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 used SURFER, it had 21 happened, perhaps knowingly, perhaps unknowingly. 22 The issue very briefly is this. Kriging 23 has a property that it recreates or that it provides 24 as an estimate at the known points, exactly the value 25 that is put in. So that if you go plug back in the 314 1 point at which you collected an observation, then the 2 estimate, the surface that it gives you value for 3 that point, is value that you estimated originally. 4 However, in SURFER, in the contouring 5 program, it does some smoothing so that it has the 6 effect of taking these observed points which would be 7 re-created exactly by the kriging, and averaging them 8 down or up as the case may be. 9 Q. Why would it average it down or up? 10 A. Well, because it is doing smoothing, it 11 has a smoothing algorithm. 12 Q. That's to make it fit into the particular -- 13 A. To get nice smooth contours, it does that. 14 Q. Otherwise you might end up with a contour 15 that starts -- I guess you couldn't draw a contour? 16 A. No, you tend to get something that looks 17 more like a lemon merangue pie, little peaks where 18 the observation -- troughs where the observations 19 are. 20 Q. Does the GEOPACK not use the isotropic 21 analysis? 22 A. Well, that particular question is not 23 related to the assumption of isotropy, that is simply 24 an artifact of the SURFER plotting program. 25 Q. Part of the SURFER program? 315 1 A. Yes, which our information was that he had 2 used and we had acquired also. 3 Q. You said SURFER is not a statistical 4 package? 5 A. That's correct. 6 Q. Is GEOPACK a statistical package? 7 A. Yes, it is a statistics package. 8 Q. You have got no smoothing in it? 9 A. Not that that is done without one knowing, 10 there's user control over those decisions which there 11 is not in SURFER. 12 Q. So if you don't do anything in GEOPACK, 13 you may smooth over just as you would SURFER? 14 A. No. 15 Q. I'm lost now. You said GEOPACK is user 16 controlled? 17 A. Yes. 18 Q. Is there smoothing? 19 A. Well, I don't know for a fact whether 20 there is or is not smoothing in GEOPACK, it is 21 possible that one could do that if one wanted to. 22 That certainly would not occur without one making an 23 explicit decision, whereas in SURFER, that is 24 happening and the user would not necessarily know 25 that it was happening. 316 1 Q. So in SURFER, you don't have an 2 opportunity to make a decision? 3 A. Exactly. 4 Q. And the decision would be what, an 5 assumption? 6 A. Yes. SURFER is a general purpose plotting 7 and analysis program that does many things of which 8 kriging is only a very small part. It has a routine 9 that allows one to do kriging. 10 Since it is only a small part, there's not 11 the amount of user control as there is in, say, 12 GEOPACK which is an entire program exclusively 13 dedicated to doing kriging and related spacial 14 analysis. 15 Q. What forms of variogram are we talking 16 about or assumptions that aren't applicable? 17 A. Well, this is another problem. I think it 18 probably says in the report it is not quite clear 19 what SURFER does about the variogram. We presumed, I 20 believe, that it probably was doing a linear one 21 because that is the easiest one to use. 22 Q. SURFER is a program, right? 23 A. Yes. 24 Q. Why couldn't someone go and look at the 25 program and determine what it is doing or not doing? 317 1 A. Because it is proprietary and you can't 2 get the source code. 3 Q. So -- 4 A. It is commercial software, all you get is 5 the executable. 6 Q. So you are not sure whether or not it does 7 smoothing? 8 A. No, we are sure it does smoothing. 9 Q. Why? 10 A. Because you can check and see whether or 11 not you can reproduce the observed value and you 12 don't, so we know it is doing some smoothing. 13 Q. So you are saying it doesn't even 14 reproduce the observed values? 15 A. Exactly. 16 Q. When kriging should -- in I guess a 17 perfect world, kriging meant that it should reproduce 18 the observed value and also other values in between? 19 A. Absolutely -- well, it should reproduce 20 the observed value exactly, that's a property of 21 kriging. 22 Q. So when you ran the SURFER program, it 23 didn't -- 24 A. When we ran the SURFER and asked to give 25 us back the values at which point we had 318 1 observations, we did not get the observed values 2 back. 3 Q. When you did SURFER, it asked you to give 4 it the observed points? 5 A. No, we ran a special case where we asked 6 it to go reproduce for us the observed values and it 7 did not. It gave us back something close to, but not 8 the same. So we know it is doing some smoothing. 9 Q. How close? 10 A. I couldn't answer you that. 11 Q. Real close? 12 A. I mean not real close. 13 Q. What's "real close"? 14 A. Well, not often the third decimal point, 15 or you would say this is just numerical round-off; it 16 is clearly different, it is going to depend on the 17 particular point. 18 Q. You don't know what assumptions are 19 applicable because you couldn't look at the SURFER 20 program or do you know what assumptions are not 21 applicable? 22 A. I guess you would have to be a little more 23 specific. 24 Q. I think I'm not being specific. 25 You originally said that there were not 319 1 assumptions that were applicable. Were you referring 2 to the SURFER program itself and the assumptions that 3 the program makes or looking at other variables or 4 assumptions that had to be made before you ran the 5 program? 6 A. These are assumptions that are hard-wired 7 into SURFER, but which the user is not apprised, 8 which may not be appropriate. 9 Q. Because we don't know what they are? 10 A. We don't know exactly what they are. They 11 are buried in the code. 12 Q. You said they are hard-wired? 13 A. Hard-wired in the code, but you get only 14 the executable, so you don't know exactly what's in 15 there. 16 Q. So the way you checked it was to run the 17 data and try to see if that program would re-create 18 the observed points, and that's -- 19 A. That was with respect to the smoothing, 20 that was the way we checked that, yes, to determine 21 that in fact there was some smoothing going on. 22 Q. What else did you check for? 23 A. Well, I think I indicated we went and did 24 the same analysis with GEOPACK, and compared our 25 results with the ones of Reddy's that are in one of 320 1 those figures there. 2 Q. Would that be figure 2A and 2 B? 3 A. Well, 2A and 2 B are one set -- that 4 simply is a comparison of our analysis using SURFER 5 and GEOPACK. And I think the issue here was that in 6 SURFER, we didn't know exactly what variogram form it 7 was fitting, although we suspected linear. 8 GEOPACK, we used linear isotropic and 9 that's just a comparison of the two sets of results. 10 They are similar, but not identical. 11 Q. Now, in the GEOPACK, you are able to elect 12 linear and isotropic -- or is linear and isotropic, 13 you wouldn't say it was synonymous? 14 A. No, isotropy is a completely different 15 issue than the form of the variogram. Linear is one 16 form of the variogram. There are spherical 17 variograms, exponential variograms and other forms. 18 All of those can be isotropic or anisotropic. 19 Q. I guess with anisotropic, there is a 20 difference? 21 A. Depending on direction. 22 Q. Angles? 23 A. Yes. 24 Q. What is figure 3 telling us here? 25 A. That's just a plot of the variogram, 321 1 that's a fitted variogram out of GEOPACK, and that's 2 the raw data, the empirical variogram from the raw 3 data. It is sort of a typical plot like you would 4 see in most analyses. 5 That assumes isotropy, so those points 6 represent an average within a distance class over all 7 angles, and I don't know if you want me to explain 8 kriging theory and variograms and so on or not, but 9 it is not particular to this analysis, the method 10 that's used. 11 Q. I think we will probably get to it a 12 little later. 13 A. Okay. 14 Q. How much of this report remains viable, 15 going further ahead and thinking ahead on the use of 16 the program GEOEAS? 17 A. Well, that part of the analysis which is 18 essentially basically just doing some detective work 19 to try to figure out what Reddy did, all of it. 20 Q. This report? 21 A. Yes. If you asked me a different question 22 having to do with whether we are using particular 23 results of that, you would probably get a different 24 answer, but in terms of the applicability of that 25 report in a detective sense, all of it. 322 1 Q. Which results are you going to use, the 2 GEOEAS? 3 A. Do you want to skip ahead four reports? 4 Q. Certainly. It is a natural progression I 5 think I notice in the reports? 6 A. Yes, there is, and there are different 7 analyses which are applied to different variables and 8 so on. 9 Q. Do you go back and look at Reddy's work 10 again using GEOEAS? 11 A. Well, for the 10 centimeter total 12 phosphorus data, I believe all of the data that we 13 have used -- I may have to correct myself a little 14 later on -- but certainly most of the data that's 15 been used is Reddy' data, the data isn't the issue, 16 it is how they were analyzed to produce spacial mass. 17 MR. LOREDO: This is not a bad time to 18 break. 19 (Luncheon recess at 12:25 p.m.) 20 21 22 23 24 25 323 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25