Pre-harvest Calibration of a Paired and Nested
Watershed Study to Evaluate Event-Based
Suspended Sediment Export
The Little Creek Study
Swanton Pacific Ranch
Davenport, CA
Presentation Outline
• Location overview
• Management history
• Little creek monitoring project
•
•
•
•
Michael C. Gaedeke
August, 2007
Project Director
Brian Dietterick
Management history
Goals
Study design
General timeline
Water quality data collected and
stat analysis
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
Looking Upstream in
North Fork ~2000
• 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
Looking Downstream in
North Fork ~1910
– 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
1
Timeline and study design
Calibration period 2001-2008
•Measure existing water quality
Treatment 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
Paired Watershed Design
•Control watershed: South Fork
•Treatment watershed: North Fork
Nested Watershed Design
•Control Watershed: Upper North Fork
•Treatment watershed: North Fork
Field data collection – stage and
streamflow
Field data collection – water quality
samples
Stage
A
Fitted Line Plot
S F Flow = 16.7 2 - 51.81 SF Stage
+ 40.2 0 SF St ag e* *2
12
S
R- Sq
SF Fl ow (f t^ 3/s)
B
• 1-hour interval samples
• Collection of 24 bottles
before swap
0.133903
99.9%
R- Sq(a dj)
10
99.9%
8
6
4
2
0
0. 7
0.8
0.9
1.0
S F St ag e (ft )
1.1
1.2
C
Fitted Line Plot
NF Flow = - 1 .476 - 1.2 21 N F Stage
+ 29.10 NF S tage**2
25
S
R-Sq
R-Sq(a dj)
D
NF Flo w (ft ^3/ s)
20
A. Flow Meter
B. Pressure
0.5 7 35 4 7
99 .6 %
99 .5 %
15
10
5
Transducer
C. FW -1 Stage
0
Recorder
0.2
0.3
0.4
D. Staff Gage
0. 5
0.6
0.7
NF St ag e (ft )
Lab Water Quality Testing
• Lab analysis of
1 hour samples
0.8
0.9
1.0
Defining the dataset – Predicting
SSC from turbidity
• Need for SSC to be predicted from turbidity
• Turbidity
– Turbidity shown to be a better predictor than flow
• Units: nephlometric turbidity
units (NTUs)
• 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.
• Suspended Sediment
Concentration (SSC)
• Units: mg/L
A
A. Turbidimeter
B. Scale
B
– Regressions assessed based on r 2, p-value, residual
plots, and fits
– Data transformations when necessary
2
Defining the dataset – storm events
• Minimum storm event size based on turbidity
– Peak must be greater than 20 NTUs
– Only samples >20 NTUs analyzed for SSC
• Define storm events based on the hydrograph
using Hewlett and Hibbard (1967) 0.05 slope
method
End event
– Separates storm flow from base flow
• Storm event ends when turbidity drops below 20
NTUs or the 0.05 slope line intersects the
hydrograph
Paired and nested analysis after
transformations
Event load calculation
Paired
• Events with complete SSC and flow
datasets used for analysis
• Calculate individual hourly loads to
determine event loads
North Fork load (l n[kg/ha])
– Must have both datasets to calculate loads
Fitted Line Plot
Nort h Fork lo ad (ln[kg /ha]) = 2. 02 3 + 0. 76 57 Sout h Fork lo ad (ln[kg /ha])
7
Regression
95% CI
6
S
R -Sq
5
1. 00230
64.8%
R -Sq( adj)
4
63.4%
Nested
3
2
Fitt ed Line Plo t
1
No rth Fo rk l oad (ln[kg/ha]) = 0 .5 0 90 + 0 .8 39 6 Uppe r N. Fo rk lo ad (ln[k g/ha])
0
-3
-2
-1
0
1
2
South Fork load (l n[kg/ha])
3
• Event loads establish the calibration dataset
– Changes in the relationship used to detect
change
4
North Fork l oad (ln[kg/ha])
– High temporal variability requires hourly sums
6
Regression
95% C I
55
S
4
95.5%
95.3%
3
2
1
0
-1
-2
-1
0
1
2
3
Upper N. Fork l oad (ln[kg/ha])
4
5
6
New regression line comparison to existing
confidence intervals for NF versus SF
Assessing detectable magnitude of
change using confidence intervals
90%
30% Increase
50%
70%
Increase in
in Existing
Exist ingNF
NF Conditions,
Conditions,NF
NFvv SF
SF
77
Data
Data
Points
Points
and
and
Data
Points
and
Data
Points
and
Associated
Associated
Regressions
Regr essions
Associated
Regressions
Associated
Regressions
Existing
conditions
Existing
Existing
conditions
conditions
Existing
conditions
NF
load
increased
50%
NFNF
load
load
increased
incr eased
70%
90%
NF
load
increased
30%
66
North Fork
Fork load
load (ln[kg/ha])
North
(ln[kg/ha])
• 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
0.382495
R -Sq
R -Sq( adj)
------------95%
C.I.
for existing
conditions
55 ------------95%
95%
C.I.
C.I.
existing
existing
conditions
conditions
95%
C.I.
forforfor
existing
conditions
44
33
22
1
1
0
0
-1
-3
-3
-2
-2
-2
-1
-1
-1
000
111
222
333
SouthFork
Forkload
load
(ln[kg/ha])
South
South
Fork
lo
ad(ln[kg/ha])
(ln[kg/ha])
444
555
3
New regression line comparison to existing
confidence intervals for NF versus UNF
30%
10% Increase in Existing NF Conditions, NF v UNF
6
NorthFork
Fork load
load(ln[kg/ha])
(ln[kg/ha])
North
6
Data Points and
Data Points and
Associated Regr essions
Associated Regressions
Existing conditions
Existing conditions
NF load increased 30%
NF load increased 10%
------existing
conditions
------- 95%95%
C.I.C.I.
for for
existing
conditions
5
5
4
4
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
33
22
11
00
-2
-2
-1
-1
0
1
2
3
4
0
1
3
4
Upper
North
Fork2load (ln[kg/ha])
Upper North Fork load (ln[kg/ha])
5
5
6
6
Other Little Creek Study
Components
– 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
Questions?
• 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
4