pre-harvest calibration of a paired and nested

1

Pre-harvest Calibration of a Paired and Nested Watershed Study to Evaluate Event-Based

Suspended Sediment Export

The Little Creek StudySwanton Pacific Ranch

Davenport, CA

Michael C. GaedekeAugust, 2007

Project DirectorBrian Dietterick

Presentation Outline

• Location overview• Management history• Little creek monitoring project

• Goals• Study design• General timeline• Water quality data collected and

stat analysis

Management history

Looking Downstream in North Fork ~1910

Looking Upstream in North Fork ~2000

Little Creek study goals and current analysis

• Scientifically document water quality and channel conditions before, during, and after single-tree and small group selection harvests

• Evaluate the effectiveness of current forest practice rules and best management practices for timber harvesting activities in maintaining existing water quality and channel conditions

• Current analysis– Analyze the storm event water quality data from the Little Creek

watershed to assess the calibration (pre-treatment phase) of the Little Creek Study.

– Describe the existing variability– Determine the magnitude of change capable of being detected in

the post-treatment period

2

Paired Watershed Design•Control watershed: South Fork•Treatment watershed: North Fork

Nested Watershed Design•Control Watershed: Upper North Fork•Treatment watershed: North Fork

Timeline and study designCalibration period 2001-2008

•Measure existing water qualityTreatment in 2008

•Harvest portion of watershed between the North Fork and Upper North Fork stations

Post-treatment 2008-2011+•Measure for potential change in water quality

Field data collection – stage and streamflow

NF Stag e (ft)

NF

Flo

w (

ft^

3/s)

1.00.90.80.70.60. 50.40.30.2

25

20

15

10

5

0

S 0.573547R-Sq 99.6%R-Sq(adj) 99.5%

Fitted Line PlotNF Flow = - 1 .476 - 1.2 21 NF Stage

+ 29.10 NF S tage**2

S F Stag e (ft)

SF F

low

(ft̂

3/s)

1.21.11.00.90.80. 7

12

10

8

6

4

2

0

S 0.133903R- Sq 99.9%

R- Sq(adj) 99.9%

Fitted Line PlotS F Flow = 16.7 2 - 51.81 SF Stage

+ 40.2 0 SF St ag e* *2

A. Flow Meter

B. Pressure

Transducer

C. FW-1 Stage

Recorder

D. Staff Gage

Stage

B

C

D

A

Field data collection – water quality samples

• 1-hour interval samples• Collection of 24 bottles

before swap

• Lab analysis of 1 hour samples

• Turbidity• Units: nephlometric turbidity

units (NTUs)

• Suspended Sediment Concentration (SSC)

• Units: mg/L

Lab Water Quality Testing

A. Turbidimeter B. Scale

AB

Defining the dataset – Predicting SSC from turbidity

• Need for SSC to be predicted from turbidity– Turbidity shown to be a better predictor than flow

• Previous research has indicated SSC versus turbidity is a variable relationship that is best defined on an event basis

• Regression analysis used to establish relationship.– Regressions assessed based on r2, p-value, residual

plots, and fits– Data transformations when necessary

3

Defining the dataset – storm events

• Define storm events based on the hydrograph using Hewlett and Hibbard (1967) 0.05 slope method– Separates storm flow from base flow

• Minimum storm event size based on turbidity– Peak must be greater than 20 NTUs– Only samples >20 NTUs analyzed for SSC

• Storm event ends when turbidity drops below 20 NTUs or the 0.05 slope line intersects the hydrograph

End event

Event load calculation

• Calculate individual hourly loads to determine event loads– High temporal variability requires hourly sums

• Event loads establish the calibration dataset– Changes in the relationship used to detect

change

• Events with complete SSC and flow datasets used for analysis– Must have both datasets to calculate loads

Paired and nested analysis after transformations

Paired

Nested

South Fork load (ln[kg/ha])

No

rth

For

k l

oa

d (

ln[k

g/

ha]

)

543210-1-2-3

7

6

5

4

3

2

1

0

S 1. 00230R -Sq 64.8%

R -Sq( adj) 63.4%

Regression95% CI

Fitted Line PlotNorth Fork load (ln[kg/ha]) = 2.023 + 0.7657 South Fork load (ln[kg/ha])

Upper N. Fork load (ln[kg/ha])

No

rth

Fo

rk l

oa

d (

ln[k

g/

ha]

)

6543210-1-2

6

5

4

3

2

1

0

-1

S 0.382495

R -Sq 95.5%R -Sq( adj) 95.3%

Regression95% C I

Fitted Line PlotNorth Fork l oad (ln[kg/ha]) = 0.5090 + 0.8396 Upper N. Fork load (ln[k g/ha])

Assessing detectable magnitude of change using confidence intervals

• Back-transformation of confidence interval into non-logarithmic numbers not valid

• Generate a synthetic dataset representing percentage increases over the original dataset

• Transform the new dataset, perform regression, and compare new regression line to original confidence interval

New regression line comparison to existing confidence intervals for NF versus SF


No

rth

Fo

rk l

oa

d (

ln[k

g/

ha

])

543210-1-2-3

7

6

5

4

3

2

1

0

Existing conditionsNF load increased 30%

Associated RegressionsData Points and

30% Increase in Existing NF Conditions, NF v SF

95% C.I. for existing conditions-------


No

rth

Fo

rk l

oa

d (

ln[k

g/

ha

])

543210-1-2-3

7

6

5

4

3

2

1

0






No

rth

Fo

rk l

oa

d (

ln[k

g/

ha

])

543210-1-2-3

7

6

5

4

3

2

1

0






No

rth

Fo

rk l

oa

d (

ln[k

g/h

a])

543210-1-2-3

7

6

5

4

3

2

1

0

-1





4

New regression line comparison to existing confidence intervals for NF versus UNF

Upper North Fork load (ln[kg/ha])

No

rth

Fo

rk l

oa

d (

ln[k

g/

ha

])

6543210-1-2

6

5

4

3

2

1

0



95% C.I. for existing conditions

10% Increase in Existing NF Conditions, NF v UNF

-------

Upper North Fork load (ln[kg/ha])

No

rth

Fo

rk l

oa

d (

ln[k

g/h

a])

6543210-1-2

6

5

4

3

2

1

0



95% C.I. for existing conditions

30% Increase in Existing NF Conditions, NF v UNF

-------

Conclusions• Suspended sediment versus turbidity relationship

has allowed for prediction of suspended sediment data– Established on an event basis

• Data thus far has yielded a sufficient number of event-specific suspended sediment loads to enable simple linear regression analysis– Nested (NF versus UNF) relationship indicates less

variability than paired (NF versus SF) relationship– Narrower confidence intervals for magnitude of change

detection for nested relationship

Other Little Creek Study Components

• Annual geomorphic surveys– Longitudinal profiles and cross sections to

detect potential sediment source/sink areas

• LIDAR mapping analysis (Russ White)– Stream channel and road features under forest

canopy– Comparison with conventional surveys

Questions?

pre-harvest calibration of a paired and nested

Documents