent with our experience with the natural
world. But just as we may wonder how much the characters in a novel
are drawn from real life and how much is artifice, we might ask the
same of a model: How much is based on observation and measurement of
accessible phenomenon, how much based on informed judgement, how much
is convenience? Fundamentally, the reason for modeling is a lack of
full access, either in time or space, to the phenomena of interest.
In areas where public policy and public safety are at stake, the
burden is on the modeler to demonstrate the degree of correspondence
between the model and the material world it seeks to represent and to
delineate the limits of that correspondence."

"Thus the primary value of models is heuristic: Models are
representations, useful for guiding further study, but not
susceptible to proof."

I'll look forward to reactions and discussion on these matters.

Cordially,

Greg McIsaac

Picking up from Cassie WEbster's comment: expert systems have explanation
features, why can't we build models with an explanation capability?

Tony Starfield

Tony Starfield writes:

>I believe a model is both a tool and a communication device. If it is
>misused, then there is something wrong with the way in which it was
>designed and/or the system within which it is used. There is a lot to be
>said for designing models for use, and the main purport of this note is to
>share with you a book that I have found illuminating: The design of
>Everyday THings by Donald A Norman, Doubleday 1988. The system within
>which a model is used is something I have not seen discussed or written
>about -- it is something we need to think about. Modeling is usurping the
>role of statistics in some areas, without the appreciation of what it can
>or cannot do. In other areas it is being grafted onto decision-making
>systems without careful thought. I'd be interested in comments on how to
>educate the players and design human systems to make proper use of models.

Along these lines, another useful book is Jonathan Casti's "Alternate
Realities -- Mathematical Models of Nature and Man" (Wiley, 1989),
particularly the concluding chapter: "How do we know? Myths, models and
paradigms in the creation of beliefs." You might be intrigued by the
subheadings of this chapter:
1. ordering the cosmos
2. models and theories
3. paradigns, revolutions and normal science
4. explanations vs. descriptions
5. predictions, simulations and models
6. "good" models and "bad"
7. validation and falsification
8. competition among models
9. science and pseudoscience
10. science, religion and the nature of belief systems

Ricardo Salvador | Internet: rjsalvad@IASTATE.EDU | "Thou art a little
1126 Agronomy Hall | BITNET: a1.rjs@ISUMVS | soul bearing about
Iowa State University | Voice: (515) 294-9595 | a corpse."
Ames, IA 50011-1010 | FAX: (515) 294-3163 | -Marcus Aurelius-

Tony Starfield writes:

>Picking up from Cassie WEbster's comment: expert systems have explanation
>features, why can't we build models with an explanation capability?

No reason at all. The explanation capacity is presumably part of the
model interface, and I agree with Tony that we should be paying more
attention to the interface and not just focus on getting the code right.

An unfortunate reality related to some of the postings in response to
Tony's is that not only do many people write models in which they don't
make the assumptions clear, but the same is true when they write
research papers and reports about the models. I've just been going
through a stack of papers and am appalled by how few of them state the
assumptions behind the models. In many cases these assumptions, once
stated, are clearly false.

Bill Silvert
silvert@biome.bio.ns.ca

> Bill Silvert wrotes:
>
> An unfortunate reality related to some of the postings in response to
> Tony's is that not only do many people write models in which they don't
> make the assumptions clear, but the same is true when they write
> research papers and reports about the models. I've just been going
> through a stack of papers and am appalled by how few of them state the
> assumptions behind the models. In many cases these assumptions, once
> stated, are clearly false.
>
A related issue is the process of getting models published. It is a real
"constrained design", with the need for brevity on one hand vs.
completeness on the other hand. The only thing worse than writing them,
in my mind, is in trying to do a good job when asked to review such a
paper.

But it is still worth the effort, and I agree completely with your last
sentence...this is indeed one of the best reasons for writing about your
work.
--
R.S. Gates, Ph.D. P.E. gates@aen.uky.edu
Associate Professor ph (606) 257-3000 x 208
208 Agricultural Engineering Building fax (606) 257-5671
University of Kentucky
Lexington, KY 40546-0276
USA

Try getting the EPA One-Liners database on disk through the Freedom of
Information Office.

David Gustafson phone: 314/537-7194
Environmental Sciences fax: 314/537-6134
Monsanto e-mail: digust@ccmail.monsanto.com

Dear Agmodels-l and Soils-l Subscribers,

I am quite pleased with the traffic I have been seeing on
these lists lately. I hope you are all enjoying this and finding
the information as useful as I do. I no longer feel the need to
'prod' subscribers into communicating their ideas to the lists
and am taking a 'hands-off' approach. If any of you feel that
there is some way in which I can make the list more effective,
please drop me a note.

Due to the increased traffic on the lists, I would like
to remind subscribers that there is an optional digested format
of these lists available. In order to receive all the posts to
the list in a tidy once-a-day package in your mailbox, send the
following message to listserv@unl.edu:

set listname mail digest

where listname refers to the list you subscribe to, in this case
either soils-l or agmodels-l.

For new subscribers who would like to see what
discussions took place before you joined the list, monthly
archives are available by sending the following message to
listserv@unl.edu:

get listname logyymm

where listname is as mentioned above and 'yy' is the last two
digits of the year and 'mm' is the number of the month. For
instance, to receive the archive file for January 1994, send the
command:

get agmodels-l log9401

for the agmodels-l archive.

Our department is about to purchase a workstation which
will be tied to the Internet. I would like to make the
workstation available for anonymous ftp so that list subscribers
can get the archive files by ftp instead of get blah blah blah. I
am also thinking of starting a models archive for public
domain/shareware agricultural simulation models. Do you think
this would be useful? I also would like to set up a World Wide
Web server with hypertext pages. Anyone have any ideas how
agmodels-l/soils-l could best use such capability? Any other
ideas will be most welcome!

Thanks again for your participation in these lists and keep on
posting!

Sincerely,

Jerome Pier
Agmodels-l/Soils-l List Owner
jp@unl.edu

Model Debaters:
I cannot help but observe that much of the debate going thru the ether
on models, user interfaces vs. underlying code and assumptions, etc,. is
somewhat like the story of the description of the elephant by the 5 blind
men. Each sees well the part he has contact with, and each is at least
partly right, but the big picture contains all of them, and a debate as to
which is more important seems at least from one perspective to be folly.

Roger E. Smith
roger@lily.aerc.colostate.edu
303 491 8263

set agmodels-l mail digest

I saw the article in science too. I have seen similar arguments before,
most modelling texts emphasize that no model can ever be "proved" to
be true. In addition others have demonstrated how a highly predictive
model can be based on false premises but nevertheless be quite useful
within a specified range.

What struck me while reading the article was the question "What about
airplanes?" Well what about them? What I'm getting at is that aircraft
design relies heavily on modelling, and very acurate modelling at that.
Airframes of the early 20th century were built without modelling by
the "eye-ball" method and then either worked or didn't. I doubt whether
the authors of the article would climb into a plane which hadn't dad
some modelling in its development. So the problem is not with verifyingt
models per se but with certain kinds of models, specifically earth
science or biological models. I do not think that these are inherently
any more "fictional" than any other models. However, I also think
that they are dealing with systems which are both more complex and
less well understood than an airplane wing. This in turn derives
partly from the difficulty of measuring aspects of the systems (measured
whole forest canopy photosynthesis lately?) and partly from
the way the modeling process is often undertaken. It is my opinion
that the development cycle of the modelling process is too long, and
that this is "rate-limiting reaction" which dictates the speed of
progress in this research. If one imagines a model as an experiment
if model creation and testing takes 1 year, then that equals 1 expmt/yr.
Compare this with a field which can produce 1 expmt/mo. which field
is going to make more rapid progress?

To address this I propose the following:

1) Whenever possible models should be simple rather than complex, addressing
simulation of only the essential system components not every possible
internal interaction and variable.

2) Model testing should be an integral part of model development rather
than an after thought near the project's end. This requires the adoption
of a generally accepted set of model testing tools, implemented in a form
which is easy to use, so that for example the test-modify-retest
cycle would be a task for an afternoon rather than a month. This would
also reduce the quantity of hype around models in which testing was
delayed until years of effort were invested and it became necessary to
"put the best face on" test results which showed minimally adequate
performance. Which leads directly to point 3.

3) Model development tools need to be radically improved. The development
environment *itself* should have an easy to use interface. Most scientist
do not have the training to be really good at programing in command-line
languages like FORTRAN and C++, and sometimes if they do gain the
knowledge they get lost in the elegance of coding and loose sight of
the science. Bringing in non-scientist programmers is problematic as
well since there is always something lost in the translation and
the programmers often don't fully grasp the nature of the problem they
are working on. There should be a way to construct models in which
the interface facilitates contact between the scientist and the
implementation of the scientific ideas, with as little work as
possible. Using the model construction program should be as far
as possible transparent, and the function of the constructed model
should be visually intuitive to another scientist not previously
familiar with it. This facilitates "inter-human portability"
otherwise known as "understanding how the damn thing works."

Given such an approach and such tools it should be possible to
rapidly create and test *several* models and select the best from
among them, and subject it to rapid iterative improvement process.
Ultimately, we should then arrive at better models as well as
the limits of current experimental understanding.

Jonathan Haskett

>
> set agmodels-l mail digest
>
>

Not to embarrass anyone but don't forget that _commands_ must be
sent to listserv@unl.edu and discussion is sent to
agmodels-l@unl.edu....

Thanks,

Jerome Pier
List Owner
jp@unl.edu

On Feb 16 Roger Smith wrote:

> Model Debaters:
> I cannot help but observe that much of the debate going thru the ether
> on models, user interfaces vs. underlying code and assumptions, etc,. is
> somewhat like the story of the description of the elephant by the 5 blind
> men. Each sees well the part he has contact with, and each is at least
> partly right, but the big picture contains all of them, and a debate as to
> which is more important seems at least from one perspective to be folly.
>
> Roger E. Smith
> roger@lily.aerc.colostate.edu
> 303 491 8263

Roger,

I'm likely one of those blind men you refer to, but I
respectfully disagree. I don't think I was insisting that my
concerns were the most important aspect of modeling, nor did it
strike me that anyone else in the discussion was blindly insisting
that their concerns were most important. I think we were sharing
concerns about different aspects of the problems of modeling.

Greg McIsaac

But of course all parts of the elephant are important, aren't they!

In all matters concerning the modeling of elephants I defer to the
expertise of Dr. Starfield who has extensive experience in this area.

Jonathan Haskett

set agmodels-l mail digest

On Wed, 16 Feb 1994 10:22:26 -0600, Greg McIsaac wrote:

>
>
>Roger,
>
>I'm likely one of those blind men you refer to, but I
>respectfully disagree. I don't think I was insisting that my
>concerns were the most important aspect of modeling, nor did it
>strike me that anyone else in the discussion was blindly insisting
>that their concerns were most important. I think we were sharing
>concerns about different aspects of the problems of modeling.
>
>Greg McIsaac

Greg:
You took my comments particularly, and they were meant generally. And I
was not intending to say any modeller was blind, but only that we are all
most concerned with the part of modelling we have most or most current
problems with. No mephor is perfect. But none of us can see modelling
from all angles. Those concerned with putting good science in models need
the view of those who are users to make the best interface, for example, if
we want to get science in the hands of decision makers.

Roger E. Smith
roger@lily.aerc.colostate.edu
303 491 8263

set agmodels-l mail digest

Please sign me off this list, too noisy!

I cannot decide whether all parts of the elephant are important unless I
am told the context. To me that is the crux of modeling. A model does
not simulate reality, it is a tool designed for a purpose. If the
purpose is to figure out how high an elephant can reach, then clearly all
parts are not important.

If one thinks in this way, then the most important information one needs
to have about a model are (a) the purpose for which it was designed, and
(b) a list of assumptions. Without both pieces of info, I cannot judge
whether the model meets its purpose, nor can I decide whether I can use or
modify it for a somewhat different purpose. Front ends and other devices
to make models user friendsly are not (or should not be) devices that
discourage thoughtful use of a model; on the contrary, they provide
ESSENTIAL information for intelligent use of a model.

I ask students in my modeling classes to make write down what they would
look for if they were asked to review a modeling paper, then I give them
a paper and ask them to review it. I reckon it is more important for them
to be able to discuss why a model (or modeling paper) is "good" or "bad"
than to learn the intricacies of modeling. This is what modeling
methodology is all about.

Tony Starfield

>Not to embarrass anyone but don't forget that _commands_ must be
>sent to listserv@unl.edu and discussion is sent to
>agmodels-l@unl.edu....

Another important point is that since the return address goes to the
entire list, users should be sure to address personal responses to
individuals, and not just use the reply function.

Consequently it is desirable to include your reply address in the body
of the message. Some mailers bury it in the header information.
It is a good idea to include your address in your signature file.

Bill
--
Bill Silvert at the Bedford Institute of Oceanography
P. O. Box 1006, Dartmouth, Nova Scotia, CANADA B2Y 4A2
InterNet Address: silvert@biome.bio.ns.ca
(the address bill@biome.bio.ns.ca is only for mailing lists)

Greetings to the AGMODELS-L. I sent this introduction out a few
days ago, but did not see it posted, so I will try again. We have
been silently reading the postings to this group, but several weeks
ago the comments of Drew Laughland, and Jerome Pier's reply, got me
interested in joining the discussions.

My name is Mike Munster (B.S. in agronomy, M.S. in plant pathology).
I am currently a research assistant at North Carolina State University in
Raleigh, NC, USA. My colleague Gyanendra "Gyani" Dhakwha has a
Ph.D. in meteorology, with experience in crop modeling. We both
work for the Agroecosystem Resource Group of the Environmental
Monitoring and Assessment Program (EMAP), a effort to describe
the condition of the natural resources of the U.S.A. Our group
is concerned with the ecological state of agricultural areas.

Gyani and I work on indicators of crop productivity, and we are
attempting to use of crop growth models in a sort of backward way
to factor out the known sources of variation (managment, soil,
weather) in crop yield as reported in our surveys. We hope that
(a) the variance is reduced and (b) subtler aspects of the system
health might be uncovered. Our data come from across the states
of North Carolina (1992) and Nebraska (1993). Models to be
tested are SOYGRO, CERES-maize, CERES-wheat, and EPIC.

Another colleague of ours says that if a model has an acronym, it
is too complicated to be trusted, so we will consider simpler ideas
that still have underlying models: nitrogen use efficiency and
water use efficiency. Drew Laughland's comments drew my
attention because I, too, need simple, regionally applicable
indices. Someone had suggested calculating nitrogen excess as
the (N applied) - (crop yield)*(concentration of N in the grain).
I have been reassured by at least one person that the grain N
concentration is consistent enough to make the index useful, at
least for maize, but I am skeptical, especially since I must
consider other grain and nongrain crops, too. Drew's suggestion
of (N applied) - (crop yield)*(minimum N rate needed) sounds
interesting. Has anyone out there done determinations of either
N concentrations in harvested products or N response curves
(seems like the minimum N rate needed might come from there)?

<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>
| MIKE MUNSTER |
<> EMAP-Agroecosystems phone 919-515-3311 <>
| 1509 Varsity Dr. fax 919-515-3593 |
<> Raleigh, NC 27606 email mike_munster@ncsu.edu <>
| USA |
<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>

unsubscribe

This just came to me from Rick Olson. I think some of this group might be
interested:

CALL FOR PAPERS
Decision Support Systems in Agriculture
A Session of "Decision Support - 2001"
Toronto, Canada,
September 12-16, 1994



In recent years, computerized decision-support systems have
become important tools in the management and analysis of
agricultural systems. Such systems are being used, or are
under development, in such diverse fields as cotton pest
management, rice production and viticulture. Their usefulness
is not surprising, because agricultural domains are tremendously
complex. Often, management revolves around multiple, and at
times conflicting, objectives. In addition, the knowledge
required to do a creditable job tends to be multi-disciplinary in
nature.

"Decision Support Systems in Agriculture" is an opportunity to
present state-of-the-art research in this exciting field, and to
exchange information and ideas with other practitioners.

Appropriate topics include, but are not limited to, the following:

GIS-based systems
Knowledge-based (expert) systems
Model/Simulation-based decision systems
Agricultural data-base management
Integrated systems

For the purposes of this session, "agriculture" is considered in
the broad sense of managed ecosystems, and thus potentially
includes such domains as aquaculture, range management and
forest pest management, as well as production agriculture.

The session will consist of eight presentations. Abstracts of no
more than one page in length should be submitted, by April 1,
to:

Dr. Richard L. Olson
USDA-ARS
P.O. Box 5367
Mississippi State, MS 39762
601-324-4367 (phone)
601-324-4371 (fax)
rolson@asrr.arsusda.gov

Due to time limitations, we may not be able to accomodate all
submitted papers. Submitters will be notified of acceptance/rejection
by April 15.

-----------------------------------------------------------------
Nicholas D. Stone ISIS Lab
Department of Entomology Phone: (703) 231-6885
202 Price Hall FAX: (703) 231-9131
Virginia Tech Internet: nick@vt.edu
Blacksburg, VA 24060 BITNET: nstone@vtvm1
-----------------------------------------------------------------

Hello,
Can anybody give me an email address for W.R.Gardner
He was at

Collge of Natural Resources
University of California
Berkley
CA
USA

in 1990.....

Thanks

Graham Lewis

I am a freelance working with cereal crop models on a very
small basis and have had *NO* experience with academic
modelling. I have recently started to market a very simple
wheat model (which is a PC implementation of the AFRC
ARCWHEAT model) in the UK aimed at end users. A lot of
farmers here are not computerised or are just becoming
computerised. Even most consultants are not very familiar
with computers. In these circumstances the user interface of
any model must be simple otherwise the user just gives up.
Rather than use FORTRAN which seems to be widely used in
academic circles I use Clipper which is like a cross between
dBase and C. Clipper is partly object oriented and I have OO
extensions for the language. There is a loss of performance
but there are considerable gains in terms of program
maintenance and flexibility.

While a complex model can never be understood by the end
user there must be a case for programs with easy user
interfaces if only to encourage the understanding and use of
models by ordinary people.

Regards
Steve

*===========================================================*
| Dr S. H. Binns Tel +44(0)482 813711 |
| 54 Park Road Email |
| Sproatley Internet Address: |
| Hull sbinns@cix.compulink.co.uk |
| HU11 4PG |
| United Kingdom |
|-----------------------------------------------------------|
| Freelance computer systems designer specialising in |
| agricultural crop simulation models. |
| |
*===========================================================*

Hello, my name is Robert Pollock. I am a graduate student
in Agricultural and Biological Engineering, and I am primarily
interested in modeling conditions in controlled horticultural
environments, such as growth chambers and greenhouses.

Many thanks to Jerome Pier for starting this discussion group.

My main concern is whether to continue programming in Fortran and Matlab
or look for a packaged software modeling environment. Packages
may become outdated (CSMP), limit access to a samll circle of users
(Ideas), or hide sources of computational error in proprietary software.

Should I be using a packaged platform (ANSYS, Patran)
or continue with Fortran ?

****************************************************************

Robert K. Pollock 803/656-4078
Greenhouse Engineering Research Fax:656-0338
Agricultural and Biological Engineering
Clemson University, Clemson, South Carolina 29634-0357

****************************************************************

> From: "Greg McIsaac" <gfm@age2.age.uiuc.edu>
> To: "Multiple recipients of list" <agmodels-l@unl.edu>

> > Date sent: Mon, 14 Feb 1994 12:16:48 -0600
> > Send reply to: <agmodels-l@unl.edu>
> > From: "" <roger@lily.aerc.colostate.edu>
> > To: Multiple recipients of list <agmodels-l@unl.edu>
> > Subject: Re: Modelling Comments
>
>important because
> > users are more sophisticated now, and Georg Hoermann claimed
>that models
> > are not used because of similar reasons, and complained that none
>of the
> > hydrologic models can be used by their biologists. Part of the latter
> > problem is that people are beginning to think of computer models of
>complex
> > processes like they think of computer games. The biologists, for
>example,
> > (not to pick on them) want a hydrologic model that will substutute for
> > hydrologic knowledge. That is in my opinion part of the problem.

This is the kind of argument I hear all day 8-). Ten years ago,
the people from our computing center said that a non-computer
scientist is not able to work with data base, graphics software or
the like.
Now the hydrologists say that biologists are not
able to work with hydrologic models, because they lack
hydrologic understanding.
If a model is not able to control common errors, the model
is wrong and not the user. I know that these two positions
are extreme positions, the truth is somewhere in between and
it applies not only to biology and hydrology.

Example: error control in input data. My word processor
always complains about wrong page margins, but soil
water models do not give an error message if your
soil physics or transpiration-data values are out of range.
("the conductivity is to high for clay soils").
They have no pre-defined soil-physics in a data base.
I know that this is not the *best* solution, but it is
better than calculating water balances without
any soil data.

Models *are* a way to transfer knowledge to other
scientists. It is a task of the author to make sure that
the knowledge is represented in an understandable
way.

BTW, I have an example of a drag&drop interface here,
but it is entirely in German language and runs only
with OS/2.

Gruss Georg

Georg Hoermann - georg@geki.toppoint.de, schorsch@pz-oekosys.uni-kiel.d400.de
Ecosystem Research Center - Kiel University
Schauenburger Str. 112 - D-24118 Kiel

As both a farmer and a modler I think you will need to make the farmer
understand the model unless you can prove the model to be very good.
Gordon

Gordon Couger
Biosystems & Agricultural Engineering
Oklahoma State Universtiy Stillwater OK

Hello, everybody:

Nice to meet you all on screen in the beginning of the new year !

My name is Hao WANG. Wang is my family name. I was born in June 1963.
I came from Nanjing University, China, there I have worked for 5 years after I
got my Ph.D degree as a Associate Professor in the Department of Atmospheric
Sciences. My teaching & research fields mainly are Mesoscale Modeling,
Boundary-layer Meterology, Micrometeorology and Climate dynamics. Now I am
visiting Iowa State University, I just have developed Shelterbelts terbulent
models, which can reproduce real aerodynamic flows around and above
shelterbelts ( both leeward and windward ), I will apply the models to
Agricultural / forest and bioclimatological system simulation. Could you give
me some data about snow drift effect, rainfall effect,
evaporation/evapotransration effect, sand and soil erosion effect ......... as
well as air flows effects ............. all the data about shelterbelts/
windbreaks. Also, could you give me suggestion about which literatures I must
read for the forcoming modeling, such as soil and vegetation evaporation
processs, could you give me a list ? I greatly appreciate your help !

Thanks a lot

******************************************************************************

Dr. Hao WANG
Dept. Geo.& Atmos. Sci.
3010 Agronomy Hall
Iowa State University
Ames, Iowa 50011
U.S.A.

Phone: (515) 294-4758

Fax: (515) 294-3163

E-mail: wanghao@iastate.edu

*************************************************************************

SEE YOU LATER !

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

!finger jd

Oklahoma has just got over 100 weather stations around the state reproting
in every 15 minutes with data taken every 5 minutes. I don't know if it is
ready for distribution yet but you might contact Ron Elliott here a Okstate
his address is elliott@olesun.agen.okstate.edu.
gordon

Department of Agriculture 22/2/94 8:18
SUBJECT: RE>Models and user interface.
I am interested in comments on user interface design for simulation models
used for tactical or strategic decision support. I would like to be
provocative, make some comments and pose some questions:

o Many users have not paid for an interface and suffer as a result.
I would argue that the price sensitivity of rural producers in their purchase
of the first computer leads them to buy cheap and nasty hardware, mediocre
software and a system, in which little or no attention has been paid to
making the interface suitable for a first time user. [I'll come clean. My
preference is for Mac/ Windows rather than text based interfaces] So you see
8088's or 286's with DOS and users needing cheat sheets to remember all sorts
of codes and commands. {what I call a RAT (remember and type) interface}
Much of their intellectual effort goes into driving the thing rather than on
the problem they are trying to solve with it.
To make matters worse, because the user feels they had such a poor return on
investment in their first, such system, they are reluctant to throw more
money at the problem to acquire a system with a decent interface. So even
when people like use come along with a better software, it is a battle to
overcome these prejudices.
People coming along later, who start with Windows or Mac don't know what all
the fuss is about. With modern applications development tools such as
Toolbook and Hypercard making good interfaces is a snap.

o Do users have time to wait for an answer?
My experience with simulators similar to AFRC/Wheat is that they run far too
slowly on microcomputers and are very complicated for rural producers to
operate.
When a simulator is finished it usually comes up with a set of deterministic
figures for yield or whatever. Then the wheat grower typically asks about
the risk involved. This implies the need for probabalistic output which
takes into account weather variability, hence Monte Carlo trials, hence more
runs, hence more time... blah!!
What some people have done, including myself, is to generate look-up tables
for a finite array of input conditions with the simulator and then write a
Windows/Mac interface which allows the user to select inputs with pop-up
menus or buttons
and see the results plotted instantly as probability curves.
Our growers seem to like exceedance probabilities. Not sure how widespread
this is?
The loss of information resulting in descretization of inputs is offset by
model accuracy, or lack thereof, anyway. All they want is some corroborating
evidence which they combine with other information. None of the growers I've
met are going to base their decisions only on a DSS.

o Interface design principles.
My bias is for Mac's so I fall back on the "Apple Human Interface Guidelines"
which is full of tips and plenty of references. I guess there are similar
books for Windows developers. These sorts of texts are well worth a read.

Glyn Rimmington
gmr@agriculture.unimelb.edu.au
------------
From: Stephen Binns <sbinns@cix.compulink.co.uk>
Subject: Models and user interface.
I am a freelance working with cereal crop models on a very
small basis and have had *NO* experience with academic
modelling. I have recently started to market a very simple
wheat model (which is a PC implementation of the AFRC
ARCWHEAT model) in the UK aimed at end users. A lot of
farmers here are not computerised or are just becoming
computerised. Even most consultants are not very familiar
with computers. In these circumstances the user interface of
any model must be simple otherwise the user just gives up.
Rather than use FORTRAN which seems to be widely used in
academic circles I use Clipper which is like a cross between
dBase and C. Clipper is partly object oriented and I have OO
extensions for the language. There is a loss of performance
but there are considerable gains in terms of program
maintenance and flexibility.
While a complex model can never be understood by the end
user there must be a case for programs with easy user
interfaces if only to encourage the understanding and use of
models by ordinary people.
Regards
Steve
sbinns@cix.compulink.co.uk

>Many thanks to Jerome Pier for starting this discussion group.
>
>My main concern is whether to continue programming in Fortran and Matlab
>or look for a packaged software modeling environment. Packages
>may become outdated (CSMP), limit access to a samll circle of users
>(Ideas), or hide sources of computational error in proprietary software.
>
>
>****************************************************************
>
>Robert K. Pollock 803/656-4078
>Greenhouse Engineering Research Fax:656-0338
>Agricultural and Biological Engineering
>Clemson University, Clemson, South Carolina 29634-0357
>
>****************************************************************

Robert -

Our approach was to quickly drop FORTRAN because of lack of reusablity
of model components, poor language structure and poor development
tools. We share similar concerns regarding simulation "packages"
which are not extensible and relatively efficient. We are currently
on two tracks. Track 1 utilizes C++ and builds on a heavilty
object-based simulation framework, adding C++ "simulation objects" as
needed. We use this for our "heavy-duty" simulations where
flexibility, dynamic configuration, custom user interfaces, external
hooks and performance are primary concerns. Track 2 utilizes a
STELLA-like rapid development tool (MS Windows-based) for doing our
simpler "quick and dirty" models and for teaching. Works out very
well for us.

- John Bolte
Bioresource Engineering
Oregon State University

unsubscribe r. gommes

Dear all!

I am working with transportation models, with special attention on the
losses that can occur between the supply and demand sides, due to bad road
conditions, lack of specialized vehicles (e.g., refrigerated cars) and
inapproapriate packaging.

The products that I am dealing are basically fresh fruits and vegetables,
for which I am trying to estimate a kind of "loss (or damage prediction)
funcion", that perhaps could be aproximated by the eventually available
"shelf lives", that could have been estimated for those products, under
different temperature and humidity conditions.

I would very much aprecciate in hearing about your experience in the matter,
recommendation on some pertinent literature, etc.

Thanks in advance!

Jose Caixeta
Institut fur Agrarokonomie
Universitat Kiel

E-mail: apm17@rz.uni-kiel.d400.de

Error capturing within models is becoming more and more important. In the
RZWQM we have now developed a routine to scan the input for common errors.
This should save us a lot of pain and agony. More error checking must be
done in the future...I am all for this approach.

+---------------------------------------++---------------------------------+
| Dr. Jon D. Hanson || Comm: (303)490-8323 |
| USDA, Agricultural Research Service || Fax: (303)490-8310 |
| Great Plains Systems Research || jon@gpsrv1.gpsr.colostate.edu |
| 301 S. Howes, P.O. Box E || FTS2000: a03jonhanson |
| Fort Collins, Colorado 80522 || |
+---------------------------------------++---------------------------------+

Greetings, this is belated introduction. I have been reading the mail for
sometime, but did not take time to introduce myself.

I am Bruce Curry, currently a visiting professor at Univ of Florida in
Agri. Engineering. Have been involved in crop modeling for a long time.
Developed, with a lot of post-doc help, a physio-based soybean simulator,
SOYMOD, in the mid-70's and have been involved in various ways since. I
took early retirement from Ohio State(Reseach Center at Wooster) in 1988 and
since have been working part time here at U of Fl on various projects all
related to using crop models and the GCM's to estimate the effect of climate
change on crop production and water use.

It is interesting to read on this listserver about concerns that are
still relevent after 25+ years, such as validation/verification, what
language, how friendly the model needs to be, etc. We as modelers and
scientists must continue to be open to new ideas like the ones we read on
this list. One must decide what the model is to be used for and by whom,
BEFORE, the designing and development starts. We must make them more user
friendly, but still based on the best science available. The threat of
misuse is real.

I have rambled enough. Do look forward to more ideas and discussion
on agmodel-l.

R. Bruce Curry, Dept. of Agr. Engineering, Rogers Hall, Univ of FL,
Gainesville, FL 32611 0570. Email - rbcurry@agen.ufl.edu

Hello,

I'm a new subscriber to Agmodels-l. After a brief archive reading
I have saw every member introduce themselves.
Than I do it..

I'm TABOUREL florence and since december 93 I've finished my study
in the French Great Scool of Agronomy (INA-PG = Institut National
of Agronomy in Paris-Grignon).
Before I was successful in a Licence and Maitrise of Biology
at St Jerome University (in South of France).
Now I'm doing a PhD in France at the National Institut of Agonomy's
Research (INRA).

My Thesis topic are the Carbon and Nitrogen flux Modeling in grass
crop.
I'm not a specialist in Modeling Process than I have some formations
at Jussieu University (PARIS) to learn the used of this "tool".
I think this group of discussion could help me to progress in
this areas because presents modelings problems and idears are
treat.
Its the best way for mee to contact specialist and non specialist
for ask them some questions. Actually I don't think I
could answer to your question and participate actively to forum
, but I'll was pleased to read your mails.
I hope my questions and my English wouldn't be too funny or
so stupid. I wish some of you answer to my mail.
Then my first question is a technical question about a computer.

*Do you know a software for compartimental analyse ,
??
What is its specifical fonction ?
Is it friendly used?
What are its limitations?
What is its price and reference?

Thanks for this news.
Your faithfully ,
FLO.

By way of introduction, I am Peter Leroy, Ground Manager of the Sydney
Cricket Ground where we are involved in other work beside the maintenance of
International sports fields. i suppose you could say that I am one of the
end users of the most peoples work in this newsgroup.

Over the last few years we have been involved in some minor r&d into hybrid
turf-grasses for for our proposed indoor natural turf-grass stadium.

One of the reasons I was particularly interested in the work that you do
Florence. :]

The soils that we work with vary from 90/10 sand loam mixes to a
montmorillonitic clay soil for the preparation of cricket wickets. These
soils are rare in Australia but are ideal for the sport of cricket. A
general particle analysis shows approximately 60% clay, 10-12% coarse sand,
10% fine sand, 10% medium sand, 5-8% organic matter and silt. This is
compacted to 500 psi with 30-60% soil moisture content. Then we are expected
to grow turf :]

Any information I can gain on the way soils react is a benefit to us as the
"end users"...

If I can offer any assistance I would be pleased to do so... otherwise I
read this newsgroup with great interest.

:peter

I have read the paper by Oreskes, Schrader-Frechette and Belitz in SCIENCE
(vol 263 pp 641-646 94) and it seems to me you could summarize it as follows:
-- all models are imperfect representatiions of truth
-- to "verify" a model means to demonstrate that it is true
-- to "validate" a model means to demonstrate that it is internally self-
consistent and (I THINK they say this) that it can be shown to mimic
reality for some specific case

Using these two definitions they then argue that no model is verifiable or
validatable. I don't know any informed individual in the known universe that
would disagree with their argument about verification. As for validation, I
am reminded of the philosopher who, when shown a field of spotted cows said
"well, of course we only know that they are spotted on this side."

Every decision we make is based on a model and the issue is not certainty but
the reliability of the assumptions we make. It is not clear to me what
relevance Orestes et al.'s demand for certainty has in the real world. Policy
WILL be made and models WILL be used--hopefully they will be as good as they
can be, but the only certainty is that they will fail these author's criteria.

R. DON WAUCHOPE, RESEARCH CHEMIST DON@TIFTON.CPES.PEACHNET.EDU
USDA AGRICULTURAL RESEARCH SERVICE (912) 386-3892 FAX 386-7225
POB 748, U. GA COASTAL PLAIN EXPERIMENT STATION
TIFTON, GA 31793
------------------------------ Cut here ------------------------------