Assessing Morphological Performance of Stream Restoration in North Carolina

Barbara Doll, PE, Ph.D., Extension Specialist NC Sea Grant and Biological and Agricultural Engineering

Department, NC State University

Acknowledgements Co-Investigators

Greg Jennings, BAE Dept., NC State University (retired) Jean Spooner, BAE Dept., NC State University Dave Penrose, BAE Dept., NC State University (retired) Joseph Usset, Statistics Dept., NC State University (Ph.D.

candidate) Mark Fernandez, NC State University, BAE Dept.

(Master’s student) Jamie Blackwell, BAE Dept., NC State University Michael Shaffer, BAE Dept., NC State University

NC Clean Water Management Trust Fund Field Work – Dave Penrose, Greg Jennings, Mike

Shaffer, Karen Hall, Mark Fernandez, Dan Clinton, Lara Rozell, Jess Roberts, numerous other students

What is stream restoration? The process of converting an unstable, altered or degraded

stream corridor, including adjacent riparian zone and floodprone areas to its natural or referenced, stable conditions considering recent and future watershed conditions (NC DWQ)

Natural Channel Design (Hey, 2006)

“Rosgen” Method

Fluvial geomorphological method for designing NATURAL STABLE CHANNELS

Analogue procedure - cross-sectional area and pattern

relationships (i.e. sinuosity) are scaled from a natural stable reference stream to determine the restoration design

High-quality “reference” streams

serve as design templates

Determine Restoration Potential

Performance Range

Reference Reaches

Disturbed Channels

Restored Streams

Research Goal: • Develop tools for measuring functional uplift to advance the practice of stream restoration.

Research Objectives Obj. 1: Develop and evaluate methods for assessing

eco-geomorphological conditions of restored streams. Obj. 2: Compare condition of restored streams to

impaired and high quality reference channels. Obj. 3 Develop a “scale” for evaluating restoration

need and performance Obj. 4: Determine if location, site selection and design

relate to the resulting condition of restored streams

Eco-geomorphological = integration of hydrology, fluvial geomorphology and ecology in river systems

Scope of the Project

Visited 156 streams between 2006 – 2012 Applied five rapid stream assessment methods Sampled macroinvertebrate communities from

85 restored streams Compiled restoration design data for 79 streams Conducted watershed assessment for 130 streams Performed extensive multivariate statistical

analyses

Obj. 1 - Develop and Evaluate Stream Assessment Tools

n=65 restored streams Quantitative/Qualitative

EGA - Eco-Geomorphological Assessment (NCSU for CWMTF)

Qualitative (visual) SPA - Stream Performance Assessment (NCSU) SVAP - Stream Visual Assessment Protocol (USDA) RCE - Riparian, Channel and Environmental Inventory –

(Peterson) RBP- Rapid Bioassessment Protocols – habitat survey

(US EPA)

Eco-Geomorphologial Assessment EGA

A. Channel Condition

C. Aquatic Insects

B. Bank and Riparian Habitat

D. Condition and Function of Structures

Evaluation Categories

Sub-Categories # of variables

Points

Channel Condition

Bedform Condition 10 20 Dominant Substrate Material 3 12 Streambank Stability 6 24

Riparian Habitat

Riparian Vegetation 5 20 Floodplain Condition 6 24

Macro invertebrates

Community Structure 5 24 Cover and Refuge 12 20

In-stream Structures

Structure Function 4 16 Structure Condition 3 12

Total Score 54 172

Stream Performance Index (SPA) • Channel bedform

• Channel pattern

• Floodplain connection

• In-stream habitat features

• Sediment transport

• Streambank Condition

• Streambank vegetation

Rapid Visual Assessment of 17 Variables; Total Points = 110

Five stream assessment methods applied at 65 restored streams – EGA, SPA, RBP, RCE & SVAP

Objective 1 – Continued Evaluate Stream Assessment Tools – Determine how well assessments predict macroinvertebrate metrics). Method: Linear regression, principal component analysis (PCA) and principal component regression (PCR) (n=65 restored streams.

Response Variable: Aquatic Insects

Upstream and in-reach sampling compiled as 6 Macroinvertebrate Metrics :

No. of dominant taxa No. of dominant EPT taxa EPT abundance Dominant taxa in common DIC (%) % shredders and predators Number of indicator taxa

Dominant Taxa

EPT Taxa EPT

Abundance

% Shredders

& Predators

Indicator Taxa

DIC

R2 p R2 p R2 p R2 p R2 p R2 p

EGA 0.24 *** 0.29 *** 0.23 *** ns 0.26 *** ns SPA 0.07 · 0.10 * 0.07 · 0.09 · 0.14 * ns RBP 0.31 *** 0.37 *** 0.33 *** ns 0.42 *** ns RCE 0.28 *** 0.29 *** 0.26 *** ns 0.31 *** ns SVAP 0.18 ** 0.26 *** 0.25 *** ns 0.33 *** ns

Some correlations revealed for Number of Dominant Taxa, No. of Dominant EPT Taxa, EPT Abundance and No. of Indicator Taxa with all five assessment scores. However, variability is high.

Hypotheses: Prediction can be improved by 1) addressing collinearity and subjective variable weights and by 2) adding watershed factors.

Principal Component Analysis (PCA)

Dimension Reduction High-dimensional data with collinearity (lots of

variables that correlate) Reveals underlying structure in the data Produces a number of independent artificial

variables (called principal components or PCs) PCs are linear combinations of the original

variables. The weighted factor for each variable reflects its relative importance in explaining the variability of the specific PC

Dominant

Taxa EPT Taxa EPT

Abundance

% Shredders

& Predators

Indicator Taxa

No. of Variables

Total No. of PC's

% Variability Explained

EGA Total Raw Points 0.24 0.29 0.23 0.03 0.26 1 PCA EGA 0.61 0.68 0.62 0.29 0.58 44 11 76.3% SPA Total Raw Points 0.07 0.10 0.07 0.09 0.14 1 PCA SPA 0.49 0.47 0.32 0.16 0.39 17 7 78.1% RBP Total Raw Points 0.31 0.37 0.33 0.03 0.42 1 PCA RBP 0.37 0.46 0.39 0.20 0.45 13 5 77.4% RCE Total Raw Points 0.28 0.29 0.26 0.01 0.31 1 PCA RCE 0.61 0.69 0.65 0.19 0.65 19 8 77.6% SVAP Total Raw Points 0.18 0.26 0.25 0.02 0.33 1 PCA SVAP 0.59 0.70 0.65 0.09 0.66 14 6 78.6%

R-squared from Linear Regression

Watershed Assessment

using GIS

0

2

4

6

8

10

12

14

16

0.0 10.0 20.0 30.0 40.0 50.0 60.0

EPT taxa vs. Impervious Cover %

0

2

4

6

8

10

12

14

16

0.0 20.0 40.0 60.0 80.0 100.0 120.0

EPT taxa vs. Developed %

0

2

4

6

8

10

12

14

16

50.0 55.0 60.0 65.0 70.0 75.0 80.0 85.0 90.0

EPT taxa vs. CN

0

2

4

6

8

10

12

14

16

0.1 1.0 10.0 100.0

EPT taxa vs. Watershed Size

0

2

4

6

8

10

12

14

16

0 100 200 300 400 500

EPT taxa vs. Time of Concentration

y = 2.54ln(x) + 14.55R² = 0.59

0

2

4

6

8

10

12

14

16

0.0010 0.0100 0.1000 1.0000

EPT taxa vs. Basin Slope

V1= Basin Slope V2=Time of Concentration V3=Watershed Size

V4 = Curve Number V5=% Developed V6=% Impervious

First three PC’s for Watershed (90% Variance Explained)

Dominant

Taxa EPT Taxa EPT

Abundance

% Shredders

& Predators

Indicator Taxa

No. of Variables

Total No. of PC's

% Variability Explained

EGA Total Raw Points 0.24 0.29 0.23 0.03 0.26 1 PCA EGA 0.61 0.68 0.62 0.29 0.58 44 11 76.3% PCA (EGA +Watershed) 0.74 0.81 0.72 0.26 0.70 50 12 77.3% SPA Total Raw Points 0.07 0.10 0.07 0.09 0.14 1 PCA SPA 0.49 0.47 0.32 0.16 0.39 17 7 78.1% PCA (SPA + Watershed) 0.66 0.68 0.53 0.20 0.56 23 8 78.8% RBP Total Raw Points 0.31 0.37 0.33 0.03 0.42 1 PCA RBP 0.37 0.46 0.39 0.20 0.45 13 5 77.4% PCA (RBP + Watershed) 0.63 0.72 0.59 0.24 0.65 19 6 77.4% RCE Total Raw Points 0.28 0.29 0.26 0.01 0.31 1 PCA RCE 0.61 0.69 0.65 0.19 0.65 19 8 77.6% PCA (RCE + Watershed) 0.77 0.82 0.72 0.17 0.73 25 9 78.2% SVAP Total Raw Points 0.18 0.26 0.25 0.02 0.33 1 PCA SVAP 0.59 0.70 0.65 0.09 0.66 14 6 78.6% PCA (SVAP + Watershed) 0.72 0.79 0.66 0.11 0.66 20 7 79.5% Watershed 0.65 0.70 0.55 0.13 0.52 6 PCA Watershed 0.41 0.43 0.34 0.09 0.40 6 2 78.6%

R-squared from Linear Regression

Conclusion 1 Rapid stream assessments ability to predict

aquatic macroinvertebrate metrics in restored streams can be improved with ordination and addition of watershed variables.

Conclusion 2 Rapid stream assessments best predict EPT,

indicator and total number taxa metrics.

Artwork by Ethan Nedeau

Obj. 2 – Compare eco-geomorphological condition of

restored streams to impaired and high quality reference channels. SPA - (NCSU) 156 Streams: 93 restored, 21 impaired, 29 reference

quality, and 13 reference streams with minor incision

Method: Use PCA and PC-based factor analysis to compare stream performance by stream condition

Stream Locations

First 3 SPA PC’s

explain 57.5 %

of variance n=156

# Variable F1 F2 F3 15 Streambank condition 0.85 0.02 0.16 17 Floodplain function 0.78 -0.01 0.05 16 Streambank vegetation 0.77 0.26 0.06 14 Sediment transport 0.72 -0.09 0.37 6 Pattern 0.64 -0.21 0.12 10 Rootmats 0.22 0.82 0.04 11 Overhanging veg -0.29 0.74 0.12 8 Leaf packets -0.11 0.71 0.15 9 Undercut banks 0.29 0.68 0.02 3 Riffles length slope 0.16 0.14 0.86 1 Riffles pools alternating 0.14 0.11 0.76 2 Riffles pools located 0.33 -0.02 0.73 4 Riffles clean material 0.02 0.18 0.62 12 Rootwads 0.02 0.38 -0.07 7 Large woody debris -0.13 0.35 0.14 5 Pools length depth 0.24 0.02 0.2 13 Boulder clusters 0.05 -0.04 0.13

Proportion Var 19% 15% 15% Cumulative Var 19% 35% 50%

Factor Scores with

Varimax Rotation Note: Varimax maximizes the sum of the variances of the squared loadings

Factor 1 – General Morphologic Condition

# Variable F1 15 Streambank condition 0.85 17 Floodplain function 0.78 16 Streambank vegetation 0.77 14 Sediment transport 0.72 6 Pattern 0.64

Conclusion: General morphologic condition of restored streams is the same as reference streams and different from impaired streams.

Streambank Condition & Vegetation

Floodplain Connection

Channel Pattern

Sediment Transport

Factor 2 - Habitat

# Variable F2 10 Rootmats 0.82 11 Overhanging veg 0.74 8 Leaf packets 0.71 9 Undercut banks 0.68

Conclusion: There is a high range of variability in habitat for recently restored streams

Stable Undercut Banks

Rootmats

In-Stream Habitat

Leaf Packs

Factor 3 - Bedform

# Variable F3 3 Riffles length slope 0.86 1 Riffles pools alternating 0.76 2 Riffles pools located 0.73 4 Riffles clean material 0.62

Conclusion: There is a high range of variability in bedform for recently restored streams

Channel Bedform

• Riffles • Steps • Pools

Conclusion 3 Restored streams are similar to reference streams in

terms of geomorphic conditions, but for bedform and habitat conditions, restored streams have a lower mean score and greater variability.

Obj. 3 – Develop a “scale” for evaluating restoration need and

performance SPA (NCSU) 130 streams: 84 restored (benthic sampling), 21 impaired

and 25 reference quality Watershed Assessment

Method: PCR, least squares and ridge regression used to predict EPT taxa (for 84 restored streams). Cross-Validation for prediction error. Use the “best” regression model to predict outcomes for reference and impaired streams.

Cross Validation indicates that Ridge Regression results in the lowest prediction error. Also, the cross-validation score is the lowest for PCR if 14 PCs are retained.

Prediction Error for 3 Regression Methods (n=84 streams)

Apply Ridge

Model to 130

Streams to predict

Total No. Dominant

EPT Values

n=21 n=25 n=84

Impaired Reference Restored

Predicted No. Dominant EPT Taxa – Ridge Regression Model

n=10 n=11 n=5 n=20 n=31 n=53

Y Intercept 4.33 BS Basin Slope 1.38 %D % Developed -1.26 BC Boulder clusters 0.88 CN CN -0.75 UB Stable Undercut banks 0.65 RM Rootmats 0.46 Pool Pools length depth -0.40 R-P Riffles pools alternating 0.39 SV Streambank vegetation -0.31 PT Pattern -0.28 R Riffles clean material 0.26

OV Overhanging veg -0.25 LP Leaf packets -0.24 RW Rootwads -0.24 LWD Large woody debris 0.21 FF Floodplain function -0.20 RPL Riffles pools located 0.18 ST Sediment transport 0.10 % % Impervious -0.10 RIF Riffles length slope 0.08 SC Streambank condition -0.07 Size Watershed Size 0.04 Tc Time of Concentration 0.04

Dominant EPT Taxa = 4.33 +1.38 *BS-1.26* %D + 0.88* BC =0.75*CN+ 0.65 UB +0.46*RM-0.44 Pool + 0.39*R-P -0.31*SV + etc….

Ridge Regression Equation

A scale for evaluating the “potential” uplift for stream restoration projects can be developed from sampling biologic communities and assessing habitat and watershed in a range of stream conditions and applying ordination and regression statistics.

Note: Macroinvertebrates are not an appropriate metric for urban streams

Conclusion 4

Obj. 4 – Determine if location, site selection and design relate to the resulting eco-geomorphological

condition of restored streams

79 restored streams -benthic macroinvertebrates & watershed assessment

Method: Use PCA, PCR and PC-based factor analysis to determine which factors correlate with benthic metrics. Use Ridge Regression to predict dominant EPT taxa.

Potential explanatory variables for restoration performance and biotic indices

Watershed

• % Impervious % Developed • Runoff Curve Number Basin Slope • Time of Concentration Watershed Size

Landscape

• Ecoregion • Valley Slope • Substrate (D50, D84, % Sand)

Design

• Bankfull Width Bankfull Mean Depth • Width/Depth Ratio Average Channel Slope • Sinuosity Bankfull Cross-Sectional Area • Entrenchment Ratio, ER

Ridge Regression Model Results Morphology +

Watershed

Conclusion: Morphology + Watershed factors explain a substantial amount of variability in EPT taxa.

Dominant EPT Taxa = 4.44 + 1.14*Sval – 1.44*%Dev. +1*Dbkf – 1.28*%Sand+ 0.91*ER- 1.06*CN+ 0.86*BS - etc….

Ridge Regression Equation – 17 Variables

Positive Negative Sval 1.14 % Developed -1.44 Dbkf 1.00 % Sand -1.28 ER 0.91 CN -1.06 Basin slope 0.86 Watershed Size -1.02 D84 0.62 Save -0.49 Wbkf 0.50 K -0.48 % Impervious 0.47 D50 -0.18 Tc 0.17 Abkf -0.16 [W/D] -0.12

Dominant EPT Taxa = 4.44 + 1.14*Sval – 1.44*%Dev. +1*Dbkf – 1.28*%Sand+ 0.91*ER- 1.06*CN+ 0.86*BS - etc….

Ridge Regression Equation After Variable Elimination

Positive Negative Basin Slope 1.66 CN -1.42 ER 1.02 K -0.46 D50 0.81 [W/D] -0.42 Svalley 0.74 Tc -0.05 Wbkf 0.74

Conclusion 1: Larger (wider) streams in steeper valleys with course substrate with un-developed watersheds have more EPT taxa

Dominant EPT Taxa = 4.44 + 1.14*Sval – 1.44*%Dev. +1*Dbkf – 1.28*%Sand+ 0.91*ER- 1.06*CN+ 0.86*BS - etc….

Ridge Regression Equation

Positive Negative Basin Slope 1.66 CN -1.42 ER 1.02 K -0.46 D50 0.81 [W/D] -0.42 Svalley 0.74 Tc -0.05 Wbkf 0.74 Conclusion 2: Wider floodplain widths indicate higher EPT taxa numbers.

Ridge Regression Model Results Morphology + Watershed after

variable elimination

Conclusion: Eliminating factors results in a reduction in the variability explained in the EPT taxa

Conclusion 5 Larger (wider) streams in steeper valleys with larger

substrate and un-developed watersheds have higher numbers of dominant EPT taxa

Conclusion 6 Larger accessible floodplain widths (higher ER

values) correlate with higher EPT taxa values.