Freshwater Creek, California...

Annual and Storm Analysis, Turbidity Threshold Sampling (TTS)

[Coho Male]


Oncorhynchus Kisutch

 

Click HERE to view a copy of the Salmon Forever Hydro Year 2000 Highlights document written by project leaders Clark Fenton and Jesse Noell.


OVERVIEW


The Turbidity Threshold Sampling (TTS) program at Upper Freshwater Creek has been successfully maintained for the second hydrologic year with the help of community volunteers. This web page serves as an overview of the data processing sponsored by Redwood Community Action Agency.

The 2000 hydrologic year was a calm year in frequency of storm events, but the magnitude of the storm events that did materialize were of greater significance than those during the 1999 hydrologic year. Three of the six major storms analyzed this year reached peak stages above 4 feet. Two of these storms were back to back in mid January 2000.

This web page also serves as a summary of the various phases of data processing at the Freshwater site. Annual estimates of suspended load based on a linear regression and a LOESS smooth regression are also presented along with six segregated storm load estimates. Stage discharge data and depth integrated sample data for the Freshwater site are also posted here. Other plots include point samples taken over both hydrologic years, and histogram plots showing consecutive hours of a given turbidity threshold exceedence. You may also download a FORTRAN90 processing routine, turb_thresh_plot, which produces a text file for conversion into a histogram plot in Microsoft EXCEL.

This analysis was carried out on a UNIX based SUN workstation using programs developed by research personnel at the USDA Forest Service Redwood Sciences Laboratory and written in SPLUS, PERL, and FORTRAN. You may email me [Ben Bray] if you would like more information on these processing programs. Some of the plots presented on this page were also developed using EXCEL. (If you have read this far you are probably very interested in this information. My advice is to go through this web page and print all the plots before reading the text in this page. And don't forget to read the important note below.

IMPORTANT NOTE: Plots on this web page are presented as thumbnails, click on them to view them in full size. ALSO, click BACK on your web browser to return to this page.

STUDY GOALS


The primary goals of the analysis were to...

INITIAL PROCESSING


Data processing consisted of a preliminary analysis of the raw data files using a series of data processing routines as mentioned above. The individual raw data files were "appended" into a single file (refered to as the .flo file becasue of the file extension). Some processing programs allow the user to adjust sensor data based on comparisons with field observations. The two plots presented here were taken from two processing programs: TTS_MAIN_PRO (left) and STORMPLOT (right). The TTS_MAIN_PRO figure shows both electronic and observed stage records. Over a five-day period we find that the electronic and observer records begin with close agreement but slowly diverge. Note also that the increased stage during a storm event may also account for a larger difference in observed and electronic records due to wave attenuation, for example.
[TTS_MAIN_PRO RAW PLOT]
Data analysis began by appending each data dump taken over the course of the 2000 hydrologic year. Appending was completed using a processing program called TTS_MAIN_PRO written by R. Field and J. Fisher, employees at Redwood Sciences Laboratory (RSL). TTS_MAIN_PRO allows the user to plot the raw stage data and allows the user to adjust the stage data based on observations recorded by personnel in the field.
[TURBIDITY RAW PLOT]
Plot From TTS_MAIN_PRO Processing Program
Raw Plot From STORMPLOT Processing Program

A program was also developed by RSL employees in the database language Paradox, which digitizes the field form information and stores the information in a format required by TTS_MAIN_PRO for turbidity threshold processing. The next step in the analysis was to construct raw plots of the entire data set. This was completed using a plotting routine written in SPLUS called STORMPLOT. STORMPLOT has a graphical user interface (GUI), which allows the user to specify subsets of the data by starting and ending date/time of the data record. After entering the record length, the user selects OK and a plot is created with stage and turbidity on the left and right vertical axes, respectively. The abscissa is date and time. After identifying suspect data or gaps in the data set, stage and turbidity data were corrected using two processing programs TTS_FIX_TURB and TTS_FIX_Q. These two programs prompt the user for the flo filename (e.g. ftr00.flo) along with starting and ending date/time criterion. The user can then select the correction method, such as interpolation or reconstruction from another site, if one exists. After correction, the stage or turbidity data is automatically coded according to the method used. Click here to view the stage, turbidity, and lab codes.

THE 2000 HY RATING CURVE UPDATE


This section presents the stage discharge summary data for the Freshwater station. Fourteen additional measurements were taken during the course of hydrologic year (HY) 2000 as shown below.

1999 Measurements

DATE TIME QUALITY RATING AVERAGE STAGE (ft) DISCHARGE (cfs)
12/17/98 1545 good 0.60 31.4
01/16/99 1635 good 1.05 67.1
01/19/99 1300 fair 1.60 160.2
01/23/99 1116 good 2.37 328.4
01/24/99 1000 good 1.41 109.8
02/06/99 921 poor 1.11 74.3
02/06/99 1009 poor 1.65 215.5
02/06/99 1054 poor 2.61 413.00
02/06/99 1150 poor 3.65 674.9
02/06/99 1306 fair 4.00 831.0
02/06/99 1524 good 3.30 423.7
02/06/99 1602 good 3.10 363.2
02/06/99 1701 fair 2.75 325.9
02/06/99 1758 fair 2.58 285.3
02/08/99 1556 good 1.42 116.5
02/15/99 905 good 1.005 68.6

Plot of Segregated 1999 and 2000 Measurements This first plot shows both 1999 and 2000 HY measurements in one figure. There were concerns that the stage/discharge rating relationship was changing due to deposition at the study site. From this plot no drastic changes in the rating relationship were shown as was expected given the amount of deposition (~ 1 ft) at the study site. Clearly there is more variance, and thus more uncertainty, in the rating relationship as discharge increases.

2000 Measurements

DATE TIME QUALITY RATING AVERAGE STAGE (ft) DISCHARGE (cfs)
11/05/99 1700 fair 0.21 0.4
11/21/99 1450 fair 0.702 21.8
01/11/00 2113 good 1.78 203.3
01/11/00 1740 good 1.98 250.1
01/14/00 1112 good 4.325 891.9
01/14/00 1031 good 4.5 958.5
01/14/00 1627 good 2.90 488.3
01/14/00 1235 good 4.025 782.9
02/14/00 1157 good 4.08 781.3
02/14/00 1252 good 4.38 869.2
02/14/00 1429 good 4.13 760.8
02/14/00 1706 good 3.75 634.9
02/14/00 1813 good 3.35 528.8
02/28/00 2046 good 1.71 163.5

Stage/Discharge Rating Curve This next plot shows the rating curve fit to the complete data set. This fit was carried out using the solver package in MS Excel. The objective was to minimize the sum-of-the-square-residual between observed and predicted data assuming a power relationship between stage (H-feet) and discharge (Q-cubic feet per second) . When the 1999 HY rating equation [Q=68.3H1.68] is compared with the combined 1999-2000 HY rating equation [Q=61.6H1.81] a slight decrease in the intercept is noted. Also the exponent of the power function has increased slightly. For hydrologic year 2000 the magnitude of events captured were greater than that in 1999, thus the rating equation has changed slightly due to the additional measurements.



DEPTH INTEGRATED SAMPLING (DIS) ANALYSIS


This year ten measurements were taken by Salmon Forever field personnel. The samples were analyzed at the Sunny Brae Suspended Sediment Laboratory for suspended sediment concentration. The final concentrations along with corresponding point sample concentrations (taken by the ISCO sampler) are presented in tabular form.

DIS DATA

DATE TIME DIS CONC. (mg/L) LAB CODE PS CONC. (mg/L)
01/11/00 1945 2.162E+02 0 2.193E+02
01/11/00 2145 1.636E+02 0 1.794E+02
01/14/00 1015 1.577E+03 0 1.626E+03
01/14/00 1100 1.335E+03 0 1.261E+03
01/14/00 1615 6.571E+02 4 5.945E+02
02/14/00 1245 2.117E+03 0 2.024E+03
02/14/00 1500 1.162E+03 0 1.007E+03
02/14/00 1700 8.845E+02 0 6.816E+02
02/14/00 1745 8.263E+02 0 6.058E+02
02/28/00 2030 5.900E+01 0 7.163E+01

The plot shown here is the linear regression fit to the 2000 DIS data set using EXCEL. From the slope of this regression (1.01) one can infer nice agreement with point samples taken by the ISCO and depth integrated samples taken by field personnel. The variance from this linear regression model is likely attributed to the level of precision in the laboratory work.

Depth Integrated vs Point Sample Linear Regression

DIS Linear Regression



A paired f-test was used to test whether we could reject the model DI = PS; a linear model with slope of one and intercept of zero (Note: DI - Depth Integrated and PS - Point Sample). The outcome of the test was that we could not reject the DI = PS model. What this boils down to is that the annual load predicted from point samples were not adjusted for the entire cross section. More DIS data will narrow the range in confidence of both the slope and the intercept. Perhaps next year ISCO concentrations may heve to be adjusted to account for a greater sediment concentration across the whole cross section due to more confidence in the intercept of the linear regression model.




ANNUAL LOAD ESTIMATION


The next part of the processing procedure is the most exiting part and so it is my favorite; determination of the annual load and storm loads for Freshwater Hydrologic Year 2000. The loads are presented first in graphical form, followed by model fits to the ftr00.sed data set. Then the loads using both the linear model and LOESS smooth model are presented in tabular form.

Annual Load With Bottle Numbers

Annual Load Estimate Without Bottle Numbers

Annual Load Without Bottle Numbers

Annual Load Estimate With Bottle Numbers

Annual Load With Bottle Numbers



This first plot shows the complete data set for hydrologic year 2000. Sample bottles show up as solid dots on the sedigraph. On the sedigraph the left scale is suspended sediment concentration (mg/L) and the right scale is turbidity (NTU). The bottom section of the plot is the hydrograph where the left scale is discharge (m3/s) and the right scale is stage (ft). The load presented on this graph is based on the linear regression model shown below.

Here we have the same plot as that on the left, only the sample bottles are actual sample bottle numbers within a given dump (refer to codes page).
Linear, Polynomial, and Cubic Fits to Data Set

Linear, Polynomial, and Cubic Fits to Data Set
Scale: 0-2200mg/L, 0-700NTU

Linear, Polynomial, and Cubic Fits to Data Set, Zoom

Linear, Polynomial, and Cubic Fits to Data Set
Scale: 0-250mg/L, 0-150NTU

Linear and LOESS fits to Data Set

Linear and LOESS Fits to Data Set
Scale: 0-2200mg/L, 0-700NTU

Linear and LOESS fits to Data Set, Zoom

Linear and LOESS Fits to Data Set
Scale: 0-250mg/L, 0-150NTU

The plot above shows the linear regression model along with the quadratic and cubic polynomial fits to the 2000 data set. You'll note that the cubic and quadratic fits plot virtually on top of each other. Though they fit slightly better on the lower end, they tend to overestimate the peak concentrations, and thusly are less desirable as compared with the linear model. Here we show the linear, quadratic, and cubic fits on a finer scale
(Turbidity Range:0-150 NTU, Suspended Sediment Range:0-250 mg/L).
This plot shows both the linear and LOESS smooth models. Just as we found last year, the LOESS smooth model is superior at fitting the high end of the turbidity-suspended sediment relationship. The linear model slightly under predicts the peak samples. Above we show the linear model against the LOESS smooth model on a finer scale. Again the LOESS model tends to follow the general trend of the data more closely than the linear model at this range. Though there is a lot of scatter in the data at this scale, one can make out a definite curve-like trend in the data set.
(Turbidity Range:0-150, Suspended Sediment Range:0-250)


Linear Regression Stats
Intercept: -91.96
Slope: 3.06
Degrees of freedom: 213 total; 211 residual
Residual standard error: 89.4
Quadratic Regression Stats
Intercept: -56.60
turb: 2.35
I(turb2): 0.0014
Degrees of freedom: 213 total; 210 residual
Residual standard error: 83.4
Cubic Regression Stats
Intercept: -46.14
turb: 2.04
I(turb2): 0.0029
I(turb3): -1.74e-06
Degrees of freedom: 213 total; 209 residual
Residual standard error: 83.2
LOESS Regression Stats
Number of Observations: 213
Equivalent Number of Parameters: 4
Residual Standard Error: 82.56
Multiple R-squared: 0.97
Residuals:
min 1st Q median 3rd Q max
-618.6 -14.58 -2.629 11.45 364

For hydrologic year 1999, 2.5 million kg of suspended sediment was estimated to have traveled down Freshwater between the dates 1/13/99 to 8/1/99. For hydrologic year 2000, the linear estimate (refer to above plot) and LOESS models both predicted an annual suspended load on the order of 4.5 million kg between 11/10/99 to 5/26/00. One cannot compare the two load estimates directly because the data set was much more complete for HY 2000, keeping in mind also that both years were very different hydrologicly. A visual test as to whether the turbidity - suspended sediment relationship is changing, of which the load estimates are fundamentally based, was to plot the HY 1999 and HY 2000 samples on a single plot. This was carried out and is shown in the section after the storm loads, which are presented next.


 
DESCRIPTOR LINEAR LOESS
NUMBER OF OBSERVATIONS 213 213
R-SQUARED 0.96 0.97
RESIDUAL STANDARD ERROR 89.4 82.6
PREDICTED LOAD (kg) 4554629 4479326
PREDICTED LOAD (kg/ha) 1322 1300
PREDICTED LOAD (ton/mi2) 378 372
Note: Freshwater Watershed Area Above Study Site: 13.3 mi2 (3445 ha)

STORM LOAD PREDICTION


The six largest storms of HY 2000 were selected based on peak turbidity and/or stage. The sum total of all storm load estimates is 4001600kg. The LOESS model predicted an annual load of 4479326kg (refer to previous section). The sum total estimated storm load is 89.3% of the LOESS annual load estimate. Just as was revealed in the HY 1999 data set; the majority of the suspended load is indefinitely mobilized during a relatively short time of the year, during significant storm events. The table below summarizes the data for each of the six storm events.
 
STORM START
DATE
END
DATE
LOAD ESTIMATE
(kg)
NUMBER OF
SAMPLES
PEAK STAGE
(ft)*
PEAK TURBIDITY
(NTU)*
1 11/29/99 1630 12/01/99 0800 101274 6 2.152 500
2 01/10/00 1200 01/12/00 2300 1311497 25 4.693 696
3 01/13/00 1000 01/15/00 1800 1061939 14 4.622 583
4 01/15/00 2130 01/17/00 0000 182757 8 2.872 331
5 02/14/00 0000 02/15/00 1330 1020729 40 4.404 628
6 02/26/00 1700 03/02/00 0000 323404 16 2.783 263
*NOTE: PEAK STAGE AND PEAK TURBIDITY MAY NOT HAVE OCCURED AT THE SAME SAMPLE TIME INTERVAL!!

PLOT OF 1999 AND 2000 ISCO SAMPLES



One plot I thought would be interesting to show, that Jack Lewis recommended, was a plot of the combined samples from HY 1999 and HY 2000. If there is a significant trend in sediment transport that could be distinguished from year to year, it should be evident on this plot. The three plots below show both HY 1999 and HY 2000 sediment samples distinguished by marker type and color. Circles have been placed around data that has either been coded for possibly corrupt suspended sediment concentration or turbidity (as explained in the codes documentation file). The plots tend to show that the data from both years are in agreement. Each data point seems to be within the observed variance of either individual data set.
Linear, Polynomial, and Cubic Fits to Data Set
Turbidity Scale: 0-800 NTU
Suspended Sediment Scale: 0-2500mg/L
Linear, Polynomial, and Cubic Fits to Data Set, Zoom
Turbidity Scale: 0-200NTU
Suspended Sediment Scale: 0-650mg/L
Linear and LOESS fits to Data Set
Turbidity Scale: 0-150NTU
Suspended Sediment Scale: 0-250mg/L

DURATION OF TURBIDITY THRESHOLD EXCEEDENCE


"Turbidity: High concentrations of suspended sediment may delay or divert
spawning runs and in some instances can cause avoidance by
spawning salmon (Smith 1939; Servizi et al. 1969; Mortensen
et al. 1976). Salmonids were found to hold in a stream
where the suspended sediment load reached 4,000 mg/L (Bell
1986). Though high sediment loads may delay migration,
homing ability does not seem to be adversely affected
(Murphy 1995). Cowlitz River chinook salmon returned to
the hatchery seemingly unaffected by the sediments derived
from the eruption of Mount St. Helens, Washington although
in the highly impacted Toutle River tributary of the
Cowlitz, coho salmon did stray to nearby streams for the
first two years following the eruption (Quinn and Fresh 1984).
... Turbid waters have been mentioned as affecting migration but
little documentation is available in the literature.
Thomas (1975) found fry migration increased as turbidity
increased. Lloyd et al. (1987) found that turbid streams
were avoided by juveniles except when the fish must pass
through them on migration routes. There is also some
evidence that diel migrations of salmonids is influenced
by turbidity. Many salmonids tend to migrate during the
evening hours (Burgner 1991), presumably to avoid
predation. However, in streams with higher turbidity,
migrations may be evenly dispersed over day and night."
MANTECH REPORT: "An Ecosystem Approach to Salmonid Conservation" TR-4501-96-6057. December 1996.

This last series of plots are histogram plots of exceedence of a given turbidity threshold. These plots have come out of an interest in research conducted by Newcombe and McDonald (1991) and Newcombe and Jensen (1996). These researchers compiled literature on the physical and biological response of fish to elevated levels of suspended sediment. Due to the nonexistence of continuous suspended sediment data to apply to the models presented by Newcombe and McDonald (1991) and Newcombe and Jensen (1997) other researchers have tried to extend the model for turbidity thresholds. This must be done carefully because the turbidity-suspended sediment concentration relationship may be well defined for a specific site, the relationship is site specific and depends upon many variables such as flow regime and geology. The greater the threshold and the longer the time of exposure the greater the biological stress, the most stressful of course being death. Clearly turbidity is a less desirable water quality variable, compared to suspended sediment concentration, from the standpoint that threshold levels will not be universal among all sites, but it is desirable from a the standpoint of data collection. Measurement of continuous turbidity and development of a relationship between turbidity and suspended sediment concentration is much more feasible for a specific site compared with continuously monitoring suspended sediment (imagine the lab work involved with that!). A well defined relationship between turbidity and suspended sediment concentration is essential in this monitoring approach, especially if management models like ones proposed by Newcombe and McDonald (1991) or Newcombe and Jensen (1997) are to be extended to turbidity thresholds. One of the first data sets that has been extensive enough to adequately apply these models is clearly here at Freshwater. The author of this web page (Ben Bray) wrote a FORTRAN processing routine that outputs histogram data (for importing into Microsoft Excel) given user defined turbidity levels. The program requires the standard .flo file for processing. DOWNLOAD a copy of the FORTRAN turbidity threshold routine EXECUTABLE or FORTRAN90 SOURCE CODE so that you may produce your own turbidity threshold histogram plot. Two examples are shown below. Instructions for use of this program may also be obtained by clicking here.

Histogram Plot: Threshold=25NTU
Turbidity Threshold: 25NTU
Histogram Plot: Threshold=80NTU
Turbidity Threshold: 80NTU
Histogram Plot: Threshold=380NTU
Turbidity Threshold: 380NTU

[Coho Banner]
| Turbidity Threshold Publications | Back to Main Page |
- Salmon Forever - Redwood Community Action Agency - Redwood Sciences Lab[Coho Banner]

Freshwater // Last updated on October 19, 2000, by Ben Bray