Site Classification for Re-calibration of the Alabama Index of Stream Macroinvertebrate Condition...

30
Site Classification for Re-calibration of the Alabama Index of Stream Macroinvertebrate Condition Ben Jessup and Jen Stamp Tetra Tech, Inc. SWPBA November 14, 2012 Lake Guntersville State Park

Transcript of Site Classification for Re-calibration of the Alabama Index of Stream Macroinvertebrate Condition...

Site Classification for Re-calibration of the

Alabama Index of Stream Macroinvertebrate Condition

Ben Jessup and Jen StampTetra Tech, Inc.

SWPBANovember 14, 2012

Lake Guntersville State Park

Why Classify?• To allow comparable judgments• Apples and oranges• Level playing field• Natural conditions

Why Classify in Alabama?

• The index of stream macroinvertebrate condition should be recalibrated

• Initial index development included samples from a limited portion of the state

• New samples have been collected from all areas of the state

• Site classification is the first step in index recalibration

Sites with benthic data collected from 1994-2005 Sites with post-2005 benthic data

Process for Site Classification

• Data compilation– Disturbance variables– Classification variables– Benthic samples and metrics

• Reference site identification– Based on the HDG– Confirmed with site data– Verified by ADEM

• Classification Analyses

Reference Site Identification

• Primary Reference Indication– HDG categories 1, 2, and 3 = Reference– HDG categories 6, 7, and 8 = Stressed

• Site data confirmationEnvironmental Variable Reference Near Ref Stressed

Specific Conductance <100.0 <200.0 >300.0Atrazine <0.06 <0.08 >0.10Chemical/Biological Oxygen Demand (CBOD5) <1.8 <2.0 >2.2Fecal Coliform (BacT) <500 <750 >1000Chlorophyll a (Chl_a) <3.0 <4.0 >5.0Total Dissolved Solids (TDS) <75 <120 >150Total Phosphorus (TP) <0.06 <0.13 >0.20Nitrate + Nitrite (NO2NO3) <0.10 <0.25 >0.40Habitat Score >67 >55 <55

Calibration dataset

• 130 reference sites• 113 stressed sites

• Benthic samples • Collected April-August from

1993-present• One sample per site used in

calibration

45

65

67

6871

75

Classification Analysis

• Find the environmental factors that affect biology

• Mapping• PCA of classification variables

– All sites • NMS Ordination of sites by taxa

– Reference only– Operational Taxonomic Units (OTU)

• Cluster analysis and DFA• Comparison of metric distributions

– Box plots and bi-plots

Temperature

Elevation

Level 4 Ecoregions

• Fall Line• Special considerations

– 65a & b – Blackland Prairies,

Flatwoods, & Margins– 68 c and d– Plateau Escarpment &

Southern Table Plateaus

PCA

• Classification variables only• All sites

Strongest drivers on each axis

Factor 1, 22%Temperature, Latitude

Factor 2, 16%Area, Gradient

Factor 4, 9%Precipitation, Day of year

Avg Temp -0.88% Cobb.,Bould. 0.64Elevation 0.84Latitude 0.86

Gradient -0.68SlopeNHD -0.59Discharge 0.56

Width 0.66Area 0.82

Factor 3, 13%Discharge, IonsDisch:Area -0.68Discharge -0.58Hardness 0.63Alkalinity 0.64

Precipitation -0.53Day of year 0.52

4 5 A4 5 B

4 5 D6 5 A

6 5 B6 5 D

6 5 F6 5 G

6 5 I6 5 J

6 5 P6 5 Q

6 7 F6 7 G

6 7 H6 8 B

6 8 C6 8 D

6 8 E6 8 F

7 1 F7 1 G

7 1 H7 1 J

7 5 I

-3

-2

-1

0

1

2

3

PC

AF

act

1

PCA F1 by ecoregion

45 65 67

68 71

75Fact

or 1

(Tem

pera

ture

, Lati

tude

)

PCA F2 by ecoregion

4 5 A4 5 B

4 5 D6 5 A

6 5 B6 5 D

6 5 F6 5 G

6 5 I6 5 J

6 5 P6 5 Q

6 7 F6 7 G

6 7 H6 8 B

6 8 C6 8 D

6 8 E6 8 F

7 1 F7 1 G

7 1 H7 1 J

7 5 I

E CO RE G IO N_ L 4

-3

-2

-1

0

1

2

3

4

PC

AF

act

2Fa

ctor

2 (A

rea,

Gra

dien

t)

45 65 67 6871

75

PCA F3 by ecoregionFa

ctor

3 (D

isch

arge

, Ion

s)

45 65 67 6871

75

PCA F4 by ecoregion

4 5 A4 5 B

4 5 D6 5 A

6 5 B6 5 D

6 5 F6 5 G

6 5 I6 5 J

6 5 P6 5 Q

6 7 F6 7 G

6 7 H6 8 B

6 8 C6 8 D

6 8 E6 8 F

7 1 F7 1 G

7 1 H7 1 J

7 5 I

E CO RE G IO N_ L 4

-4

-3

-2

-1

0

1

2

3

4

PC

AF

act

4Fa

ctor

4 (P

reci

pita

tion,

Day

of y

ear)

45 65 67 6871

75

NMS

NMS_130PA

Axis 1

Axi

s 2

ECOREGION

45 Piedmont 65 S.E. Plains 67 Rdg + Val 68 S.W. Apps71 Int. Plat.

NMS_130PA

Axis 1

Axi

s 2

ECOREGION

45 Piedmont 65 S.E. Plains 67 Rdg + Val 68 S.W. Apps71 Int. Plat. • Standardized

taxonomy• Presence/absence

NMS_130PA

Axis 1

Axi

s 2

0

200

400

600Elev_m

Axis 1r = .378 tau = .291

Axis 2r = -.442 tau = -.346

0 200 400 600

FallLine_Code

12

NMS_130PA

Axis 1

Axi

s 2

FallLine

Above Below

NMS_130PA

Axis 1

Axi

s 2

ElevGrp

<150 >150

NMS of samples by taxaNMS_130PA

Axis 1

Axi

s 2

FallLine_Code

12

NMS_130PA

Axis 1

Axi

s 2

FallLine

Above Below

Icthyoregion FL

Tennessee Valley Above

Ridge and Valley/Piedmont Above

Plateau Above

Hills and Coastal Terraces Below

Southern Plains Below

Blackland April

O'Neil and Shepard, 2007

Cluster AnalysisNMS_130PA

Axis 1

Axi

s 2

PA-5FB3

1216

NMS_130PA

Axis 1

Axi

s 2

PA-5FB3

1216

NMS_130PA

Axis 1

Axi

s 3

PA-5FB3

1216

NMS_130PA

Axis 1

Axi

s 2

4grps

12316

Discriminant Function Analysis (DFA)

F-remove p-val

PC_CobbBould 7.72864 0.00

tmax14_C 11.79340 0.00

Elev_m 6.91748 0.00

LONGITUDE 4.70966 0.004

%Correct 1 2 3 16

1 86.8 46 2 2 3

2 11.8 9 2 0 6

3 33.3 3 0 4 5

16 79.2 9 0 1 38

Total 69.2 67 4 7 52

Four clusters

F-remove p-val

PC_CobbleBoulder 17.1 0.00

LONGITUDE 5.8 0.00

StrahlerOrder 3.9 0.02

AvgWidth_ft 2.8 0.06

Three clusters

% Correct 1 2 16

1 83.0 44 4 5

2 17.2 11 5 13

16 66.6 13 3 32

Total 62.3 68 12 50

DFA

• Can predict differences between 2 groups– Elevation, % cobble-boulder

• Does not distinguish 3-4 groups very well– May have something with stream size

Metric DistributionsEP

T Ta

xa %

45 65 67 68 71

Remaining Variability

Bec

ks

EP

T T

axa

(%)

HB

I

clin

ger

taxa

EP

T T

axa

Inse

ct T

axa

Into

lera

nt

Tax

a

Ple

cope

tra

Tax

a

Tric

hopt

era

Tax

a

% S

crap

er

Tol

eran

t ta

xa

(%)

% T

oler

ant

Month -0.12 -0.13 0.12 0.06 0.05 0.10 -0.09 -0.35 0.23 0.54 0.29 0.22

Alkalinity (mg/L) -0.47 -0.35 0.27 -0.53 -0.44 -0.46 -0.46 -0.49 -0.49 0.16 0.33 0.26

Hardness (mg/L) -0.59 -0.42 0.40 -0.58 -0.50 -0.50 -0.56 -0.61 -0.47 0.21 0.43 0.34

PC_CobbleBoulder 0.44 0.42 -0.54 0.27 0.30 0.11 0.42 0.32 0.35 -0.03 -0.34 -0.44

AvgPC_Riffle 0.67 0.57 -0.59 0.48 0.53 0.35 0.65 0.52 0.51 -0.18 -0.53 -0.56

Above Fall-line

Spearman rhoReference only

Metrics can be adjusted individually using residuals of the regression slope

Remaining Variability

% E

PT

HB

I

Em

phem

ero

pte

ra

Taxa

Ple

copetr

a T

axa

% P

lecopte

ra

Tole

rant

taxa (

%)

% T

ole

rant

GRADIENT_ft/mi 0.19 -0.42 -0.08 0.48 0.59 -0.41 -0.29

Hardness (mg/L) 0.01 0.25 0.53 -0.23 -0.41 0.15 0.09

PC_CobbleBoulder 0.35 -0.31 0.36 0.27 0.20 -0.35 -0.52

AvgPC_Riffle 0.62 -0.66 0.23 0.53 0.44 -0.53 -0.64

Below Fall-line

Spearman rhoReference only

Metrics can be adjusted individually using residuals of the regression slope

Conclusions

• The strongest categorical classification variable is the Fall Line, where above and below groups are defined by ecoregions

• Variability in some metrics remains after categorical classification, based on % riffles, % cobble/boulders, and hardness

• Individual metrics can be adjusted continuously or in steps

Classification Errors

• Yes• Site classes are generalizations

– We cannot assign each site to its own class– We group for convenience and statistics– We lose specificity

• If we define using ecoregional borders, some sites will be close to the edge (transitional)

• Some sites will be natural anomalies

Abstract• In 2008, the Alabama Department of Environmental Management (ADEM) developed

macroinvertebrate indices for high and low gradient streams using 205 sites from ADEM’s 1994-2005 datasets. These sites covered a limited portion of the state. An additional 285 macroinvertebrate bioassessments were conducted statewide between 2006-2011, filling in gaps in the spatial coverage of sites. ADEM sought to confirm or refine the classification of sites for biological assessments as a first step in recalibrating the indices. Using the human disturbance gradient (which includes land uses), water quality, and physical habitat, 130 reference stream sites were identified in all of the major ecoregions. Principal components analysis (PCA) showed how natural environmental variables in reference sites differed among level 4 ecoregions and were related to stressor variables and macroinvertebrate metrics. With the basic patterns apparent in reference sites, the PCA was run again with reference and non-reference sites to help classify sites in regions with low reference representation. The PCA axes related to biological variability were used to categorize sites in groups of level 4 ecoregions with similar biologically relevant characteristics. Taxa ordination was used for further exploration of biological types and to confirm final site classes. Site classes in Alabama were based on level 4 ecoregions, with stream size, slope, and water hardness as modifiers.

Discussion Agenda

• Questions on the classification analysis• Exceptional stream types• Next Steps

– Respond to questions– Develop final classes and metric adjustments– Test index alternatives– Test index in exceptional stream types