1 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 2nd, 1994 9:16 a.m. - 12:15 p.m. 20 DEPOSITION OF DENNIS P. LETTENMAIER 21 VOLUME I 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. 2 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 3 1 INDEX 2 Witness Direct DENNIS P. LETTENMAIER 3 By Mr. Loredo: 4 4 EXHIBITS 5 NUMBER DESCRIPTION PAGE 6 1 Notice of deposition 6 7 duces tecum 8 2 CV of Dr. Lettenmaier 48 9 3 A letter dated April 149 12, 1993 from Dr. 10 Lettenmaier 11 4 A letter dated May 20, 1992 from Dr. Lettenmaier 158 12 13 14 15 16 17 18 19 20 21 22 23 24 25 4 1 Thereupon -- 2 DENNIS P. LETTENMAIER 3 was called as a witness and having been first duly 4 sworn, was examined and testified as follows: 5 DIRECT EXAMINATION 6 BY MR. LOREDO: 7 Q. Good morning, Dr. Lettenmaier. My name is 8 Jose Loredo, and I represent the South Florida Water 9 Management District. 10 If you please, for the record, state your 11 name and your address. 12 A. Dennis Lettenmaier, 3213 South Norman 13 Street, Seattle, 98144. 14 Q. Are you married? 15 A. No. 16 Q. Have you ever been deposed before? 17 A. Yes. 18 Q. When were you deposed? 19 A. It has been some time back, I can't 20 remember specifically, the past 10 years. 21 Q. Is that the last time? 22 A. Yes. 23 Q. What was the area? 24 A. It was a flood issue. 25 Q. And was it concerning Seattle, Washington? 5 1 A. Yes. 2 Q. What was the nature of your deposition or 3 the underlying theme of it? 4 A. It had to do with causation of flooding 5 and damages incurred by a property owner. 6 Q. Did you do any statistical analysis to 7 determine the causation of the flood? 8 A. That problem was not primarily 9 statistically based, it was more physical causality. 10 Q. And how did you determine the physical 11 causality? 12 A. Site visit, past background of the area, 13 some limited data analysis. It was a rather small 14 case. 15 Q. How much time did you spend on it? 16 A. Of the order of a day or two. 17 Q. And who employed you for this? 18 A. We are going way back. Inglewood Golf & 19 Country Club, as I recall. 20 Q. Did you testify at trial? 21 A. No, it settled out-of-court. 22 Q. Have you ever testified as an expert 23 witness at trial? 24 A. Yes, once. 25 Q. Can you tell us when? 6 1 A. Precisely, I would be guessing; around 2 1980. 3 Q. What was that for? 4 A. That was also a flooding case, also a 5 rather small homeowner's case. 6 Q. Was that also in Seattle? 7 A. Yes, King County Superior Court. 8 Q. Was this a physical analysis? 9 A. Yes. 10 Q. How much time did you spend on it? 11 A. Again, that was of the order of a couple 12 of days. 13 Q. I will show you a copy of a notice of 14 deposition which was served upon your counsel for the 15 petitioners, Sugar Cane Growers, and I would like you 16 to take a look at it, breeze through it, and tell me 17 if you have seen it before. 18 A. Yes, I have. 19 Q. When did you see it? 20 A. Yesterday, possibly before. I definitely 21 saw it yesterday. 22 MR. LOREDO: I would like to mark as 23 Exhibit 1, the notice of the deposition duces tecum. 24 (Lettenmaier Exhibit 1 was marked for 25 identification) 7 1 BY MR. LOREDO: 2 Q. You stated you definitely saw it 3 yesterday. Do you recall if you had seen it before 4 yesterday? 5 A. I don't recall, no. Likely it is that I 6 have seen it and not read it carefully until 7 yesterday. 8 Q. And who did you discuss it with? 9 A. Mr. Blank. 10 Q. The subpoena requires the production of 11 documents, and I would like you to point to page 7, 12 which starts with No. 1 and runs through page 11 13 which ends at No. 15, and I would like you to take a 14 look at it again and see if I can maybe refresh your 15 memory. 16 Because these were documents which were 17 subpoenaed and asked to be produced from your 18 records, and I would like to see if you had an 19 opportunity to look at that, or at least if opposing 20 counsel or someone else had asked you to put together 21 documents to produce to the respondents in this case. 22 A. Yes, they have. 23 Q. And when did you have discussions relating 24 to the production of documents in this case? 25 A. Well, to go back, there was a request for 8 1 documents nine months or so ago for a deposition that 2 did not occur. This is probably the reason I didn't 3 read this one particularly carefully because there 4 was a list of documents, and all the rest were 5 prepared for a previous deposition which was not 6 taken. 7 Q. I see. 8 A. So I updated the documents and supporting 9 material for this deposition that were requested. 10 Q. Who did you speak to when you first 11 gathered the documents for your first deposition 12 which did not take place? 13 A. Mr. Blank. 14 Q. Did you speak to anyone else? 15 A. Not that I recall. 16 Q. Did you produce all the documents to Mr. 17 Blank? 18 A. Yes. 19 Q. How did you go about going through your 20 files and your computer disks and your notes and 21 knowing what documents to produce? 22 A. I went through and I made a copy of all of 23 the computer files that I have used in the case to 24 date. That's my primary means of archiving work that 25 I've done. 9 1 I also went through the paper file and 2 produced for him all of the documents that I have 3 written associated with the case, including all 4 internal correspondence. 5 Q. Any scrap notes, notebooks? 6 A. No, I don't keep those. 7 Q. Most of your notes would be left on the 8 computer? 9 A. Exactly, or have been written up in the 10 forms that I indicated. 11 Q. So you actually produced it nine months 12 ago, and you provided the documents to Mr. Blank? 13 A. Yes. 14 Q. Did you recently update? 15 A. Yes. 16 Q. What did you update? 17 A. All of the computer files and all of the 18 reports and correspondence. 19 Q. In what media form was that, was that all 20 in computer form, was it it in letter form? 21 A. Both, both hard copy and computer tape. 22 There are two computers involved, there's a PC 23 involved and there's a UNIX work station involved, 24 two separate medium. 25 Q. PC being? 10 1 A. Personal computer. 2 Q. Okay. And the UNIX? 3 A. Microsoft DOS. 4 Q. What is the Microsoft DOS? 5 A. Microsoft DOS is the operating system used 6 by IBM compatible personal computers, such as the one 7 the gentleman has here. The medium used is the three 8 and a half inch disks, such as the ones you see 9 there. The medium that was used on the UNIX work 10 station was a 60-meg cartridge tape. 11 Q. I think we did receive that yesterday. 12 And that runs off of what? I received a diskette 13 which is about 12-by-4 -- 14 A. Magnetic tape. 15 Q. Is that read just like any tape recorded 16 machine? 17 A. Yes, most UNIX work stations will have 18 access to the cartridge drive which will read that 19 format. 20 Q. Was that the only -- I guess the cartridge 21 that you provided with respect to your work? 22 A. That and the three and a half inch disks, 23 yes. There are two magnetics, electronic media 24 involved. 25 Q. What I'm trying to ascertain is for the 11 1 UNIX work station, we received one 60-megabyte 2 cartridge. Is that the only one? 3 A. Yes, from the UNIX work station, yes. 4 There are a large number of files on that tape. 5 Q. Does the 60-megabyte also work on the DOS 6 operating system? 7 A. No. 8 Q. What system does it work on? 9 A. UNIX, U N I X. 10 Q. Is that -- 11 A. A widely used operating system used on 12 scientific computers in most major university and 13 scientific laboratories. 14 Q. Was there a reason to use that versus a 15 standard floppy disk? 16 A. Yes, because there is no floppy disk 17 medium on our work station. 18 Q. How many computer floppy disks did you 19 provide to Mr. Blank? 20 A. I believe six, I'm not certain. 21 Q. Was that recently? 22 A. Yes, within the last two to three weeks. 23 Q. Had you provided Mr. Blank any floppy 24 disks prior to that? 25 A. Yes, there were floppy disks provided nine 12 1 months ago for the deposition that didn't take place. 2 Q. What were those floppy disks? 3 A. They were all of the information 4 pertaining to this case from an associate of mine who 5 has worked on it, which is identical to the ones that 6 were provided two to three weeks ago, with the 7 exception that the ones two to three weeks ago 8 contain all of the information and computer files to 9 date, including the nine-month or so intervening 10 period. 11 Q. And the associate you are referring to is 12 who? 13 A. The name? 14 Q. Yes. 15 A. Lisa Dally Wilson whose name is on some of 16 the reports. 17 Q. Did she assist you in the work that was 18 generated? 19 A. Some of the work, yes. 20 Q. Is she a student? 21 A. No. 22 Q. An employee? 23 A. No. 24 Q. Did she receive compensation for her work? 25 A. Yes. 13 1 Q. I guess we would have to categorize her as 2 an independent contractor in this case? 3 A. Yes, she is a subcontractor to me. 4 Q. How long did she work with you on this 5 case? 6 A. Elapsed time, the beginning date? 7 Q. Yes. 8 A. Approximately, it would be the date of the 9 first report that her name is on minus about two 10 months, so probably around January 1 -- late 1992, 11 fall of 1992. 12 Q. And is she continuing to work on the 13 project now? 14 A. Yes. 15 Q. Can you tell me her background 16 educationally? 17 A. The bachelor's degree, I'm not certain of; 18 she has a master's degree in civil engineering from 19 the University of Washington and some additional 20 graduate work in civil engineering. 21 Q. Were you ever her teacher? 22 A. I was her adviser for her master's 23 program. 24 Q. Did she do a thesis? 25 A. Yes. 14 1 Q. What was that on? 2 A. Operation of storm water detention 3 facilities for water quality improvement, 4 approximately would be the title 5 Q. Operation of storm water "redemption"? 6 A. Detention facilities for water quality 7 improvements. 8 Q. Do you have a copy of her work? 9 A. Her work? 10 Q. Her thesis. 11 A. Do I have a copy with me of the thesis? 12 Q. Well, we could start with that. Do you 13 have a copy of the thesis with you? 14 A. No. 15 Q. I didn't think so. 16 Do you have a copy in Florida with you? 17 A. No. 18 Q. Do you have a copy back in Washington? 19 A. Yes. 20 Q. Could we obtain a copy of that? 21 A. Could I ask counsel as to how such a 22 request get approved? I have no objection to 23 providing a copy. 24 MR. BLANK: Probably the best thing would 25 be to send it to me and let me take a look at it. 15 1 THE WITNESS: The answer is yes. 2 BY MR. LOREDO: 3 Q. Can you tell me a little bit about it? It 4 seems to relate somewhat. Since we are dealing in 5 the Everglades with storm water treatment areas, can 6 you tell me a little bit about what this thesis was 7 about? 8 A. The scale was much different, first off. 9 This was an operation of two detention facilities, 10 one in a residential subdivision, and I'm guessing 11 some, because this was about 10 years ago, but as I 12 recall, about a hundred-acre development. 13 The work had to do with some mathematical 14 predictions and field data collection to indicate how 15 the facility was affecting water quality downstream. 16 In the first case, the residential 17 facility, it was a dry detention pond which fills 18 only during storms. The second site was an 19 industrial -- or actually the municipal transit 20 authority's operating base for a large number of 21 buses that were stored for maintenance overnight and 22 so on. 23 That was a wet detention facility, and the 24 issue there had to do primarily with oil and grease. 25 The size of that facility was I believe about 50 16 1 acres, mostly blacktop. 2 Q. I don't think I got the location, the 3 residential and industrial. Was that in Washington? 4 A. Both of them were in the Seattle 5 metropolitan area, not within the city, but within 6 King County. 7 Q. What was your conclusion? 8 A. There was some improvement achieved in the 9 water quality by the detention facilities; the dry 10 one, the effects were somewhat modest. There was 11 more effect on the wet facility. 12 Q. And how was that conclusion reached? 13 A. Based on field data collection. 14 Q. And who did the field data collection? 15 A. Ms. Dally Wilson under my direction. 16 Q. How many samples were taken? 17 A. Again, I'm going from recollection of work 18 that was done some time back, so I would have to go 19 look at the specific reports to see, but the data 20 collection period extended over two seasons, winters, 21 as I recall, and there were some number of storms, 22 approximately 10 in each case. 23 Q. When you say "two seasons," I guess you 24 started with one winter and you came back around 25 another winter and you went back to another winter, 17 1 so you could have had three winters? 2 A. My recollection is that there were two 3 winters. 4 Q. What did you do with the data that was 5 collected? 6 A. The physical data were analyzed, the 7 results were summarized in her report, which was also 8 Ms. Dally Wilson's thesis. The electronic form of 9 the data are probably around somewhere. I wouldn't 10 want to have to try to read the magnetic tapes they 11 are on, but they are archived somewhere. 12 Q. After you summarized the data, what did 13 you do with the data? 14 A. Well, as I indicated, they are in a 15 written report which is in fact the thesis that you 16 have requested. 17 Q. I guess what I'm trying to get at is what 18 was your next step? After you summarized that, I 19 guess you try to develop some kind of relationship or 20 statistical analysis? 21 A. No, this was primarily a study to document 22 the effects, that was as far as the study went. 23 Q. You never created any type of statistical 24 models? 25 A. For general prediction? 18 1 Q. Yes. 2 A. No. 3 Q. Any specific predictions? 4 A. No. This was entirely a summary for the 5 particular sites in question. 6 Q. So all you did was basically you took 7 data, you analyzed the data, and you came to a 8 conclusion that at least in the wet detention areas, 9 there was more than of an impact versus the dry? 10 A. Yes. 11 Q. Is the storm water detention facility 12 still up? 13 A. To my knowledge, yes. 14 Q. Has Ms. Wilson done any other work beside 15 the Everglades work, which we will get to, under your 16 supervision? 17 A. Yes. 18 Q. What kind of work was that? 19 A. I had a contract with South Florida Water 20 Management District to review their Lake Okeechobee 21 water quality monitoring programs. 22 Q. Besides the Lake Okeechobee work, anything 23 else -- or the Everglades work, which I plan to get 24 to that later? 25 A. In which she has worked as a subconsultant 19 1 to me? 2 Q. Yes. 3 A. No. 4 Q. Let me go back to this. 5 Looking back at Exhibit 1, the subpoena 6 duces tecum, going through it, I would like to 7 confirm that you provided all the documents and that 8 we in fact received all the documents pursuant to the 9 subpoena. 10 The first one is a current copy of your 11 curriculum vitae, or your resume. 12 I believe you did provide, or at least it 13 was received by fax, a resume. When did you provide 14 your current resume to Mr. Blank? 15 A. We got a more current one yesterday. I 16 believe the one that probably was provided you within 17 the last two weeks had been provided to him either at 18 the initiation of my work for them or possibly for 19 the prior deposition. It has a date on the last 20 page, is the way you can tell. I'm not sure what 21 date you have there. 22 Q. I have a date of August 14, '93? 23 A. That's the one from yesterday and that is 24 the most current vitae in my files in Seattle. 25 Q. When did you provide that to Mr. Blank? 20 1 A. That was faxed yesterday. 2 Q. To Mr. Blank? 3 A. Yes. 4 Q. And is it up to date? 5 A. It is not entirely up to date, vitae in 6 the academic world never are. The updates are 7 probably of minimal interest to this case, but the 8 updates would have to do with certain refereed 9 articles on about page 3, which were in a review 10 status and have been accepted for publication now. 11 Q. You have been going to the University of 12 Washington since 1975 starting as a research 13 engineer. Who was your mentor, adviser? 14 A. My dissertation adviser was Stephen 15 Burges. 16 Q. Can you spell Burges? 17 A. B U R G E S. 18 Q. What is the distinction between the 19 department of civil engineering and the department of 20 civil engineering-college of fisheries? Why that 21 distinction? 22 A. The college of fisheries no longer exists, 23 it is now the college of fisheries and ocean 24 sciences, I believe. At that time, I had a joint 25 appointment. There's a college of engineering within 21 1 which the civil engineering is a department. I had a 2 joint appointment for a short period between the 3 college of fisheries and the department of civil 4 engineering. 5 Q. What was your BA in, your major? 6 A. BS. 7 Q. Your BS, I'm sorry? 8 A. Mechanical engineering. 9 Q. Is that a four- or five-year program? 10 A. Four. 11 Q. Is it now a five? 12 A. No. 13 Q. Is Mr. Burges still at the University of 14 Washington? 15 A. Dr. Burges, yes. 16 Q. What was your thesis on? 17 A. Design of water quality networks. 18 Q. What did you say? 19 A. Design of -- it is probably titled Stream 20 Quality Monitoring Networks. 21 Q. Networks? 22 A. Yes. 23 Q. Can you tell me a little bit about that? 24 A. It had to do with the placement of water 25 quality monitoring stations on river networks, how 22 1 many stations, how far apart, how frequently data 2 should be collected for optimal detection of water 3 quality trends, given a constraint on the number of 4 samples that could be collected. 5 Q. How much of an area are we talking about 6 in the stream? 7 A. The demonstration cases, as I remember, 8 were for river networks on the order of thousands of 9 square miles. 10 Q. How long did you work on your thesis? 11 A. The actual dissertation research? 12 Q. Yes. 13 A. A year, a year and a half. 14 Q. Was that including actually writing the 15 thesis and reaching the conclusions? 16 A. Yes. 17 Q. What conclusions did you reach? 18 A. Well, conclusions -- I would have to go 19 read the thesis again to see what were listed as 20 conclusions. The results of the work were a method 21 to optimize the placement of stations and the 22 trade-off between -- designing the trade-off between 23 number of stations and how often you would collect 24 data at each station. 25 Q. What was the method? 23 1 A. The method was based on a combination of 2 the Kalman Filter to determine where stations would 3 go and statistical trend analysis to determine how 4 often samples would be collected at individual 5 stations. 6 Q. You said it was dependent on "column and 7 filter"? 8 A. K A L M A N filter. 9 Q. K A L M A N? 10 A. Kalman is the name of a person. 11 Q. Is that a method? You named it after 12 somebody? 13 A. The Kalman filter is fairly widely used in 14 control theory applications. It came from space 15 craft guidance problems. 16 Q. Kalman? 17 A. K A L M A N. 18 Q. And the second word? 19 A. Filter, F I L T E R. 20 Q. I'm sorry, I cut you off, but I was trying 21 to get that down. 22 You said it was a technique commonly used? 23 A. In space craft guidance, it was developed 24 for originally. 25 Q. Can you tell me a little bit about the 24 1 method? 2 A. Just briefly, it is a method for combining 3 information from two sources, one of which is 4 measurements and the other of which is a 5 deterministic model prediction. 6 Q. So you are combining actual measurements 7 with, you said, a deterministic model? 8 A. Yes. 9 Q. A deterministic model, is that variables 10 you create or actual data? 11 A. In this particular case, this was a model 12 that would predict how river quality for particular 13 variables would change as one went downstream given 14 certain characteristics of the river and certain 15 characteristics of the inflows, waste discharges to 16 the river. 17 Q. What sort of variables were in here? 18 A. As I recall, in that model, dissolved 19 oxygen, biochemical oxygen demand, I believe total 20 phosphorus, and I cannot remember at this point 21 whether it was total or orthophosphorus and a measure 22 of nitrogen. 23 Q. I apologize, orthophosphorus, what is 24 that? 25 A. That's simply another way of chemically 25 1 analyzing for a certain type of phosphorus, it is a 2 subconstituent of total phosphorus. 3 Q. Then you said you subsequently developed a 4 statistical trend analysis? 5 A. Yes. 6 Q. And what type of relationship did you 7 discover? 8 A. Well, the issue is not the relationship 9 that was discovered, it is how in the optimization, 10 the two components linked together, one being the 11 description of evolution of water quality down the 12 river, which was a key element in determining where 13 the stations would go, and the other was the 14 statistical trend assessment at the individual 15 stations, which was a key element in determining how 16 frequently one would sample. 17 Q. How frequently did you determine it should 18 be sampled? 19 A. That depends on the characteristics of the 20 river. 21 Q. Did you determine what the characteristics 22 of the river were to figure out how often you should 23 sample and how far apart the stations should be? 24 A. Yes, but this was a generic analysis which 25 was then applied in some particular cases. So 26 1 depending on the characteristics of the particular 2 river, the trade-offs between how often one would 3 sample and how many stations would then fall out. 4 Generally, the sampling frequencies tended 5 to be between, roughly, very roughly, once every two 6 weeks, and once every few months is where the 7 solution usually tended to lie. 8 Q. When you said generic approach, you didn't 9 apply it to any particular river? 10 A. I did do an application, yes, but it was a 11 demonstration application. In fact, I believe there 12 were two. 13 Q. And then did your demonstration concur 14 with your conclusions and opinions? 15 A. Well, that isn't the way the experiment 16 was designed. It was a development of a method and 17 demonstration application. There was no comparison, 18 for instance, with data to know what was right 19 because there is no way of determining in the absence 20 of an extremely extensive data collection program, 21 which does not exist, what the correct or true 22 solution would be. This was a demonstration 23 application. 24 Q. And in the demonstration, did you actually 25 take data from the river? 27 1 A. There was no primary data collection. 2 There were data used. They were not my data. 3 Q. Was there any quality control with respect 4 to the data used? 5 A. In what respect? 6 Q. In the collection of it, in the 7 measurements of the data? 8 A. The data were quality controlled, but that 9 was not part of my research. 10 Q. You just took someone else's data and 11 applied it to your analysis? 12 A. This was a demonstration application. 13 Q. I guess I am stuck on "demonstration 14 application." I don't know the difference. 15 A. That simply means the primary part of the 16 work was to develop a method. In order for readers 17 of the work to understand, to help them understand 18 how it might be used, there was a demonstration 19 application performed. 20 Q. This was done twice? 21 A. It was done for two separate rivers, yes. 22 Q. Which rivers? 23 A. The Spokane River was one, and as I 24 recall, the Snohomish River -- S N O H O M I S H -- 25 both Indian names, were used. 28 1 Q. Do you know if your analysis is being used 2 now by anyone? 3 A. The analysis was used in some subsequent 4 work that I did for the State of Washington 5 Department of Ecology to redesign their ambient 6 stream water quality monitoring network. 7 Q. Was that done by you? 8 A. It was done by me, yes. 9 Q. So you actually went out and took data now 10 and you have applied it? 11 A. No, again, this was based on data that 12 they had collected. I went back and determined how 13 often they should be sampling at the stations they 14 had, where the stations should be relocated, and how 15 many should be on the rivers. There's a report that 16 summarizes that work. 17 Q. Was phosphorus an issue? 18 A. That was one of many water quality 19 constituents that the agency was collecting. 20 Q. If I show you your resume, could you tell 21 me which publication? 22 A. Yes, probably. It would be a technical 23 report. 24 Q. Great. 25 (Pause) 29 1 A. The 1977 report Detection of Trends in 2 Water Quality. 3 You don't mind me making a mark? 4 Q. No, that would be fine. 5 A. This one. 6 Q. How long, how much of your time did you 7 dedicate to this? 8 A. I believe that was approximately half time 9 for approximately a year. 10 Q. When you say "half time," half of your day 11 every day? 12 A. Yes. 13 Q. Would you say a thousand hours then? 14 A. Probably a fair guess. 15 Q. Have you done any consulting work with Mr. 16 Burges, Dr. Stephen Burges? 17 A. Can I see my vitae for a minute? There 18 may be one project way back. 19 I think you will find the first two, which 20 were way back in the mid-70s, he would have had some 21 involvement in. 22 Q. Stochastic Analysis for Flow Yield 23 Evaluation? 24 A. Yes. Of the Bull Run Reservoir Water 25 Supply for the City of Portland, Oregon. 30 1 Q. And University of Washington Review of 2 River Basin Coordinating Committee Proposal to 3 Municipality of Metropolitan Seattle? 4 A. Yes, that was a very small project, 5 probably a few hours of my time in that. 6 Q. I see on your next consulting work that 7 you did a review of Washington State Ambient Stream 8 Water Quality Monitoring Network? 9 A. That is the same as the report that we had 10 just spoken about, the same project. 11 Q. Is it? 12 A. That was done in a consulting mode. 13 Q. Now, do you teach presently? 14 A. Some. My appointment is primarily 15 research. 16 Q. Have you taught? 17 A. Some. 18 Q. What kind of courses? 19 A. I have taught classes in at one point -- I 20 can't remember the exact title of the course, but it 21 was statistics for fisheries for forestry majors; I 22 have taught computer programming; I have taught fluid 23 mechanics. 24 Q. What was the last one? 25 A. Fluid mechanics; I have taught statics. 31 1 Q. What is statics? 2 A. Statics is basically the study of forces 3 on immobile objects or structures. 4 Q. The study of the forces? 5 A. The balance of forces on immobile objects, 6 i.e. bridges. That's an introductory class. That's 7 generally a second-year engineering class. 8 I have taught more recently a graduate 9 level class in rainfall analysis and prediction, and 10 another in environmental data analysis. 11 Q. Was this all at the University of 12 Washington? 13 A. Yes. 14 Q. The first part when you said you had 15 taught statistics for fisheries, the computer 16 programming, was that several years ago? 17 A. That was quite some time ago. The 18 statistics was done at the time, I believe. My 19 appointment was jointly with the college of 20 fisheries, it would have been the late 1970s. The 21 computer programming was early mid-1980s. 22 Q. And the more recent teaching, the rainfall 23 analysis and the environmental data analysis, when 24 was that? 25 A. Those have been over the last three years. 32 1 Q. Are these to mechanical engineers? 2 A. Civil engineers, my appointment is in 3 civil engineering. 4 Q. Are these one-year courses? 5 A. No, one term, one-quarter classes. 6 Q. A quarter being half a year? 7 A. No -- well, the quarter system is three 8 academic quarters plus summer, so typically a quarter 9 is of 10-weeks duration. 10 Q. Did you have any actual applications in 11 your rainfall analysis in your courses? 12 A. Actual applications as in case studies? 13 Q. Yes. 14 A. I have had them do a little bit of work 15 with radar rainfall data. I have also had them do 16 some analysis in stochastic model fitting to guage 17 precipitation theories. 18 Q. If I can pronounce it -- "stochastic," 19 what does that mean? 20 A. Stochastic? Oh, that simply means 21 probabilistic but varying in time as well, so the 22 structure of time varying uncertain data that are 23 characterized by some probability description at a 24 given time and some evolution in time as well. 25 Q. I guess that's like a seasonal probability 33 1 because it could be the rainy season versus the dry 2 season? 3 A. No, not necessarily. That is more of the 4 nature of the sequencing of rainy days and the amount 5 of precipitation that occurs in subsequent time 6 periods; for instance, the amount of precipitation 7 occurring in one day as related to the occurrence or 8 lack thereof of precipitation in the next day, and if 9 there's precipitation, how much. The time evolution 10 of the probabilities, of the probability 11 distribution. 12 Q. There's probably a separate course on 13 that? 14 A. There are a number of different courses on 15 stochastic theory, yes. That is a field in and of 16 itself. It is simply one application. 17 Q. When you said "radar rainfall," is that 18 data collected from the radar? 19 A. Yes. 20 Q. I didn't know radars can detect rainfall. 21 A. The National Weather Service has invested 22 a considerable amount of money in radar rainfall over 23 the recent past. 24 Q. The environmental data analysis, what type 25 of case studies did you have there? 34 1 A. The nature of the course has to do with a 2 number of topics in environmental data analysis, such 3 as spacial data analysis, kriging being one of the 4 topics we looked at. 5 I don't remember the particular data set 6 that I had given them to work with. I believe it was 7 some well nitrogen data. That was simply one aspect 8 of that course. Environmental trend analysis was 9 another application. They may actually have worked 10 with some of the Lake Okeechobee data in that case. 11 I would have to go back and look at the 12 course outline to see. There was some basic 13 probability estimation for different kinds of 14 environmental data involved in the course. There 15 were some aspects of graphical display and 16 interpretation of data. 17 Q. What type of environmental data are we 18 looking at? 19 A. Water quality certainly being the primary 20 one. I can't recall if I used applications with air 21 quality data or not, probably mostly dealing with 22 surface water quality, although some groundwater 23 quality issues as well. 24 I believe we did use the data, the soil 25 chemistry data from Love Canal as one of the 35 1 applications, so there would have been some soil 2 contamination issues discussed as well. 3 Q. Did everyone pass? 4 A. Yes. 5 Q. You were telling me about spacial analysis 6 and kriging. I guess I will take the opportunity to 7 learn a little about it now before I get into it 8 later. 9 Can you expand on that? 10 A. The method, how it works? 11 Q. Yes. Is spacial analysis the same as 12 kriging? 13 A. Kriging is a particular form of spacial 14 analysis. 15 Q. Can you just give me a summarization of 16 kriging? 17 A. Yes. Basically it is a data interpolation 18 that is sometimes termed a stochastic interpolator. 19 Q. Turned in to? 20 A. "Termed" a stochastic interpolator. 21 Basically what it does is it takes observations at 22 some finite number of locations, for instance, wells, 23 or in the case we have applied it to here, core data 24 that are associated with a particular set of 25 coordinates. 36 1 From those, it attempts to estimate at any 2 arbitrary point what the best estimate of the 3 variable in question is, and then how uncertain that 4 estimate is based on what it infers from the data 5 that have been collected. 6 And it is a linear, it is a linear 7 interpolator which means that the value at any 8 arbitrary point is some weighted sum of all of the 9 observations. 10 Q. So if we had two wells, you are going to 11 try to calculate at some point in the middle of it 12 what the data should be? 13 A. That would be an extremely simplified 14 case. In fact, usually kriging is not used for a 15 number of samples less than -- around 30 is a very 16 rough rule of thumb, but if one had, say, 30 wells -- 17 Q. In a line? 18 A. Not in a line, over an area, over an area 19 with some spacing and one wanted to draw, for 20 instance, contour lines, and to do that, one needed 21 an estimate of the value of whatever variable at any 22 arbitrary point, kriging would do that. 23 Q. You said "linear," and I guess when I 24 think linear -- 25 A. No, I'm sorry, "linear" simply means that 37 1 the value is a weighted sum of the values at any of 2 the points as opposed to, for instance, the values 3 squared or some arbitrary transformation. 4 Q. A little help for later. 5 Are you presently teaching? 6 A. As in this term? 7 Q. Yes. 8 A. No, I will be next term. 9 Q. What are you going to teach next term? 10 A. It will be the precipitation analysis 11 class. 12 Q. Is that rainfall, the same thing? 13 A. Yes, yes. Actually I am -- there's a 14 seminar that I have organized this term, so I guess I 15 should say that technically I guess I am "teaching." 16 Q. Participation may not necessarily mean 17 rain, though? 18 A. Well, it could be snow or sleet or some 19 other form; mostly it deals with precipitation of 20 rain. 21 Q. What's the seminar going to be about? 22 A. Well, the seminar, the one that's ongoing 23 this term is the required seminar for all master 24 students in our program. So it has a range of 25 topics, and I have simply organized the speakers who 38 1 have come in. Most of the topics have to do with 2 hydrology or water resources, one way or the other. 3 Q. Who are some of the speakers you have 4 organized? 5 A. Names? 6 Q. Yes. 7 A. Oh, Charles Howard runs a consulting firm 8 in Victoria, BC; Jeff Richey -- R I C H E Y -- from 9 our oceanography school -- there are 10 of them. I'm 10 not sure I can remember all the names even when I go 11 down through the whole list. 12 Q. Just to the best of your recollection. 13 A. To the best of my recollection, Wendell 14 Tanbgorn -- T A N G B O R N -- was one of the 15 speakers. 16 Q. Where is he from? 17 A. He runs his own consulting firm in 18 Seattle, a hydrologist. 19 Keith Loague -- L O A G U E -- from the 20 University of California; Dave Montgomery from our 21 geology department -- who gave the seminar yesterday 22 -- I wasn't there -- 23 Q. Are you going to speak? 24 A. No, not in this series. 25 Q. Is there anyone from Florida? 39 1 A. No. We don't have money to fly people in. 2 Q. Will any of the topics in the seminar deal 3 with the Everglades? 4 A. No. 5 Q. And then you said the seminar is ongoing 6 now, you are putting it together now. 7 You will teach a course next quarter in 8 precipitation, correct? 9 A. Yes. 10 Q. That's for the masters also? 11 A. Well, it is for graduate students, a 12 graduate level course, both Ph.D and masters. 13 Q. Are you advising anyone now on I guess 14 their Ph.D's? 15 A. Yes. 16 Q. Several people? 17 A. I have several students, yes. 18 Q. Not to get into all the students, but does 19 anything have to do with phosphorus trend models? 20 A. No, no. 21 Q. Anything to do with the Everglades cases? 22 A. No. 23 Q. Lake Okeechobee? 24 A. No. 25 Q. You figured if you did that we might call 40 1 them in for depos? 2 A. No, none of the work has anything remotely 3 to do with Florida. 4 Q. Is Steve Millard a speaker at the seminar? 5 A. No. 6 Q. Dr. Steve Millard. 7 Do you need a license for your consulting 8 work? 9 A. You don't need a license, no, but I do 10 have one. I am a licensed professional engineer. 11 Q. Do you do anything other than consulting 12 work? 13 A. Well -- 14 Q. -- and teaching? 15 A. I wouldn't put it that way to my 16 colleagues at the university. 17 Q. I'm sorry. 18 A. I am an employee of the University of 19 Washington 75 percent of the time. 20 Q. I guess the other 25 percent would be 21 dedicated to consulting work? 22 A. Yes. 23 Q. You have listed some professional 24 memberships. Are you active in any of those 25 organizations? I guess when I say "active," I mean 41 1 just showing up and participating a little bit? 2 A. Yes, the American Geophysical Union, I am 3 most active in, and I am on the American 4 Meteorlogical Society committee as well. 5 Q. What type of activities? 6 A. For AGU, I am the chapman conference 7 coordinator. Chapman conferences are short, 8 typically three- to five-day conferences on focused 9 topics in all areas of geophysics. I help coordinate 10 those conferences for the union. 11 I am also the deputy editor of Water 12 Resources Research. 13 Q. Water -- 14 A. Water Resources Research, which is a 15 journal. I believe I am a member of the Surface 16 Water Committee. 17 Q. What comes under geophysics? 18 A. Basically earth sciences, all areas of 19 earth sciences. And within the American Geophysical 20 Union, that is split, more or less, into areas like 21 the sort of hard earth sciences, things like 22 tectonophysics, seismology, vulcanology and so on. 23 There are eight sections, I believe, and 24 I'm not sure I could name them all, but the other 25 side is the land-atmosphere area of which hydrology 42 1 is my home. That's one of the major sections, 2 atmospheric sciences-oceanography, ocean sciences. 3 Q. I guess the ocean does affect -- let me 4 back up. 5 Hydrology, how would you define hydrology? 6 A. Hydrology is basically a water at or near 7 the land surface. 8 Q. How long have you been the deputy editor 9 for the Water Resources -- 10 A. Water Resources Research? 11 Q. Right. 12 A. About six months. I was an associate 13 editor before that. I think that's listed on the 14 first page. 15 Q. Yes. 16 The title "Referred" Publications -- 17 A. "Refereed." I hope it is not misspelled 18 there. 19 Q. No, I decided not to pronounce the next 20 "E." 21 Q. What do you mean by "refereed," you just 22 supervise? 23 A. A refereed publication is one that 24 typically is a journal publication which has been 25 reviewed as opposed to a report which would not 43 1 necessarily have undergone a formal review. 2 Q. So all refereed publications have been 3 reviewed by one or other experts in the field? 4 A. Yes, usually some number of experts, and 5 it is handled by an editorial editor or editorial 6 board of some nature. 7 Q. You have listed them in chronological 8 order? 9 A. Yes. 10 Q. Let me start with you your most recent 11 one. Hydrological forecasting. Does that have 12 anything to do with the Everglades? 13 A. No. 14 Q. What was it about? 15 A. This is the handbook chapter? 16 Q. Yes. 17 A. That's a chapter in the Handbook of 18 Hydrology which is a McGraw-Hill publication. It has 19 to do with forecasting of river flow. 20 Q. How do you forecast river flow? 21 A. It depends on the time and scale. 22 Q. The time period in a year? 23 A. Well, how far into the future you are 24 interested in forecasting, i.e., flood forecasting 25 might be a matter of only hours or so if it was a 44 1 flash flood. Seasonal forecasting for reservoir 2 operations could be a matter of months. There are 3 different methods. 4 Q. You can actually forecast for a flood? 5 A. Oh, floods are forecast all the time. 6 Q. What causes -- I mean, you think of a 7 flood that's going to be caused by rain or some other 8 dumping on the earth, something happening out there? 9 A. Yes, flood forecasts are issued, and 10 generally there's little forecasting skill before the 11 onset of the storm. Once a storm has begun, though, 12 forecasts are routinely issued warning people what 13 the river stage might be, which areas should be 14 evacuated, things like that. 15 Q. Can you forecast for a storm? 16 A. Could you clarify that? 17 Q. Well, you said the flooding was dependent 18 on a storm coming, and I'm wondering whether or not 19 you can forecast when storms are going to come based 20 on I guess prior data analysis of storms? 21 A. That's the meteorologist's problems. 22 Q. But you wouldn't do any forecasting other 23 than from the storm itself then? 24 A. Again, it depends on time scale. 25 Short-term forecasting may integrate meteorlogical 45 1 forecasts as well as information about what the river 2 flows are at that time, other information about the 3 conditions of the land surface, for instance, soil 4 moisture. 5 Q. Your Dynamic Modeling of the Spacial 6 Distribution - Precipitation in Remote mountainous 7 Areas. 8 What was that about? 9 A. That was part of a dissertation of a 10 student of mine which has to do with understanding 11 how precipitation varies in complex mountainous 12 terrain. 13 Q. In Seattle? 14 A. Well, there were applications of the work 15 to the Olympic Mountains of Washington and the Sierra 16 Nevadas of California, but the model is a physically 17 based model and is not particular to any region. 18 Q. The Trends in Stream Quality in The 19 continental U.S. 20 Does that have anything to do with the 21 Everglades? 22 A. That was the result of work in part that I 23 did when I was on leave with the U.S. Geological 24 Survey, and has to do with analysis of trends at a 25 large number of stations across the continental U.S. 46 1 Some part of those stations are in 2 Florida, and I do not recall the particular streams 3 in South Florida that were involved; one of the 4 streams draining from Lake Okeechobee; I seem to 5 recall Fish Eating Creek has one of the stations on 6 it. 7 And there may be some others in South 8 Florida, but it is not specifically directed at South 9 Florida in the national scale analysis. 10 Q. What were you trying to do or determine? 11 A. We were simply looking for the spacial 12 pattern of trends in water quality since the onset of 13 this program of the U.S. Geological Surveys in the 14 early, mid-1970s. 15 Q. Did you find any trends? 16 A. Yes, there were some trends. I would have 17 to go look at the paper to see the specific patterns 18 because there were a large number of constituents 19 that were monitored in general. There are not a 20 large number of trends in any one constituent. 21 Typically, the largest number of the 22 trends, as I recall, were in constituents like total 23 nitrogen and would be around maybe a hundred of the 24 500 stations at most. 25 Q. Do you want to take a break? 47 1 A. Sure. 2 (Recess) 3 BY MR. LORDEO: 4 Q. As interested as I am in the refereed 5 publications, we could probably sit here talking 6 about them all day. 7 What I would like you to key on -- and 8 maybe I will let you look at this -- obviously what 9 we are interested in is where you might have done 10 something similar or where there are similar 11 statistical analyses as to the SWIM Plan. 12 A. Would you like me to go through the vitae 13 and perhaps mark those that have something to do with 14 water quality and aspects that may have any bearing 15 on the case, and it would save some time? 16 Q. Yes, I certainly would. 17 A. Okay. 18 Q. I would like you to err on the side -- if 19 you even think it has something to do with our case, 20 I would like you to mark it. 21 A. None of it has to do directly with this 22 case. You won't find anything in there which has to 23 do with analysis of Everglades data or anything like 24 that. 25 Other than the work for the district, I 48 1 have not worked in South Florida before that. So you 2 are talking all about methodological work and I think 3 I have indicated the ones that might be most 4 applicable. 5 Q. I don't think I have marked this as an 6 exhibit. 7 Why don't we mark this as Exhibit 2. 8 (Lettenmaier Exhibit 2 was marked for 9 identification) 10 BY MR. LOREDO: 11 Q. All right, Doctor, I guess you went 12 through the refereed publications and the technical 13 reports? 14 A. I just went through the refereed 15 publications. If you want me to mark refereed 16 reports, I can do that as well. 17 Q. Why don't you finish. That was a previous 18 marking under technical reports? 19 A. Yes. 20 Q. If you would also look, you had some other 21 publications above that. 22 A. These are papers in reviews, but none of 23 those are applicable. 24 Q. Thank you, Doctor. 25 It will make this a lot easier. 49 1 Did we already talk about the 2 Probabilistic Methods in Stream Quality Management? 3 A. Where did that appear? 4 Q. In your first one of the refereed 5 publications. You had referred to this because you 6 had worked with Dr. Burges on it? 7 A. Yes, I think so, right. 8 Q. Why don't you tell me a little bit about 9 that publication and the methodology used. 10 A. This is the one in -- the Water Resources 11 Research bulletin is the name of the publication. I 12 have to admit the details of that, I can't recall. 13 That was probably the first publication I was 14 involved in. 15 I believe that it had to do more with 16 indicating how probability -- the role, for instance, 17 probability theory might play in stream quality 18 management. It was more in the nature of a tutorial. 19 Q. You may not recall the specifics in the 20 publication, but what role do you believe the 21 probability should play? 22 A. Well, I think it is a key element in 23 characterizing natural systems. One has to 24 understand how a particular sample might be viewed in 25 the larger picture, the uncertainty that is 50 1 associated with, say, individual measurements and 2 collections of measurements. 3 Q. I guess one of the key elements in what 4 variables affect certain data is time and other 5 factors? 6 A. Well, I think that's one aspect of it, 7 yes. 8 Q. There are other aspects? 9 A. The key aspect of probability theory is 10 that one views a sample as a particular realization 11 that can come from some statistical population, that 12 others could have come as well. One then talks about 13 probability distributions rather than particular 14 values. 15 Q. Can you repeat what the key element was 16 again? I almost had it. 17 A. Now I can't remember what I said. 18 I think the idea was that a group of 19 observations need to be characterized in terms of 20 what we cannot explain about them in terms of known 21 factors, and what's left over is what gets described 22 probabilistically. That might be the best way to 23 view it. 24 Q. Why don't we talk about the next 25 publication that you have marked in Exhibit 2 in your 51 1 resume: Detection of Trends in Water Quality Data 2 with Dependent Observations. 3 A. That's part of my dissertation work. 4 Q. We already talked about that? 5 A. Yes. 6 Q. That had to deal with how many stations -- 7 A. Yes, but that particular publication only 8 has to do with the trend detection part of the work, 9 so that's essentially the time frequency of sampling 10 issue that dealt with in that publication. 11 There's another publication that I may or 12 may not have marked, a couple down, which has to do 13 with spacial number of stations and how far apart and 14 so on. It would be in the American Society of Civil 15 Engineers, about 1976 or '77. 16 Q. Do you want to take a quick look? 17 A. Sure. 18 I don't think the spacial part of it is of 19 particular interest here, so I probably didn't mark 20 it. 21 Q. Okay. 22 A. The Design of Trend Monitoring Networks, 23 October '77, is the other part of the dissertation. 24 Q. But in that publication, there's really 25 nothing of interest to us in terms of statistical 52 1 modeling? 2 A. That has to do with more with the Kalman 3 filter part of the work. I think for this case, the 4 trend part of it is of more direct interest. 5 Q. But that's not applicable here? 6 A. It is not applicable in any of the work 7 that I have done. 8 Q. But was a trend developed? 9 A. I'm saying the trend work is applicable in 10 some of the reviews and so on that I have conducted, 11 and that's marked. 12 Q. And I will go back to your Detection of 13 Trends in the Water Quality. 14 What trend did you detect, if any? 15 A. Well, again, this is methodological. In 16 fact, I think if you review some of the publications 17 of one of your witnesses, you will see actually 18 references to that particular work, the one I marked 19 there. 20 Q. This is just physical? 21 A. Well, the one with the dependence, 22 Detection of Trends with Dependent Observations, is 23 that the one you are talking about? 24 Q. Yes. 25 A. That has to do with what the effect is of 53 1 statistical dependence between observations in time. 2 That was an issue that hadn't really been dealt with 3 in the work before then, and that paper ended up 4 being somewhat widely quoted in that area. 5 Q. But was there any modeling done? 6 A. Well, statistical and probalistic 7 modeling, yes. 8 Q. What type of modeling? 9 A. That's a difficult question, but as I 10 recall, I used what's known as a Lag One Markov -- 11 that being the name of a person. 12 Q. M A R? 13 A. K O V -- to describe the dependence in 14 time between observations. 15 Q. The Lag One Markov, is that a widely used 16 model? 17 A. In time series modeling, yes. 18 Q. Again we touched on this before, but when 19 we talk about time series, we are not looking at 20 seasonal, we are looking at other factors? 21 A. Seasonal is one of the factors that can in 22 fact has to be addressed. 23 Q. What are the other factors, variables? 24 A. The statistical dependence between 25 observations is the major one that's dealt with. 54 1 That dependency may exist regardless of the 2 seasonality. 3 Q. So like maybe a decision whether to 4 analyze data on a weekly versus a monthly basis? 5 A. That plays a strong role in that decision 6 because the statistical dependence becomes stronger 7 as the sampling frequency becomes higher, and there 8 ends up being a loss of information relative to what 9 information would be in a sample where all the 10 observations were independent, and that loss of 11 information becomes a more serious issue as the 12 samples become closer together. That's basically 13 what that paper deals with. 14 Q. You are saying the more sampling that you 15 do, the more likelihood of having lost information or 16 misinformation? 17 A. It is not so much an issue of 18 misinformation as it is redundancy in samples. If 19 you take four samples over a period of four months as 20 opposed to four samples over a period of four weeks, 21 the seasonal issues aside, there's more information 22 in the four samples taken in the four months. 23 Q. If you take 16 samples over the four 24 months versus four samples over the -- 25 A. There certainly will be more information 55 1 in any case of practical interest than the samples 2 taken over -- 3 Q. Wouldn't you get a truer model by taking 4 more samples? 5 A. Yes, but in the real world, people have 6 budgets. 7 Q. What misled me a little -- I shouldn't say 8 "misled," but seemed inconsistent, you said there's 9 redundancy in the more samples that you took, and I 10 thought -- 11 A. Typically one expresses this as a ratio of 12 something called the effect of independent sample 13 size divided by the actual sample size, and if 14 there's no redundancy, then those two are the same 15 and that ratio is one. 16 As you increase your sampling frequency 17 and start to encounter statistical dependence, that 18 ratio goes below one. 19 Q. Let's take a look at the next publication 20 which you have marked: Design Considerations for 21 Ambient Stream Quality Monitoring in the Water 22 Resources Bulletin, 1978. 23 A. That publication resulted from the work 24 that I did for the state department of ecology, and I 25 believe it was more of the nature of a case study in 56 1 how some of the earlier methods, which were described 2 more in a theoretical context, were applied to an 3 actual state-wide water quality monitoring effort. 4 Q. Was this an actual application, then, of 5 theories? 6 A. The work that was done for the state was, 7 yes. If you are asking me what was in that 8 particular paper -- I'm going from recollection, and 9 I would be more than happy to provide you copies of 10 any of those publications, incidentally, that are 11 marked there. 12 Q. Well, obviously you marked it because you 13 thought it had some type of relationship? 14 A. Well, what I have done is I have marked -- 15 as you might notice there, I have done work in a 16 number of different fields. 17 Q. Yes. 18 A. Some of which have absolutely nothing to 19 do with the Everglades. So I have gone through and 20 tried to mark the ones which have some bearing on 21 water quality and statistics of water quality, not 22 because there's any particular application in them 23 that might have to do with, say, South Florida or 24 anything like that. 25 Many of them still are fairly broad or 57 1 methodological, and none of those publications, I 2 believe, have any particulars having to do with South 3 Florida in them. 4 Q. Not particulars, but how about 5 methodology? 6 A. Well, the methodology certainly can and in 7 some cases have been applied. 8 Q. And the publication, the Design 9 Considerations for Ambient Stream Quality Monitoring, 10 which you were just referring to, what type of 11 methodology was used? 12 A. That's a nonparametric statistical 13 analysis, I think that's the key point. 14 Q. Can you tell me what that is? I don't 15 know how many times I have seen this -- nonparametric 16 system? 17 A. Nonparametric statistics. 18 Q. What does that mean? 19 A. It basically just means that you do not 20 have to assume a particular mathematical form of the 21 probability distribution. 22 Q. When would you do that? 23 A. Well, some, and I think I could probably 24 say many people in the field, would argue that that 25 is the preferred method of analyzing water quality 58 1 data for a trend certainly, for the reason that it is 2 difficult or impossible to identify a specific 3 mathematical form for a probability distribution of 4 water quality data. They tend to be statistically 5 poorly behaved. 6 Q. Let me back you up on this so I can 7 understand it. 8 When you are looking at water quality, you 9 are looking at its composition, phosphorus content, 10 different contents of the water, and are you telling 11 me that we don't want to develop a -- when we analyze 12 it, we are not looking at a formula or mathematical 13 formula of that water quality, we are just going to 14 ignore it? 15 A. No, what I am saying is this. I will just 16 take as an example some station out in the 17 Loxahatchee, and you give me all the data that's been 18 collected at that station and you say, I want to do 19 some trend analysis at that station. 20 If I do a parametric analysis, the first 21 thing I am going to need to do is to determine what 22 probability distribution fits those data. If I do a 23 nonparametric analysis, I do not require that 24 assumption, I do not require to fit that. 25 I'm saying it is very difficult to develop 59 1 a defensible parametric fit or determine what the 2 mathematical distribution should be, and most people 3 in the field would say that a nonparametric analysis 4 is preferred there because it does not rely on the 5 assumptions that otherwise must be made. 6 Q. Have you ever done a parametric 7 statistical analysis of water quality? 8 A. Oh, I'm sure I have. If you ask me when, 9 I can't recall. 10 Q. Where? 11 A. When or where. 12 Q. No whens or no wheres? 13 A. No whens or wheres, I can't recall. Most 14 of the work I have done has been nonparametric. 15 Q. In the last two years, have you done any 16 parametric statistical analysis of water quality? 17 A. I don't believe so. 18 Q. I think one of the statements you said is 19 it is almost impossible to do, or if you do it, the 20 assurance or confidence level you will have on it 21 will be very low. Do you agree with that statement? 22 A. Not necessarily. 23 Q. Why not? 24 A. We have to deal in specifics here. 25 Q. Okay. 60 1 A. I would not be willing to be held to a 2 statement that a parametric analysis could never be 3 defended. 4 Q. For water quality? 5 A. For water quality, even then I wouldn't 6 want to say that. 7 Q. Okay. 8 A. Most of the time it has problems. 9 Q. Because you can't determine -- 10 A. Let me give you an example of a case where 11 one might be able to defend it, and that would, for 12 instance, be an analysis of the temperature of a 13 water body which tends to be quite constrained, it 14 tends not to have occasional very large or very small 15 values like one might have, for instance, in 16 phosphorus measurements where most of the data might 17 be fairly nicely behaved and then there's an 18 occasional large value which doesn't seem to quite 19 fit. 20 That sort of thing doesn't happen with 21 temperature, so if someone did a parametric analysis 22 of stream temperature, I wouldn't be so concerned as 23 I would be with certain other variables. 24 Q. That would be temperature related to what, 25 though? 61 1 A. Say one wanted to do an analysis at some 2 station to see if there had been a long-term trend in 3 temperature and one did a parametric analysis, I 4 wouldn't be nearly so bothered by that as a 5 parametric analysis of, say, total phosphorus. 6 Q. That's because you would tend to have data 7 that I guess you would consider outlier? 8 A. Yes, but there's a major disagreement as 9 to what constitutes an outlier, and some people would 10 argue that other than mistakes there is no such 11 thing, and I would tend closer to that argument. 12 Q. As a statistician, you would rather use 13 all data rather than excluding any data? 14 A. Yes. The problem with parametric analysis 15 and stream values is that they cause major problems, 16 so then one is tempted to classify some of the very 17 large values as outliers and then not use them, so 18 the methods work better. I have a philosophical 19 objection to that approach. 20 Q. Now, the nonparametric statistical 21 analysis, you don't worry about creating a 22 mathematical formula for relationships between in 23 water quality? 24 A. No, I don't think that's a correct way to 25 characterize it. 62 1 Q. Can you characterize it properly for me? 2 A. All that is not required is to say what 3 the form of the probability distribution is. That 4 particular step is not required. A trend analysis in 5 some sense is saying something about a temporal 6 relationship in the values. 7 For instance, if there is a trend, the 8 values may be becoming greater or less in time. The 9 nonparametric analysis can do that. It just doesn't 10 require you to assume the form of the probability 11 distribution. 12 Q. The reason for determining the probability 13 distribution is to determine whether or not -- the 14 likelihood of the relationship? 15 A. Well, are we talking about trend analysis? 16 Q. Yes, trend analysis. 17 A. The way a trend analysis is done usually 18 is that one forms a null hypothesis and an 19 alternative hypothesis. The null hypothesis is that 20 the probability distribution is not changing with 21 time. The alternative hypothesis is that it is 22 changing in some way, for instance, the values later 23 in time, for instance, are getting to be larger. 24 What was the original question I started 25 on? 63 1 Q. Well, I was trying to figure out why we 2 wouldn't use a nonparametric, because you had defined 3 it as saying that it was a form of determining the 4 probability distribution. 5 A. Well, okay, I think what I was getting at 6 is in a nonparametric analysis, one typically -- one 7 way of doing it, and the way that most nonparametric 8 tests are used, they use the ranks of the data 9 somehow so that they look and see over time if one 10 combines all the data, and look from smallest to 11 largest, is there some tendency for the bigger 12 values, the largest rank values to occur later in the 13 time series as opposed to earlier. 14 That doesn't require that you use the 15 specific observations, it only requires that you use 16 the ranks. 17 Why would somebody use the parametric 18 analysis, I think was your question, now that I 19 recall. 20 The reason is that if one knew the 21 probability distribution, one can develop a test 22 which has -- it more powerful, is the statistical 23 word, for determining trends, i.e., if there's really 24 a trend there, it has a higher probability of 25 detecting that trend. 64 1 That is the reason people -- there 2 essentially is some additional information being used 3 about the form of the probability distribution, 4 hence, the power would be larger. 5 However, in some specific cases, you can 6 show that nonparametric tests, even when the 7 probability distribution is known, do almost as well, 8 and that they do much better if the assumption is 9 violated. That's the rationale why people like to 10 use nonparametric analysis. 11 Q. In using the parametric, you say it had 12 more power, and when you say "power --" 13 A. If the assumptions are met. 14 Q. Is that competent? How do you define 15 "power"? 16 A. No, power means -- you have a null 17 hypothesis and an alternative hypothesis. The null 18 hypothesis either might be true or the alternative 19 hypothesis might be true, we don't know, we only have 20 some data out there. 21 Q. Right. 22 A. So you have two things that could go 23 wrong. One is that you could fail to reject the null 24 hypothesis when in fact the alternative hypothesis 25 was true, that there really was a trend. The power 65 1 is actually just one minus that probability. So it 2 is the probability of detecting a trend when there 3 really is one. 4 Q. This will help me for later. 5 Q. Are there other experts out in the field 6 that would disagree with you as to using the 7 parametric analysis and the nonparametric analysis? 8 A. I'm sure there are. 9 Q. Can you name any? 10 A. No, but I think if you got a group of 11 theoretical statisticians in the room, you would 12 probably be more likely to find people who would 13 argue for the parametric analysis, and I think people 14 who probably have done more work and gotten their 15 hands on the data would tend more to the 16 nonparametric school. 17 Q. People that got their hands on the data 18 meaning? 19 A. Have actually done trend analyses of data 20 and are familiar with the characteristics of water 21 quality data. 22 Q. Would tend toward nonparametric? 23 A. Would tend toward nonparametric. 24 Q. Such as yourself? 25 A. Such as myself, and I think you could go 66 1 look at a textbook by Dennis Helsel and Bob Hirsch at 2 the U.S. Geological Survey that came out recently 3 that would, I'm sure, take that view. 4 Q. Do you recall the trend that you developed 5 in your article, under the Design Considerations for 6 Ambient Stream Quality Monitoring, do you recall it? 7 Was there a trend model done? 8 A. Well, no, that, again -- this is the Water 9 Resources Bulletin paper? 10 Q. Yes, it was. 11 A. That has to do with the work that I did 12 for the state department of ecology and that was in a 13 design context. There are two, somewhat separate 14 concerns here. One is design, which has to do with 15 where are the stations, how often do you collect 16 samples, that sort of thing. That would be used for 17 the later purpose of conducting the trend analysis. 18 That was really the concern of the state's 19 department of ecology, with design, as opposed to 20 going back after the data had been collected some 21 number of years later and analyzing them. 22 While there is some analysis, preliminary 23 analysis that one has to do to determine the 24 parameters that could be used in the design, in other 25 words, you can't do the design in the complete 67 1 absence of data, that work had to do with design and 2 not analysis. 3 Q. Moving ahead, you had marked here 4 Assessment of Environmental Impacts Part 1: 5 Intervention Analysis, and this was in the 6 Environmental Management. 7 A. Yes, that actually is a parametric, 8 describes a parametric analysis. Those papers are 9 methodological. 10 Q. So you didn't do any analysis? 11 A. I'm not sure if there's an example in 12 those papers or not, but if there is, it is of the 13 nature of here is how the stuff would be applied sort 14 of at the end of the paper. The papers are primarily 15 methodological. 16 Q. I guess you did Part 1 and Part 2. There 17 was one titled: Intervention Analysis, and the 18 second part was Data Collection. 19 Again that was methodological? 20 A. It is basically methodological. 21 Interventional analysis is a particular parametric 22 method that at that time was relatively new for time 23 series analysis, and basically those papers described 24 how the method worked. 25 Q. Using the parametric analysis, statistical 68 1 analysis? 2 A. Yes. And there were some -- I believe 3 there were some design applications there as well. 4 Q. When you say "design applications," are 5 you taking a trend modeled of some sort and applying 6 the data to it? 7 A. Well, it has to do basically with the 8 analysis of the power, the probability of detecting 9 trends, and how design factors like the sample size 10 being a key one, what role that plays, and also how 11 the true magnitude -- what magnitude of trend could 12 you hope to detect with a given sample size and given 13 certain characteristics of the time series model like 14 the dependence between samples. 15 Q. You talk about the nonparametric approach 16 within the articles? 17 A. No, that's a parametric approach. 18 Q. Did you put any warnings that maybe the 19 parametric approach should not be used in particular 20 situations? 21 A. I can't recall what the discussions were. 22 In those particular papers, the assumptions were made 23 pretty clear. 24 Q. You are saying if all the assumptions are 25 true or if you know all the assumptions, then you can 69 1 use a parametric approach? 2 A. Well, if you can demonstrate, certainly if 3 you can demonstrate that you know the form of the 4 probability distribution, there's no problem with the 5 parametric approach. 6 Q. Now it is starting to make some more 7 sense. 8 For your next publication, you have Data 9 Requirements for Kriging: Estimation and Network 10 Design, and again this is for Water Resources 11 Research? 12 A. Right. 13 Q. 1981. 14 What were you talking about here? 15 A. That was some early work that I had done 16 on kriging with a student of mine. That was actually 17 a major part of his master's thesis, and at that 18 time, kriging was a relatively new idea in the water 19 resources literature. It came out of the mining 20 geology literature and it started to be used. 21 One of the issues we were interested in 22 was how many samples in space does one need to really 23 be able to make use of this method. The problem 24 there is that in order to make the method work, you 25 have to characterize the form of something that is 70 1 called the variogram from the data, and most of the 2 literature assumed that you knew the variogram and 3 various nice things fall out. 4 We addressed the problem via some computer 5 simulations of how many stations did you really need 6 in order to be able to estimate the variogram to then 7 in turn do the analysis. 8 Q. Is the variogram like the probability of a 9 population? 10 A. No, it is analogous to a correlation 11 function in space which tells you -- instead of 12 dependence in time, it tells you dependence in space, 13 how closely related are observations at two spacial 14 locations, and that gets related to their separation 15 distance and possibly the angle, how they are 16 separated as well as the distance. 17 Q. Do you ignore time? 18 A. In that sort of analysis, that's primarily 19 a spacial analysis, and the answer is yes. It is 20 basically applicable to problems where the 21 observations are not time changing or at least not 22 changing in time very rapidly. 23 For instance, the phosphorus core data is 24 a good application. If you go back a week later, the 25 core is going to be essentially the same. 71 1 Q. You say that you develop characteristics 2 for the variogram? 3 A. No. We were interested in the problem of, 4 if one has to estimate the variogram to then in turn 5 go do the spacial analysis and kriging, how many 6 samples or how many spacial locations did you need 7 data at to be able to do that. 8 Q. You do a computer simulation to determine 9 is that? 10 A. Yes. 11 Q. Do you recall what was determined? 12 A. Well, it depends on a whole lot of things, 13 and it depends on the characteristics of the 14 underlying data themselves, if you can imagine, but 15 as a very rough rule of thumb, sort of 30 to 50 16 seemed to be what was coming out. If you had many 17 less than that, you probably shouldn't be doing it, 18 that sort of analysis. 19 Q. 30 to 50 sample sites? 20 A. Yes. 21 Q. Did it matter how many samples you took at 22 a particular site? 23 A. Well, again, the kriging theory says you 24 have one. 25 Q. One sample per site? 72 1 A. Yes. 2 Q. The use of the kriging theory is so that 3 you can estimate or determine the probability of a 4 site not actually tested? 5 A. Well, it is not a probability so much as 6 it is to estimate a value at sites that are not 7 actually sampled, and that one would like to know 8 something about the probability distribution of the 9 data or the uncertainty of that estimate. 10 Q. That's another calculation? 11 A. Well, that falls out of the kriging. You 12 get two things, you get the value and you get an 13 estimate of the variance which helps you characterize 14 the probability distribution. 15 Q. What would be a variance that would be 16 acceptable? 17 A. There is no such thing. 18 Q. I mean, if you your estimate of variance 19 came out zero, is there a zero value occurring? 20 A. The variance of zero is a deterministic 21 calculation. 22 Q. It is the other way around? 23 A. The other way around. 24 Q. 1 or a hundred? 25 A. No. The variance is not -- it depends on 73 1 the scale of the data. Variances come out in the 2 units of the data squared. Standard deviation is the 3 square root of the variance and has the same unit of 4 the data. 5 What sometimes people do is define the 6 coefficient of variation as being the ratio of the 7 standard deviation to the mean, and one could perhaps 8 get some crude characterization of how variable 9 something is from the coefficient of variation. 10 Q. Can I get you to say that one more time. 11 A. Sure. The definition of the coefficient 12 of variation? 13 Q. Yes. 14 A. The coefficient of variation is the 15 standard deviation divided by the mean. 16 Q. Okay. So you are dealing with percentages 17 then? 18 A. If you multiplied by a hundred, yes. 19 Q. I will leave this part alone for now. 20 Have you developed any computer programs 21 for kriging yourself? 22 A. The co-author on that paper, Jim Hughes, 23 did, yes. 24 Q. Have you developed any subsequently? 25 A. No. 74 1 Q. Is there a standard program used for 2 kriging? 3 A. I'm not sure there's a standard program. 4 We have used some programs that were put together by 5 EPA. 6 Q. The results in kriging, you said, were two 7 things: the value at a particular location? 8 A. Yes. 9 Q. And then the estimate of the variance of 10 that value? 11 A. Right. Essentially what you get out of it 12 -- if we go back to the application in, say, one of 13 the WCAs where it was applied to core data -- and 14 some other people have as well -- if you get a map of 15 the "best estimate," it essentially is the mean, the 16 method's estimate of the mean, the best estimate. 17 And you also get a map of the variance, or 18 if you take the square root, the standard deviation 19 which gives you an idea of how uncertain those 20 estimates are. 21 Q. Do you calculate an average variance after 22 that? 23 A. There's one value that comes out at each 24 point. 25 Q. But then you develop a map with hundreds 75 1 of points maybe -- or do you? 2 A. The map would be of the form of say a 3 contour -- contours. You are characterizing at any 4 point what the value is, and then there's another map 5 which is its variance at any point. 6 Q. You did that for WCA/1? 7 A. Those reports pertained to -- I have to 8 look in there -- WCA/1 has been done, yes, and WCA/2A 9 has been done as well based on data by one of the 10 District subcontractors. 11 Q. WCA/3, has that been done? 12 A. We had attempted to do some analysis in a 13 small part of WCA/3, but there are not sufficient 14 data there to really do a very good job. 15 Q. Have you done it for the Everglades 16 National Park? 17 A. No. 18 Q. Have you done it for the Everglades 19 agricultural area? 20 A. No. 21 Q. We will get to your reports later. 22 The next publication that you had marked 23 off in your resume was titled: Consolidation of a 24 Stream Quality Monitoring Network. 25 You did this with J.R. Stedinger and R. 76 1 Vogel? 2 A. No, I don't think so, unless that's a 3 typo. 4 Q. No? I am actually reading the wrong one. 5 A. It had nothing to do with water quality. 6 Q. D.E. Anderson and R.N. Brenner or 7 "Breen-er"? 8 A. Brenner, yes. 9 Q. What was the nature of that publication? 10 A. That was another review of the trend water 11 quality monitoring network that the municipality of 12 metropolitan Seattle had on some small streams in the 13 Seattle area. 14 Q. What methodology did you use, if any? 15 A. There was a screening method developed to 16 determine the relative importance of different basins 17 they were sampling and how many stations might go on 18 each stream channel within their small basins, and 19 then there was an analysis similar to the one done 20 earlier for the state for the time frequency 21 sampling. 22 Q. Had they put up any stations or basins? 23 A. Yes, they had an existing network, and 24 they were concerned with what the rationale was for 25 why various stations they were operating were in fact 77 1 being operated, and for whether the allocation of 2 stations made sense. 3 Q. And what did you conclude? 4 A. Well, we helped them reallocate some of 5 the stations. We felt that they were oversampling 6 some stream basins, undersampling some others. 7 Q. Did you add any any new stations? 8 A. As I recall, we did recommend that some of 9 the basins have new stations added and some of the 10 other basins have stations dropped. 11 Q. Were any models developed for this, 12 statistical models? 13 A. Not specifically for that work. There 14 were -- I mean, there were computer analyses 15 conducted, so there were programs used and possibly 16 written for some specific purposes. 17 Q. What type of computer analyses? 18 A. My recollection -- I would have to go back 19 and check the particular report that was used and the 20 paper there -- was that there was an algorithm to 21 help select and prioritize stations based on some of 22 the attributes of the particular candidate stations. 23 Q. You said there was an algorithm developed? 24 A. Yes. 25 Q. Can you tell me what that is? 78 1 A. It is not a statistical algorithm, it 2 basically was just a scheme for combining some 3 priorities in terms of different attributes that had 4 been developed by metro, by the client, in terms of 5 why they might want to sample at different locations. 6 Q. What were they sampling for? 7 A. There were a number of constituents. 8 Phosphorus was certainly one of them, nitrogen was 9 another, I believe dissolved oxygen, temperature, and 10 probably a few others. 11 Q. Were there any relationships developed or 12 stated between the constituents? 13 A. I don't believe so directly. 14 Q. Indirectly? 15 A. Well, indirectly possibly in the sense 16 that we went through the list of constituents and 17 tried to identify some of those that might have been 18 redundant, because there was a cost issue involved. 19 I think we looked at the issue of did they really 20 need to be collecting all of the constituents they 21 were collecting. 22 Q. This cost issue always plays a part in all 23 your analyses? 24 A. Well, it is certainly an issue in design. 25 Q. Did stage ever come into play in your 79 1 analysis of the stream water quality? 2 A. I believe there was some discussion of it, 3 and there was a concern that at many of their 4 stations, they were in fact not monitoring, were not 5 monitoring discharge, I believe there were 6 recommendations to that effect. 7 Q. Did you make any recommendations? 8 A. Oh, there were a whole series of 9 recommendations. That was the point, that's what we 10 had been asked to do, conduct a review of their 11 network. 12 Q. Do you know if they actually took your 13 recommendations and actually did certain things? 14 A. Implemented? Yes, my understanding is 15 most of them were implemented. 16 Q. Have you ever gone back to see what the 17 effect of implementing your recommendations were? 18 A. No, not in that particular case. 19 Q. In any particular case, have you gone 20 back? 21 A. Well, gone back in what sense? Aware of 22 what's been done and so on? 23 Q. I guess what I am really more getting at 24 is whether you compared what you thought would happen 25 to what actually happened later? 80 1 A. That's not a proper characterization of, 2 say, a design. There's no characterization of what 3 should happen, this is a question of how many 4 stations are you going to sample, where are they 5 going to be, how frequently are you going to collect 6 data. 7 Q. I understand what you are saying, and I 8 can see where my question might not be proper, but 9 would you go back and say maybe you need more 10 stations or less stations or stations somewhere else? 11 A. It would be an interesting retrospective 12 analysis to go back and see how well the 13 recommendations might have worked out. In that case, 14 no, I haven't done it. We have never been hired to 15 go back. Futhermore, that particular agency has been 16 subsumed by another agency, and there's some problems 17 of institutional memory. 18 Q. Moving on to your publications. 19 You have marked Multivariate Nonparametric 20 Tests for Trend in Water Quality. This was in Water 21 Resources Bulletin in 1988. 22 A. That's a methodological paper. It has to 23 do with how one does a nonparametric analysis for 24 more than one water quality constituent currently, 25 instead of analyzing all the constituents 81 1 individually, if one treats them at the same time and 2 does the same analysis. 3 Q. Doesn't that skew your result more or 4 aren't you more susceptible to errors when you 5 combine everything? 6 A. In what respect? 7 Q. With my little knowledge of statistics, I 8 always thought when you start combining data and you 9 did that, that you are more prone to have errors in 10 your design or your modeling. 11 A. You are not combining data in this 12 analysis. What you are doing is, for instance, to 13 take an example, where you have two constituents -- 14 Q. Phosphorus and nitrogen or something? 15 A. Phosphorus or nitrogen, it might not be a 16 particularly good example, but phosphorus and, say, 17 total dissolved solids. One way of doing a trend 18 analysis is simply to say, let's go do an analysis of 19 total phosphorus and we get some result, and let's go 20 do an analysis of total dissolved solids and we get 21 some result. 22 What this method says is you do a 23 concurrent analysis to see whether taken together, 24 there's some picture of change, instead of taking 25 them one at a time. 82 1 Sometimes, the reason for doing this is 2 that sometimes while there may not be a statistically 3 significant trend in any one particular constituent, 4 when one analyzes them as a group, there is in fact 5 something going on. 6 Q. Again, this was a paper dealing with the 7 method of it? 8 A. It is methodological. 9 Q. So there are no applications of it? 10 A. The application of that paper is in the 11 '91 national trend analysis, that we used the method 12 there. 13 Q. Do you know if you marked that? 14 A. Yes, I think so. 15 Q. The Trend in Stream -- 16 A. -- Quality in the Continental U.S., or 17 something like that. 18 Q. Talking about the concurrent constituents, 19 how do you negate the possibility of the 20 interdependence of the components being treated? 21 A. Here it is treated. That's the whole idea 22 of the analysis. 23 Q. In what way is it treated? 24 A. It is a nonparametric analysis, so it 25 knows about what it called the rank correlation 83 1 between the variables. 2 Q. You lost me. 3 A. Okay. The interdependence between the 4 variables is in a sense a correlation between them, 5 that if one value tends to be high, the other value 6 tend to be high. 7 An extreme case, if you had total 8 phosphorus as one constituent and orthophosphorous as 9 the other, perhaps they tend to vary similarly and 10 there's a correlation between them. It estimates 11 that correlation, but it does it in a rank sense, in 12 a nonparametric sense. 13 Q. Rank from 1 to 100? 14 A. No, in terms of everything is done in 15 terms of the ranks of the data as opposed to their 16 numerical value. 17 Q. Who ranks the data? 18 A. If you have a sequence of observations, 19 let's say, 4 times, 1.3, 2.1, 1.4 and so on, one 20 simply goes through and ranks those from smallest to 21 largest, and then goes and takes at the first time 22 what is the rank of the observation that was 23 collected then, not its numerical value, but its 24 rank. And then you calculate a correlation between 25 the ranks as opposed to between the raw values. 84 1 Q. So you are ranking all the phosphorus 2 constituents and maybe the dissolved solids? 3 A. Yes. 4 Q. And then you combine them? 5 A. Not combining them. 6 Q. You are doing a trend on the concurrent? 7 A. Yes, it is a multivariate analysis, but 8 instead of univariate analysis. So one has 9 essentially a matrix instead of a vector, that's all 10 that's happened. 11 Q. I think this was touched in Dr. Millard's 12 deposition -- maybe it wasn't. 13 One of the next publications you marked 14 here is Design of Soil Sampling Study to Determine 15 the Habitability of the Emergency Declaration area, 16 Love Canal, New York. 17 A. Right, he was the co-author on that. In 18 fact, he was the more senior author than I was, so 19 I'm sure anything he has told you would be -- 20 Q. Actually, it was skipped over since I 21 think it was assumed that counsel knew about it. I 22 didn't know much about it. Can you just tell me a 23 little bit about it? 24 A. About the paper? 25 Q. Well, about what was going on there, what 85 1 you did and what was involved. 2 A. Well, I was involved in the soil sampling 3 design in Love Canal over a period of about three or 4 four years. The history of that project, very 5 briefly, was that there had been a report before my 6 time done on it issued from the EPA which essentially 7 failed to review. 8 And the project was started essentially 9 all over again to attempt to characterize the area, 10 what was known as the emergency declaration area in 11 which about 800 houses had been evacuated in a ring 12 around the immediate Love Canal dump site. 13 And the question was whether those houses 14 could be reinhabited and what would be done. So 15 there was a rather extensive sampling program 16 initiated to characterize the existence of certain 17 trace chemicals in the soil, and there were a lot of 18 chemicals identified called Love Canal indicator 19 chemicals. 20 Those particular chemicals were identified 21 by another group that was outside the purview of our 22 study, but they were sampled at a number of 23 locations, and the question was in order to determine 24 whether there was a difference between certain 25 neighborhoods within this emergency declaration area 86 1 and some identified control areas, how many samples 2 would need to be collected. 3 And that was the aspect of what I worked 4 on, both an analysis of a pilot study and then design 5 of the statistical methods, and ultimately conducting 6 the analysis. I was involved with basically the 7 other people who were co-authors on that paper in 8 that study. 9 Q. You refer to Love Canal. Was there a 10 canal around there? 11 A. Well, Love Canal was in the news around 12 1980. It was the first major hazardous waste site in 13 the U.S. that got widespread public attention. There 14 was a Presidential order which evacuated the area in 15 about 1979 or 1980. 16 There was an area that was evacuated 17 following some health studies which had shown that 18 there was possible health effects to children that 19 had attended a school which would have been 20 constructed immediately on top of the waste site, and 21 people living in the immediate area. 22 Consequently, President Carter issued an 23 order which evacuated the area and directed the 24 Environmental Protection Agency to conduct a study to 25 determine what should be done. 87 1 That first study, which was to have been 2 conducted over a period of six months or so in about 3 1980 or '81, was the one that failed review by the 4 Center for Disease Control. The National Bureau of 5 Standards, the Office of Technology Assessment, was 6 involved in that. 7 At any rate, that study was set aside, and 8 EPA was directed to go back and do a more thorough 9 study to determine what should be done, and that was 10 the point at which I became involved in the early 11 mid-1980s. 12 Q. You ended up designing? 13 A. I worked on the design with the others 14 there of the soil chemistry data collection, and 15 other decisions such as how many laboratories would 16 be used, which laboratories would analyze certain 17 samples, because it turned out that there were so 18 many samples being collected that no one laboratory 19 could conduct all the analyses within the time period 20 that the quality control people indicated they would 21 have to be analyzed. 22 There were a number of those issues of 23 allocation, number of samples, how the statistical 24 analysis would be conducted that I was involved in. 25 Q. Did you design where the samples should be 88 1 taken, when? 2 A. I was not personally involved. That study 3 does describe, though, some issues. Some of the 4 other co-authors in that paper were involved in 5 questions of what order the samples would be 6 collected at, how things would be randomized to 7 assure that there was no essentially contamination of 8 the statistical results by the order in which the 9 samples were collected. 10 Q. I think you referred to this before as the 11 application of the methodology of a prior article 12 which this one is now titled: Trends in Stream 13 Quality in the Continental U.S.? 14 A. Right. 15 Q. And that was the application -- 16 A. The multivariate analysis was one aspect 17 of that. 18 Q. Thank you. 19 What did the multivariate analysis show? 20 A. Well, to describe the paper very briefly, 21 that was one aspect of it. We applied the seasonal 22 Kendall's test -- Kendall is a name -- the method 23 which has been developed by some people at the U.S. 24 Geological Survey for the National Stream Quality 25 Accounting Network Stations throughout the U.S., of 89 1 which I believe about 400 of their 500 or so stations 2 had sufficient length of record, and passed some 3 other screening criteria. 4 There was a large number of constituents 5 that were grouped. I think there were as many as 30 6 or 40 constituents that they routinely analyzed. 7 Those were grouped into certain areas, and I can't 8 remember things like trace metals -- I can't remember 9 exactly what we call them. Again I will get you a 10 copy of the paper. 11 We conducted the seasonal analysis 12 individually, and then we did a generalization of the 13 multivariate method to include seasons as well, and 14 applied it to each of the groups. 15 So that in addition to the analysis for 16 each of the individual constituents in each group, 17 for instance, common ions -- they had a list, and 18 sodium would certainly be one, and I think potassium 19 was in there and some others, those were analyzed 20 individually, and then we also analyzed the group 21 using a multivariate test. 22 This was done for all the stations in the 23 country, and then there were maps developed to show 24 where trends in these things were increasing and 25 where they were decreasing and where there were 90 1 significant changes. 2 Q. This was like a concurrent -- you grouped 3 constituents together. Did you group them all 4 together or did you decide which constituents to 5 group together? 6 A. Again, we had 30 to 40 constituents. We 7 determined six or seven groups that logically fit 8 together; common ions I remember was one, trace 9 metals, I remember the other -- I can't remember the 10 others. 11 Q. Was the data ranked? 12 A. No, we simply said these are typically 13 similar, that it makes sense to analyze common ions 14 together, but not to lump in common ions with, say, 15 nutrients which was another group. I think nutrients 16 just had total phosphorus and I think total nitrogen. 17 So we would go within each group and do 18 the analysis individually, which of course then 19 doesn't matter whether it is in a group or not, but 20 then we would analyze the group using the 21 multivariate at the time. 22 Q. In a nonparametric approach? 23 A. It was all nonparametric approach, the 24 seasonal Kendall test is a nonparametric method. 25 Q. Then you developed trend models from them? 91 1 A. Those are statistical trend models. 2 Q. Were there maps? 3 A. They are maps that show simply whether 4 over that period of record there was statistically an 5 increase, uptrend, downtrend or no significant trend. 6 Q. Did you find any trend? 7 A. Well, there's a whole paper full of 8 papers. It is a little different to characterize, 9 and I can't remember the details well enough. I 10 think I indicated earlier that with either groups or 11 individual constituents with the most trends, one 12 would have maybe a hundred or so out of the 400 or so 13 stations. 14 So I think around 25 percent was about the 15 most that we would find trends in. It is a 16 relatively short record, about 10 years -- a little 17 longer. 18 Q. When you do the nonparametric approach, do 19 you get a variant or indicator of how much confidence 20 you can give that particular trend model that you 21 have developed? 22 A. I don't know quite what you mean. 23 Q. What I am trying to get at is when you do 24 a parametric approach, you get a probability of the 25 population? 92 1 A. You get a probability of distribution, but 2 in the end, you get -- the analysis is the same 3 thing, the result of the analysis is the same thing. 4 You could have a statistically significant trend or 5 you don't, at some significance level. 6 Q. Do you quantify the statistical result 7 that you have reached in terms of confidence -- I 8 guess I am getting back to alpha and beta? 9 A. Well, sure. The analysis is all conducted 10 with some P values and significance level. So if if 11 the significance level is .05, which is one that's 12 often used, that is just the type 1 error 13 probability, which is the probability of rejecting 14 the null hypothesis when in fact it is really true, 15 which is a finding a trend when there really isn't 16 one. That's the thing you fix. 17 Then with that fixed, you go and see 18 whether or not you have a significant trend. 19 Q. You do that at the end? 20 A. That happens in any statistical analysis. 21 The difference, one of the things that often falls 22 out of a parametric analysis which does not in a 23 nonparametric analysis is the magnitude of the trend. 24 The nonparametric analysis says there's 25 one there, and there either is or is not one there. 93 1 The parametric analysis may give you an estimate of 2 how big. 3 Q. Moving on to your technical reports, I 4 point you to Design of Monitoring Systems for 5 Detection of Trends in Stream Quality. 6 A. That's 39? 7 Q. Yes, No. 39. 8 A. That's my dissertation. 9 Q. We already talked about that. 10 The next one, is this also part of your 11 thesis: Detection of Trends in Stream Quality: 12 Monitoring Network Design and Data Analysis? 13 A. That was the work done for the state 14 department of ecology which we also discussed. 15 Q. Moving right along. 16 Aquatic Monitoring: Data analysis and 17 network Design Using Regionalized Variable Theory. 18 It looks like the same student? 19 A. That's the same -- the Water Research 20 paper in '81 is basically that, the distillation of 21 that. That in fact is his master's thesis. 22 Q. Have you continued to work with Mr. 23 Hughes, H U G H E S? 24 A. Hughes is now on the faculty in the 25 biostatistics department of the university. 94 1 Q. Has he helped in any of the Everglades 2 work? 3 A. No. 4 Q. Have you talked to him about it? 5 A. No -- well, I had a discussion at one 6 point before Steve Millard was brought in to ask 7 whether he would be interested when we needed some 8 statistical help, and his answer was no, that he was 9 too busy and he suggested that Steve Millard, whom I 10 also knew, might be able to do it. So the only 11 discussion that ever was had with him had to do with 12 his availability. 13 Q. The technical report titled: Routine 14 Streams and River Water Quality Trend Monitoring 15 Review? 16 A. About '82? 17 Q. March of '82. 18 A. That's the work for metro which was in 19 that Water Research Bulletin paper which we 20 discussed. 21 Q. Can you just -- 22 A. There's a paper there with Anderson and 23 Brenner and we discussed prioritizing the different 24 attributes and so on. 25 Q. And this is the same? 95 1 A. This is the same. The report has all the 2 details and the paper has more of a summary. 3 Q. Maybe this next one we talked about also: 4 Operation of Detention Facilities for Urban Stream 5 Quality Enhancement? 6 A. We talked about that because that's Lisa 7 Dally's thesis. 8 Q. That's what I thought. 9 Other Technical Reports: Design of 10 Nonradiological Aquatic Sampling Programs for 11 Newclear Powerplant Impact Assessment Using 12 Intervention Analysis? 13 A. The same stuff as is in that Part 1 and 14 Part 2 we discussed -- Environmental Management, or 15 whatever the journal was -- about 1980 or so with 16 Hipel and McLeod. I'm not sure I have marked it on 17 the refereed publications. 18 Q. I think we talked about Part 1 and Part 2. 19 Can you tell me a little bit more to jog 20 my memory? 21 A. This has to do with the intervention 22 analysis, and we discussed that it is a parametric 23 method. 24 Q. Okay. We got into talking about 25 parametric versus nonparametric? 96 1 A. Right. 2 Q. And I might have taken you off that. 3 A. I mentioned that the intervention analysis 4 was a method that had been suggested in the 5 statistical literature, and we explored it for use in 6 environmental trend analysis basically. 7 To be perfectly honest, I probably would 8 not use that method in most problems I have come 9 across because it makes some assumptions that I 10 consider to be too restrictive, but it was something 11 that looked interesting, and we did some 12 investigation into it at that point. 13 Q. When you are saying assumptions that are 14 too restrictive, assumptions imposed upon you or 15 assumptions that you had to make? 16 A. Assumptions of the model. 17 Q. That you had to make? 18 A. That you have to make. You have to make 19 the parametric assumptions to start with. You need 20 to know -- the data needs to be normally distributed. 21 The other problem with that method is that it doesn't 22 like missing data very well. 23 So it likes a nice long time series where 24 you have observations every month, it doesn't like it 25 if those observations are missed. I think we had 97 1 done some analysis in those papers to look at how one 2 might deal with problems -- I take that back. 3 There's another paper having to do with 4 interventional analysis data, and that paper assumes 5 all the data are there, and that's a real problem. 6 Q. Stormwater Quality Evaluation for 7 Metro-Transit Operating Base. 8 This is a Lisa Dally? 9 A. Yes. I can't remember exactly why that 10 was a separate report. I think what happened was we 11 got an additional small contract to do some more work 12 for metro, and that would have been the report that 13 had to do with the follow-on work. This was all 14 basically performance of oil-water separators. 15 Q. This is where you tested between the dry 16 ponds and wet ponds? 17 A. I think the dry versus wet stuff was in 18 the earlier report that you mentioned, and Metro East 19 operating base, that is the transit facility, that's 20 a wet facility. 21 Q. Did you test for anything other than oil 22 and grease? 23 A. That was the main thing they were 24 interested in. 25 Q. Nothing else, no other constituents? 98 1 A. I don't remember if there were other 2 constituents tested. In the follow-up, the second 3 report you have mentioned, I think not. I think it 4 was entirely oil and grease. 5 Q. There's nothing to discuss on Other 6 Publications. You haven't marked anything under 7 Other Publications in your resume? 8 A. I don't think I saw anything on there that 9 really was water quality related. 10 Q. You did look at that page then? 11 A. Yes, I did look at t