Guidelines for Setting Turbidity Thresholds Rev 12/30/2003
The following guidelines are designed to collect a few samples in small storms
and more, but not too many, in large storms. “Few” and “too many” are subjective
terms and it is up to each investigator to define their desired range of sample
abundance by setting the number of thresholds. The guidelines are based on
simulations with data from Caspar Creek (Lewis, 1996) and seven winters of
experience suggesting that they meet our objectives (i.e. to accurately and
economically estimate suspended sediment loads for, on average, the 6 largest
storm events each year). However, with different environments or different
objectives, these guidelines may not be optimal. For example, if one had
a special interest in sampling the first few (likely small) events of the
wet season, then an extra threshold or two might be temporarily added near
the low end of each threshold scale. If relatively more emphasis is to be
placed on low flows in general, then the square root scale might be replaced
with a logarithmic scale. However, any alteration that places more emphasis
on low turbidity conditions will result in more samples (and higher costs),
unless the number of thresholds is reduced at the same time.
Using the threshold calculator applet
The TTS web page has a turbidity threshold calculator that can be used
with any web browser. For this to work properly you will need the Java
2 plug-in from Sun Microsystems. Click on "download the JRE".
- Using the Sensor Maximum slider, set the maximum NTU reading that your
sensor can record. This value can be determined by calibration and may
not be the same as the nominal range given by the manufacturer. The manufacturer
should be able to provide the necessary calibration information, however.
- Set N, L, and U on the Rising Threshold sliders, based on the criteria
described below under Rising thresholds.
- Set N, L, and U on the Falling Threshold sliders, based on the criteria
described below under Falling thresholds.
- If desired, test the thresholds as described below under Simulating TTS.
- Install the thresholds in the TTS Campbell program as described below
under Entering thresholds in the Campbell TTS program.
- Determine the lowest non-zero threshold, L. This should be a value that
is above typical inter-storm turbidity values. In small streams it should
also be a value that is expected to occur only after the stage rises enough
to submerge both the turbidity sensor and the pumping sampler intake.
- Determine the highest threshold, U, within the range of your turbidity
- Determine the number of thresholds, N, between L and U (including both
L and U).
- Use the threshold calculator applet, or manually calculate
thresholds as follows.
- Compute d = (U0.5–L0.5)/(N-1)
- The thresholds between L and U are (L0.5+d)2, (L0.5+2d)2,
… , (L0.5+(N-2)d)2
- Because of the way the algorithm is written, additional thresholds are
needed at 0 and above the sensor measurement range, e.g. 9999.
- The complete set of rising thresholds to be assigned is: 0, L, (L0.5+d)2,
(L0.5+2d)2, … , (L0.5+(N-2)d)2,
- The procedure is similar to that for rising thresholds, except (1) guidelines
for determining L, U, and N are slightly different, (2) no threshold is
needed above the sensor’s range, and (3) thresholds are assigned in descending
order in the Campbell TTS program.
- L should be a value that is at or above typical inter-storm turbidity
values. In small streams it should be a value that is expected to occur
before the stage falls enough to expose either the turbidity sensor or the
pumping sampler intake. It is best to choose a different L for falling
turbidity than for rising turbidity. Otherwise it is likely that, in a
small storm event where only the lowest rising and falling thresholds are
exceeded, only two samples would be collected, both at nearly the same turbidity.
- N should be higher than that chosen for rising thresholds.
- N and U can be altered in a trial-and-error fashion to minimize redundancy
between rising and falling thresholds. However, because samples are not
taken precisely at the threshold turbidity, but occur only when the threshold
has been passed for two intervals, the risk of re-sampling the same turbidity
is not all that great a concern. Samples are least likely to occur precisely
at thresholds in rising turbidity conditions and at high turbidity levels
in general, i.e. when turbidity tends to change most rapidly.
- The complete set of falling thresholds to be assigned is: U, (L0.5+(N-2)d)2,
…, (L0.5+2d)2, (L0.5+d)2, L,
Sensor range 0-500
Rising (L=10, U=450, N=10):
0 10 27 51 84 125 174 231 296 369 450 9999
Falling (L=15, U=475, N=17):
475 427 382 340 300 262 227 195 165 137 112 90 70 52 37 25 15 0
Sensor range 0-1000
Rising (L=15, U=900, N=10):
0 15 46 94 158 240 338 453 585 734 900 9999
Falling (L=20, U=950, N=17):
950 851 758 670 587 510 439 372 311 256 206 161 122 89 60 37 20 0
Sensor range 0-2000
Rising (L=20, U=1850, N=10):
0 20 77 170 300 467 670 910 1187 1500 1850 9999
Falling (L=30, U=1900, N=17):
1900 1698 1507 1328 1160 1004 858 724 602 491 391 302 225 159 105
62 30 0
TTS can be simulated with any existing Campbell TTS data file and a few functions
written in the R programming language. If you have already installed R for
making data plots, then it is simple to add the simulation functions. These
functions and instructions for installation and usage can be obtained
from TTS web page. The package provides the capability to read a raw TTS
data file and determine when sampling would have occurred for a hypothetical
set of thresholds. A plot can be produced that shows the record of stage,
turbidity, and simulated pumped samples. Many of the other program parameters
specified in subroutine 7 of the TTS program can be altered as well.
thresholds in the Campbell TTS program
- Copy the thresholds into subroutine 7 (lines 196-200 in TTS Rev 4.1).
The rising thresholds are entered from lowest to highest and must include
the values 0 and 9999. The falling thresholds are entered from highest
to lowest and must include 0 (but not 9999). Eight thresholds are entered
per Bulk Load statement in the Campbell program. The ninth parameter of
the Bulk Load statement is the starting address in memory where the previous
8 values will be stored.
- The number of thresholds must be entered in subroutine 7, where maxrindex and maxfindex are assigned (lines 194
and 195 in TTS Rev 4.1). The total number of non-zero thresholds is
entered as parameter 2 in these statements, i.e. N+1 for rising thresholds
and N for falling thresholds. In all of the above examples, the values
entered are 11 and 17, respectively.