A Statistician s Take on the Benthic Index of Biotic Integrity

Size: px
Start display at page:

Download "A Statistician s Take on the Benthic Index of Biotic Integrity"

Transcription

1 A Statistician s Take on the Benthic Index of Biotic Integrity Grace Chiu, Peter Guttorp PIMS Postdoctoral Fellow, Statistics & Actuarial Science, SFU, and Visiting Scholar, Statistics, U of Washington Chair and Professor, Statistics, U of Washington November 5, 2003 TIES 2003, Page 1

2 Outline: 1. Photos of impacted Puget Sound Lowland Streams TIES 2003, Page 2

3 Outline: 1. Photos of impacted Puget Sound Lowland Streams 2. Biomonitoring and the B-IBI TIES 2003, Page 2

4 Outline: 1. Photos of impacted Puget Sound Lowland Streams 2. Biomonitoring and the B-IBI 3. More on the B-IBI TIES 2003, Page 2

5 Outline: 1. Photos of impacted Puget Sound Lowland Streams 2. Biomonitoring and the B-IBI 3. More on the B-IBI 4. Statistician s take?? current efforts TIES 2003, Page 2

6 Outline: 1. Photos of impacted Puget Sound Lowland Streams 2. Biomonitoring and the B-IBI 3. More on the B-IBI 4. Statistician s take?? current efforts Revamping the IBI definition the SOBIBI Exploratory and bootstrap analyses TIES 2003, Page 2

7 Outline: 1. Photos of impacted Puget Sound Lowland Streams 2. Biomonitoring and the B-IBI 3. More on the B-IBI 4. Statistician s take?? current efforts Revamping the IBI definition the SOBIBI Exploratory and bootstrap analyses 5. Insight TIES 2003, Page 2

8 Two Puget Sound Lowland Streams TIES 2003, Page 3

9 Two Puget Sound Lowland Streams TIES 2003, Page 3

10 Two Puget Sound Lowland Streams TIES 2003, Page 3

11 Which Stream Is Healthier?? TIES 2003, Page 4

12 Which Stream Is Healthier?? Check bugs... TIES 2003, Page 4

13 Which Stream Is Healthier?? Check bugs... TIES 2003, Page 4

14 Which Stream Is Healthier?? Check bugs... Rock Creek Juanita Creek TIES 2003, Page 4

15 Which Stream Is Healthier?? Check bugs... Rock Creek Juanita Creek benthic invertebrate diversity and abundance = function of stream health (biotic integrity) TIES 2003, Page 4

16 Which Stream Is Healthier?? Check bugs... Rock Creek Juanita Creek benthic invertebrate diversity and abundance = function of stream health (biotic integrity) Benthic Index of Biotic Integrity (B-IBI) TIES 2003, Page 4

17 Biomonitoring TIES 2003, Page 5

18 Biomonitoring track impact of human activities on ecological systems TIES 2003, Page 5

19 Biomonitoring track impact of human activities on ecological systems need to develop a report card measure TIES 2003, Page 5

20 Biomonitoring track impact of human activities on ecological systems need to develop a report card measure e.g. B-IBI TIES 2003, Page 5

21 Biomonitoring track impact of human activities on ecological systems need to develop a report card measure e.g. B-IBI compare report cards over time and/or over space TIES 2003, Page 5

22 The Benthic IBI (B-IBI) TIES 2003, Page 6

23 The Benthic IBI (B-IBI) Developed by Kerans and Karr (1994) TIES 2003, Page 6

24 The Benthic IBI (B-IBI) Developed by Kerans and Karr (1994) Modeled after the original IBI (a.k.a. fish IBI) developed by Karr et al. (1986) TIES 2003, Page 6

25 The Benthic IBI (B-IBI) Developed by Kerans and Karr (1994) Modeled after the original IBI (a.k.a. fish IBI) developed by Karr et al. (1986) measures diversity / abundance of fish species instead of benthic creatures TIES 2003, Page 6

26 Main Idea of the -IBI ( = B or FISH) TIES 2003, Page 7

27 Main Idea of the -IBI ( = B or FISH) identify attributes / metrics which best reflect diversity / abundance TIES 2003, Page 7

28 Main Idea of the -IBI ( = B or FISH) identify attributes / metrics which best reflect diversity / abundance e.g. total taxa (species), % 3 most dominant taxa, etc. TIES 2003, Page 7

29 Main Idea of the -IBI ( = B or FISH) identify attributes / metrics which best reflect diversity / abundance e.g. total taxa (species), % 3 most dominant taxa, etc. metrics should be sensitive to impact of urbanization on environment TIES 2003, Page 7

30 Main Idea of the -IBI ( = B or FISH) identify attributes / metrics which best reflect diversity / abundance e.g. total taxa (species), % 3 most dominant taxa, etc. metrics should be sensitive to impact of urbanization on environment combine identified metrics -IBI TIES 2003, Page 7

31 Main Idea of the -IBI ( = B or FISH) identify attributes / metrics which best reflect diversity / abundance e.g. total taxa (species), % 3 most dominant taxa, etc. metrics should be sensitive to impact of urbanization on environment combine identified metrics -IBI For 10-metric Benthic IBI developed for PSL: VERY POOR POOR FAIR GOOD EXCELLENT TIES 2003, Page 7

32 Details: CHOOSING METRICS TIES 2003, Page 8

33 Details: CHOOSING METRICS start with large number of candidate metrics apply various statistical diagnostics to discard irrelevant and redundant metrics TIES 2003, Page 8

34 Details: CHOOSING METRICS start with large number of candidate metrics apply various statistical diagnostics to discard irrelevant and redundant metrics as deemed appropriate by Kerans & Karr TIES 2003, Page 8

35 Details: CHOOSING METRICS start with large number of candidate metrics apply various statistical diagnostics to discard irrelevant and redundant metrics as deemed appropriate by Kerans & Karr statisticians have recognized issues with choice of metrics we shall not discuss them here TIES 2003, Page 8

36 Details: SCORING METRICS TIES 2003, Page 9

37 Details: SCORING METRICS i.e. standardizing metric values measured on different scales across different streams TIES 2003, Page 9

38 e.g. metrics correlated with stream size TIES 2003, Page 10

39 e.g. metrics correlated with stream size (Reference (least impacted) Sites. From Karr et al ) TIES 2003, Page 10

40 e.g. metrics correlated with stream size graph includes repeated observations (Reference (least impacted) Sites. From Karr et al ) TIES 2003, Page 10

41 e.g. metrics correlated with stream size graph includes repeated observations trisect area under graph (by eye or regression or pre-assigned percentiles) (Reference (least impacted) Sites. From Karr et al ) TIES 2003, Page 10

42 e.g. metrics correlated with stream size graph includes repeated observations trisect area under graph (by eye or regression or pre-assigned percentiles) compare sampled site (AVERAGED OVER REPLICATES) to graph get metric score for site (Reference (least impacted) Sites. From Karr et al ) TIES 2003, Page 10

43 e.g. metrics correlated with stream size graph includes repeated observations trisect area under graph (by eye or regression or pre-assigned percentiles) compare sampled site (AVERAGED OVER REPLICATES) to graph get metric score for site metrics with small ranges don t score 3 (Reference (least impacted) Sites. From Karr et al ) TIES 2003, Page 10

44 e.g. metrics correlated with stream size graph includes repeated observations trisect area under graph (by eye or regression or pre-assigned percentiles) compare sampled site (AVERAGED OVER REPLICATES) to graph get metric score for site metrics with small ranges don t score 3 similar scoring (Reference (least impacted) Sites. From Karr et al ) mechanisam for corresponding benthic metrics TIES 2003, Page 10

45 Details: SCORING METRICS i.e. standardizing metric values measured on different scales across different streams TIES 2003, Page 11

46 Details: SCORING METRICS i.e. standardizing metric values measured on different scales across different streams subjective definition of reference sites weird discrete scale cutpoints very subjective requires recalibration in presence of spatial or temporal changes TIES 2003, Page 11

47 Details: SCORING METRICS i.e. standardizing metric values measured on different scales across different streams subjective definition of reference sites weird discrete scale cutpoints very subjective requires recalibration in presence of spatial or temporal changes COMPUTING THE B-IBI: TIES 2003, Page 11

48 Details: SCORING METRICS i.e. standardizing metric values measured on different scales across different streams subjective definition of reference sites weird discrete scale cutpoints very subjective requires recalibration in presence of spatial or temporal changes COMPUTING THE B-IBI: B-IBI = sum of metric scores TIES 2003, Page 11

49 Details: SCORING METRICS i.e. standardizing metric values measured on different scales across different streams subjective definition of reference sites weird discrete scale cutpoints very subjective requires recalibration in presence of spatial or temporal changes COMPUTING THE B-IBI: B-IBI = sum of metric scores Q: Really use B-IBI to monitor stream health over time?? TIES 2003, Page 11

50 Our Efforts TIES 2003, Page 12

51 Our Efforts MAIN GOAL: device a statistically oriented metric scoring mechanism that is unaffected by spatial / temporal changes TIES 2003, Page 12

52 Our Efforts MAIN GOAL: device a statistically oriented metric scoring mechanism that is unaffected by spatial / temporal changes SIMPLE STANDARDIZATION! TIES 2003, Page 12

53 Our Efforts MAIN GOAL: device a statistically oriented metric scoring mechanism that is unaffected by spatial / temporal changes SIMPLE STANDARDIZATION! score = metric value metric mean metric SD adjusted for stream size (regression), if necessary TIES 2003, Page 12

54 Our Efforts MAIN GOAL: device a statistically oriented metric scoring mechanism that is unaffected by spatial / temporal changes SIMPLE STANDARDIZATION! score = metric value metric mean metric SD adjusted for stream size (regression), if necessary SO BIBI = sum of all scores TIES 2003, Page 12

55 Our Efforts MAIN GOAL: device a statistically oriented metric scoring mechanism that is unaffected by spatial / temporal changes SIMPLE STANDARDIZATION! score = metric value metric mean metric SD adjusted for stream size (regression), if necessary SO BIBI = sum of all scores reference sites unnecessary (as long as some unimpacted sites are included in the study) TIES 2003, Page 12

56 Our Efforts MAIN GOAL: device a statistically oriented metric scoring mechanism that is unaffected by spatial / temporal changes SIMPLE STANDARDIZATION! score = metric value metric mean metric SD adjusted for stream size (regression), if necessary SO BIBI = sum of all scores reference sites unnecessary (as long as some unimpacted sites are included in the study) continuous scale which requires no recalibration TIES 2003, Page 12

57 Our Efforts MAIN GOAL: device a statistically oriented metric scoring mechanism that is unaffected by spatial / temporal changes SIMPLE STANDARDIZATION! score = metric value metric mean metric SD adjusted for stream size (regression), if necessary SO BIBI = sum of all scores reference sites unnecessary (as long as some unimpacted sites are included in the study) continuous scale which requires no recalibration non-subjective TIES 2003, Page 12

58 SOBIBI vs B-IBI on PSL Streams TIES 2003, Page 13

59 SOBIBI vs B-IBI on PSL Streams 1997 & 1998 data, taken from Morley (2000) TIES 2003, Page 13

60 SOBIBI vs B-IBI on PSL Streams 1997 & 1998 data, taken from Morley (2000) study region: Puget Sound Lowland (PSL) 16 watersheds (> 16 sites), 10 metrics TIES 2003, Page 13

61 TIES 2003, Page 14

62 o o TIES 2003, Page 14

63 Thornton Creek, one of the first basins in the Puget Sound region to be developed, is now one of the most heavily urbanized. The headwaters of this creek drain one of the oldest shopping malls in the nation. Fourteen of the 16 study basins typically had one or two B- IBI monitoring sites. Two basins (Little Bear and Swamp Creek) were sampled at nine and eight sites respectively (Figure 4). Multiple sites were selected to examine variation in biological condition within these heterogeneous basins. N Swamp Creek W S E Little Bear Creek Sammamish River # Kilometers Figure 4. Distribution of study sites ( ) along Swamp and Little Bear Creek. These two basins are in close proximity to each other and are of similar size, gradient, and geology. They differ primarily in the extent and pattern of urbanization relative to the stream channel. Note that only sites on the main channel of these streams were considered for within basin analysis and that sites immediately below restoration projects were excluded from land cover analysis. Diagnostic evaluation. The relationships between B-IBI and substrate and flow features were tested across a sub-set of sites selected for land cover analysis. Substrate data were provided by a concurrent study at 18 invertebrate monitoring sites in 1997 (C.P. TIES 2003, Page 15

64 SOBIBI vs B-IBI on PSL Streams 1997 & 1998 data, taken from Morley (2000) study region: Puget Sound Lowland (PSL) 16 watersheds (> 16 sites), 10 metrics TIES 2003, Page 16

65 SOBIBI vs B-IBI on PSL Streams 1997 & 1998 data, taken from Morley (2000) study region: Puget Sound Lowland (PSL) 16 watersheds (> 16 sites), 10 metrics Morley provides explicit scoring criteria based on previous work by Karr & Co.: TIES 2003, Page 16

66 SOBIBI vs B-IBI on PSL Streams 1997 & 1998 data, taken from Morley (2000) study region: Puget Sound Lowland (PSL) 16 watersheds (> 16 sites), 10 metrics Morley provides explicit scoring criteria based on previous work by Karr & Co.: metric ) total taxa [0,14) [14,28) ) % dominance >75 (55,75] [0,55].. 10) % predators [0,4.5) [4.5,9) 9 TIES 2003, Page 16

67 SOBIBI vs B-IBI on PSL Streams 1997 & 1998 data, taken from Morley (2000) study region: Puget Sound Lowland (PSL) 16 watersheds (> 16 sites), 10 metrics Morley provides explicit scoring criteria based on previous work by Karr & Co.: metric ) total taxa [0,14) [14,28) ) % dominance >75 (55,75] [0,55].. 10) % predators [0,4.5) [4.5,9) 9 TIES 2003, Page 16

68 SOBIBI vs B-IBI on PSL Streams SAMPLING: 3 replicates per site TIES 2003, Page 17

69 SOBIBI vs B-IBI on PSL Streams SAMPLING: 3 replicates per site metric value = average over replicates (B-IBI & SOBIBI) TIES 2003, Page 17

70 SOBIBI vs B-IBI on PSL Streams SAMPLING: 3 replicates per site metric value = average over replicates (B-IBI & SOBIBI) except # long-lived taxa and # intolerant taxa: metric value = sum over replicates (B-IBI only) TIES 2003, Page 17

71 SOBIBI vs B-IBI on PSL Streams SAMPLING: 3 replicates per site metric value = average over replicates (B-IBI & SOBIBI) except # long-lived taxa and # intolerant taxa: metric value = sum over replicates (B-IBI only) measure of human influence: % urbanized area TIES 2003, Page 17

72 SOBIBI vs B-IBI on PSL Streams 1997 B IBI TIES 2003, Page 18

73 SOBIBI vs B-IBI on PSL Streams 1997 B IBI corr = B IBI SOBIBI 1997 SOBIBI TIES 2003, Page 19

74 SOBIBI vs B-IBI on PSL Streams 1997 B IBI corr = B IBI SOBIBI 1997 SOBIBI SOBIBI retains general dist n shape highly correlated with B-IBI continuous scale no tied sites TIES 2003, Page 19

75 SOBIBI vs B-IBI on PSL Streams 1998 B IBI corr = B IBI SOBIBI 1998 SOBIBI TIES 2003, Page 20

76 SOBIBI vs B-IBI on PSL Streams 1998 B IBI corr = B IBI SOBIBI 1998 SOBIBI O TIES 2003, Page 20

77 SOBIBI vs B-IBI on PSL Streams 1998 B IBI corr = B IBI SOBIBI 1998 SOBIBI O gap SOBIBI gives clearer distinction of healthy sites?? TIES 2003, Page 20

78 SOBIBI vs B-IBI on PSL Streams : BIBI % urban TIES 2003, Page 21

79 SOBIBI vs B-IBI on PSL Streams 1997 : BIBI / o: SOBIBI (rescaled) % urban TIES 2003, Page 22

80 SOBIBI vs B-IBI on PSL Streams 1997 : BIBI / o: SOBIBI (rescaled) : BIBI % urban % urban TIES 2003, Page 23

81 SOBIBI vs B-IBI on PSL Streams 1997 : BIBI / o: SOBIBI (rescaled) 1998 : BIBI / o: SOBIBI (rescaled) % urban % urban TIES 2003, Page 24

82 SOBIBI vs B-IBI on PSL Streams 1997 : BIBI / o: SOBIBI (rescaled) 1998 : BIBI / o: SOBIBI (rescaled) % urban % urban SOBIBI and B-IBI tails coincide TIES 2003, Page 24

83 SOBIBI vs B-IBI on PSL Streams 1997 : BIBI / o: SOBIBI (rescaled) 1998 : BIBI / o: SOBIBI (rescaled) % urban % urban SOBIBI and B-IBI tails coincide GOOD! TIES 2003, Page 24

84 SOBIBI vs B-IBI on PSL Streams 1997 : BIBI / o: SOBIBI (rescaled) 1998 : BIBI / o: SOBIBI (rescaled) % urban % urban SOBIBI and B-IBI tails coincide GOOD! SOBIBI indicates smaller disparity among mid-ranked sites TIES 2003, Page 24

85 SOBIBI vs B-IBI on PSL Streams 1997 : BIBI / o: SOBIBI (rescaled) 1998 : BIBI / o: SOBIBI (rescaled) % urban % urban SOBIBI and B-IBI tails coincide GOOD! SOBIBI indicates smaller disparity among mid-ranked sites SOBIBI missed REAL disparity? TIES 2003, Page 24

86 SOBIBI vs B-IBI on PSL Streams 1997 : BIBI / o: SOBIBI (rescaled) 1998 : BIBI / o: SOBIBI (rescaled) % urban % urban SOBIBI and B-IBI tails coincide GOOD! SOBIBI indicates smaller disparity among mid-ranked sites SOBIBI missed REAL disparity? or... TIES 2003, Page 24

87 SOBIBI vs B-IBI on PSL Streams 1997 : BIBI / o: SOBIBI (rescaled) 1998 : BIBI / o: SOBIBI (rescaled) % urban % urban SOBIBI and B-IBI tails coincide GOOD! SOBIBI indicates smaller disparity among mid-ranked sites SOBIBI missed REAL disparity? or... B-IBI too variable?? TIES 2003, Page 24

88 A Bootstrap Study TIES 2003, Page 25

89 A Bootstrap Study GOAL: Compare distr nal properties between B-IBI and SOBIBI TIES 2003, Page 25

90 A Bootstrap Study GOAL: Compare distr nal properties between B-IBI and SOBIBI METHOD: resample organisms from (observed) field samples TIES 2003, Page 25

91 A Bootstrap Study GOAL: Compare distr nal properties between B-IBI and SOBIBI METHOD: resample organisms from (observed) field samples obtain 10,000 bootstrap samples TIES 2003, Page 25

92 A Bootstrap Study GOAL: Compare distr nal properties between B-IBI and SOBIBI METHOD: resample organisms from (observed) field samples obtain 10,000 bootstrap samples DATA: Morley s 1997 data TIES 2003, Page 25

93 TAXON ID CLING? INTOL? PRED?... COUNT IN (FIELD) SAMPLE 1 Y N N N Y Y Y Y Y TOTAL COUNT TIES 2003, Page 26

94 A Bootstrap Study means: SOBIBI means: B IBI e 15 0e00 2e TIES 2003, Page 27

95 A Bootstrap Study means: SOBIBI means: B IBI e 15 0e00 2e note different scales for SOBIBI and B-IBI TIES 2003, Page 27

96 A Bootstrap Study means: SOBIBI means: B IBI e 15 0e00 2e note different scales for SOBIBI and B-IBI B-IBI distr n JAGGED! TIES 2003, Page 27

97 A Bootstrap Study means: SOBIBI means: B IBI e 15 0e00 2e note different scales for SOBIBI and B-IBI B-IBI distr n JAGGED! from metric discretization? TIES 2003, Page 27

98 A Bootstrap Study means: SOBIBI means: B IBI e 15 0e00 2e note different scales for SOBIBI and B-IBI B-IBI distr n JAGGED! from metric discretization? TIES 2003, Page 28

99 A Bootstrap Study means: SOBIBI means: B IBI e 15 0e00 2e note different scales for SOBIBI and B-IBI B-IBI distr n JAGGED! from metric discretization? B-IBI large bias!! TIES 2003, Page 29

100 A Bootstrap Study means: SOBIBI means: B IBI u 3 l 2e 15 0e00 2e note different scales for SOBIBI and B-IBI B-IBI distr n JAGGED! from metric discretization? B-IBI large bias!! VERY POOR POOR FAIR GOOD EXCELLENT TIES 2003, Page 30

101 A Bootstrap Study SDs: SOBIBI SDs: B BIBI TIES 2003, Page 31

102 A Bootstrap Study SDs: SOBIBI SDs: B BIBI ENORMOUS BIAS for B-IBI!! TIES 2003, Page 31

103 A Bootstrap Study SDs: SOBIBI SDs: B BIBI ENORMOUS BIAS for B-IBI!! need rescaling to compare variabilities TIES 2003, Page 31

104 A Bootstrap Study SDs: SOBIBI SDs: B BIBI ENORMOUS BIAS for B-IBI!! need rescaling to compare variabilities examine ranges TIES 2003, Page 31

105 A Bootstrap Study Ranges: SOBIBI Ranges: B IBI TIES 2003, Page 32

106 A Bootstrap Study Ranges: SOBIBI Ranges: B IBI note true range of 38 never attained by bootstrap samples TIES 2003, Page 32

107 A Bootstrap Study Ranges: SOBIBI Ranges: B IBI note true range of 38 never attained by bootstrap samples give B-IBI handicap TIES 2003, Page 32

108 A Bootstrap Study Ranges: SOBIBI Ranges: B IBI note true range of 38 never attained by bootstrap samples give B-IBI handicap divide SOBIBI by smallest observed range (27) divide B-IBI by largest possible range (40) TIES 2003, Page 32

109 A Bootstrap Study SDs: SOBIBI SDs: B BIBI SDs: SOBIBI/27 SDs: B IBI/ TIES 2003, Page 33

110 A Bootstrap Study SDs: SOBIBI SDs: B BIBI SDs: SOBIBI/27 SDs: B IBI/ sd = sd = TIES 2003, Page 34

111 A Bootstrap Study minima: SOBIBI maxima: SOBIBI minima: B IBI maxima: B IBI l u 4 l 4 u 5 l 5 u TIES 2003, Page 35

112 A Bootstrap Study B IBI worst site (true=th1) B IBI best site (true=ro1) LB4 MI1 TH1 BB2 LB1 RO1 SOBIBI worst site (true=th1) SOBIBI best site (true=ro1) MI1 TH1 RO1 TIES 2003, Page 36

113 A Bootstrap Study B IBI 2nd worst site (true=mi1) B IBI 2nd best site (true=bb2,lb1) BS1 LB3 LB4 MI1 TH1 BB1 BB2 BB3 LB1 RO1 SW3 SOBIBI 2nd worst site (true=mi1) SOBIBI 2nd best site (true=lb1) MI1 TH1 BB2 BB3 BB5 JE1 LB1 TIES 2003, Page 37

114 A Bootstrap Study Corr(SOBIBI,%urban) Corr(B IBI,%urban) Corr(SOBIBI, B IBI) TIES 2003, Page 38

115 A Bootstrap Study Corr(SOBIBI,%urban) Corr(B IBI,%urban) Corr(SOBIBI, B IBI) both yield same info on stream health TIES 2003, Page 38

116 A Bootstrap Study Corr(SOBIBI,%urban) Corr(B IBI,%urban) Corr(SOBIBI, B IBI) both yield same info on stream health But TIES 2003, Page 38

117 A Bootstrap Study Corr(SOBIBI,%urban) Corr(B IBI,%urban) Corr(SOBIBI, B IBI) both yield same info on stream health But SOBIBI MUCH MORE PRECISE!!! TIES 2003, Page 38

118 Insight TIES 2003, Page 39

119 Insight SOBIBI has non-subjective scoring TIES 2003, Page 39

120 Insight SOBIBI has non-subjective scoring SOBIBI drastically reduces time spent on TIES 2003, Page 39

121 Insight SOBIBI has non-subjective scoring SOBIBI drastically reduces time spent on choosing reference sites TIES 2003, Page 39

122 Insight SOBIBI has non-subjective scoring SOBIBI drastically reduces time spent on choosing reference sites (re)calibration of metric scoring mechanism TIES 2003, Page 39

123 Insight SOBIBI has non-subjective scoring SOBIBI drastically reduces time spent on choosing reference sites (re)calibration of metric scoring mechanism SOBIBI has higher specificity for most and least impacted sites TIES 2003, Page 39

124 Insight SOBIBI has non-subjective scoring SOBIBI drastically reduces time spent on choosing reference sites (re)calibration of metric scoring mechanism SOBIBI has higher specificity for most and least impacted sites SOBIBI is much less variable TIES 2003, Page 39

125 Insight SOBIBI has non-subjective scoring SOBIBI drastically reduces time spent on choosing reference sites (re)calibration of metric scoring mechanism SOBIBI has higher specificity for most and least impacted sites SOBIBI is much less variable thus year-to-year difference in SOBIBI for fixed site likely indicates changing stream health TIES 2003, Page 39

126 Insight SOBIBI has non-subjective scoring SOBIBI drastically reduces time spent on choosing reference sites (re)calibration of metric scoring mechanism SOBIBI has higher specificity for most and least impacted sites SOBIBI is much less variable thus year-to-year difference in SOBIBI for fixed site likely indicates changing stream health not necessarily so for B-IBI TIES 2003, Page 39

127 Farewell This presentation is available in PDF format at weblink THANK YOU! TIES 2003, Page 40