232

 

1 DIVISION OF ADMINISTRATIVE HEARINGS

DEPARTMENT OF ADMINISTRATION, STATE OF FLORIDA

2

SUGAR CANE GROWERS COOPERATIVE )

3 OF FLORIDA; ROTH FARMS, INC.; and )

WEDGWORTH FARMS, INC., )

4 Petitioners, )

vs. )DOAH Case No. 92-3038

5 SOUTH FLORIDA WATER MANAGEMENT ) 92-3039

DISTRICT, an agency of the State ) 92-3040

6 of Florida; et al., ) 92-6796

Respondents. ) 92-6797

7 - - - - - - - - - - - - - - - - - x 92-6799

FLORIDA SUGAR CANE LEAGUE, INC., ) 92-6800

8 UNITED STATES SUGAR CORPORATION; )

and NEW HOPE SOUTH, INC., )

9 Petitioners, )

vs. )

10 SOUTH FLORIDA WATER MANAGEMENT )

DISTRICT, an agency of the State )

11 of Florida; et al., )

Respondents. )

12 - - - - - - - - - - - - - - - - - x

FLORIDA FRUIT AND VEGETABLE )

13 ASSOCIATION; LEWIS POPE FARMS; )

W.E. SCHLECHTER & SONS, INC., )

14 and HUNDLEY FARMS, INC., )

Petitioners, )

15 vs. )

SOUTH FLORIDA WATER MANAGEMENT )

16 DISTRICT, an agency of the State )

of Florida; et al., )

17 Respondents. )

- - - - - - - - - - - - - - - - - x

18 100 Southeast 2nd Street

Miami, Florida

19 March 3rd, 1994

9:40 a.m. - 12:25 p.m.

20

DEPOSITION OF DENNIS P. LETTENMAIER

21 VOLUME III

22 Taken before BARNET I ABRAMOWITZ, court

23 reporter and Notary Public in and for the State of

24 Florida at Large, pursuant to Notice of Taking

25 Deposition filed in the above cause.

 

233

 

1

APPEARANCES

2

3 ON BEHALF OF THE PETITIONERS FLORIDA SUGAR CANE

LEAGUE, INC., UNITED STATES SUGAR CORP., and

4 NEW SOUTH HOPE, INC.

5 EARL BLANK KAVANAUGH & STOTTS , P.A.

One Biscayne Tower - Suite 3636

6 Two South Biscayne Boulevard

Miami, Florida 33131

7 BY: ROBERT H. BLANK, ESQ.

8

ON BEHALF OF THE RESPONDENT SOUTH FLORIDA WATER

9 MANAGEMENT DISTRICT

10 POPHAM HAIK SCHNOBRICH & KAUFMAN

International Place - 41st Floor

11 100 Southeast 2nd Street

Miami, Florida 33131

12 BY: JOSE A. LOREDO, ESQ.

13

ON BEHALF OF THE RESPONDENT-INTERVENOR

14 UNITED STATES OF AMERICA

15 THOMAS A.W. FITZGERALD, ESQ.

Assistant United States Attorney

16 99 Northeast 4th Street

Third Floor

17 Miami, Florida 33132

18

19

20

21

22

23

24

25

 

234

 

1 INDEX

2 Witness Direct Cross

DENNIS P. LETTENMAIER

3

By Mr. Loredo: 235

4 By Mr. Fitgerald 404

5 EXHIBITS

6 NUMBER DESCRIPTION PAGE

7 5 A document entitled 298

Analysis of Uncertainty

8 in STA Requirement

9 6 A document entitled 335

Estimation of WCA-2A

10 Sediment Phosphorus

Deposition, etc., dated

11 October 15, 1993

12 7 A document entitled 336

Review of Kriging methods

13 as Applied to Estimation of

Soil Phosphorus in WCA-2A,

14 dated January 5, 1993

15 8 A document entitled 346

Integration of Richardson

16 and Patrick Data into

Spacial Analysis of WCA-2A

17 and WCA-3 CM Phosphorus

Core Data, dated June 7,

18 1993

19 9 A document entitled 354

Estimation of Soil

20 Phosphorus and Topography

in the WCAs, dated May 5,

21 1993

22 10 A document entitled 359

Estimation of Soil

23 Phosphorus in WCA-2A

24

25

 

235

 

1 Thereupon --

2 DENNIS P. LETTENMAIER

3 was called as a witness and having been previously

4 duly sworn, was examined and testified as follows:

5 DIRECT EXAMINATION (Continuing)

6 BY MR. LOREDO:

7 Q. Good morning, Doctor.

8 A. Good morning.

9 Q. Just to remind you, you still under oath.

10 We will try to pick up where we left off

11 yesterday.

12 Going back over my notes, what did you do

13 in preparation for this deposition?

14 A. Preparation of this deposition, put

15 together the computer files and the mag tape that you

16 have, reviewed my files.

17 Q. And when you say you reviewed your files,

18 what did that consist of?

19 A. Sending copies of everything that was in

20 them, other than correspondence from other people to

21 me or other reports prepared by others that are

22 public information, to Mr. Blank.

23 Q. Did you have any discussions with anyone

24 other than Mr. Blank in preparation for your depo?

25 A. Perhaps you would like to rephrase that.

 

236

 

1 Q. Did you speak to anyone other than Mr.

2 Blank in preparation for your deposition?

3 A. Could you explain "in preparation"? In my

4 mind, there was no preparation other than to send

5 documents.

6 Q. Did you have any discussions with anyone

7 regarding your upcoming deposition as to -- to be

8 more specific, did you speak to Dr. Millard about the

9 deposition?

10 A. There was some discussion of it. In the

11 context of other -- there were no specific

12 discussions having to do with the deposition in

13 particular, but the subject did come up in some other

14 conversation we had, yes.

15 Q. Can you tell me the nature of those

16 discussions?

17 A. I think he made some mention of the fact

18 that when his deposition was taken, there was some

19 mention made of correspondence to and from me and

20 that there might be some question regarding that.

21 And I think I told him I couldn't recall

22 exactly what conversation that was, and I would deal

23 with it at that point; that I didn't know

24 particularly what conversation it was, nor was I

25 going to take the time to find out because I was busy

 

237

 

1 with other things.

2 Q. Was the conversation more as to specific

3 correspondence between the both of you?

4 A. I think simply there was mention that

5 there had been some correspondence to or from me -- I

6 don't recall which -- which there had been discussion

7 of in his deposition, and he thought I should know

8 that question might come up.

9 Q. Did you go into the context of the letters

10 or communications between yourselves?

11 A. I think he mentioned it, but I can't

12 remember.

13 Q. Do you know anything else that you spoke

14 about other than the correspondence?

15 A. No -- well, excuse me, in any context

16 having to do with his deposition?

17 Q. Yes.

18 A. I think he made some mention of the

19 deposition itself, some aspects of it, sure. It

20 didn't have to do with mine. He recounted some

21 aspects of his deposition.

22 Q. Can you tell me what your understanding of

23 his account of the deposition was?

24 A. I think the only thing he mentioned was

25 that there had been some question as to whether his

 

238

 

1 opinions were "final."

2 Q. Did you have any response to that?

3 A. No.

4 Q. Did he tell you whether or not his

5 opinions were final?

6 A. I think there was some question as to what

7 constituted a final opinion.

8 Q. From your understanding, what constitutes

9 a final opinion?

10 A. Well, I guess I don't have a firm

11 understanding in a legal context what constitutes a

12 final opinion, and I think most scientists never have

13 what they would classify as a final opinion because

14 they believe, and I do as well, that there is always

15 -- the scientific method allows that as new

16 information becomes available, one's opinion might

17 always change.

18 Q. You might never have a final opinion in

19 this case. Is that possible?

20 A. No, because my understanding is any final

21 opinion would be conditional upon the data that would

22 be available to that point, and that would be a

23 scientific engineering judgment based upon all the

24 information available to that point.

25 Q. Do you have any other things you might

 

239

 

1 remember in your discussions with Dr. Millard

2 regarding his deposition or your upcoming deposition?

3 A. Not that I can recall.

4 Q. Did you talk to anyone else that might

5 have been deposed in this case?

6 A. That's been deposed at any time? I have

7 talked to John Davis, but there's been no discussion

8 regarding deposition.

9 Q. What were your discussions with Mr. Davis

10 about?

11 A. We have had numerous discussions, mostly

12 regarding data and analyses that were ongoing,

13 exchange of information and so on, a good bit of

14 which we talked about yesterday.

15 Q. Did you ever discuss any of your opinions

16 with Dr. Davis?

17 A. Which opinions?

18 Q. Regarding the limits level in Loxahatchee,

19 the limits in the Park, any of your analysis in the

20 WCAs?

21 A. We have had numerous discussions over the

22 last year and a half.

23 Q. Well, I would like to probe your mind, and

24 for you to tell me what you talked about.

25 A. I'm not sure where to start.

 

240

 

1 Q. Let's start with the limits and levels in

2 the Loxahatchee and its derivation.

3 A. Well, there's certainly been discussions

4 regarding the entry and access data collected.

5 Q. And what were those discussions about, the

6 nature of those discussions?

7 A. There have been many discussions, some of

8 them over the quality of the data.

9 Q. Have you had any problem with the quality

10 of the data collected in the entry and access?

11 A. I think we discussed that yesterday. I

12 have reviewed some of the data that were collected,

13 and in particular data that were collected at the

14 same time and in the same helicopter entries,

15 technical definitions -- that's how they got to the

16 site.

17 But at the same time, it is my

18 understanding that the Department of Justice

19 consultant and the ESP rep, which in most or all

20 cases was John Davis, and in one case, I believe the

21 Water Management District person or persons took

22 samples at the same time, I have seen the data, the

23 results of the analysis of the data that were

24 collected, and John and I have discussed how and why

25 those numbers might be so different.

 

241

 

1 Q. Well, tell me how.

2 A. Well, it is clear, we talked about this

3 yesterday --

4 Q. We will probably be asking you questions

5 either just to see whether you are going to testify

6 the same as yesterday, and so --

7 A. That's fine, I am happy to go over the

8 same ground if you would like, but I believe I

9 mentioned one date in October, it might have been the

10 20th, I can't recall in particular, when there were

11 samples collected by the Department of Justice's

12 consultant, ESP and the District; that there were

13 substantial differences between the Department of

14 Justice numbers and the ESP numbers across all of the

15 sites visited, whereas the Water Management District

16 numbers and the ESP numbers were quite similar.

17 There had been discussion as to why that

18 had been the case, and I know what John Davis's

19 opinion of why that is the case is.

20 Q. What is his opinion?

21 A. His opinion is that sufficient care was

22 not taken by the Department of Justice consultant to

23 avoid withdrawing sediment and other detritus into

24 the sample, and that both he and the Water Management

25 District were quite careful to avoid doing that and

 

242

 

1 this made a big difference in the concentrations.

2 Q. There was some terminology I didn't

3 understand there.

4 Can I back you up to the first part of

5 your sentence?

6 A. Sure.

7 Q. Can you repeat that? I understand there

8 was a problem with the department's sampling versus

9 ESP and the District's, and there was a word in there

10 you used?

11 A. Detritus, sediment?

12 Q. Detritus, what does that mean?

13 A. Detritus is just surficial material that

14 might be withdrawn into a sample, for instance, parts

15 of leaves, dead leaves, that sort of material.

16 Q. Other than talking to Dr. Davis about the

17 quality, did you have discussions, other discussions

18 with Dr. Davis as to opinions you might have had or

19 he might have had?

20 A. Again, you would have to be more specific.

21 Q. As to the derivation of the limits in

22 Loxahatchee.

23 A. I'm sure we have had discussions at one

24 point or another, and I'm not sure I can recall the

25 exact detail.

 

243

 

1 Q. In general, did you have any recent

2 discussions with Dr. Davis about your opinions?

3 A. Regarding the limits in the Loxahatchee?

4 Q. Yes, let's stay there for now.

5 A. Only that the most recent data, that the

6 models that were derived based on the data in the

7 early and mid-'70s needed to be rederived using the

8 most recent data that, in fact, Dr. Millard has been

9 involved in, and I have reviewed some aspects of that

10 work.

11 Q. Are you saying to possibly to create a new

12 model based on recent data?

13 A. The question is whether the model on which

14 the limits were derived will hold up in light of the

15 more recent data, which almost doubles the sample

16 size.

17 Q. When you say "hold up," I guess you are

18 going to test the model with new data, is that it?

19 A. There's several ways of going about this,

20 but basically what the model says, as we discussed

21 yesterday, is that any observation taken out there --

22 and actually it is a certain type of mean over a

23 number of stations, mean of the algorithms as I

24 recall, the appendix E model -- depends on stage, and

25 it depends on what statisticians call a dummy

 

244

 

1 variable, which essentially means whether you are in

2 the base period or you are in the subsequent period.

3 So a particular question is whether the

4 relationship with stage, for instance, holds up given

5 the new data.

6 Q. Has that analysis been performed?

7 A. Dr. Millard performed some of that

8 analysis, yes.

9 Q. Are there any preliminary opinions?

10 A. I have an opinion based on the results

11 that I have seen, yes.

12 Q. Can you tell us your opinion?

13 A. Yes, that the new data are not compatible

14 with the old model, particularly as regards to the

15 dependence on stage.

16 Q. How did you arrive at that?

17 A. Well, I have reviewed various plots that

18 essentially superimpose the new data on the old

19 model, and the relationship with stage did not appear

20 to hold.

21 Q. When you say "superimpose," you took a

22 graph and you put one on top of another?

23 A. I didn't take, others have, the

24 information that I have reviewed.

25 Q. Is this work you are reviewing or is this

 

245

 

1 your own work?

2 A. This is work I was reviewing, Dr. Millard

3 has actually done.

4 Q. Was anyone else involved in this work?

5 A. In that particular aspect of the work?

6 Q. Yes.

7 A. No.

8 Q. Where are the plots?

9 A. I'm sorry?

10 Q. Where are the plots? You said --

11 MR. BLANK: "Where are," you are saying?

12 Q. Yes. The various plots that you

13 superimposed, where are those plots?

14 A. Dr. Millard has those. I don't have a

15 copy.

16 Q. He never gave you a copy?

17 A. No, he has them in his files.

18 Q. When you reviewed his work that was

19 superimposed, did you review it at his office, did he

20 provide you a copy of his work?

21 A. I think he brought it in. I can't recall

22 where the meeting was.

23 Q. But he never left with you a copy?

24 A. No, I didn't ask for a copy.

25 Q. You are opining that the new data is not

 

246

 

1 comparable and there's no particular dependence on

2 the stage?

3 A. That's not what I said.

4 Q. Okay, can you repeat it, then.

5 A. I didn't say there's no dependence on

6 stage, that's something I haven't made a final

7 opinion on. The relationship with stage that was

8 determined from the previous model does not appear to

9 hold, and it may be that there is no relationship

10 with stage.

11 Q. You are saying the relationship pursuant

12 to the model, actually pursuant to the model, that

13 there is a relationship with stage, does not hold up,

14 is that what you are telling us?

15 A. The previous model said that essentially

16 there are two effects that determine what a value

17 would be in this logarithmic mean. One is whether or

18 not you are in the baseline period, and the second is

19 stage, level of stage, and I believe that's an

20 average over three stages in WCA-1.

21 When you plot the more recent data versus

22 stage, the nature of the relationship is there's an

23 inverse dependence; that at higher stages, there's

24 lower concentration. That was generally what we

25 observed in the earlier data.

 

247

 

1 There's at least one observation in the

2 most recent data where the stage is very low and the

3 concentration is very low as well.

4 Q. What is that one situation?

5 A. There's at least one.

6 Q. Do you know how many there were?

7 A. How many sample dates?

8 Q. Well, you said there were data or there

9 were relationships, an inverse relationship between

10 stage and concentration levels?

11 A. That was determined from the prior data of

12 about, I believe, 14 sampling dates, including the

13 baseline period.

14 Q. Does the new data have that same inverse

15 relationship?

16 A. That's what I just indicated, that it

17 appears not to.

18 Q. It appears not to?

19 A. It appears not to.

20 Q. I thought you said at least one of them

21 appears not to, am I incorrect? I am trying to

22 paraphrase and understand what you are telling me,

23 and obviously if I am incorrect in what you said, I

24 would like you to tell me.

25 The previous data said there was an

 

248

 

1 inverse relationship between stage and concentration.

2 The new data, are you telling me that the

3 relationship -- there is no relationship, you are

4 saying that the concentration in stage, there is no

5 relationship with the data, it is coming out, some

6 are higher, the stage can be high and the

7 concentration can be low, and you are also finding

8 the stage low and the concentration low?

9 A. That is essentially correct. One has to

10 understand you have very small sample sizes here

11 which is really the basis for the problem.

12 Q. The small sample size, are we talking the

13 baseline period is small or the new data is small?

14 A. Both, but the new data have essentially

15 doubled total sample size. The original sample size

16 on which the model was based and the limits, is

17 exceptionally small from a statistical standpoint.

18 Q. You feel since the new data is double the

19 sample size, that it would be more -- I guess you

20 could determine more of an accurate model from the

21 new sample size or at least give new confidence to

22 the new sample versus the old?

23 A. Or the combined sample.

24 Q. Put them all together?

25 A. That's one way of doing it.

 

249

 

1 Q. When you put them all together, you run

2 into a lot more problems, don't you?

3 A. You have another problem with

4 understanding how one models the most recent data, if

5 one has to add another effect.

6 Q. The effect being the old data?

7 A. One way one could do it is to add an

8 effect that says that the data, either baseline --

9 they are mid-'80s, post-baseline or they are recent,

10 but then one is faced with the problem of

11 understanding, if they are recent data, why it is

12 that they should be different.

13 Q. That would be the baseline period?

14 A. From the post-baseline mid-1980s period.

15 Q. I was going over your letter dated May

16 20th, 1992 to Rick Burges, which we have marked as

17 Exhibit 4. Who is DPL & Associates?

18 A. DPL is me, those are my initials.

19 Q. Is that a company that you set up?

20 A. Yes, but I'm a sole proprietor.

21 Q. Is your contract with Peeples, Earl &

22 Blank with DPL & Associates or individually for

23 yourself?

24 A. Well, I have no current contract, we

25 discussed this yesterday. Whether the original

 

250

 

1 contract was with DPL & Associates or me in person,

2 I'm not certain. The entities are the same, so.

3 Q. Is there anyone else working with you at

4 DPL & Associates?

5 A. No. As I indicated, Ms. Dally Wilson has

6 been working under my direction, but she has

7 subcontracted to me and is not an employee of mine.

8 Q. Has Dr. Millard subcontracted with you?

9 A. No.

10 Q. You may not know this, but he directly

11 contracted with Peeples, Earl?

12 A. That's a question you would probably have

13 to ask Mr. Blank.

14 Q. In the letter it says: "I think these

15 comments more or less address your questions 1

16 through 5," I believe those questions posed to you

17 from Peeples, Earl.

18 I will give you the letter to look at, but

19 is there a way you can tell me what those questions

20 were?

21 A. I would have to look at their --

22 Q. Well, they haven't produced their letter.

23 A. Well, this is a long time ago.

24 Q. I understand that.

25 I would assume that those questions --

 

251

 

1 where I'm getting to is I want to start focusing on

2 what the scope of your work or the initial scope of

3 your work as addressed to you originally was.

4 A. I don't quite see where you are looking.

5 Q. The last page.

6 (Pause)

7 A. Without seeing the correspondence that had

8 the questions 1 to 5 in them, there's no way I can

9 recall. That was almost two years ago.

10 Q. One of the first things you bring up in

11 your letter, or at least it is a statement you made:

12 "I suspect that the estimate used was the sum of the

13 instant concentration times cumulative flow during

14 the interval between samples."

15 Can you tell me what were you thinking

16 about there?

17 A. Well, that's a load estimate or a flow

18 adjusted concentration, but I don't recall what that

19 pertains to.

20 Q. I will let you read it.

21 (Pause)

22 A. It says: "Flow weighted concentrations are

23 used." This I believe is in the Park, but --

24 Q. That was going to be another question.

25 Do you know whether or not you were

 

252

 

1 referring to the Park, Loxahatchee?

2 A. Well, since Loxahatchee doesn't use flow

3 adjusted concentrations, in my recollection, it

4 probably pertains to the Park, but it is a general

5 issue which pertains to other aspects of this

6 analysis as well, and it has to do with how loads are

7 estimated.

8 Q. I think we talked a little about this

9 yesterday.

10 You have a preferred method for

11 determination of loads?

12 A. I have indicated I think in the letter

13 there what the preferred method is based on my

14 experience.

15 Q. Can you expound on that a little bit for

16 my edification?

17 A. Sure. I think I reference some work in

18 there by the U.S. Geological Survey branch of systems

19 analysis at their headquarters in Reston, and there

20 are some papers that have appeared in Water Resources

21 Research documenting that method.

22 Q. Are you the editor for Water Resources

23 Research?

24 A. No, I'm a deputy editor.

25 Q. Can you expound on the preferred method

 

253

 

1 for me?

2 A. Well, the method that has been detailed is

3 what is sometimes termed the rating curve method, and

4 the rating curve method is that one estimates a

5 relationship between flow and concentration or load

6 -- actually, the relationship goes the other way

7 around.

8 The dependent variable is load or

9 concentration and the independent variable is flow.

10 One develops that relationship based on all

11 concurrent observations of concentration or load and

12 flow, then goes back --

13 Q. Can I stop you for one second. You are

14 saying load or concentration?

15 A. Yes, load is concentration times flow.

16 Q. Okay. You can keep going because that was

17 confusing.

18 A. One develops that relationship, and then

19 to estimate the long-term load, one applies that

20 relationship to all of the flow values because flow

21 is generally measured continuously and concentration

22 is not.

23 Q. Can you measure load by itself, is there a

24 way to calculate load?

25 A. I just told you what the calculation is:

 

254

 

1 Concentration times flow.

2 Q. So there's no way to just estimate -- is

3 load the mass of the pH of the phosphorus?

4 A. Load is the flux of concentration through

5 some cross-section. If one wrote it down in

6 mathematical form, it is an integral over an area and

7 an integral over time of the concentration times

8 velocity essentially.

9 Q. Velocity?

10 A. Times velocity.

11 Q. How does velocity fit in there?

12 A. Because if you have a cross-section -- if

13 you are computing a flux, it is how fast whatever it

14 is you are trying to compute the flux of is going

15 through the cross-section times the amount of it

16 that's there. The amount essentially is the

17 concentration. You have the velocity in there as

18 well.

19 Q. And you come up with a load, and load is

20 in metric tons?

21 A. Per some unit of time.

22 Q. Area is also a variable here?

23 A. You are integrating over area, so that

24 would go away.

25 Q. So it doesn't become a factor?

 

255

 

1 A. You have integrated over it, so when one

2 says take flow times concentration, the flow is

3 essentially the velocity integrated over the

4 cross-section. It is like an average velocity times

5 an area, that's why the area goes away.

6 Q. Now that I understand that, can I put you

7 back to the relationship between flow -- well,

8 wouldn't there always be a relationship between flow

9 and load then?

10 A. Yes, because flow is a component of load.

11 That's why some people develop the relationship

12 between concentration and flow. We discussed this

13 yesterday.

14 Q. You don't think there's a relationship

15 between concentration and flow?

16 A. In most of the data that I have examined

17 in South Florida, there is at most a weak

18 relationship.

19 Q. You did say that yesterday.

20 That's based on your physical analysis of

21 the data?

22 A. That's based on data analysis, yes.

23 Q. And when we say "data analysis," you just

24 looked at all the data?

25 A. In the computer programs you have, there

 

256

 

1 are a number of directories full of files, analyses,

2 plots and so on that have examined those relationship

3 for number of stages.

4 Q. I would have enjoyed reviewing and

5 preparing for your deposition and having time to

6 review all of this. Unfortunately, I didn't have

7 enough time to look at it. So some of my questions

8 may not seem to show an understanding of your

9 diskettes because I didn't have an opportunity to

10 look at them.

11 That's why I have to sit and probe on you.

12 I am sure all your work is there, and we will be able

13 to all figure it out when we have time to look at it,

14 but your deposition is now, so I will do the best I

15 can.

16 To get you back on the rating curve

17 method, the rating curve method develops a

18 relationship between the flow and the load?

19 A. The relationship can be either between the

20 load and the flow or the concentration and the flow.

21 Since the load is just the flow, which is the

22 independent variable multiplied by the concentration,

23 essentially you add one to the power of the

24 relationship.

25 Q. Didn't you say something about all

 

257

 

1 concurrent variables?

2 A. I believe those comments -- although I'm

3 not certain, but I believe those comments are related

4 to the Everglades Park, and in particular would be

5 applicable to things like the 12 structures where

6 there are instantaneous measurements of

7 concentration, but continuous measurements of flow.

8 Q. You think that's inappropriate?

9 A. No, what I have said is the concern that

10 is outlined in that paragraph has to do with some

11 mention somewhere in the document that I was

12 reviewing, there is mention made of use or there is

13 use in an analysis of a flow adjusted concentration.

14 And my concern was that apparently the

15 method used to estimate the flow adjusted

16 concentration, which is just load divided by flow, my

17 concern was that the method that was used to estimate

18 load may have been inappropriate.

19 Q. I guess from your point of view, it is

20 comparing apples and oranges because one is using the

21 continuous flow and the other one is just

22 instantaneous?

23 A. No, the concern was that my preference is

24 to use the rating curve method to estimate load. I

25 do not believe that the District has estimated loads

 

258

 

1 using the rating curve method. In fact, I'm

2 essentially certain that they have not.

3 I speculated in the letter as to what I

4 thought their method might have been because they did

5 not detail it in the information I was given to

6 review.

7 Q. I need to understand your rating curve a

8 little bit more.

9 A. Sure. Would you like specific references

10 to the method?

11 Q. That would help.

12 A. I am sure we can find them for you, they

13 are in Water Resources Research, public documents,

14 and that is the standard method used by the U.S.

15 Geological Survey, the Department of the Interior

16 agency.

17 Q. Can you give me an example using Park

18 data, give me an example? What I am trying to do is

19 grasp it and acknowledge what you are trying to do in

20 paragraph one, the criticism you put in paragraph

21 one.

22 Maybe if you can take me through an

23 example using the rating curve method, I can

24 understand how your measurement of load is done and

25 maybe I can relate it back.

 

259

 

1 A. You have to understand I have not in that

2 document made an estimate of load myself or flow

3 adjusted concentration. I merely commented how I

4 thought that was done in some documents I was

5 reviewing.

6 Q. I understand.

7 A. If you would like me to lead you through

8 the method again?

9 Q. Yes.

10 A. The method --

11 Q. Well, I would like you to lead me through

12 the rating curve method.

13 A. Yes, the rating curve method, you would

14 like me to lead you through that?

15 Q. Yes, I would.

16 A. The method is to take all of the dates on

17 which concentration measurements were taken, and in

18 this particular case, we are assuming that the

19 samples are collected via grab samples, which is, in

20 fact, to my recollection, the case at the 12

21 structures.

22 Q. Grab samples?

23 A. Grab samples as opposed to composite

24 samples.

25 Q. Grabbing it versus.?

 

260

 

1 A. A grab sample is essentially an

2 instantaneous sample which is taken via, essentially

3 as it sounds, grabbing some sample, or perhaps if it

4 is the U.S. Geological Survey taking the sample, by

5 taking samples across the cross-section to represent

6 variation, but still it is taken essentially

7 instantaneously.

8 Q. You mentioned something the U.S.

9 Geological Survey would do. They would take

10 different samples in an area?

11 A. What the U.S. Geological Survey attempts

12 to do in general, and I'm not certain whether they --

13 first off, I'm not certain whether they have been

14 involved in the sample collection pertaining to these

15 particular samples or not, but their method is to

16 take samples across the cross-section, to divide the

17 cross-section of a stream into panels and take

18 samples across and then to combine them into one

19 sample which represents the cross-section.

20 Q. Would they average that?

21 A. Well, the averaging is done because the

22 sample is composited and one concentration is taken.

23 The water is put together.

24 Q. So you end up taking an average?

25 A. Essentially, the average is done for you.

 

261

 

1 Q. Would it have been better to take an

2 analysis of each cross-section sample?

3 A. Probably not, but I don't believe the

4 Water Management District does any of that anyway.

5 So that's a moot issue.

6 Q. I understand, but I figured you would get

7 more accurate data because maybe there's a different

8 concentration on one side of the stream versus the

9 other for who knows what reasons?

10 A. No, when you are compositing, that's

11 exactly what's being done. The sample has averaged

12 the different components as opposed to going and

13 taking samples at each of the places across the

14 cross-section, sending them to the laboratory and

15 averaging laboratory results -- it is just an

16 expensive, it simply adds expense.

17 Regardless, to get back to where I was,

18 whether the sample is collected in that manner or is

19 collected, say, in the middle of, in or near the

20 middle of the canal stream or whatever channel you

21 are dealing with, that is taken -- what I refer to

22 grab sample, it means that is essentially taken at

23 some instant of time or over in a very short period

24 of time.

25 Those samples -- well, the discharge that

 

262

 

1 occurred when the sample was taken -- and this

2 usually is a daily value, but that can be debated --

3 is then, all of those pairs are used to develop what

4 is known as the rating curve.

5 The rating curve is simply the

6 relationship between those concentrations and the

7 flow that was occurring at the time that the sample

8 was collected.

9 Q. Y-X access?

10 A. X-Y access, yes.

11 Q. And plot them, and you develop a curve?

12 A. There are various ways of developing the

13 curve, but that curve is the rating curve.

14 Q. Does discharge affect concentration?

15 A. Where?

16 Q. Well, I think we talked about this. When

17 you say discharge, is that --

18 A. Discharge and flow are interchangeable.

19 Q. The same thing, so you are saying the same

20 thing?

21 A. Yes.

22 Q. Flow is being measured continuously?

23 A. What is measured continuously is stage,

24 but that is converted to flow, so there is a

25 continuous flow record, yes.

 

263

 

1 Q. So stage is directly related to flow or

2 are we measuring the same thing?

3 A. Yes, and I should be a little careful in

4 this region, it is not necessarily always stage

5 that's measured, in some cases there's some other

6 independent variable from which some flow record is

7 created. That's usually done by the U.S. Geological

8 Survey, in some cases by Water Management District.

9 Q. How is flow measured?

10 A. We discussed this yesterday. The standard

11 method in free-flowing streams is to measure stage,

12 and that is typically measured via some stage

13 recorder which records the level of water in, say,

14 the river continuously, and then there is a stage

15 discharge relationship developed by going out and

16 sampling across the cross-section, the velocity, in

17 splitting the cross-section off into different

18 panels, measuring the velocity in each of those

19 panels, multiplying out, essentially numerically

20 integrating the velocity times the incremental area

21 until you have covered the whole area, that gives you

22 an instantaneous discharge measurement.

23 The stage is known at the point that was

24 done, those are done over range of stages and a curve

25 is developed. That curve is then used to apply to --

 

264

 

1 that curve is assumed to apply at all times so that

2 the continuous stage readings are then converted to

3 discharge.

4 Q. And discharge is flow?

5 A. Yes.

6 Q. The same thing?

7 A. That being said, that method is not always

8 applicable in situations where one has canals and so

9 on. The USGS in some cases, I believe, the Water

10 Management District, uses other methods.

11 Q. Of determining flow?

12 A. Of determining flow.

13 Q. Is it a relationship with stage or is the

14 stage not even looked at?

15 A. I cannot say for certain what is used at

16 all of the structures at which flow records are

17 created in South Florida. In some cases, pump

18 revolutions might be used. There are other methods

19 of estimating flow in difficult situations.

20 Q. I had some idea about like a little meter

21 by the pump discharge just clocking away and that's

22 your flow?

23 A. Well, that's one sway. In fact, they can

24 rate the pumps and use pump revolutions in some

25 cases.

 

265

 

1 Q. Now, you talked about this yesterday, and

2 I guess this goes in the hydrological field. Are

3 there other factors that can affect stage and affect

4 flow, which I guess is precipitation, rain, other

5 factors.

6 Does that weigh in?

7 A. Other factors than what?

8 Q. That would affect stage?

9 A. I guess I am a little confused by the

10 question.

11 Q. Okay. Are there other factors looked at

12 when looking at stage in determining what flow is?

13 Is there any reason to worry about the other factors

14 because they are always constant, such as

15 evaporation?

16 A. I don't understand. When we are talking

17 about a discharge measurement, we are talking about a

18 particular point.

19 Q. Okay.

20 A. Evaporation is something that occurs over

21 an area. You are mixing apples and oranges.

22 Q. I see what you are saying.

23 So I guess the variable I am talking about

24 then would affect the concentration?

25 A. Perhaps.

 

266

 

1 Q. Perhaps?

2 A. I don't believe there's any claim that

3 flow is the only thing that would affect

4 concentration.

5 Q. Now, when you said in your letter: "I

6 suspect that estimate used was the sum of the

7 instantaneous concentration times cumulative flow,"

8 have you been able to look further whether or not you

9 confirmed your suspicions?

10 A. My suspicion now is that it was done a

11 little differently than what I indicated there.

12 Q. How differently? I thought we were

13 talking about that. You did say they were

14 instantaneous concentrations times cumulative flow.

15 Isn't that what they are doing?

16 A. Well, that's what I said in that letter,

17 that I suspected they were doing. They did not

18 document where their flow adjusted concentration came

19 from. Since then, I have had access to some of the

20 programs that the District uses.

21 I do not know if those programs were used

22 in particular for the flow adjusted concentrations

23 that were mentioned in the report, but if that is a

24 standard District method, it operates a little bit

25 differently than what I indicated, than what I said I

 

267

 

1 suspected was the case in that letter.

2 Q. When you say "operates a little

3 differently," in what way differently?

4 A. Apparently what they do is to interpolate

5 between concentration measurements in time.

6 Q. So it is really not instantaneous then?

7 A. Well, it is not clear that it comes out

8 much different in the end, but they do interpolate.

9 Q. When you do interpolate, you do it at two

10 different points, right?

11 A. In time.

12 Q. And you come out with in interpolator --

13 A. It is like assuming that the

14 concentrations will vary linearly between the two

15 dates -- the concentration was collected,

16 measurements were collected every two weeks, and you

17 are halfway in between, then it will turn out that

18 the concentration they estimated is just the average

19 of the two. It is a weighted average anywhere else

20 in between.

21 Q. Do you agree or disagree with that method?

22 A. Well, I don't think that's a very good

23 method.

24 Q. I guess you would use a rating curve

25 method?

 

268

 

1 A. I believe the rating curve method is

2 preferred, yes.

3 Q. Because you will end up with a

4 relationship. When you end up with a rating curve,

5 you will have a curve, and you develop a formula of

6 that relationship, of the concentration and the times

7 it was calculated?

8 A. Yes.

9 Q. And you say that's better than doing the

10 interpolator method?

11 A. The reason that's better is, for instance,

12 if there is a relationship with flow -- for instance,

13 say the concentration went up with flow and there

14 were some very high flow in, say, one of the days in

15 between the measurements, the averaging method would

16 simply say it was the average of the two observations

17 which might, for instance, correspond to low flows.

18 It would completely miss the high flow and hence high

19 load which would have occurred in between.

20 Q. And the rating curve would pick it up?

21 A. The rating curve will pick that up because

22 it knows there's a relationship between concentration

23 and flow, so essentially it will infer that a high

24 concentration would have occurred associated with the

25 high flow.

 

269

 

1 Q. In your rating curve, you would use all

2 data, you wouldn't pick anything out because it was

3 picked up -- I remember reading somewhere, I don't

4 think it has to do with water or flow levels, but

5 picked in a alligator path, I guess it affected the

6 data?

7 A. These are only concentrations that are

8 being collected at the same location as the discharge

9 measurements, i.e., in the case of the 12 structures,

10 these are structures, concentrations that are

11 associated with water moving through the structures.

12 That's not at an issue.

13 Q. So you are saying every test should count

14 and there's no reason to ever have an outlier?

15 A. No. One could argue, and I in fact would

16 not include times at which there was no flow in the

17 relationship, and I would develop separate

18 relationships depending on the direction of the flow.

19 Q. When you say "no flow," is that because

20 there's no water now?

21 A. They don't always have the gates open or

22 don't always have the pumps running. But on some

23 occasions, if it is on a sampling date, the District

24 has collected a sample, regardless of whether there's

25 water moving.

 

270

 

1 Q. And you would develop a formula or do

2 something with the data -- if there is no stage or

3 flow, I presume, I thought stage and flow were equal

4 and we are talking about no water at a particular

5 structure --

6 A. We are talking about no water moving.

7 Q. No water moving -- well, if no water is

8 moving, is there water there?

9 A. Generally, yes.

10 Q. There's always water at all the stations?

11 A. Generally they have a location where they

12 are collecting the concentration measurement, and

13 that is collected at that same location, for

14 instance, immediately downstream or downstream of the

15 prevailing flow direction from the structure might be

16 where the sample is collected.

17 There's a pool there, so regardless of

18 whether or not the water is moving, they can still

19 take a sample, and often do.

20 Q. If the stage is very low --

21 A. Well, we are really talking about a stage

22 difference at this point.

23 Q. What difference?

24 A. Between the water level above the

25 structure and below the structure. There's one way

 

271

 

1 it might be done. The stage issue and how that is

2 used to compute flow, is somewhat dependent on the

3 particular structure.

4 Q. What dependencies are we talking about?

5 A. Well, in most cases at the locations that

6 have been concerned to my analysis, the U.S.

7 Geological Survey is computing the flows. I don't

8 have full knowledge at every structure precisely how

9 they are doing that, and it does vary from structure

10 to structure.

11 One way of doing it is to relate, as I

12 indicated, the discharge to stage, and I said in fact

13 it is really the difference in stage between above

14 and below because that's what determines how much

15 water moves; and gate opening. It depends on the

16 structure and the configuration of the structure.

17 That really makes no difference for the

18 analysis that we are discussing here because we are

19 using the USGS computed flow records, however they

20 might do that.

21 Q. How much time did you spend reviewing the

22 SWIM Plan?

23 MR. BLANK: At what point, counsel? Are

24 you talking about prior to writing this letter, or?

25 Q. I'm talking prior to writing this letter,

 

272

 

1 let's start with that.

2 A. Prior to writing that letter, a few days.

3 Q. How much time did you spend reviewing the

4 SWIM Plan during this case?

5 A. I'm not sure what you would consider to be

6 reviewing the SWIM Plan.

7 Q. How much time did you actually pick up the

8 plan and read it?

9 A. And stared at the letters on the page?

10 Q. Yes.

11 A. I would have a very difficult time

12 determining how much time I actually spent looking at

13 the letters on the page.

14 Q. How about reading appendix E of the SWIM

15 Plan?

16 A. I have read it. I could not recall the

17 number of hours I have actually spent reading the

18 plan.

19 Q. Did you read it more than once?

20 A. I have looked at various parts of it more

21 than once, yes.

22 Q. Have you read appendix F?

23 A. Yes, I read appendix F.

24 Q. Yesterday, you weren't sure. The reason I

25 brought it up is I know one of the things you relied

 

273

 

1 on in your things you reviewed -- and I believe

2 appendix F was about modeling, if I am correct?

3 A. Again, I have difficulty recalling what is

4 in appendix E and what is in appendix F. E, I

5 believe, is the limits and F is --

6 Q. Models.

7 A. -- models -- there are lots of models,

8 that doesn't tell me too much.

9 Q. How many hours did you put into this case

10 to date?

11 A. It is easier for me to think in terms of

12 days, probably an average of four a month over the

13 last year and a half, probably 70 days, plus or

14 minus.

15 Q. Have you done any other consulting work

16 for agricultural industries?

17 A. Ever?

18 Q. Yes.

19 A. If you handed me a copy of my vitae, on

20 the last page there is a section -- I believe not,

21 but I would have to look at the section on consulting

22 there to make sure that I don't mislead you.

23 Q. Okay.

24 (Pause)

25 A. The closest thing, there would be some

 

274

 

1 work having to do with the Newlands project in

2 California, but that was a water quantity issue, and

3 in fact, as I recall, that was only a few days review

4 of a work plan.

5 Q. They were called Newlands?

6 A. N E W L A N D S, near Reno.

7 Q. What year was that?

8 A. 1985.

9 Q. How about any work for sugar companies?

10 A. No.

11 Q. Would you take a look at paragraph 1 of

12 Exhibit 3, and it says here in the last sentence:

13 "information with yellow tabs is not to be copied."

14 A. Paragraph 1.

15 Q. Under A 1, the very first one, it says:

16 information under yellow tabs, do not copy, and I'm

17 paraphrasing.

18 Do you know what that relates to?

19 A. No.

20 Q. No idea?

21 A. No. That was almost a year ago.

22 Q. I was trying to figure out what you yellow

23 tabbed, if you provided diskettes, and I couldn't

24 figure it out.

25 A. Well, there were paper copies provided as

 

275

 

1 well, certain information, and the yellow tabs

2 obviously pertain to the yellow copies.

3 Q. And I was trying to figure out what those

4 copies are, were, or what you didn't want copied?

5 A. I have no idea at this point.

6 Q. Do you think Lisa would, Ms. Wilson?

7 A. Remotely possible.

8 Q. We received correspondence from counsel

9 dated March 1. It said you were going to rely on the

10 Burns & McDonnell Everglades Protection Project

11 Historical Phosphorus Loads to the Everglades

12 Agricultural Area dated February 4th, 1994.

13 In what way do you intend to rely on that

14 document?

15 MR. BLANK: I think you mischaracterized

16 what I said. I think I said may or could.

17 MR. LOREDO: Okay.

18 BY MR. LOREDO:

19 Q. "Has reviewed and may rely." I will

20 restate the question.

21 Let's start with have you reviewed that?

22 A. I don't recall what's in that report. If

23 I can look at the report, I can probably tell you

24 whether I have reviewed it.

25 MR. BLANK: What was the date on it?

 

276

 

1 MR. LOREDO: February 4th, 1994.

2 MR. BLANK: That's the one right in front

3 of you.

4 THE WITNESS: This one? This is the one

5 with page 4 on it.

6 BY MR. LOREDO:

7 Q. Okay.

8 There are two parts of the report, you

9 have a 2 and a 3?

10 A. Coincidentally since I have not had a

11 chance to review it yet, I brought it with me, these

12 are both parts. I thought if I was here a little

13 early, I would read them.

14 Q. I will let you take a look at this letter.

15 Take a look at No. 4. The following report is a

16 report prepared by Dr. Walker. I want to know if you

17 have reviewed those reports or have intentions to

18 rely on any of those reports.

19 A. This is No. 4?

20 Q. Yes.

21 (Pause)

22 A. Yes. The items under No. 4, yes, I have

23 seen those.

24 Q. You have?

25 A. Yes.

 

277

 

1 Q. Do you intend to rely on them?

2 A. You would have to tell me what in a legal

3 sense you mean by "rely on."

4 Q. Are you using any of the information

5 provided therein for your work, or for the opinions

6 you are going to expound at the hearing?

7 A. I have reanalyzed some aspect of some

8 implications of what some of the assumptions and

9 models he has developed might imply, so in that

10 sense, I guess you could say they are being relied

11 on.

12 Q. Do you agree or disagree with those

13 assumptions and models?

14 A. I don't agree with all of them, but I'm

15 not sure the issue is necessarily, should be

16 characterized as being questioning his assumptions.

17 Q. You don't question another expert's

18 assumptions?

19 A. No, I didn't say that.

20 Q. What did you say?

21 A. I said I don't believe the issues as I

22 perceive them necessarily pertain solely or even in

23 large part to his assumptions.

24 Q. How do you perceive the issue?

25 A. I perceive the issue -- I perceive the

 

278

 

1 major engineering and scientific issue in many of

2 these analyses having to do with uncertainty.

3 Q. There's always uncertainty in these types

4 of analysis, is there not?

5 A. In natural systems, there's always

6 uncertainty, yes.

7 Q. Are there any other major perceived issues

8 other than uncertainty?

9 A. Well, one can always question the form of

10 the models that are used.

11 Q. Anything else?

12 A. I think between the form of the models

13 used and the uncertainty, I think basically you

14 capture the major issues, yes.

15 Q. Have you reviewed Reddy's February 1994

16 reports?

17 A. I am aware of those reports, I have

18 glanced through them. I don't recall the details of

19 how they are different. I need to look at a copy of

20 the report.

21 Q. You say you glanced through. Did you

22 spend a lot of time, did you make yourself familiar

23 with it enough where you can --

24 A. I glanced at it quickly.

25 MR. BLANK: You said February 1994?

 

279

 

1 MR. LOREDO: Yes.

2 BY MR. LOREDO:

3 Q. It would be unfair for me to ask you any

4 questions on your opinion as to that report?

5 A. I think that would be a correct statement.

6 MR. FITZGERALD: For the record, I would

7 like to make a statement, too. I think it is unfair

8 that we not be in a position to ask questions

9 regarding those, although I understand you may not

10 have had the opportunity to review this as yet.

11 But I would request from counsel that if

12 in fact the witness subsequently reviews either the

13 Reddy material or the February materials in Burns &

14 McDonnell, and in any way modifies his opinions,

15 thoughts, analysis or performs additional work based

16 on any of those materials identified in the March 1

17 letter from Earl, Blank & Kavanaugh that we have been

18 looking at, that delineates additional materials the

19 witness reviewed and/or may rely on, that we

20 immediately be apprised of that fact so we have the

21 opportunity in a timely way to redepose the witness

22 briefly about just that area.

23 MR. BLANK: I will stipulate to that, and

24 assuming it will be reciprocal with regard to any new

25 data that consultants receive. In other words, if

 

280

 

1 your consultants obtain data subsequent to their

2 deposition which they analyze and in turn change

3 their opinions based on that data or other reports,

4 that we be notified also. I think that's the

5 understanding among counsel anyway.

6 MR. FITZGERALD: Even beyond the

7 understanding of counsel, I sort of feel that's

8 inherent in the hearing officer's December --

9 MR. BLANK: I agree as long as it's

10 reciprocal here.

11 MR. FITZGERALD: The reason I draw a

12 distinction is the witness has already been provided

13 them and they are identified in the letter as having

14 been reviewed and may or may not be relied on, so in

15 this case, we may be in a slightly different

16 situation.

17 In reading the letter last night, I saw

18 the "reviewed and may rely," and I took the reviewed

19 in its common meaning, that the witness had already

20 absorbed them and was ready to go forward, but with

21 that understanding I am comfortable.

22 THE WITNESS: If you had a copy of the

23 Reddy report, I could probably shed more light on it

24 because my own recollection was there were numerous

25 documents in them, and how it pertains to earlier

 

281

 

1 documents and particularly some of Reddy's earlier

2 work which we have used rather extensively, would be

3 more apparent to me if I saw the document.

4 BY MR. LOREDO:

5 Q. Have you had an opportunity to determine

6 what was done in Estimating Inflows into the Shark

7 River Slough?

8 MR. BLANK: Counsel, are you back on Mr.

9 Lettenmaier --

10 THE WITNESS: We are back to May 20th.

11 MR. BLANK: What paragraph?

12 MR. LOREDO: Paragraph 2.

13 THE WITNESS: So the question now is?

14 BY MR. LOREDO:

15 Q. Have you had an opportunity to determine

16 what was done in estimating the inflows in the Shark

17 River Slough?

18 A. I don't recall exactly what that question

19 pertained to, but I have in fact subsequently

20 reviewed additional work which has some maps that I

21 think have a better idea.

22 Q. What work did you review?

23 A. Some work of William Walker's in

24 estimating trends. In fact I think I indicated

25 yesterday I have some of his computer files.

 

282

 

1 Q. What was done to estimate the inflows?

2 A. I believe there are discharge measurements

3 at S-333 and the S-12 structures.

4 Q. Did the reverse flow occur?

5 A. I cannot recall in some cases; yes, on

6 occasion. Not very often, as I recall.

7 Q. Do you know if that was handled, if

8 reverse loads were handled?

9 A. To the best of my recollection, in his

10 analysis, he does not consider days on which flow was

11 occurring in a reverse direction. It is also my

12 recollection that's a relatively minor issue, i.e.,

13 it does not occur on many days.

14 Q. Are we talking both at S-333 and S-12?

15 A. Yes.

16 Q. You believe at the S-12, there are reverse

17 flows?

18 A. I do not believe that occurs very often,

19 if ever, at the 12 structures.

20 Q. Going over paragraph 3 -- I think we

21 talked about some of this, and I apologize to have to

22 be rereading this, but, as you know, I got this

23 yesterday.

24 Can you expound on what you were telling

25 us in paragraph 3?

 

283

 

1 A. Well, I think the problem had to do there

2 with the fact that the model essentially compares the

3 concentration with what it infers would have occurred

4 under the same flow in the baseline period, and then

5 attempts to determine whether that's significantly

6 greater or not.

7 The concern had to do is when this is done

8 on an annual basis, there are a lot of different ways

9 in which one can get the same flow. It could occur

10 fairly uniformly over the whole year or much of the

11 flow could occur in a short period of time. I was

12 concerned that when one simply uses an annual flow,

13 that that can be -- one can be misled rather

14 seriously.

15 Q. You don't do anything to develop a

16 confidence in that or not, or actually totally reject

17 the idea?

18 A. The only additional work I have done

19 having to do with Shark River in the Everglades, to

20 which this pertains, the ENP to which that particular

21 comment pertains, is a review very recently of the

22 analysis which William Walker performed which we

23 received in electronic form just a few weeks ago.

24 That's a long answer to a way of saying

25 that I have not done much additional analysis of that

 

284

 

1 particular issue since that letter was written in May

2 '92.

3 Q. Are you going to do it or have you

4 reviewed Walker's tapes?

5 A. Yes, I indicated I have.

6 Q. Do you plan to do more work on it now?

7 A. I indicated yesterday that I had in fact

8 reanalyzed some of his relationships and that pending

9 receipt of the additional data from the District,

10 which Mr. Blank informed me has now occurred, and

11 when they are provided to me in a form that they are

12 readily usable, yes, I will.

13 Q. The additional data will have that much

14 effect on your analysis?

15 A. I suspect yes.

16 Q. Why?

17 A. Because the original period of analysis

18 only went through 1989, and there are four additional

19 years of data that have been collected since then.

20 Q. What's robust regression?

21 A. Robust regression is a means, method of

22 doing regression which is not overly influenced by

23 extreme values.

24 Q. Such as in a parametric approach?

25 A. Well, regression is by definition a

 

285

 

1 parametric approach.

2 Q. So when you do nonparametric, there's no

3 regression analysis?

4 A. Well, as I indicated yesterday, most

5 regression -- most parametric techniques -- most

6 nonparametric techniques are unable to give

7 quantitative values, instead they simply tell you

8 whether things have changed or not. When one is

9 attempting to define a competence limit, usually some

10 form of parametric analysis is indicated.

11 Standard regression theory is based on

12 something called least squares. Least squares has an

13 undesirable property that a single value that is

14 extreme in some sense, can unduly influence the

15 relationship.

16 Q. Is that why there may be a chance of

17 having more outliers?

18 A. Well, what some people would term

19 outliers, some statisticians would term an outlier,

20 other statisticians would say is simply another

21 observation from the statistical population. The

22 question is how poorly behaved statistically

23 distributions are handled in an analysis.

24 Q. Wouldn't a statistician try to ascertain

25 why that data is so skewed, would they do that?

 

286

 

1 A. Statisticians aren't particularly known

2 for delving into the details of the data.

3 Q. Wouldn't they try to ascertain what a

4 physical relationship should be and then --

5 A. Statisticians don't deal with physics.

6 Q. Or character relationships?

7 A. If you told a statistician that the data

8 came from two different statistical populations, then

9 he or she would do the analysis quite likely

10 differently than if you said they were all

11 characterizing one statistical population.

12 Q. I am looking at your paragraph 4 now about

13 the elimination of some observation and data. Do you

14 believe they were justifiable?

15 A. My own philosophy is, as I indicated

16 yesterday, is not to ignore data unless there are

17 demonstrable mistakes or in some other way can be

18 demonstrated not to come from the same statistical

19 distribution. In general I'm opposed to eliminating

20 data.

21 Q. In Dr. Millard's deposition, we had a

22 discussion about the Bayesian analysis?

23 A. Yes.

24 Q. What is that?

25 A. You will ask me to discuss Bayes' theorem?

 

287

 

1 Q. Sure, in terms as layman as possible.

2 A. Not having been present at Dr. Millard's

3 deposition, I can't re-create for you or tell you

4 exactly what he was referring to. There's a whole

5 field of Bayesian statistics, and Bayes' theorem

6 basically just says if you have some prior

7 information, how a posterior distribution which goes

8 and takes that prior information and then updates it

9 with some additional information, how that would be

10 determined. That's in a layman's sense.

11 What statisticians sometimes call a

12 frequentist approach is simply that you take all of

13 the data and you do your analysis, that you don't

14 have a prior or prior distribution.

15 Q. You would have two different populations

16 or two different distributions in the Bayesian

17 theory?

18 A. Well, not necessarily. Bayesian

19 statistics is a method. If one thinks things should

20 be a certain way when one starts and then one wants

21 to modify that opinion based on additional data, it

22 gives a formalism for how that updating will be done.

23 That's a very general view. I don't know

24 what his particular comment would have been regarding

25 Bayesian's --

 

288

 

1 Q. It doesn't matter what his comment was, I

2 wanted to know what your understanding was.

3 A. As I indicated, a method for updating.

4 Q. Do you agree with the Bayesian analysis?

5 A. In general, I'm not much of a fan of

6 Bayesian analysis. In some applications, it might be

7 appropriate.

8 Q. Do you feel that you should put it all

9 together and develop your analysis from that?

10 A. That's my general view.

11 Q. You strongly question the elimination of

12 the September 24th, 1980 observation. There's no

13 basis that's given for this decision.

14 Have you looked further into that or did

15 you make any further analysis into that or make more

16 inquiries?

17 A. No.

18 Q. You don't agree with it, I guess?

19 A. Well, we discussed the reasons that I

20 don't agree with eliminating data.

21 Q. You also said you don't agree with the

22 Bayesian analysis?

23 A. I didn't say that.

24 Q. Did you or don't you?

25 A. I said I generally don't prefer Bayesian

 

289

 

1 methods, but I don't quite understand what that has

2 to do with the rejection of a particular value.

3 Q. You don't understand how the Bayesian

4 analysis has to do with the rejection of a particular

5 value?

6 A. That's what I just said.

7 Q. I thought Bayesian analysis -- what they

8 did was they start with an idea of how things,

9 relationships should be, and I guess they kind of try

10 to work the data and analysis into that and have two

11 different distributions to try to make the data

12 correspond actually where you are trying to get to?

13 A. Bayesian analysis is a whole field of

14 statistics and decision theory as well. You could

15 probably fill a library with books on Bayesian

16 analysis.

17 And that's why I indicated I don't know,

18 when you bring up something that Dr. Millard might

19 have mentioned in his deposition, I would have to

20 have a lot more specifics. Perhaps he had mentioned

21 something with some Bayesian method of outlier

22 analysis, I don't know, it would be speculating.

23 Q. I don't want you to speculate on what his

24 testimony or what he said or what he understands, it

25 is your understanding and how it relates to the

 

290

 

1 modeling that was done here.

2 But I guess I can say, true or false, that

3 you don't believe that it is appropriate to be used

4 here, the Bayesian analysis?

5 A. I didn't say that.

6 Q. False, you don't agree?

7 A. The question is not at all clear.

8 Q. Okay.

9 A. As I just indicated, you could fill

10 libraries with books on Bayesian analysis. There are

11 numerous specific examples, techniques and all the

12 rest which fall under that general category, in fact,

13 telling me what the specific method is that one might

14 apply to this particular data value. I can't tell

15 you whether I would agree with that or not. I might

16 or I might not.

17 Q. I understand.

18 MR. LOREDO: Let's take a break.

19 (Recess)

20 BY MR. LOREDO:

21 Q. I want to talk a little bit about your

22 report titled: Analysis of Uncertainty in STA Area

23 Requirement.

24 Q. Where did you get in table 1 the TP load,

25 or how did you arrive at that?

 

291

 

1 A. Well, the first place I would be inclined

2 to look would be at appendix F of the SWIM Plan

3 because that's referenced here particularly.

4 Q. Do you want to look at appendix F?

5 A. I can look at that, sure.

6 (Pause)

7 A. If it didn't come from there, it came from

8 a program we were provided from the District.

9 I think you will find if you go convert

10 units here out of table 3 in the SWIM Plan, you will

11 probably get 1.234 cubic kilometers per year and they

12 are using acre feet per year.

13 Q. The question was the TP load, the 111, how

14 do you arrive at the 111?

15 A. That, again, is someone else's number, and

16 I believe that's an appendix F number. The

17 precipitation in the the ET, in appendix F, were

18 exactly the numbers here, and I think if we went

19 through the rest of them, we would find all these

20 numbers came out of appendix F as well.

21 The program that was used here was in fact

22 provided by the District and was recoded by ESP,

23 which I think is indicated in the report, and they

24 had their software hard-wired in the program, which

25 is my understanding of an appendix F value, and a

 

292

 

1 cursory review seems to suggest that's the case.

2 Q. The Monte Carlo simulation, is that yours,

3 where did that come from?

4 A. Are you talking about the general method

5 or the particular program?

6 Q. Well, both.

7 A. The general method is a widely used method

8 in applied statistics for estimating probability

9 distributions which cannot easily be derived

10 analytically, i.e., in a closed mathematical form. I

11 have used it extensively in my area of work, but it

12 is blindly used by many others.

13 Q. How did you use it here, just plugged in

14 the numbers, and this is in a program?

15 A. Well, no. Do you want me to describe the

16 analysis?

17 Q. Thank you.

18 A. The question was how uncertain would the

19 predicted load requirements be given certain sources

20 of uncertainty in the assumptions made to do the

21 design.

22 The design was made based on the method in

23 appendix F put forward by the District and as encoded

24 in a programming language called Stella, I believe by

25 the District. I'm not certain of the exact genesis

 

293

 

1 of that code, you would have to ask the people at ESP

2 for the details.

3 One of the people at ESP, Maria Mao in

4 particular, took the Stella code and recoded it into

5 Fortran.

6 Q. I have heard of that before, Fortran?

7 A. Fortran, those of us who are back in the

8 dark ages programmed in Fortran. Fortran I

9 understand, so when she recoded it in Fortran, I

10 could go do some analysis of my own.

11 She verified that Fortran version gave the

12 same numbers as the Stella version that was started

13 with. I was provided with a Fortran version of the

14 code.

15 Q. Why not use the Stella version?

16 A. Because I don't have a Stella compiler,

17 and then I would have to program in Stella and it is

18 sort of asking why don't we do this deposition in

19 German?

20 Q. Spanish would work, but not German.

21 A. It wouldn't work very well for me, you

22 wouldn't get very good answers. And if I decided to

23 program in Stella, I wouldn't get very good answers

24 either without spending a lot of time on it, so it

25 was done in Fortran.

 

294

 

1 Q. Which version of Fortran was used?

2 A. Which compiler? Fortran 77.

3 Q. Is that the year?

4 A. Well, Fortran 77 is the most commonly used

5 current version of Fortran.

6 Q. You said it is antiquated?

7 A. Well, that's just like getting

8 statisticians in the room and asking them about

9 Bayesian analysis. If you get a bunch of computer

10 programers in the room and you tell them you are

11 using Fortran, they will look down their noses at

12 you.

13 If you go up to the District office and

14 ask what computer language they use, you will find

15 most people using Fortran. You will find most

16 consulting firms using Fortran. In a sense, it makes

17 no difference as long as you get the same -- it is a

18 question of convenience.

19 If programming is done properly, you will

20 get the same numbers regardless what computer program

21 you use.

22 Q. Now that you have it in Fortran, then what

23 did you do?

24 A. Then I identified the inputs with which I

25 believe there was substantial uncertainty associated,

 

295

 

1 such as -- you have the list of them there, but

2 settling rate is obviously one of them, the hydraulic

3 load is another, concentration of the load, and

4 there's a table, evapotransforation, and so on.

5 I identified, based on best judgment --

6 and in some cases where I could -- some cursory

7 analysis of what reasonable uncertainty levels would

8 be associated with those inputs.

9 Then I built a computer program which

10 would sample from a probability distribution, which I

11 believe in all cases was normal, values of each of

12 those variables that could occur consistent with that

13 uncertainty, ran the program to determine what the

14 load requirement would be, and then I repeated that

15 some numbers of times, I think a hundred for

16 different independent draws of random numbers.

17 Q. Did you assume one large STA for the

18 purposes of the calculation?

19 A. It was done the same way that appendix F

20 does, I think that's the assumption.

21 Q. What did you do about the water supply

22 bypass?

23 A. I did exactly what the District did in

24 their model.

25 Q. Did you follow every assumption?

 

296

 

1 A. The code is theirs.

2 Q. So you made the same assumptions for water

3 retention?

4 A. The same assumptions were made.

5 Q. The same assumptions for --

6 A. The attempt in this analysis was not to

7 critique the District's model, it was to say if that

8 model is correct, what are the implications for

9 uncertainty.

10 Q. Did you say what the levels of certainty

11 were?

12 A. I wanted to know what the uncertainty was

13 in the area required given their model and given

14 reasonable assumptions about uncertainty.

15 Q. You are saying given the reasonable

16 assumptions. We are talking about the assumptions by

17 the District or did you make any additional

18 assumptions?

19 A. These are my assumptions about the

20 certainty of their model. That was the first phase

21 of the analysis. If you read down through the memo,

22 you will see that the Department of Justice

23 consultant had in fact -- I became aware of somewhat

24 later on after I began this -- done a similar

25 analysis with slightly different assumptions.

 

297

 

1 But I reran my analysis making consistent

2 assumptions about the uncertainty to see how much

3 difference that would make, and there are a set of

4 results in that report that reflect those assumptions

5 as well.

6 Q. What was your conclusion?

7 A. I think there's some tables in there that

8 have the results that show you what the inferred

9 probability distribution of the area requirement is.

10 Q. What would be your testimony at the

11 hearing regarding the uncertainty for STA?

12 A. My testimony would be based on those

13 assumptions and those tables of results, that the

14 confidence interval, depending on what confidence

15 interval one wanted to choose is as given in the

16 table, i.e., if one wants a 90 percent confidence

17 limit, that the area would range from whatever the

18 bounds are listed in those tables.

19 Q. Which table are you talking about, No. 3?

20 A. There are two sets of tables, one for my

21 assumed uncertainties in the inputs and the other for

22 Walker's assumptions.

23 The initial Monte Carlos are in 3 and

24 shows if you randomize all of them and if you wanted

25 a 90 percent confidence interval, you would go from

 

298

 

1 .05 to .95 on cumulative probability, and I think I

2 have said it goes from roughly 7,000 to 66,000 acres,

3 and if you make Walker's assumption, then you go from

4 something like 14,000 to 47,000.

5 Q. Which table are you looking at?

6 A. I am looking at table 4 described in here

7 on pages 5 -- page 5 essentially, how his analysis

8 that was presented in the stage meeting or his

9 assumptions about the uncertainty were incorporated

10 in my model.

11 MR. LOREDO: Let's go ahead and mark this

12 as an exhibit. It will make someone's life down the

13 line a lot easier.

14 This is Exhibit 5.

15 (Lettenmaier's Exhibit 5 was marked for

16 identification)

17 BY MR. LOREDO:

18 Q. What made you assume to use a hundred runs

19 in the Monte Carlo?

20 A. It is a reasonable number. You are not

21 going to get a much different answer if you use more.

22 Q. Didn't you tell me before, though, the

23 more sampling -- and one of the reasons that you

24 believe the data that you are going to get for

25 another four-year period, the more sampling will

 

299

 

1 probably affect your analysis?

2 A. Sure.

3 Q. Why wouldn't another hundred runs of the

4 Monte Carlo be better?

5 A. Oh, it would be better, it just won't give

6 you much different answers. It is a different

7 situation here, it is a different situation if we

8 know the distributions. In Monte Carlo I am God, I

9 set them. In the real world I don't know what they

10 are.

11 I can do a thousand, it will burn up some

12 more computer time, and in fact, those confidence

13 intervals will be -- my knowledge of them will be a

14 little more precise. It might turn out that instead

15 of 7,000 to whatever I quoted you there, maybe it

16 will be 6800, 6200, 6500, something like that.

17 It is not going to be three times, twice

18 or anything like that. You can in fact work out,

19 which I didn't bother to there, you can work out the

20 precision of those percent files based on knowing

21 that there were a hundred Monte Carlo runs made.

22 Q. In Exhibit 5, titled The Analysis of

23 Uncertainty in STA Requirement, the document we have

24 been talking about, you stated that "None of the

25 above were based on direct analysis of historical

 

300

 

1 data (see recommendations below), but rather are

2 based on a 'first guess' of the distributions."

3 What are you talking about there?

4 A. Well, you need to know -- you need to

5 characterize the uncertainty numerically somehow to

6 do the Monte Carlo analysis.

7 Q. That's why you said you were God, you

8 characterized them?

9 A. Exactly. But you have to know not just

10 what the means are, which the District values were

11 used for those, the ones in your question where it

12 came from and so on, but you also have to know the

13 standard deviations and it comes from a normal

14 distribution. Those had to be essentially guessed or

15 initial guessed to determine them.

16 Basically I indicated later on in the

17 report, how my guesses compared with Walker's guesses

18 and I ran his guesses through my model.

19 Q. A lot of guessing going on here.

20 A. He is guessing, I'm guessing, everybody is

21 guessing.

22 Q. You ran his guesses?

23 A. Yes, in table 4, the results are there.

24 A. I should point out that not all of those

25 inputs are necessarily -- they are better than

 

301

 

1 guesses, they are approximations, and in some cases,

2 for instance, the precipitation, mean precipitation

3 of evapotransforation are based on actual

4 observations, and one can work out what the standard

5 deviation is because it is the standard error of the

6 mean.

7 So for some of those values that are based

8 on direct observations, there's not quite so much

9 uncertainty associated with it, not so much

10 uncertainty associated with the specification of the

11 standard deviation.

12 You can also determine how important those

13 are by looking at the results in the tables. Some of

14 those particular uncertainties you specify turn out

15 not to make much difference. Even though you don't

16 know them very well, the results don't depend on them

17 very much anyway.

18 Q. I guess on your table on page 497 where

19 you have the range from about 7,000 to about 66,000,

20 is that the case of a worst case scenario?

21 A. I don't think so. 7,000 -- which table

22 number is that?

23 Q. I'm looking at table 3, and I guess letter

24 A after, you have "Consideration of all source of

25 variability leads to a wide range of area

 

302

 

1 requirements: the range from .05 to .95 percentage

2 points (inner 90th percentile confidence bound) about

3 7,000 to 66,000 acres."

4 That's what I am referring to.

5 A. Yes, and you are asking whether that's a

6 worst case?

7 Q. Yes.

8 A. Not necessarily.

9 Q. Why not?

10 A. Because you have assumed the model is

11 correct, and when you are dealing with uncertainties,

12 the inputs to the model, the model might be wrong,

13 too.

14 Q. Then you stated the mean is close to the

15 predicted value used in the District's input with all

16 inputs fixed of 24,900 acres?

17 A. It should be. If I go and take -- they do

18 a deterministic analysis and they put in

19 deterministically the mean values that I have listed

20 there and they run essentially the model once for

21 those means and they get out 24,000 whatever it is

22 acres, I go draw those things from probability

23 distributions which have means which are equal to the

24 particular numbers that the District used, and then I

25 get a distribution of the areas.

 

303

 

1 I should get -- the mean of my

2 distribution should be somewhere close to what their

3 number is. It won't be exactly the same. In fact,

4 it should not be exactly the same unless the model is

5 linear, which it is not.

6 Q. I got lost.

7 A. Okay.

8 Q. You said that the mean from your predicted

9 values was 24,900 acres. Is that what you are

10 telling me here?

11 A. Is that in the report?

12 Q. Yes.

13 A. Where are you?

14 Q. Paragraph A?

15 MR. FITZGERALD: The second paragraph from

16 the bottom.

17 A. Mean is close to the predicted value using

18 the Water Management District inputs for all inputs

19 fixed at 24,900 acres.

20 That just means if I go run the model

21 once, their Stella model once, and go take the inputs

22 to be the mean values of all these inputs and run it

23 once, I get 24,900.

24 If I go look at the mean from these

25 probability distributions here, which is close to but

 

304

 

1 not exactly the 50th percentile, I get something that

2 isn't that far different.

3 If you look down the column, that's 50

4 percentile there, you see a lot of numbers which are

5 not a lot different than 24,900. That's all that

6 says; if it weren't, you would be worried.

7 Q. That would be the probability that would

8 happen 50 percent of the time under your analysis?

9 A. No, it is the middle of the distribution.

10 Q. So it would be like on top of a bell

11 curve?

12 A. Exactly. It is essentially the point

13 corresponding to the mode -- not exactly that either.

14 If the distribution were symmetric, it would be the

15 point corresponding to the mode.

16 Q. Did you assume any co-variance between the

17 items in table 3 and 4?

18 A. No.

19 Q. Do you think it is possible that the range

20 that you have calculated of 7,000 to 66,000 acres is

21 because you didn't assume a co-variance?

22 A. That will make some difference, but it

23 will only make much of a difference for those

24 variables that are identified as being more

25 sensitive, those being the load, the K value, and I

 

305

 

1 can't remember what the third one is.

2 Q. I'm not sure, what do you mean by K value?

3 A. That's the settling rate. The ones that

4 make the most difference here are the phosphorus

5 load, the settling rate, and I thought there was a

6 third one, but there appears not to be.

7 Q. Those are the most sensitive constituents?

8 A. I'm sorry, hydraulic load.

9 They make the most difference: The

10 hydraulic load, the phosphorus load, and the settling

11 rate. So you would have to hypothesize there was a

12 co-variance between those to be able to make much

13 difference.

14 In other words, if you said, well, there's

15 a correlation between ET and annual precipitation,

16 and there might be some negative correlation that you

17 get less ET in years that are rainy, that won't make

18 very much difference at all to the analysis.

19 If you go and assume that there is some

20 correlation between the settling rate and the

21 phosphorus load, that would be much more likely to

22 make some difference, but you would have to have some

23 justification, which is not apparent and was not

24 apparent to me what that justification would be when

25 I did the analysis.

 

306

 

1 Q. You weren't told of any justification?

2 A. Well, that would be based on physical

3 principles, and I can see no physical principle that

4 would suggest that.

5 Q. Was the output of the STA acreage required

6 normally distributed?

7 A. Strictly? No, it won't be. Whether it is

8 approximately normal is another question.

9 Q. How probable then is the range of 7,000 to

10 66,000 acres?

11 A. I just gave you the numbers, they are

12 listed on the table.

13 Q. Table 3 you are referring to?

14 A. Yes. There's no assumption about

15 probability distribution in there and the outputs.

16 Whether or not it is normal has nothing to do with

17 that table.

18 Q. Starting on page 7, the source code

19 listing for Monte Carlo simulation program?

20 A. Yes.

21 Q. Page 502, the middle of the page where it

22 says "modified DP 11," what is that?

23 A. My notation that, as I indicated to you,

24 that Maria Mao at ESP took the Stella code and

25 converted it to Fortran, and I took the Fortran

 

307

 

1 program which was developed to be a once-through

2 analysis, as I indicated, you put the numbers in for

3 each of these values and it tells you an acreage out.

4 I put that in a loop so I could do my Monte Carlo

5 iterations a number of times.

6 Q. So that was the date you actually started

7 to do it?

8 A. Well, whether that's the date I started or

9 ended or some time in between, I couldn't really tell

10 you. It is approximate. It just gives me an

11 indication for my own information of roughly when I

12 made the changes, when I worked on it.

13 Q. So then you worked on this when you were

14 still working with the District?

15 A. That's possible.

16 Q. But I thought you weren't doing any work

17 with the Everglades or at least trying to stay away

18 from doing any work in the Everglades?

19 A. Well, the District had the final form of

20 our report back in I think it was August of '92. Any

21 work that was ongoing with the District was dotting

22 I's and crossing T's on the report. The final panel

23 review of that project and all the rest was in the

24 summer of 1992.

25 Q. You originally had done this project for

 

308

 

1 the District?

2 A. No, I didn't say that. What I said was

3 that if there was any work continuing for the

4 District, it was simply incorporating review comments

5 in the report, that the final review on the District

6 work had been conducted in the summer of 1992.

7 Q. What I would like to do at this time is go

8 through the reports that are produced, and if you can

9 kind of give me an overview of what your conclusions

10 are and your basis for those conclusions, starting

11 with the report titled: Review of Kriging Methods as

12 Applied to Estimation of Soil Phosphorus in WCA-2A.

13 A. Yes.

14 What is the question?

15 Q. What I would like you to do is summarize

16 it for me in your understanding of that what that

17 report reflects, your conclusions and your basis for

18 the conclusions.

19 A. Well, I can read through this if you want.

20 I think it is in the report.

21 Q. It would help me. I have read it and I

22 have someone else looking at it, but unfortunately we

23 didn't have sufficient time to do a complete analysis

24 so I can properly question you on it, but what I

25 would like to do is see if I can cover most of it so

 

309

 

1 I wouldn't have to call you back at a later time to

2 depose you on questions regarding that report.

3 A. Well, I think as indicated in the

4 background section here, we mention that there's some

5 work done for the District by Reddy.

6 Q. What was the purpose of that report?

7 A. Whose report, Reddy's?

8 Q. No, the report --

9 A. That's what I'm getting to. It says here

10 in the Reddy -- if you want to type this paragraph

11 in, it characterizes exactly what was done and why.

12 Q. Do you want to read it and then tell me in

13 your own words?

14 A. But these are my own words.

15 Q. I thought Lisa drafted that?

16 A. No. She may have made the initial draft.

17 All of these reports here I have been through and

18 most of the words are mine. I can either tell you in

19 words or I can read you what I wrote then. It is

20 probably more accurate to read you what I wrote then.

21 Q. I would rather you summarize from what you

22 believe that report says than read in. Obviously you

23 can read it.

24 A. The concern as indicated in the last part

25 of the background was there was an issue that Reddy

 

310

 

1 had used some kriging in his analysis, it appeared

2 that he had used SURFER -- this is a direct quote:

3 "The motivation for reanalysis of the

4 Reddy data was in part that SURFER is not a

5 statistical package. Therefore, a number of

6 variables that must be specified for application of

7 kriging are essentially hard-wired in the program and

8 the selected values are not necessarily known to the

9 user."

10 Our concern was then that he had gone and

11 apparently used SURFER to do his kriging, and there

12 are some things buried in that program we wanted to

13 better understand, what was buried in the program and

14 what implications those things might have to the

15 numerical values that he had come up with in his

16 report.

17 Q. You didn't agree with using SURFER?

18 A. We did not know what assumptions were made

19 in SURFER. We suspected that some of them were not

20 applicable. So in this report, we essentially

21 attempted to go re-create Reddy's contour maps -- and

22 there's one of them in particular, I believe

23 somewhere here, the figure number is mentioned -- our

24 figure 1 is his figure something or other -- I'm not

25 sure which figure number this is out of his '91

 

311

 

1 report.

2 But our concern was that we would like to

3 be able to reconstruct his analysis so we understood

4 what assumptions were made, and then determine

5 whether those assumptions were really in fact

6 appropriate.

7 And we did some analysis in here to look

8 at the total amount of phosphorus that would be

9 contained, total mass of the upper 10 centimeters --

10 I guess this is over the whole 2A area -- I am not

11 reading directly now. The only thing I read you

12 directly was what the purpose of the report was, as I

13 quoted to you.

14 We have different model runs in here,

15 different assumptions about the form of the

16 variogram. We were using a program at that point

17 called GEOPACK which I think we discussed yesterday.

18 Q. GEOPACK you said had bugs and you really

19 didn't like GEOPACK?

20 A. Well, we will get on to that in the

21 subsequent reports, but at the time we were doing

22 this analysis for what was done here, which was all

23 an isotropic analysis, as I recall --

24 Q. Can I stop you with that?

25 A. Sure.

 

312

 

1 Q. I think I might have looked up the word

2 "isotropic," and you can expound on that.

3 A. Isotropic just means that the spacial

4 dependence is the same in all directions.

5 Q. Is it isotropic or "trophic"?

6 A. Isotropic, T R O P I C.

7 Q. So the spacial distance is the same in

8 each --

9 A. No, it says that the dependence between

10 two observations that are taken some distance apart

11 does not depend on the angle between them, it just

12 depends on the distance.

13 Q. So you can have four points, and the

14 isotropic says it doesn't matter, these two points

15 are the same as these two points, the relationship?

16 A. It says statistically the dependence

17 between them -- and then you have to talk about what

18 dependence means, but the dependence itself is

19 determined solely by separation distance and not by

20 the angle of the separation. That's what isotropic

21 is.

22 So in this report, we went through some

23 analyses using GEOPACK, attempting to re-create what

24 Reddy had done. I believe in this report, we may

25 also, since he had used SURFER -- there are different

 

313

 

1 variations and decisions one has to make to do this

2 which are fairly technical considerations. And we

3 are essentially trying to determine what he had done,

4 what it amounts to, and what the sensitivity of some

5 of these assumptions were to the total phosphorus

6 mass over the area.

7 Q. And what did you conclude?

8 A. Well, there's a table there. We were able

9 to, more or less, reconstruct his results. We were

10 quite convinced that he had in fact used an isotropic

11 analysis, and we were reasonably convinced that he

12 had done some smoothing of the results as well.

13 Q. Smoothing of the results?

14 A. Yes.

15 Q. That sounds like an accountant changing a

16 number from one to the other. When you talk about

17 smoothing, did he make inappropriate assumptions --

18 A. We are not certain that he knew or had

19 explicitly made a decision to do that, but we were

20 quite sure that because he had us