1
1 DIVISION OF ADMINISTRATIVE HEARINGS
DEPARTMENT OF ADMINISTRATION, STATE OF FLORIDA
2
3 SUGAR CANE GROWERS COOPERATIVE )
OF FLORIDA; ROTH FARMS, INC., and )
4 WEDGWORTH FARMS, INC., )
Petitioners, ) DOAH Case No. 92-3038
5 v. )
SOUTH FLORIDA WATER MANAGEMENT )
6 DISTRICT, an agency of the State )
of Florida; et al., )
7 Respondents. )
- - - - - - - - - - - - - - - - - - x
8 FLORIDA SUGAR CANE LEAGUE, INC.; )
UNITED STATES SUGAR CORPORATION; )
9 and NEW HOPE SOUTH, INC., )
Petitioners, )
10 v. ) DOAH Case No. 92-3039
SOUTH FLORIDA WATER MANAGEMENT )
11 DISTRICT, an agency of the State )
of Florida; et al., )
12 Respondents. )
- - - - - - - - - - - - - - - - - - x
13 FLORIDA FRUIT AND VEGETABLE )
ASSOCIATION; LEWIS POPE FARMS; )
14 W.E. SCHLECHTER & SONS, INC., )
and HUNDLEY FARMS, INC., )
15 Petitioners, )
v. ) DOAH Case No. 92-3040
16 SOUTH FLORIDA WATER MANAGEMENT )
DISTRICT, an agency of the State )
17 of Florida; et al., )
Respondents. )
18 - - - - - - - - - - - - - - - - - - x
100 S.E. 2nd Street
19 Miami, Florida
April 8, 1994
20 8:40 a.m. - 4:10 p.m.
21 DEPOSITION OF ED DOWNING
22 Taken before THOMAS R. NEUMANN, Registered
Professional Reporter and Notary Public in and for
23 the State of Florida at Large, pursuant to Notice of
Taking Deposition filed in the above cause.
24 - - - - - - -
2
1 APPEARANCES
2
ON BEHALF OF THE PETITIONERS SUGAR CANE GROWERS
3 COOPERATIVE OF FLORIDA, ROTH FARMS, INC., AND
WEDGWORTH FARMS, INC.
4
HOPPING, BOYD, GREEN & SAMS
5 123 South Calhoun Street
P.O. Box 6526,
6 Tallahassee, Florida 32314
BY: CAROLYN S. RAEPPLE, ESQ.
7
SOUTH FLORIDA WATER MANAGEMENT DISTRICT
8
POPHAM, HAIK, SCHNOBRICH & KAUFMAN, LTD.
9 4000 International Place
100 S.E. 2nd Street
10 Miami, Florida
BY: GREGORY CESARANO, ESQ.
11
ON BEHALF OF THE RESPONDENT-INTERVENOR
12 UNITED STATES OF AMERICA
13 TOM WATTS-FITZGERALD, ESQ.
ASSISTANT U.S. ATTORNEY
14 99 N.E. 4th Street
Miami, Florida 33132
15
16
17
INDEX
18 Witness Direct Cross Redirect Recross
ED DOWNING
19 By Mr. Cesarano: 4
By Mr. Watts-Fitzgerald: 33
20
3
1 EXHIBITS
2
3 NUMBER PAGE
4
5 Exhibit 10 through 12..............................5
6 Exhibit 13........................................20
7 Exhibit 14........................................20
8 Exhibits 15, 16..................................121
9 Exhibit 17.......................................142
10 Exhibit 18.......................................143
11 Exhibit 19.......................................143
12 Exhibit 20.......................................143
13 Exhibit 21.......................................144
14 Exhibits 22 and 23...............................144
15 Exhibit 24.......................................145
16 Exhibit 25.......................................145
17 Exhibit 26.......................................145
18 Exhibit 27.......................................145
19 Exhibit 28.......................................146
20 Exhibit 29.......................................146
21 Exhibit 30.......................................146
22
23
24
25
4
1 Thereupon --
2 ED DOWNING
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. CESARANO:
7 Q. Good morning, Mr. Downing.
8 A. Good morning.
9 Q. We are continuing the deposition that we
10 left off last week. The rules are the same. You
11 tell me if you don't understand my questions, all
12 right?
13 A. Yes, sir.
14 Q. Since we last met, what have you been
15 doing?
16 A. Working on the classification for the 1985
17 image.
18 Q. What have you been doing? Tell me
19 specifically how you have been doing on the
20 classification.
21 A. By using the techniques as described in the
22 previous deposition, 1985 and '84 that was used as
23 the source for training sample selection in the 2A
24 study area and identified those sites by the training
25 samples set up, the classification schemes produced,
5
1 the classifications and continued, of course, the
2 analysis that I talked about also.
3 Q. Did you do anything on the other images,
4 the '87 or '89 or '91 images?
5 A. Unsupervised classification into 36 and 99
6 categories, no. We did nothing additional with
7 those.
8 MS. RAEPPLE: Off the record.
9 (Discussion off the record.)
10 MR. CESARANO: Mark these Exhibit 10.
11 (The documents referred to were thereupon
12 marked Exhibit 10 through 12 for
13 Identification.)
14 BY MR. CESARANO:
15 Q. We have marked these three documents
16 Exhibits 10, 11, 12 to your deposition. Counsel
17 produced these to me yesterday.
18 Could you explain what those are, please?
19 A. Give me one second. Document No. 12 is the
20 table indicating the area per class in the
21 interpreted TM image for 1993 document.
22 11 is a similar category map for the --
23 this is the entire Everglades study area. This is an
24 area that includes 2A and an area from Loxahatchee
25 down to the small portions of the national Everglades
6
1 park, the hectares -- class categories in hectares.
2 And No. 10 is the similar area calculation
3 table for the 1985 Landsat thematic paper image.
4 Q. When we were last taking your deposition we
5 discussed the differences in part between the
6 vegetative classes you were using on your map and the
7 classes that Mr. Rutchey used on his map. Have you
8 done any further work or analysis or attempted to
9 make those classes comparable to each other?
10 A. No. We weren't able to. We just decided
11 to use our classes with the definition that we used
12 in terms of percent cover and had no comparison. No
13 maps comparing Rutchey's classes to our classes.
14 Q. I'm going to try to make at least an
15 initial attempt to see if there is any way we might
16 be able to determine which classes of Mr. Rutchey's
17 would fall into at least your cattail classes. Did
18 you attempt to do anything like that?
19 A. Preliminarily we tried to compare roughly
20 our classes to his in some of the tables we produced
21 early on in the study.
22 Q. In looking at comparing Exhibits 10 and 12
23 which you have just identified, I note that on both
24 of those exhibits you have a class predominant
25 cattail, and we have already discussed the definition
7
1 of that, I believe 95% or greater cattail in an area,
2 correct?
3 A. We estimated roughly in the area of 95% or
4 greater to be a cattail category.
5 Q. I see in both exhibits in Exhibit 10 class
6 No. 2, predominant mix cattail/other. According to
7 class No. 3 in Exhibit 12, predominant mix
8 cattail/other?
9 A. No. The classes in the 1985 image, several
10 of the classes in this category in 1993 are combined
11 to form these classes in the '85 image. It's not a
12 one-to-one comparison by class name. If you notice,
13 there are fewer classes in '85 than there are in '93.
14 The total acreage would be the same, but the classes
15 themselves are inclusive of other classes.
16 Q. Could you explain to me which ones have
17 been combined, please?
18 A. The predominant cattail in the 1985 image
19 is predominant cattail, same definition for
20 predominant cattail in the '93 image. You go down to
21 the predominant mix cattail/other, and that includes
22 all the mixes available in the '93 category,
23 predominant mix cattail open water. All the mix
24 categories there of cattail and other would include
25 other in the predominant mix.
8
1 Q. Let me understand that. Category No. 2 in
2 the 1985 table is equivalent to categories No. 2, 3
3 and 4 in the Exhibit 12 table?
4 A. You are talking about between 10 and 11 or
5 10 and 12?
6 Q. Those are both 2A. Why don't you just tell
7 me which ones are included in category No. 2 in
8 Exhibit 12.
9 Q. Category 1 predominant cattail equals -- in
10 the -- let's start with the -- in the reference in
11 this case would be the '85 image?
12 A. In the '85 image predominant cattail and
13 predominant cattail class 1 are the same definition
14 of class. All the mix categories 2, 3 and 4 are the
15 classed or combined into the predominant mix cattail
16 other category No. 2 in the '85 image.
17 There is no equal mix category in the '85
18 image.
19 Q. Why is that?
20 A. To determine an equal mix from the air
21 photos we used we thought was not practical, so we
22 did away with that category and decided either it was
23 a mix with predominance toward either category. We
24 found no equal mix, equivalent equal mix category in
25 the '85 image. So equal mix 5, there is no
9
1 equivalent from '93 to '85.
2 Q. Category 3 in the '85 image predominant mix
3 sawgrass/other would include class 6 in the '93 image
4 which is predominant mix sawgrass/cattail, class 7
5 predominant mix sawgrass/other and 9?
6 A. Yes, and 9. Predominant sawgrass/open
7 water -- no. No. Not 9. It wouldn't be 9.
8 Predominant mix sawgrass cattail class 6 in the '93
9 image matches the predominant mix sawgrass class 3 in
10 '85.
11 Class 7 predominant mix sawgrass/other is
12 classed into class 3 in the '85 image predominant mix
13 sawgrass/other. Sparse sawgrass is equal to sparse
14 sawgrass. In other words, class 8 in the '93 image,
15 sparse sawgrass may contain periphyton. And class 5
16 in the '85 spars sawgrass may contain periphyton or
17 comparable.
18 The predominant sawgrass categories,
19 predominant sawgrass class 9, '93 image, predominant
20 sawgrass open water would fall into the predominant
21 sawgrass in the '85 image, which is class 4.
22 Predominant sawgrass in the '93 image falls into
23 predominant sawgrass in the '85 image. In other
24 words, that class 10 in the '93 image is included in
25 class 4 in the '85 image.
10
1 The slough is equivalent to slough, class
2 11 is equivalent to class 6. Class 12 in the '93
3 image is equivalent to class 7, which is broad leaf
4 vegetation/tree island. Class 13 in the '93 image
5 open water is equivalent to class 8 open water in the
6 '85 image.
7 And clouds and shadows in both the class 14
8 and clouds and shadows in the '93 image is equivalent
9 to class 9 clouds and shadows in the '85 image.
10 Q. Now, I want to ask about why you were
11 unable to determine whether or not there was an equal
12 mix of sawgrass and cattail -- or put differently,
13 why you were unable to identify that class from
14 aerial photography that you were using.
15 Why was that?
16 A. That class requires a very detailed view of
17 the site that we were unable to obtain in the 1985
18 photography. It was too close to call.
19 Q. Why are the total hectares approximately
20 100 different between the two images?
21 A. Apparently in digitizing the 2A study area
22 we were slightly off in delineation of the study area
23 in the '85 image. It was about 100 hectares
24 different.
25 Q. You told me about the accuracy evaluation
11
1 that you are presently engaged in at the time that I
2 last took your deposition. My notes reflect that at
3 that point you had gone through some 26 points. How
4 many total points were evaluated for the accuracy
5 assessment for the 1993 image?
6 A. I believe that number was 84.
7 Q. When was that accuracy assessment
8 completed?
9 A. It was completed during the deposition the
10 last time.
11 Q. What was the percentage of accuracy?
12 A. I believe that was 70% or 72%.
13 Q. Do you have the error matrix on that?
14 A. Yes.
15 Q. I would like to see it, please.
16 A. I'm sorry. I don't have that error matrix
17 with me today. It's being faxed this morning.
18 Yesterday when we finished this project -- the day
19 before when we finished this project and compiled all
20 of that NASA through -- we are members of the VIP
21 program. We were involved in the program using some
22 of the scanners that NASA has available.
23 NASA sent a film crew to our office
24 Thursday morning at 8:00. They came in and cleaned
25 out our work area, set up cameras and did interviews
12
1 for two and a half hours. And in the process, they
2 scattered stuff -- at least the records we had neatly
3 piled were moved to a different area by the film
4 crew.
5 And Pam Green is reassembling that and
6 sending me the current error matrixes. I apologize
7 for that. It was unexpected. They were supposed to
8 have been out in the field filming field work and
9 then would film in my work area Friday. But it
10 rained, so they came into my work area. But my staff
11 right now is finding that and they are faxing that to
12 us. We will have it in a few minutes.
13 MR. WATTS-FITZGERALD: Before the
14 accusations begin to fly, or I read it in the
15 paper, the Department of Justice had nothing to
16 do with the timing of this. It was actually
17 supposed to have occurred much sooner. Please
18 convey that.
19 THE WITNESS: If you know the documentary
20 thing they are doing on CNN, just remembered
21 they disrupted my work schedule by doing that.
22 BY MR. CESARANO:
23 Q. Where is it being faxed to?
24 A. Here in this office.
25 Q. To my office?
13
1 A. Yes.
2 MS. RAEPPLE: To whose attention?
3 THE WITNESS: Ed Dowling's office.
4 BY MR. CESARANO:
5 Q. Were you expecting that Pamela Green would
6 fax that first thing this morning?
7 A. Yes.
8 Q. Well, we will move off that point until it
9 gets here.
10 We were discussing with Dr. Lodge yesterday
11 some of the aerial photography, and I believe the
12 grid indicated that the scale on the national high
13 altitude program aerial photography is close to 1 to
14 65,000?
15 A. 1 to -- the scale in the photography wound
16 up being a surprise. That photography ordering to
17 NASA specifications when they sent it to us was
18 supposed to have been 1 to 58,200. But when we photo
19 reduced the quad sheets to match the photo, it didn't
20 come out to that ratio.
21 So in conversations with NASA about that
22 photography, well, obviously it was not at that
23 scale. So we matched the reduced quad sheets and
24 matched the photography and it turned out to be 1 to
25 64,900.
14
1 Q. Tell me how you went about reducing the
2 quad sheets and turning them into transparencies.
3 A. We took the in-house photography and used
4 that as a control for the quad sheet. So we matched
5 the quad sheet reduction to the NHAP, photography.
6 Q. And to your knowledge was it an exact
7 match?
8 A. It was a good map. I've never seen an
9 exact map in photo reduction.
10 Q. When you say a good map, can you quantify
11 how close?
12 A. The features were within acceptable
13 distances that we decided were necessary to do this
14 photo interpretation.
15 Q. And what were those tolerances that you
16 decided?
17 A. We attempted to match reasonably the scale
18 of the photography so when we over laid a 500 meter
19 grid we could identify the patterns within that 500
20 meter grid.
21 Q. Based solely on the reduction of the quad
22 sheets, can you tell me -- can you quantify the
23 locational error in the map?
24 A. Can I quantify the locational area in the
25 reduced quad sheets?
15
1 Q. Exactly.
2 A. I did not quantify the area in the reduced
3 quad sheets.
4 Q. But by reducing it like that, it does
5 introduce the possibility of error, does it not?
6 A. Yes.
7 Q. And the color infrared photography itself
8 has a margin of error, does it not?
9 A. Yes.
10 Q. Can you quantify that for me?
11 A. No.
12 Q. The process that Dr. Lodge used in manually
13 lining up the photography with the quad sheet also
14 has the potential to introduce error, does it not?
15 A. Yes, it does.
16 Q. Are you able to quantify that amount of
17 error?
18 A. We have not yet quantified that amount of
19 error.
20 Q. Do you believe that you will?
21 A. I don't believe it will be necessary.
22 Q. Why not?
23 A. Because we depended on photo recognition
24 within a 500 by 500 meter quad. We did not -- we
25 used a photo recognition pattern recognition within
16
1 the map and the quad sheet.
2 Q. Am I correct in stating that the total
3 locational error is a function of the margin of error
4 in the photo added to the margin of error in the grid
5 added to the margin of error introduced in the
6 process?
7 A. That would be correct if you assume you
8 were dealing in total positional error. We were
9 dealing in relational error. Those are different
10 types of errors.
11 Q. Would you explain that, please?
12 A. Positional is accuracy of each map in
13 relation to the point on the surface of the map and
14 coordinated space.
15 We were dealing more with relational
16 accuracy, which is once we identified an identifiable
17 pattern in the photo, we align those various overlays
18 and transparencies to that point. So relationally
19 within the 500 meter grid we identified areas and
20 patterns.
21 Q. In your opinion how does the total error as
22 we have just discussed relate to or compare to the
23 pixel size in the thematic mapper image of 30 meters?
24 A. As I stated already, we did not use
25 positional accuracy as the control in this photo. We
17
1 used relative accuracy. Relative positional
2 accuracy.
3 Q. Isn't it important to consider positional
4 accuracy when you are talking about extent of
5 vegetative cover?
6 A. It's important to consider positional
7 accuracy if you are basing any data you might add or
8 subtract to that map on simply a statement of
9 positional accuracy, which is not what we do.
10 Q. Talking about relational accuracy, is that
11 what you said?
12 A. Yes.
13 Q. Talking about relational accuracy and the
14 use of the color infrared photography. The dates of
15 the infrared photography were, I believe, March,
16 early March of 1984, mid February of 1980 and mid
17 February of 1985 --
18 A. There were two NHAP methods that we used in
19 the study area.
20 Q. You were using those to verify or to define
21 the vegetative classes in the November 1985 satellite
22 imagery, correct?
23 A. Yes.
24 Q. Would you not expect that in the period of
25 time between March 3, 1984 and November 2, 1985 that
18
1 there would be some significant changes in the
2 relationships of the various vegetative classes as
3 shown in the infrared photography between the
4 satellite imagery?
5 A. The relationships between the species --
6 I'm sorry.
7 Q. Between the grouping of the classes. We
8 are talking about relational accuracy. That's what
9 I'm talking about, the relationship of one species
10 location to another. In other words, it would
11 change, would it not, over that period of time?
12 A. What now? You are talking about relational
13 accuracy in terms of species communities. That's
14 really not the context I was describing relational
15 accuracy in.
16 If you could explain a little more about
17 what you mean. I'm not certain --
18 Q. As I understand it, Dr. Lodge would attempt
19 to locate various vegetative or various features in
20 the infrared photography and compare that position or
21 that spot, identify it and then locate it in the
22 satellite imagery.
23 Is that more or less correct, or you tell
24 me what your understanding of it was.
25 A. He identified the feature in the
19
1 photography and then we identified the feature
2 sometimes with his assistance in the raw image.
3 What was your question?
4 Q. How did you employ this relational or this
5 relationship, as you have described the relational
6 accuracy in that process?
7 A. The NHAP photography is color IR
8 photography. Color IR produces very good
9 differentiation in patterns in vegetation community
10 patterns. Those same patterns were apparent in the
11 raw satellite image.
12 By positioning Dr. Lodge's focus into a 500
13 by 500 meter grid as best we could determine through
14 the method I described, he would sketch out the
15 patterns that he recognized the various plant
16 communities to exhibit within that area.
17 We would take that sketch that he had
18 prepared for the patterns he saw in the infrared
19 photography, compare those patterns in the 500 by 500
20 meter grid on image which we produced in rectifying
21 the image.
22 If we could see a pattern that he had
23 described in the photo, then that became the pattern
24 we recognized in the image. From that, we selected
25 the training sample.
20
1 Is that what your question is about?
2 Q. Yes. That's what I was getting at.
3 Wouldn't you expect those patterns to change between
4 the period of time of IR photography or satellite?
5 A. Dr. Lodge has extensive knowledge of the
6 Everglades, so he would recognize any obvious shifts
7 in vegetative patterns from, say, one period to the
8 other. We had to assume, and I'm sure Dr. Lodge
9 discussed it, that he would have some recognition of
10 what was going on in those patterns and those
11 classes, those species locations.
12 MR. CESARANO: Take a short recess.
13 (Thereupon, a brief recess was taken,
14 after which the following proceedings
15 were had:)
16 MR. CESARANO: Mark that Exhibit 13.
17 (The document referred to was thereupon
18 marked Exhibit 13 for Identification.)
19 MR. CESARANO: Mark this as Exhibit 14.
20 (The document referred to was thereupon
21 marked Exhibit 14 for Identification.)
22 BY MR. CESARANO:
23 Q. We were talking about the relational
24 accuracy vis-a-vis the positional accuracy?
25 A. Did I say relational -- relative accuracy.
21
1 I'm sorry. Relative accuracy.
2 Q. And that is based on the relationship of
3 patterns, vegetative patterns to each other?
4 A. No. Relative accuracy really doesn't
5 relate specifically to vegetative mapping. Relative
6 accuracy is a mapping term that relates to the
7 accuracy of one feature's relationship to another.
8 Q. In this case, it's the vegetative patterns
9 or vegetative classes?
10 A. It's the vegetative classes, the air boat
11 trails, the tree island heads, the canal center
12 lines. It's any feature within the area you happen
13 to be attempting to bring those features together
14 into.
15 Q. Isn't it important to know the location or
16 the positional accuracy of the features at the outset
17 in order to have any confidence in the relative
18 accuracy?
19 A. It's absolutely necessary to at some point
20 to establish some type of positional accuracy when we
21 transferred the points onto the image which was
22 geographically correct.
23 Q. And the places where you established
24 positional accuracy were where?
25 A. In the image when we rectified the image.
22
1 Q. And that is based on known locations of
2 particular features?
3 A. The rectification of images -- I think we
4 went through this process already.
5 Q. We don't need details.
6 A. When you would select your ground control
7 points and then compare those known points to points
8 in the image and then geographically stretch or
9 rectify the image to that set of known points, that's
10 called rectification, and that's what we did with
11 that image.
12 Q. You did that with the satellite image?
13 A. With the satellite data, yes.
14 Q. Was it necessary to do that with the
15 quadrangle sheets?
16 A. No. Our methods did not require that we
17 rectify the quadrangle sheets.
18 Q. As you move farther away from the
19 rectification points in the image by using the
20 relative accuracy process, is there a greater margin
21 for error the farther away you get from a known
22 point?
23 A. You wouldn't do that. The points you are
24 referring to, I guess, are the ground control points,
25 the known points?
23
1 Q. Yes.
2 A. Those points were used to rectify the
3 image. The methods we used to establish the
4 geographic locations of features in the photograph
5 were all based on the relationship of those features
6 to each other. So moving from a point, you would
7 continually readjust as you move. In other words, --
8 Q. What is the margin of error that -- can you
9 quantify the margin of error in the rectification
10 process that you used?
11 A. Yes.
12 Q. What was that?
13 A. Let's see. What was it? That's part of
14 the process you would always go through when you
15 rectify an image as part of our digital files we were
16 provided. The error was very low. And frankly, I
17 don't remember -- it's in the data we have given you,
18 it's part of the record. It was well within the
19 acceptable limits for rectification of this image.
20 Q. What are acceptable limits?
21 A. The errors down to .003 you are definitely
22 down to within -- the accuracy requirements you would
23 need for us is the raw data, 25 meter by 25 meter
24 sample data set. I don't remember what ours were.
25 They were very low.
24
1 Q. Was it within .003?
2 A. I don't remember.
3 Q. But .003 is acceptable?
4 A. The software we used allowed us to quantify
5 that in terms of materials, I can't remember how many
6 meters that was.
7 Q. Plus or minus?
8 A. So many meters, yes. It's on our digital
9 tape.
10 Q. My recollection, Mr. Rutchey's report
11 indicated his was within 3 to 7 meters?
12 A. What, for his rectification?
13 Q. I think you are right. Not for
14 rectification.
15 A. You are thinking of positional accuracy.
16 He used the same process. RMS error generation and
17 converted that to meters and wound up with "X" number
18 of meters.
19 Q. Mr. Rutchey's rectification was plus or
20 minus eight meters. Do you recall if yours was
21 better or worse than that?
22 A. I don't recall.
23 Q. We spoke briefly about the tasseled cap
24 analysis that you ran initially in the 1985 image, I
25 believe?
25
1 A. Yes.
2 Q. Have you run a normalized different
3 vegetation index on the '85 image or the '93 image?
4 A. In terms of within the tasseled cap did we
5 use the coefficients that a tasseled cap normally
6 would use? Yes, we have. "Normalize" is really not
7 the right term to use there.
8 Q. What's was the right term?
9 A. Applying coefficients that are applied.
10 Q. What is the difference between those
11 results and the greeness components on the tasseled
12 cap?
13 A. What's the difference between what, now?
14 Q. Those results. You said it's not proper to
15 use "normalized," but however you phrased it.
16 A. I wouldn't use the term "normalized" to
17 apply the coefficient to a tasseled cap image mainly
18 because the normalization has taken on in this
19 proceeding.
20 We applied coefficients -- a standard set
21 of coefficients that are common in tasseled cap.
22 Those coefficients were multiplied by the brightness
23 value.
24 Q. That's a different process than the
25 tasseled cap process?
26
1 A. No. Tasseled cap is a process of applying
2 coefficients to images to reduce the amount of data
3 being increased or maintain the majority of
4 information in that data. It does that by applying a
5 set of coefficients.
6 Q. Well, in doing that process is it effective
7 to distinguish between cattail and sawgrass?
8 A. It's effective to distinguish between
9 spectral response of various species of plants and
10 ground and water.
11 Q. And in doing the map, we are concerned with
12 cattail and sawgrass, are we not?
13 A. Yes, we are.
14 Q. In your opinion, is it effective to
15 distinguish between cattail and sawgrass?
16 A. I think that it could be, yes. It could be
17 one of the two that can assist you in distinguishing
18 between the cattail and sawgrass.
19 Q. Is it similar to tasseled capping that it
20 shows vigorousness of growth?
21 A. Is what? Is it -- is what similar?
22 Q. This other process that we have just been
23 discussing using the ratios.
24 A. The coefficients are part of the tasseled
25 cap analysis. That's what tasseled cap is, it's a
27
1 set of ratios that you apply to bands in order to
2 compress data so there is no difference in what I
3 discussed and a tasseled cap. That is a tasseled
4 cap.
5 Q. Just so I'm sure I understand, that is the
6 analysis that measures total greenness of plants,
7 chlorophyl or vigorousness of growth?
8 A. I never heard it described as total
9 greeness. One component of tasseled cap analysis is
10 greeness, yes.
11 Q. Greeness values from low to high?
12 A. Low to high, yes.
13 Q. You mentioned that the difference in
14 hectares as we saw on the tape resulted from
15 digitizing the data.
16 Were you talking about the satellite
17 imagery or did you also digitize the color infrared
18 photography?
19 A. No. When I say the differences in
20 selecting out the study area, cutting it out of the
21 image.
22 Q. Did you digitize the color infrared
23 photography?
24 A. No.
25 Q. I want to understand where I went wrong. I
28
1 asked you about the normalized difference in
2 vegetation index. You said that was not the proper
3 term or description?
4 A. Based on the fact that "normalized" has
5 been used in all proceedings so far to refer to some
6 kind of adjustment to radio metric response, things
7 of that sort, I want to make sure the record
8 indicates that I'm not talking about performing a
9 normalization of historical image.
10 Tasseled cap by definition does not require
11 they do anything except apply tasseled cap images,
12 and that's what we did.
13 Q. Now I understand that. Did you understand
14 what my question was by using that term, however?
15 A. Maybe I didn't. Maybe you ought to ask it
16 again.
17 Q. As I understand it. This process is one in
18 which you determine a ratio between two bands in the
19 TM image. I think band 3 and 4, the distance and the
20 greater the ratio or the higher the number, the
21 greater the distance between the trough and the peak.
22 Are you understanding me?
23 A. Yes. I understand what you are saying, I
24 think. Continue.
25 Q. And by developing and normalizing that
29
1 ratio you arrive at a scale of brightness values
2 from -- that indicate very little vegetation or
3 greeness to the top, which is quite a large amount?
4 A. Okay. Yes. What you do is, you apply a
5 set of -- not just simply the difference between two
6 bands. You neutralize some bands by adding and
7 subtracting some bands from others and things of that
8 sort and come up with -- this has been going on since
9 the late '70's, but yes.
10 You do wind up -- I guess if you want to
11 use the term "normalized," you wind up with a new set
12 of values that contain a majority of the information
13 that was in that original set of readings but
14 compressed into a smaller space. And it is
15 reflectance values if you want to refer to that as
16 that.
17 Q. Did you do that?
18 A. Yes.
19 Q. And is that what you were describing as
20 part of the tasseled cap analysis?
21 A. Yes. That process is within the tasseled
22 cap analysis, applying those coefficients.
23 Q. Now I think I understand.
24 A. Okay.
25 Q. Counsel brought with her yesterday this
30
1 document here with the date at the bottom of April 5,
2 1994. I presume this is the latest?
3 A. This is the final.
4 Q. This is the final?
5 A. Uh-huh.
6 Q. The answer is yes?
7 A. Yes.
8 Q. When we were last together, we had some
9 discussion about some banding that appeared in the
10 image. Did you do any further -- anything further
11 with the data to try to correct or account for that
12 banding?
13 A. As I stated before, banding in my opinion
14 had no significant influence on this image. We've
15 done nothing to correct banding in this image. In
16 fact, there is no noticeable banding except in the
17 plot. I'm not convinced whether that's a plot or a
18 banding.
19 Q. What did you do with this image from the
20 last time that we were together?
21 A. Simply refined the categories, just came up
22 with the best final product that we could arrive at.
23 Q. You mentioned you are not sure whether the
24 banding is in the image or from the plotter?
25 A. I noticed no banding that had any effect on
31
1 this image. The banding, in my opinion, is not an
2 issue. The banding did not affect the classification
3 of my opinion.
4 Q. This next other document is I believe what
5 you called the entire study area?
6 A. Yes.
7 Q. You are saying, in your opinion it's a
8 non-issue, but I think it's rather clear in this
9 image. My question is, why in your opinion would the
10 banding be more distinct in the image of the entire
11 study area than the image of just 2A?
12 A. We made no adjustments for banding.
13 Q. So you don't have an opinion or don't know
14 why it would be more distinctive in the entire study
15 area than in the 2A map?
16 A. I don't think it's any more distinctive in
17 the entire study area than it is in 2A. Defective
18 banding, in my opinion, did not affect the
19 classification to any significant amount.
20 Q. Let's talk about Exhibits 13 and 14 for a
21 minute, please.
22 A. Okay.
23 Q. Could you just tell me what the
24 relationship between Exhibits 13 and 14 are, please?
25 A. Exhibit 14 is a resume, my resume.
32
1 Exhibit 13 is also an older version of my resume.
2 Q. What's the difference between the two?
3 A. This older resume, Exhibit 13, is a resume
4 that ends with my work with the state of Mississippi.
5 And this resume is a generalized, more current resume
6 of my work since then.
7 Q. And Exhibit 14 doesn't list any reports or
8 publications as Exhibit 13 does. Have you published
9 anything since the date of Exhibit 13?
10 A. Yes. Well, a few papers. Nothing that
11 applies directly to this project.
12 Q. Other than working on the 1985 image, and I
13 believe you said you refined the 1993 image somewhat
14 in the last week or so since we were last together,
15 have you done anything else in connection with this
16 project?
17 A. We have done the '85 interpretation and the
18 tasseled cap. That has been our work this week.
19 Q. I do want to get into the accuracy
20 assessment.
21 A. I apologize for that not being here. It
22 should be here. I don't know why it's not here.
23 MR. CESARANO: Other than that, I do want
24 to wait for that to come in, do you have
25 anything to go into at this point?
33
1 MR. WATTS-FITZGERALD: As long as you
2 understand he is reserving having to come back.
3 MS. RAEPPLE: Yes.
4 THE WITNESS: Would you like me to call to
5 the office?
6 MR. WATTS-FITZGERALD: Yes. We will -- we
7 are doing that. Is it my understanding we are
8 getting two accuracies, one for '93 and one for
9 '85.
10 THE WITNESS: Yes.
11 MS. RAEPPLE: Go ahead and call your
12 office.
13 (Thereupon, a brief recess was taken,
14 after which the following proceedings
15 were had.)
16 CROSS EXAMINATION
17 BY MR. WATTS-FITZGERALD:
18 Q. Good morning. I'm Tom Watts-Fitzgerald
19 representing the United States in this matter.
20 Same ground rules. If you don't understand
21 the question or need to take a break, just speak up.
22 Addressing for a moment and clarifying in
23 my mind -- the documentation may be en route -- I
24 understand the two matrixes reflecting the error
25 assessment or accuracy assessment for the 1983
34
1 imagery -- 1985 imagery and 1993 imagery that you
2 have produced will be here shortly and we will look
3 at that then.
4 There were worksheets, as that term is
5 used, by Dr. Lodge yesterday which I think you
6 referred to today that may have included traces of
7 vegetative patterns that were utilized in connection
8 with the 1985 image development.
9 Are those en route, to your knowledge?
10 A. Yes. Those are en route, to my knowledge.
11 Q. Am I correct in understanding that there
12 would be one of those for every accuracy assessment
13 point to the tune of roughly 85 of these things?
14 A. One of which one?
15 Q. One of those worksheets.
16 A. There are two worksheets. There are a set
17 of worksheets that were used in support of the
18 classification of the 1985 image, and there are a set
19 of worksheets, different type of worksheet for the
20 accuracy assessment for the '85 image. Both of those
21 are either in en route in Fed Ex or being faxed here.
22 Q. For the accuracy assessment worksheets
23 there are 84 of those or something of that order, one
24 for each point, the tables?
25 A. The total number would be listed in error
35
1 matrixes.
2 Q. Why are there less than half as many for
3 '85?
4 A. There were fewer classes and the classes
5 were broader in definition.
6 Q. By going to broader classes, that appears
7 to be reducing the apparent variability within the
8 imagery when you look at the color depictions; is
9 that true?
10 A. Reduce the variability in the image? You
11 mean the number of classes depicted in the image?
12 Q. Yes.
13 A. Obviously, yes.
14 Q. It reduces your ability to make vegetative
15 distinctions from that image to some degree?
16 A. In that the vegetative classes are
17 different since we classed, as I described earlier,
18 some classes into smaller classes, yes.
19 Q. And since you did that for the reasons you
20 stated, some of which related to the resolution
21 capabilities of the color IR from 1985, are you able
22 within a reasonable degree of certainty to compare
23 your set, the '85 imagery and '94 imagery as you
24 produced it, and make an assessment of acreage
25 changes and vegetative alterations over that time
36
1 period?
2 A. Within the limits of the data, yes.
3 Q. What could you see as the limits of the
4 data?
5 A. The fact that we have classes and the
6 interpretation that we did in '93.
7 Q. Can you make a direct comparison?
8 A. You can. Those that are directly
9 comparable that I described earlier. The others you
10 have to assume the new class is a combination of the
11 classes that were combined in the '93 image to give
12 you that same basic vegetative mix.
13 Q. Can you compare your two products to
14 Dr. Jensen's normalized time series to the various
15 years and assess vegetative alterations in comparison
16 to his assessment?
17 Can we put them side by side and reasonably
18 be able to analyze the results?
19 A. Well, since we have questions that have
20 been stated in the previous deposition about the
21 methods of Mr. Rutchey and Dr. Jensen's one-to-one
22 comparison as we stated would be inappropriate
23 because the classes are different.
24 Q. Is it my understanding that you will offer
25 no testimony at the time of hearing attempting to
37
1 make a one-to-one comparison between the work you
2 have performed and the work performed by Dr. Jensen
3 and Mr. Rutchey?
4 A. Would you define "one-to-one".
5 Q. It's the term you just used. You tell me.
6 A. You are the one that said would I do a
7 one-to-one comparison. I told you.
8 Q. No, we can read back the question. What I
9 said was, can you do a comparison. You said you
10 can't do a one-to-one.
11 A. Okay.
12 Q. What did you mean by that?
13 A. That the classes are -- as I stated earlier
14 in my deposition, the classes that Mr. Rutchey and
15 Mr. Jensen used, in our opinion, are inconsistent.
16 Therefore, there is no way to do a one-to-one
17 comparison. The classes are not the same. Our
18 classes are defined as a percent of the vegetative
19 class that we put is predominant in that class.
20 Q. I understood from your earlier deposition
21 that some of your concerns about the methodology
22 employed by Mr. Rutchey and then indirectly -- at
23 least for the '91 imagery -- by Dr. Jensen concerned
24 some of the hardware that was employed in the field;
25 is that correct?
38
1 A. That's correct.
2 Q. Have you resolved those hardware issues in
3 your mind?
4 A. Yes.
5 Q. So, in fact, you now know that the
6 equipment had the capacity to store the data during
7 the field operations?
8 A. I'm satisfied that I understand it fully.
9 My concerns about how they used the Trimble
10 equipment, if that's what you mean -- I'm not sure
11 what you mean.
12 Q. I had understood your earlier testimony to
13 be you were not confident that the equipment that was
14 used in the field in fact had the capability to store
15 the data points.
16 A. Storage was one of the concerns I had.
17 Q. Do you still have that concern?
18 A. My major concern does not involve storage.
19 Q. What are all of your concerns regarding the
20 equipment used in the field?
21 A. As stated earlier, those concerns -- as I
22 stated earlier, those were my concerns. The
23 procedures used by Mr. Rutchey are still of concern
24 to me.
25 Q. What procedures?
39
1 A. Procedures used to locate his accuracy
2 assessment once in the field.
3 Q. What about his procedures cause you
4 concern?
5 A. He had no communication to the base
6 station. He had no way to do post processing in the
7 field. Therefore, how could he have determined where
8 he was in the field when he went out with a set of
9 coordinates looking for an accuracy assessment point.
10 Q. Is it your expert opinion that one cannot
11 record the data in the field and do the differential
12 corrections and post processing afterwards?
13 A. No, it is not my opinion. That is not my
14 opinion.
15 Q. So that is possible?
16 A. Is it possible to post process data?
17 Q. Yes.
18 A. Certainly.
19 Q. I don't understand. I guess I would have
20 to ask you to elaborate what your concern was about
21 his inability to process the data in the field as one
22 of your major concerns.
23 A. According to his testimony, his paper, he
24 did no post processing in the field.
25 Q. Why is that a problem?
40
1 A. Because if you don't post process data with
2 a single hand-held free channel unit, if you don't
3 have some communication with the base station or some
4 way to adjust your position real time, then he could
5 not be exactly on the point he thought he was at
6 until he did post processing.
7 Q. And then once one had done the post
8 processing later back in the lab, for example, one
9 would know where you were while you were in the
10 field?
11 A. That's true.
12 Q. Even if that point was not what you were
13 looking for when you went out?
14 A. That is right.
15 Q. It's not a question of not knowing
16 necessarily where he really was, your concern is he
17 wasn't at the time in the field where he thought he
18 wanted to be?
19 A. He more than likely was not at the point he
20 wanted to be in while he was in the field.
21 Q. Your team used the Magellan receiver for
22 their field work for the training site selection
23 initially?
24 A. Yes, we used the Magellan.
25 Q. That did not have the capacity to store
41
1 points?
2 A. The Magellan unit doesn't store enough
3 points by itself.
4 Q. That was what Magellan model?
5 A. NAV Pro 4000.
6 Q. Which has limited storage capacity but not
7 sufficient for the work you were doing in a given
8 day?
9 A. Correct.
10 Q. Within the remote sensing community, what
11 is the percent accuracy considered acceptable -- I
12 should probably qualify that and say for this
13 purpose, user accuracy?
14 A. Within the user community, what is the --
15 Q. Within the remote sensing community?
16 A. What is the --
17 Q. Level percentage of accuracy that would be
18 considered acceptable for a map product like these
19 classified from Landsat TM data.
20 A. I think that depends on the project and the
21 person doing the work. I don't know that there is a
22 standard map accuracy. You are talking about map
23 accuracy?
24 Q. Yes.
25 A. Yes. I don't know there is a standard
42
1 acceptable level of accuracy.
2 Q. In your view, then, is there a certain
3 level of professional judgment and project specific
4 characteristics?
5 A. It has to be up to the person doing the
6 classification, yes.
7 Q. Do you, in fact, intend at the time of
8 hearing or between now and then to make an effort to
9 compare your classification products for '85 and '93
10 with Mr. Rutchey and Dr. Jensen's work?
11 A. As I stated before, we have done some
12 preliminary viewing of their products with ours. We
13 have done no extensive analysis. If I were requested
14 to do that I would do it.
15 Q. How would you go about it, given the fact
16 that you within your two products don't have
17 uniformity of classifications and do not have cross
18 uniformity with their work?
19 A. My classes are -- '85 classes are a sub set
20 of the '93 classes. There is some uniformity. It's
21 just a sub set. We would have to come up with an
22 acceptable class comparison between Mr. Jensen's and
23 Mr. Rutchey's work in order to do further comparison
24 of the two products, three products.
25 Q. You don't plan to do that unless requested
43
1 by counsel?
2 A. If I'm requested, I will. Right now we
3 don't plan to do that.
4 MR. WATTS-FITZGERALD: Counsel, if the
5 witness does that, I think we are entitled to
6 some notice because that would be a major new
7 opinion.
8 BY MR. WATTS-FITZGERALD:
9 Q. You have no final opinion on that at this
10 point?
11 A. On what?
12 Q. The comparison of your work with Dr. Jensen
13 and Mr. Rutchey's.
14 A. We have general impressions of our work
15 versus theirs, but we have no digital overlay
16 comparison yet. I haven't been asked to do that.
17 Q. Have you discussed with anyone production
18 of a digital overlay of any sort either on the
19 product that Dr. Jensen generated or Mr. Rutchey and
20 your work?
21 A. Yes.
22 Q. Have you taken any steps to produce such --
23 I assume by "overlay," you mean something like mylar
24 acetate that would go over --
25 A. I'm referring to a digital overlay, doing a
44
1 digital overlay and platting the result. You could
2 produce a hard copy overlay, but I would do it in a
3 digital environment.
4 Q. Have you begun that process?
5 A. We did some preliminary work several weeks
6 ago where we took some of Jensen's products and
7 brought them up, comparing them to ours, looking at
8 them in different windows. We did not overlay the
9 two images but looked at the differences between his
10 classes and ours.
11 Q. That was two separate images?
12 A. Yes, two separate windows.
13 Q. Was any preliminary work done through
14 software programs isolated into a single image,
15 various categories of classes, to see where there was
16 non-conformity or conformity?
17 We did some preliminary work on that. And
18 then because of the volume of data, we took it off
19 the system. We have not reinstalled it.
20 In a visual environment we set up a way to
21 look at two images on the screen, but we have done no
22 overlay.
23 Q. Do we have that on the electronic data
24 provided, I guess, yesterday?
25 A. That was a set of -- it wasn't dated. It
45
1 was a set of algorisms. You should have it on the
2 back-up tape. If you don't, we will supply one.
3 Q. What's the file name or directory for that?
4 A. I don't remember.
5 Q. You testified in response to counsel this
6 morning that with respect to area 2A the acreage
7 differential that appears in the Exhibits, I think,
8 11 and 13 was as a result of digitizing error in one
9 of the images. But I'm not sure that you said
10 specifically which of the images you determined was
11 incorrect.
12 A. Well, incorrect, I mean the digitizing, the
13 area digitized were slightly different. Obviously,
14 the acreage showed up differently. The area was
15 digitized to a slightly larger extent than '93. '85
16 had slightly more acreage than '93.
17 In eight years we lost a hundred hectares.
18 If it was my acreage, I would probably care about it.
19 Q. Did you analyze to determine where it went
20 or what portion of the image that was attributable
21 to?
22 A. No. Acreage was lost in the digitizing of
23 the study area. We included 130 hectares in one
24 image that was not included in the other.
25 Q. I have no reason to believe within the
46
1 bounds of 2A over that eight year period it grew. So
2 where did the 100 acres come from?
3 I mean, is it strictly an artifact of the
4 digitizing process? Is it an imagery problem? How
5 do you account for the differential?
6 A. Since we didn't use coordinated geometry to
7 digitize the area of 2A. To start with, digitizing
8 has in it always a percent of error, if you want to
9 call it that. You wouldn't take much difference in
10 digitizing an area of interest in one year's image to
11 another to find out less than one-tenth of one
12 percent or less than that error in the total acreage
13 of the study area.
14 Q. I take it within your professional opinion,
15 that's within the acceptable range of errors for that
16 type of process?
17 A. I would certainly think so, yes.
18 Q. We may have covered this last time around,
19 but I didn't see it in my review of the available
20 information from the earlier deposition.
21 But prior to this particular case, what
22 work had you done in the vegetative mapping field in
23 the Everglades?
24 A. None.
25 Q. Had you done any vegetative matching for
47
1 the wetlands prior to this case?
2 A. In those various industrial site studies, I
3 did using MSS data, I think a little TM data. In my
4 career I have done wetlands identification and
5 wetlands mapping mainly in terms of industrial site
6 development.
7 Q. Did you use TM data for those wetland site
8 evaluations?
9 A. I have to look at the reports and see --
10 no, I would not have used that. I used MSS on those
11 studies.
12 Q. The accuracy assessment points for '93 you
13 said you have 84 of those for the '93 image. I
14 assume -- I shouldn't assume. Are those 84 all
15 within area 2A?
16 A. Yes.
17 Q. There are no accuracy assessment points for
18 the balance of the image that shows the entire study
19 area as you have referred to it?
20 A. That's correct.
21 Q. And I noted yesterday, and it's in front of
22 us today, that that December '93 production dated
23 April 4 for the entire area has a warning or at least
24 a caveat, if you will, on it that there is no
25 accuracy associated with the areas other than WCA-2A?
48
1 A. I believe it says no accuracy assessment.
2 Accuracy is verified within Water Conservation Area
3 2A only.
4 Q. I translate that -- tell me if it's correct
5 or not -- that that means no one is making any
6 representations as to the accuracy of the image
7 beyond the boundaries of 2A?
8 A. It means the accuracy has not been
9 assessed.
10 Q. You are drawing a distinction. Tell me
11 what the accuracy is beyond 2A.
12 A. I expect the accuracy to be close to 2A.
13 Q. Quantify that.
14 A. I can't do that until we do an assessment.
15 Q. Are you going to do that?
16 A. If I'm asked to do that, I will.
17 Q. Do you have any present plans to do an
18 accuracy assessment?
19 A. No.
20 Q. How long would it take if you were asked to
21 do an assessment for the balance of the area depicted
22 in that image?
23 A. That would depend on the availability of
24 aerial photography for the entire study area.
25 Q. Where did you acquire the NHAP photography
49
1 for the '83 image?
2 A. For the '83?
3 Q. I'm sorry, '85. I keep saying that.
4 A. The '85 NHAP photography was acquired from
5 the NASA spec photography. We ordered it through the
6 base, St. Louis facility. I'm not sure where it was
7 actually shipped from, from the U.S. government.
8 Q. Now, for the '93 image you used BDA's
9 photography for 2A?
10 A. Yes.
11 Q. Was any effort made at the time you were
12 seeking to acquire CIR photography in the time frame
13 of the December '93 satellite pass to determine what
14 other CIR was available?
15 A. I'm sorry, did I make that investigation?
16 Q. Yes.
17 A. Yes.
18 Q. Was there other CIR available?
19 A. I couldn't find any, no.
20 Q. So when you said a few minutes ago it would
21 depend on the availability of CIR, I take it you are
22 really saying you can't because there is no CIR known
23 to exist?
24 A. No. Right now we are unable to find some.
25 But the way I understand it, the photography
50
1 available through Breedlove, Dennis, goes beyond 2A,
2 so I could go beyond that.
3 Q. What's your understanding of the rest of
4 the geographic coverage?
5 A. It goes outside of 2A.
6 Q. Do you know where?
7 A. Well, I have seen some indices that goes up
8 to Loxahatchee a little bit and over into 3A or 3.
9 Q. If that's the extent of the CIR imagery
10 available, would you be able to assess only those
11 portions of the areas beyond 2A that fell within the
12 scope of the vegetative coverage?
13 A. We would have to have some type -- to use
14 the same method we have to have some type of
15 photography available for the study area.
16 Q. Are any other methods available to you?
17 A. You could hold out a percent of the
18 original ground truth sites from the classifications,
19 classify using that set of ground truth points, and
20 then check your classification against your own
21 ground control points to see if, in fact, you
22 classified the image as to what you held out.
23 Q. You would go back and regenerate your image
24 with less ground truth points?
25 A. That's one option you would have, yes.
51
1 That could be done.
2 Q. Aren't you kind of robbing Peter to pay
3 Paul there? I understand your reasoning, I think, in
4 not using a ground truth point employed for
5 generation of the image or training in the image for
6 error assessment because of the bias that induces and
7 the statistical problems that would be incurred in
8 doing something like that.
9 But you in a sense end up under training
10 your algorism to generate the image as you cut back
11 at the front end of the process to reserve them for
12 later use as assessment points. You don't have that
13 many to begin with for an area that size?
14 A. Is there a question there?
15 Q. Is that true or not?
16 A. Is all of that true?
17 Q. Yes.
18 A. Yes. Of course it's true.
19 Q. Speaking of numbers and points, how many
20 accuracy points do you expect to have total for the
21 '93? Are there any more that you will be assessing?
22 A. At this point we generated 200 or 240
23 points. We possibly could assess --
24 Q. For accuracy?
25 A. For accuracy.
52
1 Q. And you used 84. Are you going to use any
2 more?
3 A. If instructed to do so, we certainly could.
4 We maintain that set of points for that reason.
5 Q. The 84 that you used, how were they
6 selected out of the 240?
7 A. In sequential order.
8 Q. And the 200 that were generated, that was
9 done with using what selection program?
10 A. Stratified random sampling.
11 Q. What measures, if any, did you take when
12 you determined to use only the first 84 to insure
13 that each of the classes was adequately represented
14 for statistical purposes in the reduced sample that
15 you chose to employ?
16 A. Since the sample was generated through
17 random numbers, since we went through the random
18 numbers sequentially we wound up with a random set of
19 the entire set.
20 Q. Who would instruct you to use additional
21 points out of the 200? Where would that decision
22 come from?
23 A. It would come from my people that hired me
24 to do this project.
25 Q. At this point you have intentions to use
53
1 additional points. That's something you are capable
2 of doing given time, but no current plans, per se?
3 A. Yes, correct.
4 Q. With regard to the accuracy assessment for
5 the 1985 image, I think you said there were 34 points
6 for that?
7 A. 34 or 37, I believe.
8 Q. How were those points selected?
9 A. Exactly the same method, random stratified
10 samples. There were 104 over 100 points. I believe
11 104 in total.
12 Q. On what basis did you determine that the 37
13 or whatever the number is, in the 30's was, that were
14 ultimately used and were identified by Dr. Lodge was
15 sufficient for statistical purposes in analyzing the
16 accuracy of the area encompassed by WCA-2?
17 A. It was just my judgment.
18 Q. Are you familiar with the body of
19 literature in the remote sensing field which
20 describes how one should go about the process of
21 assessing the appropriate number of accuracy
22 assessment points?
23 A. Yes.
24 Q. What factors did you take into account in
25 making your professional judgment that 30 was
54
1 sufficient?
2 A. Just begin the complexity of the map and
3 the major classes that we were dealing with. Major
4 classes were cattail and sawgrass.
5 Q. You mentioned in your original deposition a
6 concern because you felt that Mr. Rutchey -- and
7 perhaps Dr. Jensen, as well -- had an initial bias in
8 their assessment because they set out focused on
9 cattail, Mr. Rutchey, and sawgrass, Dr. Jensen.
10 Do you recall saying something along those
11 lines?
12 A. Stated some concerns about their method and
13 the fact about comparing the two.
14 Q. My understanding from Dr. Lodge's testimony
15 is that basically you set out to do the same thing,
16 focusing essentially on sawgrass and cattail; is that
17 not true?
18 A. That we did the same thing that Mr. Rutchey
19 or Dr. Jensen did? They did different things.
20 Q. I'm not talking methodology. I'm talking
21 mind set. You set out along with Dr. Lodge with the
22 purpose of specifically analyzing vegetative changes
23 with respect to cattail and/or sawgrass just as they
24 did, although you chose a different methodology to do
25 it?
55
1 A. Well, your question, if I understand your
2 question, yes. We set out to determine the
3 relationship between sawgrass and cattail, but we
4 feel that Dr. Jensen's methods and Mr. Rutchey's
5 methods were different. Therefore to say it was like
6 them, I guess you mean in terms of did they have a
7 goal of determining -- are you saying that both had
8 the goal of determining sawgrass and cattail?
9 Q. In both your effort and their efforts
10 looking at them collectively, yes.
11 A. Yes. The emphasis of the study was on the
12 relationship of sawgrass and cattail within 2A, yes.
13 Q. How many training samples were there for
14 the '85 image?
15 A. The acreage -- no, that wouldn't be it.
16 Yes. Let me see the tables with the acreage,
17 Exhibit 12, which is the '93 image.
18 If you are looking down at the bottom, it
19 says based on original 53 classes for December 10,
20 you see at the bottom 53 classes down at the bottom.
21 Each one, it says "based on original of each number
22 of classes."
23 Q. How do the classes tell me how many
24 training samples?
25 A. Because we produce the number of classes in
56
1 the original classification equal the number of
2 training samples. There were 35.
3 Q. You did map productions here of imagery for
4 36 training sites and 99 training sites -- I'm sorry,
5 classes. You just said classes equal training sites?
6 A. No. You are misunderstanding.
7 Q. Straighten it out for me.
8 A. What are you confused by?
9 Q. I asked you how many training samples you
10 had or training sites, if you will, for the '85
11 image. You just talked about the '93 image.
12 A. You said '85.
13 Q. I said '85 to start with. I'm not asking
14 how many classes. I understood you to say in the '93
15 image classes equals training sites and the numbers
16 don't match. I understood there were 116 roughly
17 training sites for the '93 image from your testimony
18 last time and doing all of the R, H and E's --
19 whatever that meant. Is that right, 116?
20 A. 116 observations were made.
21 Q. Training sites?
22 A. Okay. Wait. Let me understand what you
23 are asking again. Ask me one more time. I'm not
24 understanding.
25 Q. Back up. I'm looking at '93 now. You had
57
1 about 116 training sites for '93?
2 A. For the entire study area.
3 Q. When you generated the class map --
4 classified map for '93 for the entire study area, how
5 many classes did you have when you went through the
6 classification process initially?
7 A. You mean before recoding the classes into
8 general categories?
9 Q. Yes.
10 A. The map is down here and that show 36 and
11 99 classes.
12 Now I understand what you are asking.
13 There are two processes here for each year. If I'm
14 understanding correctly. A little history, I guess
15 for each year we produced an unsupervised
16 classification for 2A. 99 classes unsupervised and
17 36 classes unsupervised. That was a separate process
18 from the supervised classification.
19 Q. On the unsupervised, why did you do that?
20 A. On the unsupervised -- I thought you were
21 talking about supervised.
22 Q. We are going to go back to the supervised
23 on the unsupervised. Why did you do that?
24 A. On the unsupervised classifications we
25 produced 99 classes and 36 classes of unsupervised
58
1 classification of TM data for all of those years
2 because we wanted to see what type of spectral
3 significant classes there were in those images.
4 In other words, if you break it into 99 how
5 many classes and where are they spatially in the
6 image. If you break it into 36, where are those 36
7 clustered images. So it was to gain a feel for the
8 study area.
9 Q. But according to the printout dates, that
10 was done the end of last month. At least the
11 platting was done. When did you actually conduct
12 that exercise?
13 A. Unsupervised classification?
14 Q. Yes.
15 A. Most of that was completed in the week of
16 my deposition -- the deposition in Tallahassee,
17 Florida a week ago.
18 Q. How will you, if at all, be relying on
19 those unsupervised classification efforts in
20 formulating your opinions of testimony in this case?
21 A. They will provide an indication of spectral
22 diversity in the study areas to support analysis and
23 conclusions in the area.
24 Q. In an unsupervised classification you as
25 the operator get to set arbitrarily the number of
59
1 classifications the program should look for?
2 A. Correct.
3 Q. So if I told it I want 200 classes or 230,
4 all I have to do is set into the algorism program the
5 range of spectral signature that I want and it will
6 attempt to classify every pixel in the digitized
7 database into that number of classes; correct?
8 A. Correct.
9 Q. So the fact it's 99 or 36, per se, has no
10 significance. Those are just the numbers you chose?
11 A. No significance?
12 Q. It tells you nothing about ground truth, it
13 tells you nothing about what's really there, it just
14 assigns pixel by pixel into your set number of
15 categories subject to black spots if something falls
16 outside of the range you specified in the program?
17 A. Okay. First you make a statement that it
18 gives you no information about the image. In my
19 opinion, it does give you information about the
20 image, but it does not tell you -- it will not tell
21 you a certain class is cattail. It will tell you
22 that there are, based on what your input is, subsets
23 to this image, it will tell you what spectral classes
24 fall in that set of bands that you created.
25 Q. So if you tell it, give me 99 and here is
60
1 the specs for each of the 99, and you have an
2 incredibly homogenous area, the program could end up
3 coming back saying, sorry, there are only four
4 classes out there?
5 A. Repeat that, please.
6 Q. If I tell it I want the program to classify
7 an image in the 99 classes, unsupervised
8 classification, it will go pixel by pixel through the
9 entire image and will try to fit them into the 99 if
10 it can?
11 A. Correct.
12 Q. Some may not fit in the 99 because you are
13 specifying the range for each of those
14 classifications, are you not?
15 A. In an unsupervised classification in the
16 software -- Imagine software package, you tell it to
17 do 99 classes out of the image it will be produced.
18 Q. It will divide up your entire spectral
19 range in that image in the 99?
20 A. Yes.
21 Q. Same thing if you do it with 36?
22 A. Yes.
23 Q. After producing that work, what opinions or
24 conclusions did you derive from reviewing the
25 unsupervised class maps?
61
1 A. It gave us a feel for the effect of fire,
2 which was obvious in the unsupervised
3 classifications. It gave us a feel for where the
4 greater diversity existed in the image.
5 Q. How did you use or employ that information
6 or that sense impression that you developed in
7 producing the final product for the 1993 and 1985
8 imagery?
9 A. It served as reference material.
10 Q. So it doesn't directly affect the
11 classified products for '93 and '85?
12 A. That's correct.
13 Q. General background?
14 A. Yes. It was not directly affecting the
15 classifications, correct.
16 Q. Did you do any accuracy assessment based on
17 the imagery that was produced through the
18 unsupervised classification classes?
19 A. No.
20 Q. Did you make any effort, you or those
21 working with you at your firm, to take the
22 unsupervised classes the, 36 and 99 and identify them
23 as to vegetative type?
24 A. Yes, we did.
25 Q. There is no legend on the map that appears
62
1 to do that. Where does that appear?
2 A. It doesn't. We began -- as I stated in my
3 deposition last week, the original plan was to take
4 the 36 class or the 99, whichever we decided was most
5 appropriate, and use that in a hybrid classification
6 process. Hybrid classification process normally
7 combines an unsupervised and supervised
8 classification. But we did not use that. We did not
9 use that process on the '85 image.
10 Q. Did you use it on the '93 image?
11 A. No, we did not. It was supervised.
12 Q. So there are no hybrid products available?
13 A. That's correct. We did not do a hybrid
14 classification.
15 Q. Did you start that work and stop it or you
16 didn't begin it?
17 A. We actually never did a hybrid
18 classification. Wait. Let's see.
19 At one point when we collected about eight
20 training samples from the NHAP photography, we did
21 combine those signatures to the 36 class unsupervised
22 signature table and run the classification. But we
23 did not use that, did not ever produce it as a
24 product. Just that it was a test.
25 Q. Does that appear on the back-up tape you
63
1 brought us?
2 A. We didn't save it. It was not a
3 work-product.
4 Q. Moving now to the 1985 product, how many
5 training samples or sites were there for the 1985
6 image?
7 A. I believe there were 37.
8 Q. How many accuracy assessment points for the
9 1985 image?
10 A. Wait, let's see.
11 Q. Did you reverse the numbers? 37 roughly
12 is, in fact, the accuracy assessment number?
13 A. Just a second.
14 Q. It's not always me.
15 A. You are right. It's not. '85 -- I cannot
16 remember right now how many training samples were
17 used in the '85 image. My mind is blank. I can't
18 remember right now. Those documents are being faxed
19 here. We will have them shortly. I apologize for
20 that.
21 Q. Without the actual number, can you tell me
22 how the training sites were selected for the '85
23 imagery?
24 A. Yes. Exactly. The process was very, very
25 similar to the selection process used in the '93
64
1 image. You want me to elaborate some more?
2 Q. It was my understanding in '93 they went
3 out in the field. First, you couldn't very well go
4 out in the field in '85?
5 A. Okay. The field operations became
6 photogrammetric operations. The process, in other
7 words, the process of identifying samples were
8 similar in that we weren't in a helicopter, we were
9 perusing the NHAP photography. We found areas that
10 had recognizable vegetative patterns, zeroed in on
11 those and used those as training samples.
12 In other words, Tom would make his notes,
13 as he told you yesterday. From that, we would
14 identify the same area in the image and then select a
15 training sample from the image that met the photo
16 mission, -- if you would, observations.
17 Q. What criteria were used to decide how many
18 training sites you needed for each class that you
19 were attempting to generate?
20 A. We attempted to find some number of
21 training sites in each class and attempted to
22 stratify those per our knowledge of the image in
23 terms of most of it is sawgrass, or at the center
24 there might be cattails, something of that sort.
25 You select according to expected patterns
65
1 to some degree.
2 Q. So that was done more or less in a
3 supervised as opposed to unsupervised fashion?
4 A. Absolutely supervised.
5 Q. Because you were working from CIR
6 photography with a relatively small scale in
7 comparison to what you used for '93 with the loss of
8 resolution, et cetera., how do you account for the
9 inevitable variability within an area in terms of its
10 spectral signature?
11 A. Variability in an area was more difficult
12 to interpret. Therefore, that's why we reduced the
13 number of classes we could identify in the available
14 photography.
15 Q. Was there any class or is there any class
16 in your '85 imagery for which only one training site
17 was available or used?
18 A. I don't think so. I think we had at least
19 more than one in all classes, all field observations.
20 I can't remember one that only had one training site.
21 Q. You would be able to tell that --
22 A. When we see the worksheets, those are
23 clearly listed.
24 Q. You told Mr. Cesarano this morning that the
25 satellite image was rectified?
66
1 A. Yes. How did you determine from your
2 satellite image -- I think you were taking about '93
3 at that point, maybe this applies to '85.
4 Were both '85 and '93 geo rectified?
5 A. The '93 image was rectified and the '85 was
6 rectified to the '93 image.
7 Q. How many ground cross points did you use
8 for '93 to do that?
9 A. That is in the digital data. I'm trying to
10 remember. That is on the tape we gave you. I
11 believe there were about 34 or 30 something ground
12 control points.
13 Q. Is that throughout the entire study area?
14 A. Throughout the entire study area.
15 Q. How many within 2A, if you can recall?
16 A. I would have to go back and look at the
17 records to see.
18 Q. How did you determine the geographical
19 position of the control points?
20 A. Determine -- you mean where they were?
21 Q. Yes.
22 A. We used USGS quad sheets.
23 Q. What's the positional accuracy of quad
24 sheets?
25 A. The positional accuracy is stated by U. S.
67
1 geological survey to be plus or minus 40 feet.
2 Q. Were you working off the full sized quads
3 at this point or had you had to reduce them already?
4 A. Which image are you talking about?
5 Q. '93.
6 A. We used the full size quads.
7 Q. You said you geo rectified '85 to '93?
8 A. Yes.
9 Q. So you did not use the quad sheets for that
10 purposes for the '85 imagery?
11 A. No, we did not go to the paper quads to
12 rectify the '85 image to the '93 image. We used
13 image to image rectification.
14 Q. Because you used the quads for '93, the
15 geophysical position accuracy was the limiting factor
16 for the '93 image, or are there other factors that
17 affect the geo rectification accuracy of the image?
18 A. Our geographic rectification was based upon
19 USGS 24,000 quads. They were the control points from
20 which we selected our ground control points.
21 Q. I understand that. I'm not sure that quite
22 answered the question I had in mind.
23 A. What was the question?
24 Q. The question I had in mind is that the only
25 error input, if you will, or were there -- could
68
1 there have been other sources of positional error in
2 your process not solely attributable to the error
3 inherent in the quads?
4 A. Let me make sure I understand what you are
5 asking me. Are you asking what error might be
6 introduced in the rectification process other than
7 the error in the quad?
8 Q. Yes.
9 A. The ability of the operator to identify a
10 feature in the quad and to identify the same feature
11 in the image. Obviously, it has a potential for
12 error.
13 Q. I assume if you selected a ground control
14 point, it's because you could find it in both. Is
15 that a fair assumption?
16 A. Yes.
17 Q. The ability of the operator to find it is
18 not a factor. You used it and you found it?
19 A. No. I mean, the ability of an operator
20 always has an influence on the potential to introduce
21 error. Ground control points are usually the
22 intersection of two vectors. You have to estimate
23 that intersection so that includes the accuracy of
24 the digitizing tables, the eye of the operator. So
25 there is potential error always for operator
69
1 interface to the process.
2 Q. In your expert opinion what is the error of
3 the '93 image -- the total error, excuse me?
4 You were fitting the '93 image to the quads
5 in that entire process, what's the assessment of the
6 overall proportional error?
7 A. That's in the digital data. That's
8 normally stated. It was very small. I don't
9 remember the number. It's in the digital data
10 supplied to you. You are welcome to look at the
11 files.
12 Q. With regard to the '85 satellite image, am
13 I correct in believing there is an error inherent in
14 fitting the '85 to the '93?
15 A. There is potential error, but we used the
16 same observed points in the '93 to find that same
17 intersection, if you would, in the '85 and then
18 rectified those two together.
19 Q. What role, if any, did the reduced quads
20 play in the rectification process?
21 A. None.
22 Q. So the error inherent in fitting the '93
23 image to the full sized quads would be carried
24 forward in the effort to fit the '85 image to the '93
25 image. You can't be more accurate?
70
1 A. No, you can't. You couldn't increase
2 accuracy.
3 Q. Theoretically you could increase the
4 accuracy but you wouldn't want to say that in looking
5 at your error you would look at the total range?
6 A. Yes, that is correct.
7 Q. What is your assessment of that error in
8 '85, the '85 imagery?
9 A. We didn't do a quantitative assessment. We
10 viewed the overlay of the two images and the overlay
11 of the control points and saw no error, no increased
12 error that affected our interpretation.
13 Q. So you have no specific estimate of the
14 total location error on the '85 image as the
15 methodology for '93?
16 A. We could certainly run an RMS error on the
17 '85 image. I don't know we did that. We probably
18 did do that. I just haven't looked at it. I could
19 probably produce that if you wanted it.
20 Q. What's your pixel size for '93?
21 A. 25 meters re-examined to 25 meters.
22 Q. That means that's not what you got in the
23 original digitized data?
24 A. We got it at 25 meters. We opened it,
25 re-examined it.
71
1 Q. The scanners in the satellite don't produce
2 its data in that?
3 A. That's correct.
4 Q. So you received reprocessed data to begin
5 with?
6 A. Yes.
7 Q. Any error induced by that reprocessing
8 effort?
9 A. There could be potential for some error.
10 Q. Can you quantify it for me?
11 A. No.
12 Q. Can anybody?
13 A. I would imagine EOSAT could probably do it.
14 Q. You never seen in the literature that talks
15 about that?
16 A. I can't recall any papers. I'm sure there
17 have been people that attempted to do that.
18 Q. Didn't I see in your earlier testimony a
19 criticism that Dr. Jensen had done re-examining and
20 was using a different pixel size than the original
21 data?
22 A. I don't think you saw that. I don't
23 believe I said that.
24 Q. Well, I guess my question is, since you did
25 it you must think it's an acceptable procedure?
72
1 A. I didn't. You misunderstood what I said.
2 Q. Forget Dr. Rutchey for the moment. You
3 used the re-examined knowledge or a process set of
4 data, preliminary process, to get it into the
5 exercise you wanted to employ for some reason?
6 A. We used the standard pixel size from EOSAT
7 Corporation. We did not resample the pixel size. We
8 did not change the resolution of the data.
9 Q. It had already been done for you?
10 A. A scanner does not produce a grid. A
11 satellite produces a single reference based on the
12 central point of a scanning mechanism.
13 Every device, every mechanism for
14 establishing the satellite's recognition of a pixel
15 location on the ground is modified through some
16 re-examining method. You have to do that.
17 A pixel -- in fact, the instant field of
18 view of a scanner is usually beyond the pixel size
19 produced in the data delivered from EOSAT. So every
20 data set, every scanned data set has to be
21 re-examined in some fashion.
22 So, yes, everything has to be re-examined.
23 It's not unusual, in fact it's required that the data
24 be re-examined in some fashion in order to be usable.
25 Q. Routine operation in the field?
73
1 A. Required operation by the provider of the
2 digital data, EOSAT, SPOT, CASEY, anybody else.
3 Q. Would you agree that the overall error for
4 the '93 image is probably in the range of plus or
5 minus 20 to 30 meters?
6 A. What now? Restate that. I'm sorry.
7 Q. Would you agree that the overall error in
8 the '93 image for WCA-2A is plus or minus 20 to 30
9 meters, it's on that order?
10 A. The overall error in the '93 image -- you
11 mean the image itself, the raw data?
12 Q. Yes.
13 A. I'm sorry. Now I'm trying to understand
14 exactly what you are asking. Are you asking if I
15 think that the data from EOSAT --
16 Q. The rectification error overall for the '93
17 image.
18 A. Was within how much?
19 Q. Plus or minus 20 to 30 meters.
20 A. Our rectification?
21 Q. Yes.
22 A. I would say my recollection is better than
23 20 meters. The RMS error that we generated will tell
24 you what the estimated error is and it's in the
25 digital tape. RMS error is standard product. And to
74
1 my recollection, I can't remember exactly what it
2 was, but I would say it's much better than 30 and
3 possibly better than 20. Is that what you are
4 asking?
5 Q. Yes.
6 A. Okay.
7 Q. You described for Mr. Cesarano in some
8 degree the last time and a little more today doing
9 the KT transformation, but I don't understand how
10 that's reflected in your work.
11 Can you explain for me how the KT formation
12 has assisted you in producing either the '93 or '85
13 imagery you will be utilizing in the case since you
14 went to great trouble to produce it?
15 A. You said the KT?
16 Q. What you called tasseled cap.
17 A. You called it --
18 Q. KT.
19 A. Which is what? What's KT?
20 Q. You never heard it referred to that way,
21 after the two guys invented --
22 A. You are talking about the researchers of
23 tasseled cap?
24 Q. You said tasseled cap was one narrow
25 application of a technique.
75
1 A. You are talking about their techniques or
2 the tasseled cap?
3 Q. Where did tasseled cap come from?
4 A. That has been around a long time.
5 Q. The question is, where did it come from?
6 A. Arum did a lot of research.
7 Q. I am sorry, who?
8 A. Arum. I seen it used at NASA's Earth
9 Research Center.
10 Q. You are not answering the question. Who
11 developed the technique that you are calling tasseled
12 cap?
13 A. I don't know who developed it. My
14 knowledge of tasseled cap is based on numerous papers
15 and mention of tasseled cap over the years.
16 Q. Which of those papers sanctioned or in any
17 way suggested the validity of employing tasseled cap
18 in a sub tropical wetland environment?
19 A. I wouldn't know which paper recommended
20 that or didn't recommend that.
21 Q. Is it fair to say, then, you are aware of
22 no paper that, in fact, identifies such use in the
23 past?
24 A. I can't think of a paper that I read that
25 used tasseled cap in a sub tropical environment.
76
1 Q. Have you ever used tasseled cap in a sub
2 tropical environment to assist in the analysis of
3 spectral imagery?
4 A. No.
5 Q. Whose idea was it to use it in this case?
6 A. Tasseled cap was suggested by Dr. Roy
7 Stein.
8 Q. Who is he?
9 A. He is a professor at Sanford University.
10 Q. What's his involvement in the case?
11 A. He has been employed by some of the parties
12 in the case to just look into it and see what's going
13 on. He has been asked to form an opinion about
14 certain pieces of this case.
15 Q. How do you know that?
16 A. He has told me he was employed to
17 investigate this process.
18 Q. Have you consulted with Dr. Stein?
19 A. Yes, I have.
20 Q. Has he reviewed your work?
21 A. He has reviewed our procedures. We
22 discussed our procedures with him and spent one day
23 looking at that, and he has reviewed it.
24 Q. When was the one day looking and reviewing?
25 A. About three weeks ago.
77
1 Q. Did he provide you any comments on your
2 work?
3 A. Yes.
4 Q. In writing?
5 A. No, just comments.
6 Q. Has he produced or written products, to
7 your knowledge, resulting from his review of your
8 work?
9 A. Nothing that I have seen.
10 Q. As a result of your communications with him
11 either by person or phone or otherwise did you alter
12 your work plan in any fashion?
13 A. We introduced the tasseled cap, yes.
14 Q. Did Dr. Stein -- I presume it's Dr. Stein?
15 A. Yes.
16 Q. Did he explain to you the environment in
17 which the process you referred to as tasseled cap was
18 developed?
19 A. No, we did not.
20 Q. As I understood your testimony the last
21 time, you had concerns over the use of the Eckert
22 technique by Dr. Jensen to enable him to do his time
23 normalized series?
24 A. Yes.
25 Q. And one of your concerns, as I understood
78
1 it, was that Dr. Eckert developed his technique in a
2 semi arid ecosystem; is that correct?
3 A. It was the area, yes.
4 Q. And that you were concerned -- deeply
5 concerned, I might say, if that's a fair
6 characterization -- over the transference of that
7 technique from a semi arid environment to the sub
8 tropical environment of South Florida?
9 A. Because it required targets, as I clearly
10 stated. It required clear targets over the study
11 period.
12 Q. Are you aware of any literature that
13 suggested that the same target must be used
14 throughout a time normalized series under the Eckert
15 technique?
16 A. My understanding of the technique was -- my
17 recommendation was to use a consistent target.
18 Q. From whence do you derive this
19 understanding? Have you ever yourself employed the
20 Eckert technique?
21 A. No.
22 Q. Who told you that you got to use a
23 consistent target throughout?
24 A. It seems like I remember in the paper that
25 the target should be consistent.
79
1 Q. Are you aware of anything suggesting that
2 you can't use different targets?
3 A. I'm not aware of a paper saying that you
4 can -- what? I'm sorry.
5 Q. Are you aware of anything saying that you
6 cannot use targets that are not consistent throughout
7 the entire time normalized series?
8 A. No.
9 Q. Were you aware that the technique you
10 referred to as tasseled cap was invented or developed
11 in the study of agricultural lands in the midwest for
12 large homogenous tracks of crop?
13 A. They were used in South Carolina and three
14 other states, from what I was aware of.
15 Q. Let me ask you the question again so you
16 can answer the one I asked. Were you aware that it
17 was developed in the midwest based on assessments and
18 studies of large area homogenous crops in
19 agricultural lands?
20 A. You say "developed." I'm not aware that
21 tasseled cap is a single development. Tasseled cap
22 is an ongoing development, from what I read.
23 Q. What are you relying on for that?
24 A. I have read a few papers on tasseled cap
25 and discussed it with Dr. Roy Stein. It's an ongoing
80
1 discussion.
2 Q. Did you read these papers in connection
3 with preparing to employ the technique in this case?
4 A. Yes.
5 Q. Please identify the papers you read since I
6 don't see them in the materials.
7 A. Just papers and literature. The titles of
8 the papers, is that what you want?
9 Q. Yes.
10 A. I would have to find them. I got two
11 papers I read on tasseled cap.
12 Q. Do you have copies of those in your
13 materials at your office?
14 A. Yes, I have copies of the reports, if you
15 would like those.
16 Q. If you relied on those and employed them in
17 assisting you in applying this technique in this
18 case, I certainly do.
19 MR. WATTS-FITZGERALD: Counsel, I request
20 they be provided.
21 MS. RAEPPLE: Have they not been provided?
22 THE WITNESS: I thought they were. I'm
23 surprised if they aren't. A copy of the entire
24 file was produced.
25 Research is an ongoing proposition, but I
81
1 would be surprised they are not in the file you
2 already got. If they are not, I'll provide
3 them.
4 MR. WATTS-FITZGERALD: I indexed the
5 material I got. I don't recall --
6 THE WITNESS: I'll provide them, two
7 studies out of ASPRS one. They are discussions
8 of tasseled cap.
9 BY MR. WATTS-FITZGERALD:
10 Q. What's ASPRS?
11 A. American Society of Remote Sensing
12 Photogrammetry and --
13 Q. Are you a member?
14 A. No, I'm not.
15 Q. I take it you are not certified in
16 photogrammetric interpretations by any professional
17 organization?
18 A. No, I'm not.
19 Q. The publication listed in Exhibits 13 and
20 14, your CV and update, I didn't notice if the update
21 had anything which -- if any of those are peer review
22 journal publications.
23 A. I have done no peer review journal
24 publication except for the review through URISA. No,
25 I'm not in peer review publication.
82
1 Q. What is URISA exactly?
2 A. Urban and Regional Information Systems
3 Association.
4 Q. What is it?
5 A. It's a large group of people interested in
6 information systems, have been around for quite a
7 while.
8 Q. They have a publication?
9 A. Yes.
10 Q. You have contributed to that?
11 A. I presented many papers, as the resume
12 indicates. They have an annual conference similar to
13 ASPRS.
14 Q. Do they produce written results?
15 A. Usually they do.
16 Q. There are outlines of presentations?
17 A. Every year conference proceedings are
18 published.
19 Q. But those are not peer review, they reflect
20 whatever was presented?
21 A. It's reviewed by a group at URISA but not a
22 peer review.
23 Q. You said the tasseled cap could be a tool
24 to distinguish between sawgrass and cattail. That's
25 what you said this morning?
83
1 A. Yes.
2 Q. In what fashion in your estimation could it
3 be such a tool?
4 A. Several ways. One is to actually perform a
5 classification on a tasseled cap image and then you
6 see what those classes are.
7 Another method would be to compare
8 brightness values in a tasseled cap image to a
9 classified image brightness green. You compare
10 brightness, greenness or wetness values to a
11 classified image.
12 Q. So that's how it could be a tool?
13 A. Yes.
14 Q. Have you used it as a tool in either of the
15 two faxes you have described?
16 A. We produce the tasseled capped image
17 combining '85, '87, '89 and '93. I have used various
18 combinations of wetness, brightness and greenness
19 among those images and compared those images to our
20 classification for '85 and '93.
21 Q. I need to break that down. I'm not sure I
22 understand what you did. You merged together the
23 scatter plots, if you will, of brightness and
24 greenness and wetness for all of those years into a
25 single product?
84
1 A. We produced one 12 band image that included
2 '85, '87, '89 and '93.
3 Q. Using what satellite data?
4 A. The '85 the '87 PM image, '89 and '93 PM
5 image.
6 Q. Did you use all bands?
7 A. No. We used the tasseled cap modeling
8 capabilities with the software to produce three
9 bands. In each of those three images, the bands
10 being where -- the new bands being brightness,
11 greeness and wetness.
12 Q. When you say you produced the 12 band
13 image, what form did that take?
14 A. It's a digital image.
15 Q. Did you print that out?
16 A. It's on that tape right there. There are
17 photographs of it.
18 Q. The photographs are taken off the screen?
19 A. Yes.
20 Q. The question was, did you hard copy it?
21 A. No, we produced no hard copy of tasseled
22 cap at this point.
23 Q. Do you plan to?
24 A. If requested to do so, I will.
25 Q. And then as I understood what you said, you
85
1 compared that 12 band image to your '85?
2 A. Various components of the 12 bands image to
3 our classified '85 image and classified '93 image.
4 Q. What components of the 12 band image did
5 you compare to the '85 imagery?
6 A. Greeness and brightness and wetness.
7 Q. What were your conclusions as a result of
8 that comparison?
9 A. Our conclusion is that there is no
10 substantial change in greeness between '85 and '93.
11 Q. You have added in '93. You jumped ahead a
12 little bit. As I understand it, you did the same
13 comparison to your '93 classified image?
14 A. The tasseled cap image is a combination of
15 all of those years. So the comparisons are intrinsic
16 in the image.
17 Q. If you are comparing the 12 band image that
18 includes the two years in which -- and the same
19 imagery data that you have for '85 and '93, aren't
20 you already biased towards limiting any variation or
21 change in the greeness values?
22 A. Could you repeat that question? I do not
23 understand it.
24 Q. You took the three bands from your '85 TM
25 satellite range and put that in the 12 band?
86
1 A. I took the 36 bands excluding thermal and
2 combined that to a three band tasseled capped image.
3 Q. For '85?
4 A. Used that as part of the 12 band composite
5 image.
6 Q. The three bands are -- used tasseled caps
7 out of '85 are derived from the '85 data which is
8 used to generate your classified '85?
9 A. From the raw '85, the unclassified '85.
10 Q. Doesn't that inevitably suggest that you
11 are minimizing any variance? You are taking '85 and
12 '93 and using the same tasseled cap converted as part
13 of the comparison?
14 You don't understand what I mean?
15 A. No, I don't understand.
16 Q. I have reviewed my notes on the materials
17 provided. I can't find any indication of the fact of
18 the publications you referred to.
19 A. At the first break I'll find my copies.
20 Q. I don't know if you need to make the copies
21 if you can get me titles and reference.
22 A. You want that? Okay.
23 Q. In the efforts you described to determine
24 the 12 band image, what original TM data channels
25 were used to produce the 12 band image?
87
1 A. We used all bands in each image except the
2 thermal band.
3 Q. You said earlier that this could be a tool
4 in distinguishing between cattail and sawgrass based
5 on what you just said regarding how you employed it.
6 I don't understand you to say that you, in fact, did
7 that or attempted to do that of your analysis of the
8 '93 image, but you only attempted to determine if the
9 greeness values showed the significant shift, and you
10 concluded the answer to that was no.
11 Am I correct in my understanding?
12 A. You said a lot in that statement. Are you
13 saying -- I'm not sure what you are asking. Please
14 tell me what you are asking about the greeness.
15 Q. As you applied the analysis or the
16 comparison of your TC image to the '85 and '93
17 classified images, did you form an opinion
18 specifically about any alterations in sawgrass and
19 cattail over that time span, those species?
20 A. From the tasseled cap, singularly?
21 Q. Yes.
22 A. No.
23 Q. You told us three times now that you could
24 use the tasseled cap analysis to make such a
25 distinguishing judgment?
88
1 A. That it could be a tool in making the
2 judgment, yes.
3 Q. Have you done that?
4 A. Yes.
5 Q. How?
6 A. Visually we have compared -- we have
7 produced numerous combinations of brightness,
8 greeness and wetness from the 12 band image, and we
9 have produced those and observed greeness in
10 particular and then compared that greeness to our '85
11 classification and the '93 classification.
12 Q. And did I understand you to say the
13 greeness values from your tasseled cap '85 to '93
14 don't show any significant change?
15 A. There is not a major significant change in
16 the areas that -- the significant change -- there is
17 change I guess is significant up to the user. It
18 shows change in areas that we did the classification
19 on to see, in fact, what the greeness was at that
20 point. It showed changes in greeness. It indicates
21 greeness values in those two years.
22 Q. Can you quantify the changes you observed?
23 A. We did not attempt to quantify the changes
24 in the tasseled cap image.
25 Q. Do you plan to?
89
1 A. We could if asked.
2 Q. I take it that's a no. The question is, do
3 you plan to. Not could you do it. If somebody asked
4 you, do you plan to do it?
5 A. I'm doing what my client asked me to do. I
6 do what I'm asked to do. I can do it or not do it.
7 I'm not sure if I understand.
8 Q. Let me ask the question. Has anybody asked
9 you to do that?
10 A. No, not yet.
11 Q. Do you of your own sua sponte plan to do
12 that?
13 A. No.
14 MR. WATTS-FITZGERALD: Counsel, if you ask
15 him to do that, obviously we feel entitled to
16 certain notice on that further deposition.
17 MS. RAEPPLE: Certainly.
18 BY MR. WATTS-FITZGERALD:
19 Q. In your analysis where you found some
20 unquantified as yet shift in greeness values for 1985
21 imagery to the 1983 imagery --
22 A. '93.
23 Q. -- how much of that greeness shift can you
24 attribute within a reasonable degree of scientific
25 certainty to expansion of cattails over that time
90
1 frame?
2 A. How much of the changes in greeness can I
3 attribute to expansion of cattail?
4 Q. Yes.
5 A. The acreage figures that we produced would
6 be that combination. Those are digital files that
7 can easily be overlaid to each other. If you want to
8 quantify it pixel by pixel, we can do that.
9 Q. Let me make sure I'm clear on this. The
10 original calculations you are talking about, that's
11 based on the tasseled cap?
12 A. No, it's based on the classification.
13 Q. You are answering a question I didn't ask.
14 I'm talking about tasseled cap, based on tasseled cap
15 and the shifting greeness values, which has nothing
16 to do with your accuracy assessment.
17 In your acreage assessment in Exhibits 10,
18 11, 12, you have said you have observed a greeness
19 value shift but you can't quantify it, but there is a
20 shift over those years '85 to '93.
21 You told me the tasseled cap is a tool to
22 distinguish between cattail and sawgrass in the
23 greeness for area shifts.
24 How much of it is attributable to the
25 increase in cattail?
91
1 A. I did not tell you that tasseled cap would
2 quantify species. I said it was a tool used to help
3 assist us in quantifying the species attributable to
4 a particular species. That's what I said. Tasseled
5 cap is not -- we did not assign vegetative classes to
6 the tasseled cap images.
7 Q. Do you plan or did you attempt that in any
8 fashion thus far?
9 A. No, we have not.
10 Q. Has anyone directed you to do that?
11 A. No.
12 Q. Do you have any plans to do that at this
13 juncture?
14 A. At this point, no.
15 Q. Would you have to go through the same
16 process with color IR or would you conduct that
17 effort by reference to your classified '85 and '93
18 images?
19 A. You could do it either way.
20 Q. Which way would you do it?
21 A. I would compare it to the classified
22 images.
23 Q. It's easier. You have already got that
24 digitized, right?
25 A. We do have that.
92
1 Q. Cut Dr. Lodge out of the loop, he doesn't
2 have to sit there on the light table.
3 A. I would compare it to the classified
4 images, yes.
5 Q. So as of right now, as I understand it --
6 correct me if I'm wrong -- you have no opinion based
7 on tasseled cap as to the relative change or
8 distribution within those greeness values for '85 and
9 '93 of cattail and sawgrass?
10 A. That's correct. We did not classify the
11 tasseled cap based on species type.
12 Q. So what opinions have you formed and
13 developed based on your tasseled cap analysis that
14 are final at this time?
15 A. Tasseled cap similar to the unsupervised
16 classifications we did increase our general knowledge
17 of the study area.
18 Q. And as I understand it, it has been
19 suggested to you that there has been some change in
20 the coverage of certain species. The extent of that
21 is not quantified in terms of, say, acreage?
22 A. Not quantified in the tasseled cap but in
23 the '85 or '93 image, yes.
24 Q. We were speaking in terms of cattail and
25 sawgrass. But there are, according to Dr. Lodge's
93
1 book, something on the order of 190 other vegetative
2 species in the ecosystem at issue here.
3 How would you distinguish out from the
4 tasseled cap analysis as you have done it thus far,
5 unclassified other species of flora that have their
6 own associated greeness and brightness values?
7 A. If we did that, we would do classification
8 of the tasseled cap image.
9 Q. That's the only way to distinguish that
10 out?
11 A. We could overlay it to our existing
12 classification and see if there was correspondence.
13 Q. Have you done that, even on the screen?
14 A. No.
15 Q. Other than the photos, we have all of your
16 tasseled cap analyses on this back-up data?
17 A. That 150 megabyte tape contains the
18 complete tasseled cap model that we used.
19 Q. What system was the 150 meg produced on?
20 A. A SUN micro station work station. SPARC,
21 2, 150 megabyte, quarter inch tape unit.
22 Q. What is the tape format?
23 A. Probably TAR format. TAR back up. If it's
24 not, call us and we will tell you. I think it's TAR.
25 Q. Assuming for the sake of argument for this
94
1 question that the quantification of the greeness
2 change is relatively minor. Do you have any way
3 currently on the work you have already completed to
4 assess the greeness values you had developed from the
5 TC effort?
6 A. William Thomas or Bill Thomas, that's the T
7 you are talking about, TC tasseled cap. KT is the
8 other one.
9 Q. Do you have any way of assessing within a
10 reasonable scientific certainty whether or not the
11 mix of cattail and sawgrass species within the
12 greeness value has altered the proportion of those
13 species, if you will, from '85 to '93 based on the
14 work already completed?
15 A. Based on what we already completed today,
16 is that what you are asking?
17 Q. Yes.
18 A. Yes, I believe it did.
19 Q. On what do you base that belief?
20 A. As previously stated, we could overlay the
21 tasseled cap brightness categories and assign species
22 typed to the tasseled cap.
23 Q. In your answer just now you said "could."
24 My question was "done." You have not done that then?
25 A. I thought you said "could we."
95
1 Q. No. I am assuming the difference is very
2 small. Can you right now based on the work already
3 done tell me what the ratio shifts are or what shift,
4 if any, in the relative composition within the
5 greeness values is accounted for from '85 to '93 by
6 cattail and sawgrass?
7 A. By "can," you mean have I produced that
8 product already? My interpretation of "can" was
9 could I take the data I got and come up with that
10 answer.
11 Q. It's a semantic thing. Put the question
12 this way.
13 Tell me right now what the ratio shifts
14 were based on the greeness evaluation from '85 to
15 '93.
16 A. I understand your question now. We have
17 not done that as of right now.
18 Q. It was the way I asked the question. We
19 just look at the words a little differently.
20 A. Okay.
21 Q. The software used to produce the tasseled
22 cap analysis was still ERDAS?
23 A. ERDAS, yes.
24 Q. What version?
25 A. This model was run in the current release,
96
1 which was 8.1.
2 Q. The demonstration tape that was produced as
3 part of your discovery, it showed Mr. Erickson
4 playing with the monitors doing his golly, gee whiz
5 stuff. That was a later version not yet released?
6 A. That was one, the beta version, probably
7 8.02.
8 Q. Earlier for Mr. Cesarano you were talking
9 about the 2A image and said that you refined the
10 categories as part of your very recent work?
11 A. Yes.
12 Q. Were you talking about '85 or '93?
13 A. '93.
14 Q. How did you refine the categories?
15 A. We observed that we could break out
16 predominant -- which one was that?
17 Q. The next one down.
18 A. We determined in reviewing these classes
19 that we could break out predominant cattail open
20 water and predominant sawgrass open water. We added
21 those to the classes that we exhibited in the legend.
22 Q. Did that require you to go back to the
23 field notes?
24 A. Let's see. We reviewed field notes
25 continuously through this. We probably did, yes.
97
1 Q. And who actually conducted that exercise to
2 class the categories?
3 A. I did.
4 Q. Assisted by anyone?
5 A. Pamela Green and Terry Jackson.
6 Q. Dr. Lodge was not one of those?
7 A. Dr. Lodge has been involved in quite a bit
8 of discussion. Yes. Dr. Lodge would have had some
9 input into this, yes.
10 Q. When I look at the image, is it fair to say
11 that everything I see on the image is as a result of
12 platting out of the digitize data that has been
13 classified?
14 A. I am not sure I understand. As a result of
15 what?
16 Q. As I understand the image, it is intended
17 to convey through the color schemes and
18 classification process information to the reviewer?
19 A. Yes.
20 Q. So everything depicted on there has
21 significance to the viewer, even if it's down to the
22 cloud category which tells you what was under it?
23 A. Yes.
24 Q. If I see banding on any of these images
25 that implies to me it's in the digitized data or it's
98
1 an artifact of the platting; is that correct?
2 A. I would say that's right.
3 Q. What is the inclination to its orbital path
4 for the satellite of the sensors employed to produce
5 these two images, the '93 images?
6 A. What?
7 Q. Inclination of the sensors to the orbital
8 path of the satellite?
9 A. Inclination of the sensors to the orbital
10 path --
11 Q. What satellite are we talking about?
12 A. Landsat 5.
13 Q. Landsat 5's sensors, are they normal to the
14 footprint of the satellite?
15 A. Are they off NADER, are they pointed other
16 than straight down? Is that what you are asking?
17 Q. Yes.
18 A. They are pointed straight down. Landsat 5
19 does not have a pointing capability. Therefore, the
20 sensors are always aligned in the same direction from
21 the satellite, if I understand your question.
22 Q. Then if there were -- if banding were
23 present you would expect it to be consistent
24 throughout the areas that you mapped if there were
25 banding as a result of the sensor?
99
1 A. Expected to be consistent within that
2 image, yes.
3 Q. If you look at the image for the entire
4 study area, do you see banding in what would be the
5 Loxahatchee National Wildlife Refuge?
6 A. I do see what could be interpreted as
7 banding in the upper portion of Loxahatchee. If you
8 are speaking of this area here, is that what you are
9 talking about, here?
10 Q. The entire area of Loxahatchee.
11 A. I do not see indications of banding in the
12 southern area or 2A.
13 Q. When you hold the paper chart and look at
14 it end on, very close to end on, you still don't see
15 it?
16 A. We printed these on a -- these are color
17 electrostatic plats, and the color electrostatic
18 platting that we are using prints in rows and our
19 observation of the image on the screen was that there
20 was no significant banding.
21 I can see what you are talking about there
22 and that does have the appearance in the upper
23 Loxahatchee of possible banding in the image, but we
24 never observed it in the image on the screen and I
25 cannot tell you where that came from.
100
1 Q. I guess the problem I have when I look down
2 at WCA-3A on that image I see bands from whatever
3 source running more or less parallel east/west. And
4 when I look at Loxahatchee on the same image,
5 certainly in the northern half of it above the cloud
6 cut out, I see what appear to be bands running
7 northwest, slightly southeast.
8 If there is an artifact of your platter, I
9 would expect them to be consistent throughout the
10 entire range of the imagery. It was platted at one
11 time?
12 A. This is what you are talking about, right
13 here.
14 Q. Show me what you are talking about in the
15 image.
16 A. I see the banding effect right here. Since
17 I think you are wrong on the satellite you are not
18 interpreting what I'm talking about, the platter
19 effect which is straight horizontally.
20 Q. Down in 3A we are talking east, west or --
21 A. All right. Let me see this down here. In
22 this part of the image, what are you talking about,
23 right here.
24 Q. East/west, there are triangles, if you
25 will, in the imagery?
101
1 A. You mean this little light thing here?
2 Q. No. I see it continuously.
3 A. Where do you see it?
4 Q. I see it like every 16 lines.
5 A. Every 16 lines?
6 Q. How many sensors are there on a Landsat?
7 A. There are 7, 7 bands. I can see something
8 in this image. I do not know -- I cannot attribute
9 why that's there.
10 Q. And we can both see something in the
11 Loxahatchee part of the image?
12 A. I see something up there, yes. There is
13 some.
14 Q. Do those striations which I use, because it
15 doesn't have a normative value what the cause is,
16 they don't all run in the same direction, which is
17 roughly parallel to those in 3A, and it seems to me
18 this is an artifact of the printer. You just
19 suggested your printer prints in rows, which I assume
20 is roughly east/west normal to the paper?
21 A. Yes.
22 Q. That would not seem to account for
23 striations for Loxahatchee. I want to know what they
24 are. Are they an artifact, are they in the data or
25 is there some problem in the classification we have
102
1 not considered?
2 A. Is your question, do I know what that is?
3 Q. Yes, or can you explain that for me.
4 A. The artifact you are observing is something
5 either in the data of the platter, it looks like from
6 the angle of it. It's in the data. At least in
7 northern Loxahatchee during our classification
8 process we observed no banding that caused us
9 difficulty with the classification.
10 Frankly, I don't see -- I don't observe
11 that banding at that angle referred to, I don't
12 observe that in the southern study area. I don't see
13 it. I see perfectly horizontal lines and I do see
14 that slightly off horizontal artifact you referred to
15 here. But I do not see that here.
16 Q. Did you process the portion of the data
17 involving Loxahatchee in some way different from the
18 balance of the imagery as a result of the cutting
19 into the middle of Loxahatchee?
20 A. That was not separated out as an area of
21 interest or digitized out in any fashion from the
22 classification that occurred in this portion of the
23 study area around 2A. It was not treated separately.
24 Q. Other than the tasseled cap effort, has any
25 classification been done on images other than '85 and
103
1 '93?
2 A. Yes. We did an unsupervised classification
3 on '87 and '89.
4 Q. As I understand what you said, while you
5 classified them to the 36 and 99 class, no effort was
6 made or expended to identify vegetative species with
7 those classes?
8 A. Correct.
9 MR. WATTS-FITZGERALD: He has a call to
10 make and I'm at a good stopping point. It's a
11 quarter to 12. I would like to review these
12 materials before we continue. So if we can take
13 an early lunch and get back early --
14 MS. RAEPPLE: Do you want to take a short
15 lunch?
16 MR. WATTS-FITZGERALD: Let's come back at
17 12:45.
18 (Thereupon, a lunch recess was taken,
19 after which the following proceedings
20 were had.)
21 BY MR. WATTS-FITZGERALD:
22 Q. Your imagery for 1985, that is complete and
23 final as I understand it; is that correct?
24 A. Yes. Yes, it is.
25 Q. Did you yourself do any of the photo
104
1 interpretation in the process of developing the
2 classified images?
3 A. No, not directly. I looked at the photos
4 and gained as much familiarity as I could. But that
5 was left for Dr. Lodge.
6 Q. You stated you were going to compare Ken
7 Rutchey's procedures and results to yours or to your
8 effort.
9 Other than the limited discussion we had
10 about comparing acreages and classes and whether you
11 could overlay them or not, how you would do that?
12 Have you done anything else to analyze and compare
13 your methodologies of the final product?
14 A. To Ken Rutchey?
15 Q. Yes.
16 A. You know the GPS procedures. I stated some
17 differences in those, the way he produced classes.
18 We looked at those things, yes, if that's what you
19 are asking, yes.
20 Q. What is your opinion on the level of
21 positional error that might exist in Ken Rutchey's
22 work based on your understanding of his GPS
23 procedures?
24 A. What is my estimation of his accuracy?
25 Q. Yes.
105
1 A. I think in those -- it's really hard to say
2 post processed data. I would say he may be close to
3 what he said he was when he post processed the data.
4 My question would be how he obtained
5 information about the site when he was doing the
6 accuracy assessment. If he did have to come back and
7 post process the data and had to move that location
8 to another location, how he correlated that field
9 research with the point in terms of defining, you
10 know, the vegetative mix around the helicopter.
11 Q. You think he did all of his accuracy
12 assessment field work from the helicopter?
13 A. He stated he did a mixture of boats and
14 helicopters. He wasn't sure of the percent.
15 Q. The field points, the 230 points that were
16 generated by his random stratified sampling program
17 for accuracy assessment, are you aware of any size
18 constraints that would act to limit the possibility
19 of being in an erroneous field position in the way
20 they could affect the result prior to post
21 processing?
22 A. Am I aware of the positional error that
23 might have occurred -- one more time.
24 Q. How the point was defined, the accuracy
25 assessment point, to minimize any potential problem.
106
1 A. You mean the size of the matrix he chose
2 around the area?
3 Q. Yes.
4 A. I'm aware that he chose a matrix around the
5 area.
6 Q. Do you know how large?
7 A. I believe he chose -- let's see. I believe
8 that was 3 by 3 he chose in the field work. I'm not
9 certain.
10 But the non-post processed GPS data could
11 easily be 100 feet. I think he mentioned in the
12 paper that SA was sound, so he had to adjust those
13 sites. I assumed SA was sound.
14 Q. You have the downloaded satellite almanac.
15 Was it on or not?
16 A. The almanac doesn't tell you whether it's
17 on or not.
18 Q. Where would you look to find that?
19 A. The government, the defense mapping agency
20 will not tell you when SA is on and when it's off, to
21 my knowledge. The reason they use selective
22 accessibility, SA, is to intentionally confuse the
23 signals from the satellite. If they issued a -- the
24 purpose of SA is to confuse the signal. If you
25 generate the almanac in advance of the mix so there
107
1 would be no way to know when you generate the almanac
2 if SA is not on. The user is not notified when the
3 SA is turned on or not.
4 Q. In your accuracy assessment effort did you
5 exercise any techniques to attempt to account for
6 variability and results induced by the phrenology of
7 the flora as a result of the time interval between
8 the satellite passes and the color IR photography?
9 A. Other than Tom Lodge's knowledge of the
10 plants themselves, we took no mechanical means in the
11 image to adjust for that. Tom would do that in the
12 photo interpretation.
13 Q. How do you understand him to have done
14 that?
15 A. His knowledge of the plant material. He
16 can recognize cattail in sawgrass communities because
17 of the difference in the IR response.
18 Q. Even in the 1 to approximately 65,000 CIR
19 from 1985?
20 A. Yes.
21 Q. Well, I thought you told me that the reason
22 you classed classes was that fine distinctions became
23 very difficult to make in mixed communities in that
24 photography.
25 A. Yes.
108
1 Q. You testified you had 116 total observation
2 sites. This is your testimony last week. Would that
3 have been training sites, the total training sites,
4 116 --
5 A. Yes.
6 Q. -- we couldn't remember before lunch?
7 A. Yes.
8 Q. That extended throughout the area of study?
9 A. Yes. Some of those were removed. If, in
10 fact, the -- if the GPS was not sufficient or we
11 could not locate where it was with certainty, some of
12 those we did not use. So 116 original observations
13 to be used in this process. And through the process
14 of producing classified image, some number dropped
15 out.
16 Q. Do you know how many?
17 A. It's on the additional tape I sent you. I
18 don't know exactly how many.
19 Q. More or less than 10%? Do you have any
20 feel for it?
21 A. I guess it would be 10%.
22 Q. The approximate 12 to 18 observations you
23 testified last week were identified in the
24 Loxahatchee refuge were made from a helicopter in the
25 air; is that correct?
109
1 A. Yes.
2 Q. Were you on Board the helicopter at the
3 time?
4 A. Yes, I believe so.
5 Q. How high was the helicopter off the ground?
6 A. We were limited to 500 or 1,000 feet. The
7 pilot set the altitude. Whatever the limit that the
8 Park said was the limit. I think it was 500 or 1,000
9 feet.
10 Q. What's your assessment of the inaccuracies
11 induced in the GPS coordinates as a result of
12 attempting to figure the position of those ground
13 truth points or air hover points, whatever you want
14 to call them?
15 A. It could be substantial.
16 Q. Substantial is an adverb or adjective. Can
17 you qualify that for me?
18 A. It could be a matter of 20 meters or 100
19 meters.
20 Q. That's a differential correction?
21 A. You are talking about the accuracy of
22 hovering over where you think you are with GPS.
23 Q. I'm assuming so. Maybe I'm wrong here.
24 When you testified those were observation points, you
25 used those to train your algorism for classification?
110
1 A. Yes.
2 Q. So you needed a GPS position to go with
3 that?
4 A. Yes.
5 Q. You took that while hovering for some
6 period of time in the air?
7 A. Yes.
8 Q. They were manifestly going into a number of
9 sources of error in that.
10 Were you able to hover in a fixed spot for
11 the same period of time that was employed at the
12 ground stations to allow averaging of the
13 constellation mix?
14 A. How long? Enough to allow averaging. I
15 would not call the quality of the hovering point
16 equal to the quality of the ground.
17 Q. That's what I understand. I'm asking, can
18 you quantify that either pre or post -- post
19 processing?
20 A. We didn't quantify them. You probably
21 could, yes. We did not quantify the error in the
22 hovering points.
23 Q. With so relatively few air hover truthing
24 points in Loxahatchee in comparison to WCA-2A, is the
25 classification for Loxahatchee largely a result of
111
1 the extrapolation of spectral signatures from other
2 areas within the study area?
3 A. Other signatures outside of the study area
4 did have an influence on the Loxahatchee
5 classification, yes.
6 Q. And in the way the algorism functions and
7 the way you trained the algorism, would the fact that
8 the bulk of -- I think the majority, almost all of
9 the tracking sites were in 2A, meaning that 2A would
10 have the biggest impact in the training regime on 1?
11 A. That's possible, yes.
12 Q. If the vegetative communities in those two
13 adjacent areas are, in fact, significantly different
14 based on whatever natural factors and unnatural
15 factors are operating, does that have the potential
16 for substantially reducing the reliability of the
17 classified image for Water Conservation Area 1?
18 A. If the vegetative communities were
19 different would it reduce the reliability of the
20 classification in Loxahatchee?
21 Yes, it would.
22 Q. In the two matrixes you have provided for
23 accuracy assessment, I see no footnotes or any
24 commentary that addresses that or attempts to account
25 for it.
112
1 Was any effort made to account for that or
2 consider that facet?
3 A. Notes in the accuracy assessment table.
4 Q. There is nothing in the matrix. You have a
5 warning on one of the charts unverified, outside of
6 2A, or something to that effect.
7 I have seen nowhere in the materials any
8 effort. Of course I haven't been able to go through
9 the tape, but any effort to take into account the
10 effects of signature extrapolation, was that done?
11 A. No.
12 Q. Do you have any current plans to do that?
13 A. I have not been directed to do that.
14 Q. As I understood from your earlier testimony
15 there were 58 ground sites, training sites in 2A?
16 A. Okay.
17 Q. If you had used all 116, assuming for the
18 moment, we don't know exactly how many fell out that
19 would have been half of the total sites initially and
20 then some fell out, do you know if any of the 58 in
21 2A fell out of the training sites?
22 A. A few in 2A did fall out.
23 Q. It was across the area?
24 A. Yes.
25 Q. You stated with respect to Dr. Jensen's
113
1 report to the Water Management District, Exhibit 8
2 from the last session, that in your view it was
3 impossible to do an accuracy assessment on his work
4 and certainly none accompanied it.
5 Are you still of that view?
6 A. I'm sorry, which document.
7 Q. Dr. Jensen's report to the Water Management
8 District.
9 A. Yes.
10 Q. Are you still of the view that it is
11 impossible to do an accuracy assessment for such an
12 effort?
13 A. On the historic MSS data that he
14 classified?
15 Q. Yes.
16 A. As I stated before, if he had photography
17 for that time, then he could do an accuracy
18 assessment in the same fashion we did. I believe I
19 stated he did not do an accuracy assessment.
20 Q. You expressed a concern with respect to
21 Dr. Jensen's work that he use the two different
22 platforms with very little overlap in the band of
23 those platforms.
24 You were at Dr. Jensen's initial
25 deposition, were you not?
114
1 A. One day of it.
2 Q. And have you had an opportunity to review
3 the transcript of that deposition since then?
4 A. No, I have not.
5 Q. Do you recall his testimony regarding the
6 extent of band overlap?
7 A. Yes.
8 Q. Do you remember the reference to his
9 textbook -- actually it may have been the draft of
10 his revised textbook, the second edition, regarding
11 the actual band waive length for those respective
12 platforms quantifying on delineating the overlap?
13 A. I remember a discussion about band overlap,
14 yes.
15 Q. I guess my question, then, would be, my
16 notes indicate you said there was very little overlap
17 in the bands. Is it not a fact that in some of the
18 bands there was perfect overlap?
19 A. My statement, if I remember correctly, was
20 that the band 3 and 4 in MSS do not match band 3 in
21 SPOT, and they don't.
22 Q. That's taking into account the renumbering
23 of the bands?
24 A. The bands in SPOT do not only have a
25 partial overlap with bands 3 and 4 in MSS.
115
1 Q. Do you disagree with Dr. Jensen's
2 assessment, expressed during his deposition at which
3 you were present that the overlap was sufficient for
4 his purposes and sufficient to allow use of it as a
5 tool in satellite imagery processing?
6 A. He said it was sufficient in his work to
7 adjust signatures between the two satellites, not
8 just in general image processing.
9 Yes, I disagree with that.
10 Q. And on what do you base your assessment
11 that the overlap was insufficient?
12 A. The overlap, I didn't say, was
13 insufficient. I said there was an overlap.
14 Q. You just told me you don't think it was
15 appropriate for him to do it that way. On what basis
16 are you asserting that the overlap that did exist
17 which you have acknowledged existed to some degree
18 was not sufficient to allow that to appropriately be
19 used as it was by Dr. Jensen?
20 A. Because it is only a partial overlap
21 between one band and two bands. It's confused also
22 by the pixel size of the two satellites.
23 Q. Is it ever appropriate to use less than
24 perfect band overlap? Is there a point at which you
25 would be comfortable in using data such as that that
116
1 doesn't have a perfect identity of band width?
2 A. Are you speaking of image interpretation?
3 Are you speaking of trying to transfer a set of
4 spectral signatures from one satellite to another?
5 There is a difference.
6 Q. Start with image interpretation.
7 A. Image interpretation, you could probably
8 account for some degree of band overlap or lack of
9 total band overlap.
10 Q. How much?
11 A. I don't know.
12 Q. You never researched it?
13 A. No, I have not researched it.
14 Q. Are you aware of any literature addressing
15 that?
16 A. Addressing bands? Most of the tasseled cap
17 mentioned something about that, talks about the fact
18 between MSS and TEM band overlap or the combination
19 of those two satellites is usable. They make no
20 mention of SPOT.
21 Q. Anything to suggest that SPOT is not
22 usable? I mean, if you can overlap the other two
23 methodologies?
24 A. The difference in my statements, as I tried
25 to make clear in the deposition before, was that in
117
1 order to produce a set of signatures that are going
2 to be transferred from one satellite to the other,
3 the number of variables should be decreased.
4 When you have two bands in the MSS
5 satellite platform that have marginal overlap with
6 one band in the SPOT platform you have to make an
7 assumption of what something would look like within a
8 band that is different at different portion of the
9 spectral -- of the reflectance arena than the band
10 from which you produce the original signature.
11 Q. You just expressed a value judgment on the
12 overlap by characterizing it as marginal overlap. I
13 wanted you to tell me on what basis do you make the
14 judgment about the differences in overlap.
15 When is it only marginal, when is it
16 adequate? That's the area I'm trying to illuminate
17 and I'm not getting that through to you -- or at
18 least the answer doesn't seem to address what I'm
19 trying to find out.
20 A. I'm not sure I understand your question.
21 Q. Is there a number in your mind at which you
22 have only marginal overlap between two bands, whether
23 it be TM and MSS or TM and SPOT or MSS and SPOT?
24 Is there a point where you are saying
25 that's marginal in my mind, I wouldn't use it or it
118
1 can't be used?
2 A. I have never quantified what percent band
3 overlap between SPOT and MSS is appropriate because,
4 as I stated earlier, I did not attempt to do any
5 signature transformation between two satellite
6 platforms.
7 Q. You never attempted to do that?
8 A. No.
9 Q. Are you aware of anybody in the literature
10 doing that?
11 A. The only person as I stated before that I'm
12 aware of doing that is Dr. Jensen.
13 Q. You mentioned tasseled cap and some
14 literature on that discussing TMM and MSS.
15 Dr. Jensen didn't use tasseled cap, that's not
16 directly applicable?
17 A. Is what directly applicable?
18 Q. The literature you were mentioning or the
19 discussion of tasseled cap and the use of TMM and
20 MSS.
21 A. I'm sorry, I don't understand what you are
22 asking. Does it have anything to do with the Eckerd
23 method of using creating historic signatures?
24 Q. No. We are talking about your comment
25 about the literature in tasseled cap discussing the
119
1 use of the two different methodologies. Let's leave
2 that for a second.
3 When you were preparing for your
4 deposition, did you review all of the documents that
5 were provided in your production of materials?
6 A. Is that what he gave me?
7 Q. That's what you gave me. Who provided you
8 the article on the expert witness and litigation?
9 A. That's a corporate piece of literature.
10 Q. Have you read it?
11 A. Frankly, no, I haven't.
12 Q. What are the wave ranges -- wave length
13 ranges in each band of the Landsat satellite?
14 A. You have to provide that -- which Landsat,
15 which one, MSS or TM?
16 Q. I'm going to ask both. Start with TM.
17 A. I don't know those offhand.
18 Q. Same for MSS?
19 A. What?
20 Q. Same answer for MSS, you don't know the
21 band?
22 A. I know the band. I mean, I know red,
23 green, blue bands are the ones I use the most. I
24 don't know the frequency response.
25 Q. Is the same thing true with respect to
120
1 SPOT?
2 A. Can I quote you the point to point in each
3 band, is that what you are asking?
4 Q. Yes.
5 A. No, I can't quote that offhand.
6 Q. What did you mean when you said -- I
7 started to ask this earlier today and we got side
8 tracked. Dr. Rutchey's map was biased towards
9 cattail and Jensen's is biased towards sawgrass?
10 A. Based upon my interpretation of their
11 deposition.
12 Q. Were you aware Dr. Jensen refined his work
13 and substituted or altered his report to reflect
14 additional classification efforts?
15 A. No, I wasn't aware of that. He changed his
16 classification?
17 Q. Well, what was it about his work that held
18 you to believe there was a bias towards sawgrass?
19 A. His own statements about his methods used.
20 I described those to my understanding in the last
21 deposition. Has his work been revised?
22 Q. You have to wait until you depose me for me
23 to answer questions.
24 Did you try and find any map photography
25 for use with your '85 image from different seasons?
121
1 Were there any other available high altitude
2 photography?
3 A. In the NAP photography?
4 Q. Yes, other than the two that you used.
5 A. It seems like the NAP started after '85.
6 Q. One of the imageries was before the data?
7 A. NHAP and there is NAP. There are different
8 ones.
9 Q. You used NAP or NHAP?
10 A. National Altitude -- NAP, was a later
11 program that started after '85. We looked at both
12 indices and NHAP was the only one available that I
13 remember for the year '85.
14 MR. WATTS-FITZGERALD: Can we mark this as
15 Exhibit 15 and 16. 15 will be the accuracy
16 assessment matrix for 1985, and 16 the accuracy
17 assessment for 1993, which is also a matrix but
18 that's not the formal title on the papers that
19 were provided to us at lunch break.
20 (The documents referred to were thereupon
21 marked Exhibits 15, 16 for Identification.)
22 BY MR. WATTS-FITZGERALD:
23 Q. I read this to suggest that for 1985
24 imagery the assessment was based on a total of 39
25 assessment points; is that correct?
122
1 A. Yes.
2 Q. And for '93 the total number was 84
3 accuracy assessment points?
4 A. Yes.
5 Q. How did you statistically arrive at those
6 numbers as being adequate to determine the accuracy
7 of the two maps?
8 A. My professional judgment.
9 Q. No statistical test whatsoever?
10 A. On the image, you mean?
11 Q. No statistical test to determine such as
12 that -- are you familiar with Mr. Rutchey's
13 description how he developed his value for the
14 minimum number of points necessary to demonstrate map
15 accuracy?
16 A. As described in his report?
17 Q. Yes.
18 A. I remember reading that.
19 Q. He referenced Cochran's paper from 1978 on
20 binomial probability formulas and suggested that the
21 number of points needed to -- just for 2A to have an
22 85% map accuracy level with an error of plus or minus
23 35% was a minimum of 204. Do you agree with that?
24 A. Do I agree with that recommendation?
25 Q. With that assessment, yes.
123
1 A. That would be a nice number.
2 Q. And he, in fact, used 241 assessment
3 points, which would exceed the number, would it not?
4 A. Rutchey used a number exceeding that?
5 Q. In an accuracy assessment, is it fair to
6 say the greater number of assessment points employed
7 and verified, the greater confidence can be placed on
8 the imagery?
9 A. If the procedure supports the precision of
10 location of those points.
11 Q. Procedures being equal, then, the more the
12 merrier in terms of accuracy?
13 A. Procedures being equal to a point, the more
14 the merrier, as you say.
15 Q. I understand if you check 100% of pixels,
16 you have perfect accuracy.
17 You are expressing on Exhibits 15 and 16 a
18 confidence level in the accuracy of these two maps
19 that range from 70 to 72; the 70 being for 1993 and
20 the 72 being for 1985?
21 A. Yes.
22 Q. In your professional expertise, do you
23 believe that that level of accuracy meets the
24 standard in your field or in the field of remote
25 sensing vegetative map classifications?
124
1 A. I don't know there is a standard -- that
2 there is a standard accepted accuracy level for the a
3 map.
4 Q. I understood from your earlier testimony
5 and also understood you to say it's based on a
6 professional judgment assessment and somewhat project
7 specific.
8 What I'm asking you to do is to apply those
9 two bases for evaluation and tell me if someone in
10 your field objectively evaluating those two
11 parameters which you have described would find these
12 two accuracy assessments such that they would propose
13 a high degree of reliability on these two maps.
14 A. I think that any person in my field
15 reviewing these accuracy tables will assess the
16 accuracy at what we said it is, and that is the
17 accuracy of the map. They would say the accuracy of
18 the map is what we said it is.
19 Q. As I understood your testimony, and I want
20 to verify now -- I got the matrix in front of me.
21 Both accuracy assessment point selections of 39 and
22 84 were part of a larger field of accuracy assessment
23 points generated by a stratified random sampling
24 program?
25 A. Yes.
125
1 Q. What effect does it have by taking only 39
2 and 84, respectively, you have a significant number
3 of classes within each of those two assessments with
4 two or less sampling points?
5 A. What is the significance of the fact there
6 were two or less?
7 Q. How does that affect the result?
8 A. How do two or less samples being tested
9 affect the map accuracy result?
10 Q. I looked through these. I see a total of
11 four sites out of -- four classes out of a total of
12 20 that had no sampling points in it at all and the
13 significant additional number more than that that had
14 one or two -- which in the abstract seems like a very
15 small number.
16 A. It's a small number.
17 Q. That makes me think statistically it's a
18 very low power analysis.
19 A. Statistically since it was a stratified
20 random sample the areas that were very low acreage in
21 the classification could have very few possible
22 potential accuracy assessment signs associated with
23 them. So that would be lower.
24 Q. Is there any way for me to tell -- looking
25 at Exhibits 15 for 1985, does that suggest that on
126
1 the wire access, which is the photo taken as true
2 here for class No. 7 showing 20 correctly identified
3 interpreted image, that's what the shadow means,
4 right?
5 A. Yes, that's correct.
6 Q. And then six were incorrectly interpreted
7 for a total field of 26 true No. 7's, which were
8 predominant sawgrass. That's how the matrix works?
9 A. Yes.
10 Q. You have two error columns next to the
11 total of 26, one of 23.08, one 9.09. Is the first
12 one an omission error?
13 A. There is an error on the label. Those
14 columns, the first column is an omission. The next
15 column is commission, and the final column is
16 accuracy.
17 Q. I see.
18 A. The '93 has the correct column titles for
19 some reason, '85 does not.
20 Q. So what this implies to me is for class 7,
21 which you say because this is a stratified random,
22 must be the class with the greatest acreage?
23 A. Stratification is based upon pre-dominance
24 of that classification.
25 Q. So in the classification with the greater
127
1 acreage within the image, this is for just 2A, you
2 have a 76.92% accuracy?
3 A. In the sawgrass category.
4 Q. In the predominant sawgrass category?
5 A. Yes.
6 Q. You have 100% accuracy in the predominant
7 cattail, of which there were two sample points,
8 accuracy points?
9 A. Yes.
10 Q. Now, exactly what does the 33.33 commission
11 error mean for the predominant cattail accuracy
12 assessment?
13 A. For predominant cattail in '85?
14 Q. Yes.
15 A. Which column?
16 Q. The first true photo, predominant cattail
17 No. 1. You had a total of 2.0 omission error,
18 commission error 33.33%.
19 A. You see that column under 1? What that 33
20 means is that one point was interpreted in the
21 photography as a different category. 33%. Those are
22 our percentages.
23 Q. If you had extended this effort and
24 continued beyond the 39 to the field of assessment
25 points, would you have expected to have assessment
128
1 points in all classes?
2 A. Yes, we would. And they would be based
3 upon the classes in the classified image.
4 Q. By its ratio?
5 A. By the stratified random sampling, yes.
6 Q. Did you do a KAPA analysis for this?
7 A. No, we have not done one yet.
8 Q. Does that mean you are going to?
9 A. Well, the situation is this. We generated
10 a large number of potential stratified random
11 samplings that could be used. We used those points
12 in these accuracy assessments.
13 Once we closed the table in the imaging
14 software, it closes the table. You can't add to it,
15 you can't modify to it. KAPA is one of the functions
16 in that table.
17 So until I'm directed to close the table,
18 close the accuracy assessment, then I haven't run a
19 KAPA.
20 MR. WATTS-FITZGERALD: Counsel, do you plan
21 to direct the firm or the witness to conduct
22 that analysis because --
23 MS. RAEPPLE: Unless prohibited by the
24 hearing officer, yes, I would like him to do
25 further accuracy assessment.
129
1 If that occurs, we will provide it to you
2 and provide opportunities for further inquiry.
3 MR. WATTS-FITZGERALD: We will deal with
4 that when it happens. But in good faith, as you
5 did yesterday or the day before with Dr. Jensen,
6 we are obviously going to oppose that this has
7 to stop at some point.
8 BY MR. WATTS-FITZGERALD:
9 Q. Was the methodology for the production of
10 the accuracy assessment for 1993, Exhibit 16, the
11 same as that which you have described for 1985?
12 A. Yes.
13 Q. On the Y axis, Exhibit 16, the numbers
14 range from 1 to -- I can't read the last one, 18, but
15 they are not obviously sequential. The missing
16 classes are as a result of the class collapse --
17 A. Yes.
18 Q. -- over the time of the analysis?
19 A. Yes.
20 Q. On 16 I see a commission error for
21 predominant cattail in excess of 50%. What does that
22 signify to you?
23 A. Which figure are you looking at now?
24 Q. Exhibit 16, the figure for 1993, the
25 predominant cattail, which is the first line in the
130
1 numerology to the right of the graph. You said you
2 have total of 7?
3 A. Commission error column?
4 Q. Yes. Commission error is 53.33?
5 A. The photo interpretation, the truth found
6 and error.
7 Q. I think what was throwing us off, the total
8 column on the right on both pages, should that be
9 labeled total correct? That's really what that is,
10 isn't it, and then the total number of samples per
11 class is at the bottom of the box?
12 A. No, I don't think so.
13 Q. Let's look at position 1-1 in the box.
14 It's shaded. It says 7. The 7 is carried over to
15 the right to the total line.
16 If I go down the vertical column of 1, I
17 see in class 3, 1 was identified. It was 4. 2 were
18 identified. Class 7, 5 were identified. Those have
19 to be commission errors to stand up with 53.33?
20 A. Yes.
21 Q. So that implies that one category 3 which
22 is equal mix, one category 4 predominant mix, one
23 category 7 predominant sawgrass were identified into
24 the predominant cattail area which were not?
25 A. In the photo, yes. That is what that
131
1 indicates.
2 Q. And the photo is the true?
3 A. Photo is the true.
4 Q. Which means the image was wrong?
5 A. Well, yes. If you use the photo as ground
6 truth.
7 Q. You did, did you not?
8 A. Yes.
9 Q. You took as gospel whatever is in those
10 photos?
11 A. In order to do an accuracy assessment you
12 have to come up with something to compare your data
13 to.
14 Q. Since you weren't comparing to ground
15 truthing efforts directed at the accuracy assessment
16 points, you had to do this boot strap operation from
17 the photos?
18 MS. RAEPPLE: Object to the form.
19 THE WITNESS: Yes.
20 MR. WATTS-FITZGERALD: I shouldn't use the
21 term boot strap, that's a term of art, if that
22 was the objection.
23 MS. RAEPPLE: That was the objection.
24 THE WITNESS: Restate it again.
25 BY MR. WATTS-FITZGERALD:
132
1 Q. So this implies for '93 that your system
2 did a pretty good job placing in the predominant
3 cattail category on the image everything that is
4 predominant cattail. You are calling it 100%
5 accurate?
6 A. That's a normal statistical value, yes.
7 Q. The rest of it, the commission error says
8 that's true, but watch out because a lot of areas are
9 going to be listed as cattail that aren't?
10 A. It's saying they aren't in that class, yes.
11 There may be other mixes of cattail.
12 Q. But they are not predominant cattail as you
13 defined them?
14 A. Yes.
15 Q. In fact, 5 out of the 8 are going to be
16 predominant sawgrass?
17 A. Five out of 15.
18 Q. 5 out of 8 are incorrectly identified. I'm
19 giving you the 7.
20 A. Okay.
21 Q. Now, in '85, this methodology did not do so
22 well for predominant sawgrass and broad leaf
23 vegetation, tree islands because you got a fairly
24 significant percentage omission errors for both of
25 those classes.
133
1 You have 23.08 for predominant sawgrass and
2 50% -- you have a 100 omission error for predominant
3 sawgrass and other. So that model or that image does
4 not seem to do too well in the sawgrass range. It
5 looks like you misclassified it with a fairly high
6 degree of regularity based on the accuracy
7 assessment?
8 A. Is that a question?
9 Q. Yes. Am I interpreting that correctly?
10 A. It has a 76% of accuracy, which is higher
11 than the overall map accuracy. There are several
12 ways to look at data. You have to look at all of
13 these calculations in context.
14 Q. I was looking at all of the -- I was
15 looking for broad leaf vegetation upward. I see 50%
16 accuracy for broad vegetation/tree island. I find
17 that surprising. I find according to Dr. Lodge
18 that's one of most prominent features identifiable
19 from the imagery and CIR. I see only 76% for
20 sawgrass and zero for predominant sawgrass/other.
21 Zero for predominant cattail/other, which would
22 include, I assume, predominant cattail with some
23 sawgrass because of the way that class was classed.
24 So for the class number 3, 6, 7, 8, -- 3,
25 6, 7 and 9 the accuracy is very poor only for
134
1 predominant cattail. And as far as sawgrass with
2 periphyton, it may contain periphyton, is that right?
3 Do you recall Ken Rutchey's discussion as
4 sort of an author's observation in his report of the
5 influence of the brightness value of periphyton on
6 interpretation?
7 A. What kind of observation?
8 Q. About the influence of the brightness value
9 of periphyton communities?
10 A. You said he made some type --
11 Q. Author's observation.
12 A. Okay, go ahead.
13 Q. Do you recall that discussion in his
14 report?
15 A. Yes.
16 Q. Did you or Dr. Lodge make any effort to
17 consider the possible implications for your work of
18 the phenomenon he reports?
19 A. The reflectance of periphyton?
20 Q. Yes.
21 A. Yes, we did.
22 Q. What did you do?
23 A. Most cases where periphyton was present, it
24 was sparse sawgrass community.
25 Q. That was in your field visits for purposes
135
1 of identifying training sites that occurred during
2 what time of the year?
3 A. Our field samples were taken at the same
4 period the satellite passed over, roughly the winter.
5 Q. But for the 1985 effort you were working
6 from CIR imagery. That was roughly 10 months
7 distant. And February, March working back to
8 November when the communities -- periphyton
9 communities, at least based on what Mr. Rutchey says
10 in his report could be expected to be somewhat
11 different, was any effort made with regard to the '85
12 to take that factor into account in interpreting the
13 photo imagery?
14 A. We attempted to identify periphyton with
15 the sparseness of vegetation within it. We tried to
16 take into account the effect of periphyton.
17 Q. What does the KAPA value tell you with the
18 accuracy assessment calculations?
19 A. It would tell you how much randomness
20 influenced that classification.
21 Q. At what point does the randomness exceed
22 the acceptable limits in your profession when
23 conducting a statistical analysis and what would you
24 do if it exceeded those acceptable limits?
25 A. If it exceeded acceptable limits? There
136
1 are two divisions.
2 Q. I'm not allowed to do that. Let me start
3 with the second one first.
4 If the KAPA value you are ultimately
5 calculating, assuming that that occurs, exceeded the
6 acceptable norm or limits within the field of
7 expertise that you are going to proffer opinions
8 from, what do you do -- how do you cure the problem?
9 Can you cure the problem?
10 A. Once you complete --
11 MS. RAEPPLE: Could you just ask one
12 question at a time? Objection to the form.
13 MR. WATTS-FITZGERALD: You can still
14 answer.
15 MS. RAEPPLE: You asked three questions.
16 Which one do you want answered.
17 MR. WATTS-FITZGERALD: I have had
18 difficulty throughout getting the witness to
19 answer the question I do ask. I try to ask it
20 several ways so that we eventually get around.
21 It just takes a lot of time.
22 BY MR. WATTS-FITZGERALD:
23 Q. Can you adjust after the fact when the KAPA
24 value exceeds acceptable limits?
25 A. I don't believe so, no.
137
1 Q. Can you go back, include additional
2 assessment points in an effort to improve the
3 accuracy of your work?
4 A. The way we are set up on this project which
5 is using imaging software, once you close the table,
6 once you run the accuracy assessment in the imagining
7 software you cannot modify in any fashion that table
8 ever again.
9 If the KAPA coefficient is run on the
10 table, it's set in stone. You have to generate a new
11 set of stratified random samples and go through the
12 whole process again if you wanted to redo it at that
13 point.
14 Q. Would merely redoing it but including
15 additional accuracy assessment points -- instead of
16 stopping at 39 you go to 80 for 1985, for example,
17 will that affect the KAPA value?
18 A. Well, it could, I suppose, yes. I never
19 tried that, I don't know.
20 Q. If the accuracy assessment is valid in
21 expressing the accuracy of the image, that's what
22 it's trying to do, don't you have to go back and
23 improve the image itself to get better accuracy?
24 A. I don't know.
25 Q. If the accuracy assessment is valid and you
138
1 have a 70% accuracy in your image -- I'll give you a
2 hypothetical.
3 If you end up doing an accuracy assessment
4 that says you have a 40% accurate image, assuming 40%
5 also falls below the acceptable norm, is there any
6 alternative other than to go back and improve your
7 image?
8 A. You are talking about once you close the
9 table and -- you know, you said you have difficulty
10 having me answer questions. I'm not intending to do
11 that.
12 Q. I wasn't suggesting you were.
13 A. All right. The methods we are using, we
14 generate a table. Each of those tables, we generate
15 more random samples that we used. That is because I
16 went and had the option if the powers that be want
17 more points, we can do that. But once we closed the
18 table, then that table can never be reopened because
19 that's a quality control check.
20 So if I understand your question, if you
21 did a map, if we did this map and closed the tables
22 and it was 40% accuracy, the level of the map is 40%,
23 then the only way you could redo that accuracy
24 assessment would be to regenerate a set of points and
25 redo the whole process. Am I answering your
139
1 question?
2 Q. You actually answered the earlier one, but
3 that's good, too.
4 If your accuracy assessment is right, no
5 matter how many points you used and your accuracy is
6 only 40% based on the accuracy assessment and it's
7 valid, that's a given that implies that your image is
8 no good, that the image as platted, the result of all
9 of the work is substandard, it's deficient, it should
10 not be used?
11 A. If 40% were an acceptable accuracy level?
12 You discussed it as being acceptable.
13 Q. It's not acceptable.
14 A. If it was not acceptable and you agreed it
15 was not acceptable and you ran your assessment it was
16 less than 40% or 40%?
17 Q. And your accuracy assessment is right, you
18 have to assume it's right?
19 A. Yes. In that case you could add a thousand
20 points and assess every pixel, but it would not
21 improve the accuracy of the classification, if I
22 understand the question correctly. I think I do.
23 Q. I'm trying to figure out if they tell you
24 to go back and do more points, you said you closed
25 the tables in order to do this?
140
1 A. No. We have not closed the tables yet.
2 That's what I said.
3 Q. These are drafts?
4 A. No, these aren't drafts. These are real
5 valid tables based on number of points we have now.
6 The table ERDAS is still open. If I closed it right
7 now, it would produce exactly this result. This is
8 the same data, same set of formulas.
9 But the table in the ERDAS software is
10 still open. If I was directed to continue selecting
11 samples, I could add more samples to it. There would
12 be no closes classes where there were no samples. So
13 your questions about that could be answered and we
14 could continue.
15 I apologize for the confusion of not
16 understanding what you were talking about.
17 Q. I mentioned on Exhibits 15 and 16 certain
18 classes that have no samples or no accuracy
19 assessment points. What can you tell, if anything,
20 based on this assessment methodology about those
21 classes?
22 A. Those classes from this table, you can't
23 say.
24 Q. So if you express any opinion about those
25 categories in the two respective images, they would
141
1 be based on what?
2 A. My opinion the image.
3 Q. In your professional judgment?
4 A. Yes, it would be a professional judgment
5 call, obviously.
6 Q. And you would feel likewise constrained,
7 then, not to attempt to extrapolate these values
8 assuming they prove up after any further work or stay
9 the way they are to areas beyond 2A in a quantitative
10 sense?
11 A. You mean, would I speculate on the accuracy
12 in all of the study area, is this accuracy based on
13 the assessment 2A? I would speculate it would be
14 very similar. I wouldn't speculate that it would be
15 the same.
16 Q. As we discussed earlier, to the extent any
17 sub area within the study area has substantial
18 variation in its foliage from that found in 2A you
19 would have to diminish the accuracy values or your
20 assessment of the accuracy values, would you not?
21 A. If the foliage were different?
22 Q. The vegetation -- hypothetically, if
23 Dr. Lodge would tell you that the flora of Water
24 Conservation Area 1 is markedly different in its mix
25 and composition from the foliage in Water
142
1 Conservation Area 2 for whatever reason, would you be
2 confident in taking these quantitative values for 2A
3 and assuming or rendering a professional judgment
4 that they pretty closely, if not exactly, reflect the
5 conditions and accuracy of the image in Water
6 Conservation Area 1?
7 A. I would think there would be a high degree
8 of correlation between the accuracy in Loxahatchee
9 and the entire duty area to the 2A figures.
10 Q. Even the Loxahatchee had such a minimal
11 number of ground truthing points?
12 A. Yes.
13 MR. WATTS-FITZGERALD: I have no further
14 questions, but I want to mark exhibits.
15 I would like to mark Exhibit 17 the
16 supervised classification Landsat Thematic
17 Mapper Multispectral image for December 10, 1993
18 of the entire study area. It's got a prepared
19 date of April 4, 1994, depicting essentially the
20 northern most point of Water Conservation Area 1
21 to approximately ten kilometers of Everglades
22 National Park in the south.
23 (The document referred to was thereupon
24 marked Exhibit 17 for Identification.)
25 MR. WATTS-FITZGERALD: Exhibit 18, the map
143
1 which indicates it was produced April 5, 1994
2 scaled 1 to 50,000, Water Conservation Area 2A
3 from the supervised classification of Landsat
4 Mutispectral Thematic Mapper scene from
5 December 10, 1993.
6 (The document referred to was thereupon
7 marked Exhibit 18 for Identification.)
8 MR. WATTS-FITZGERALD: Exhibit 19 will be a
9 grid map produced and discussed by Dr. Lodge
10 which underlies the work described by the
11 witness today depicting construction lines and
12 other features utilized in the preparation of
13 their classified imagery for WCA-2A.
14 Exhibit 19 not being labeled other than
15 with the number of the exhibit.
16 (The document referred to was thereupon
17 marked Exhibit 19 for Identification.)
18 MR. WATTS-FITZGERALD: Exhibit 20 is photo
19 interpretation registration grid for NHAP 9 by 9
20 at scale 1 to 64,923.
21 (The document referred to was thereupon
22 marked Exhibit 20 for Identification.)
23 MR. WATTS-FITZGERALD: Both 19 and 20 would
24 relate to the 1985 imagery because of the scale.
25 MS. RAEPPLE: For Exhibit 19 and 18 you
144
1 will be making copies and substituting copies
2 and returning the originals to me; is that
3 correct?
4 MR. WATTS-FITZGERALD: Yes. Exhibit 21
5 will be Water Conservation Area 2A classified
6 image supervised classification of Landsat
7 Multispectral acquired November 2, 1985, the map
8 production date April 5, 1994, scale 1 to
9 50,000.
10 (The document referred to was thereupon
11 marked Exhibit 21 for Identification.)
12 MR. WATTS-FITZGERALD: Exhibits 22 and 23
13 are large grids about 4 feet by 5 feet with
14 construction lines and some features drawn in in
15 colored pencil depicting portions of WCA 2A that
16 have previously been described in Dr. Lodge's
17 initial deposition.
18 (The documents referred to were thereupon
19 marked Exhibits 22 and 23 for Identification.)
20 MR. WATTS-FITZGERALD: We are marking
21 Exhibit 24 a portion of a USGS quadrangle from
22 Fort Lauderdale to northwest 7.5 series showing
23 the southeast portion of the Loxahatchee Refuge
24 and the north central area of Water Conservation
25 Area 2 A. No date.
145
1 (The document referred to was thereupon
2 marked Exhibit 24 for Identification.)
3 MR. WATTS-FITZGERALD: Exhibit 25 is
4 Fort Lauderdale to the southeast, photo revised
5 1983 7.5 minute series.
6 (The document referred to was thereupon
7 marked Exhibit 25 for Identification.)
8 MR. WATTS-FITZGERALD: 26 is
9 Fort Lauderdale to the southwest dated 1973, no
10 revision date on it. 7.5 showing the central
11 area of WCA-2A and a portion of levee 38 east.
12 (The document referred to was thereupon
13 marked Exhibit 26 for Identification.)
14 MR. WATTS-FITZGERALD: 27 is Cooper City
15 northeast, Florida photo revised 1983, 7.5
16 minute series showing the eastern portion of
17 Water Conservation Area 2.
18 (The document referred to was thereupon
19 marked Exhibit 27 for Identification.)
20 MR. WATTS-FITZGERALD: 28, Andytown,
21 Florida. 7.5 minute series photo revised 1983
22 showing the 20 mile bend and the southeastern
23 corner of Water Conservation Area 2A where it
24 intersects 2B and WCA 3A.
146
1 (The document referred to was thereupon
2 marked Exhibit 28 for Identification.)
3 MR. WATTS-FITZGERALD: 29 is Everglades 1
4 southeast portion there of 7.5 minute series.
5 Date is 1974. No revision date.
6 (The document referred to was thereupon
7 marked Exhibit 29 for Identification.)
8 MR. WATTS-FITZGERALD: These are all
9 photographs, Exhibits 29 and 24. 26 is, as
10 well.
11 Exhibit 30 is Fort Lauderdale to the
12 northwest 1974, 7.5 minute photo showing the
13 northern portion of 2A and Loxahatchee.
14 (The document referred to was thereupon
15 marked Exhibit 30 for Identification.)
16 MR. WATTS-FITZGERALD: Finally, Exhibit 31
17 depicts the northeastern corner of Water
18 Conservation Area 2A, southern most portion of
19 Water Conservation Area 1, and adjacent portions
20 of Coral Springs. And because of the cutting of
21 the top of the map, none of the legends that
22 otherwise appear directly there on other
23 versions of Exhibit 25, Fort Lauderdale to the
24 southeast but it has various sites labeled
25 numerically and alphabetically in 2A.
147
1 MS. RAEPPLE: Mr. Watts-Fitzgerald, as for
2 Exhibits 19, 20, 22 through 31 we will need the
3 originals returned to us and we would appreciate
4 you having copies made, substituting copies as
5 exhibits to the deposition and returning them at
6 your earliest convenience.
7 MR. WATTS-FITZGERALD: I'm not sure that
8 attempts to replicate the work can be done off
9 the copies of some of these. I may not be able
10 to return then all immediately, if we are going
11 to use the photos. This is something we have to
12 talk about, getting the photos from you later.
13 But I understood that you were producing
14 the 15 meg tape and the tasseled cap as part of
15 the production. We are not marking them as
16 exhibits.
17 MS. RAEPPLE: You can keep those. As for
18 the color infrared, we will retain the
19 originals. If you need access to them for
20 whatever purpose, let us know and we will
21 provide that -- including if you want to take
22 them long enough to make copies.
23 In addition, Mr. Downing advised me at
24 lunch in particular that he overlooked providing
25 today in digital form the '85 image which you
148
1 have in hard copy. If you want it in digital
2 form we can send that to you next week.
3 MR. WATTS-FITZGERALD: We would request
4 that.
5 THE WITNESS: I'll make you one.
6 MR. WATTS-FITZGERALD: We got over the
7 lunch break all of the worksheets by Dr. Lodge
8 that were used for '85 and '93?
9 MR. CESARANO: I believe it is. That's all
10 that came in.
11 MR. WATTS-FITZGERALD: I have nothing
12 further.
13 MR. CESARANO: I have nothing further.
14 MS. RAEPPLE: No questions. We do not
15 waive.
16 (Thereupon the taking of the
17 deposition was concluded.)
18 - - - - - - - - -
149
1
2 I, ED DOWNING, do hereby certify that I
3 have read the foregoing deposition and that the same
4 is a true and accurate transcript of my testimony,
5 except for attached amendments, if any.
6
7
8
9 ----------------------------------
10
11
12
13
14
15 The signature above of ED DOWNING was
16 subscribed and sworn to before me this day of 1994.
17
18
19
20
21 -----------------------------------
22 Notary Public
23 My commission expires
150
1
2 CERTIFICATE OF OATH
3
4
5 THE STATE OF FLORIDA )
6 COUNTY OF DADE. )
7
8
9 I, the undersigned authority, certify that
10 ED DOWNING personally appeared before me and was duly
11 sworn. WITNESS my hand and official seal this 15th
12 day of April, 1994.
13
14
15 ___________________________________
16 Thomas R. Neumann, RPR
17 Notary Public - State of Florida
18 My Commission Expires: June 19, 1994
151
1 CERTIFICATE
2
3 THE STATE OF FLORIDA )
COUNTY OF DADE )
4
5
I, Thomas R. Neumann, Registered
6 Professional Reporter, do hereby certify that I was
authorized to and did report said deposition in
7 stenotype; and that the foregoing pages, numbered
from 1 to 148, inclusive are a true and correct
8 transcription of my shorthand notes of said
deposition.
9
I further certify that I am not an attorney
10 or counsel of any of the parties, nor am I a relative
or employee of any attorney or counsel or party
11 connected with the action, nor am I financially
interested in the action.
12
The foregoing certification of this
13 transcript does not apply to any reproduction of the
same by any means unless under the direct control
14 and/or direction of the certifying reporter.
15 Dated this 15th day of April, 1994.
16
______________________
17 Thomas R. Neumann, RPR
18
THE STATE OF FLORIDA )
19 COUNTY OF DADE. )
20
The foregoing certificate was acknowledged
21 before me this 15th day of April, 1994 by
Thomas R. Neumann, who is personally known to me.
22
23
________________________________
24 Notary Public - State of Florida
My Commission expires:
25
152
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15