Manual Compositing Methods for Urban Storm-Runoff Samples

In Reply Refer To:                             October 20, 1980
EGS-Mail Stop 412

Quality of Water Branch Technical Memorandum No. 81.03

Subject: Manual Compositing Methods for Urban Storm-Runoff Samples

One of the objectives of the Urban Hydrology Studies Program 
(UHSP) is to "determine the magnitude and frequency of storm-
runoff loads of water-quality constituents from typical urban 
watersheds". In order to meet this objective, it is obvious that a 
number of samples must be collected throughout each storm-runoff 
event. This is particularly true if loads are to be determined for 
those constituents associated with the sediment being transported. 
Pumping samplers usually must be utilized to collect enough 
samples to meet this objective.

Because of high analytical costs, it will not be practical to 
determine all constituents of interest for each of the samples 
collected during all storm events. Individual samples for several 
runoff events at each site will be analyzed, however, for the 
other events the individual samples will have to be composited and 
the single representative composite sample analyzed for the 
required constituents to determine the runoff-event loads.

The problem is to determine the volume of each individual sample 
to include in a composite sample that represents the quality of 
the flow during the entire event. We will assume that whenever a 
pumping sampler is installed, the pumping system and intake 
location are designed such that a sample pumped into a container 
is representative of the discharge during the pumping cycle. This 
means that the sample of the water-sediment mixture is truly a 
sample of the flow in the cross section during those few seconds 
when the sample is being pumped into the container.

The recommended method of compositing is based on the following 

. A continuous record of stage is available during the event,

2. A stage-discharge relation exists (rating curve),

3. A stage-triggered pumping sampler is operating properly; 
samples can be collected at equal time intervals or at preset 
stages, or at equal increments of flow,

4. Pumping-sampler efficiency is 100 percent; that is, all 
sediment entering the intake is delivered to the container at the 
same concentration,

5. The exact time of collection of each sample is known 

6. The exact gage height at the time of collection can be recorded 
or determined,

7. The total volume of the composite sample required by the 
laboratory is known,

8. The sampler pumps virtually the same volume for each sample, 
(not critical).

The discharge-weighted volume of each sample to be used in the 
storm-runoff composite sample is computed as follows:

Vi = qi ti x VT                           (l)
     _qi ti


Vi = volume of sample to be added to composite. 

qi = instantaneous discharge (cfs) at the time of sample 

ti = time interval (min)--equal to one-half the time since the 
previous sample plus one-half the time to the next sample. For the 
first sample, the time interval is from start of storm-runoff 
event to one-half the time to the second sample. For the last 
sample, the time interval is from one-half the time to the 
previous sample to the end of the event.

VT = volume of composite sample required by laboratory.

The example illustrated below (Fig. l) is an 18-hour storm-runoff 
event. The sampler was programmed to start pumping the first 
sample at a gage height of 1.1 ft, and to pump a sample at O.l-ft 
increments of gage-height change (rising or falling) until the 
gage height falls below 1.1 ft. This sampling scheme resulted in 
the collection of 10 samples. Table 1 is an example computation of 
the volume of each sample required to make up a 2000-mL discharge 
weighted composite sample. This method is based on the mid-
interval method of subdividing a day to determine the suspended-
sediment discharge for a day, which is described in detail in TWRI 
Book 3, chapter C3, pp 49-50.

Figure 1.--Hydrograph of 18-hr runoff event showing times of 
sample collection.

The method described above for determining subsample volumes for 
compositing probably is entirely adequate for this type of 
sampling program. It should not he used, however, to determine the 
final mean discharge for the runoff event. For instance, the final 
mean discharge computed by conventional methods for this 18-hr 
event is 10.3 or 10.4 cfs, depending on whether l5-min, 30-min, or 
6O-min gage-heights are used.

Sampling according to change in stage (discharge) or, if possible 
at equal increments of flow, is the most accurate method of 
sampling runoff events such as those encountered in the Urban 
Hydrology Studies Program. This is true regardless of whether 
samples are to be composited or analyzed individually. Use of the 
term "discharge-weighted" may infer that discharge is the only 
variable to consider when compositing. In fact, the method 
described above is based entirely on the distribution of flow 
during the runoff event. Keep in mind that we are proportioning 
water-sediment mixtures.

Representativeness of composite sample--How should the "accuracy" 
or representativeness of a composite sample be judged? The program 
objective being addressed here is the determination of "loads of 
water-quality constituents. This means that a constituent load for 
a runoff event can be accurately computed by multiplying the 
concentration of the composite sample times the mean discharge 
times a conversion factor (0.0027 for 24 hours). In order to 
verify the representativeness of a composite sample, the following 
steps should be taken:

1. Use the method illustrated above (mid-interval method) for 
determining volumes of subsamples to use in the composite.

?. Use the cone splitter to obtain the volumes required (see 
Quality of Water Branch Technical Memorandum No. 80.17).

3. Analyze a portion of each sample for sediment concentration 
(0.1 part should be sufficient).

(Missing--Table 1.  Example of computation of subsample volumes to 
be composited using the mid-interval method of subdividing.)

4. Analyze a portion of each sample for dissolved-solids 
concentration (lOO-mL aliquot from 0.1 part).

5. Plot these concentrations on the hydrograph at the correct 
sampling times and draw a smooth curve through the points.

6. Use one of the two recommended methods of subdividing to 
determine the discharge-weighted loads for the runoff event (TWRI 
Book 3, chapter C3, pp 49-52)

7. Divide the loads by the mean discharge and by the conversion 
factor (0.0027 x hours/24) to determine the mean discharge-
weighted concentrations for the event.

8. Compare the concentrations of the composite sample with those 
determined in step 7.

Step 1 has already been illustrated in Table 1. Assume that steps 
2, 3, and 4 have been completed. Figure 2 illustrates step 5.

Figure 2.--Sediment and dissolved-solids graphs based on 10 

Typically, the peak suspended-sediment concentration as well as 
the peak dissolved-solids concentration occurs before the 
discharge peak (see TWRI Book 3, chapter C3). During the first 
minutes of the passage of a typical urban storm-runoff event, the 
quality of the mixture of water and sediments is changing rapidly. 
The sediments that are entering the stream from various sources 
and distances upstream are mixing with those sediments that are 
already in the stream. As the flow increases, more sediments are 
being scoured from the streambed and banks and brought into 
suspension in the flowing mixture. The distribution of various 
sizes of sediments is continuously changing with the changes in 
turbulence and depth of flow. Although not completely documented, 
there apparently are relatively rapid transfers of some chemical 
constituents from solution to sorption on sediments; particularly, 
the metals and organic compounds. Therefore, it is important that 
samples he collected more frequently during the period from 
beginning of flow event to the peak flow.

Based on the concentration graphs and the hydrographs (Fig 2), the 
mean-interval method of subdivision was used to determine the most 
accurate sediment and dissolved-solids loads for the 13-hr event 
(step 6). The loads were 13.1 tons of sediment and 4.5 tons of 
dissolved solids. The mean discharge-weighted sediment 
concentration (step 7),

        13.1 tons
C =  _______________
     10.3 cfs x 0.00202

  = 630 mg/L suspended sediment

The mean dissolved-solids concentration,

        4.5 tons
C =  ______________
     10.3 cfs x 0.00202

  =  216 mg/L dissolved solids

This means that the sediment concentration of the composite sample 
should be equal to 630 mg/L and the dissolved-solids concentration 
equal to 216 mg/L in order to determine the correct loads. If the 
sediment concentration of the composite sample is not close to 630 
mg/L the computed sediment load will he incorrect, and all of the 
"total" concentrations (and loads) of those chemical constituents 
associated with sediment will also be incorrect.

Pumping samplers must always be programed to sample according to 
some preset schedule. The example illustrated above (Table 1) is 
one of the most commonly used sampling schemes. Three other 
sampling schemes are described below along with the scheme 
illustrated in Table 1. The major differences are either in the 
method of triggering the initiation of sampling for the flow 
event, or in the time interval between samples collected during 
the flow event.

                       Sampling Schemes

1. First sample at Ght of l.l ft and a sample collected at 0.l-ft 
increments of gage-height change (rising or falling) until gage 
height falls below 1.1 ft. (See Table 1)--samples collected.

2  First sample at Ght of 1.1 ft and a sample collected at l-hr 
intervals until gage height falls below 1.1 ft--samples collected.

3. First sample collected at start of event (triggered at Ght = 
l.0l ft) and a sample collected at l-hour intervals until same Ght 
reached--l9 samples collected (first and last samples not used in 

4. First sample collected when discharge reaches a point 25 
percent above base-flow discharge (for this station base flow was 
assumed to be 5 cfs, Ght = 1.0 ft, so the sampler was triggered to 
start sampling at a Ght = 1.12 ft, when the rated Q = 6.25cfs) and 
a sample collected at l-hr intervals; last sample collected at 
same Ght on recession--l3 samples collected.

Three methods of compositing are described below:

Compositing Methods

A. Mean-discharge-weighted--each sample is assumed to represent 
the mean discharge for it's time interval, ti. The volume of each 
sample required for the composite is computed according tn 
equation (1), except that mean discharge for the time interval is 
used instead of the instantaneous discharge.

B. Mid-interval--same as described in equation (1).

C. Short-cut--time is not used in the computation. The 
instantaneous discharges at the times of sample collection are 
summed and the ratios of each instantaneous discharge to the sum 
of the instantaneous discharges are used to compute volume of each 
sample to be used in the composite .

Composite concentrations for sediment and dissolved solids were 
computed using each of these compositing methods and the four 
sampling schemes and the results shown in Table 2. All of the four 
sampling schemes gave good results when the samples were 
composited using either method A or B. The shortcut method gave 
erratic results and therefore is not recommended.

(Missing--Table 2.  Comparison of mean concentrations of composite 
samples for storm-runoff event using various sampling schemes and 
compositing methods)

Summary and Conclusions

The representativeness of any composite sample is dependent upon 
many factors, but is most heavily dependent upon l) the spacing of 
samples collected relative to the hydrograph, and 2) the number of 
samples collected during the event. Generally, relatively fewer 
samples can be collected provided they represent the significant 
periods of the event; the rising stage and the peak(s). This is 
true whether the samples are analyzed individually or are 

Regardless of whether samples are to be composited or not, the 
selection of an appropriate sampling scheme for the types of 
runoff events that occur at a data collection site is extremely 
important. No method of compositing can yield correct 
concentrations of constituents if the samples are not collected at 
the proper times during the event. For instance, as shown in Table 
2, sampling scheme No. 3 resulted in the collection of 17 samples, 
but it was not statistically any more accurate than sampling 
scheme No. 1 where only 10 samples were collected.

Compositing method C, the shortcut method, obviously yields such 
erratic results that it should not be used. It is obvious, too, 
that this method yields better results as more samples are 
collected. If samples were collected at 15-min or 30-min 
intervals, this method may prove to be adequate at some sites.

Inasmuch as both instantaneous gage-heights and times of sample 
collection must be known before any compositing can be done, 
compositing methods A or B could easily be used in every instance. 
For the example shown, method B (equation l) gave good results for 
both sediment and dissolved solids, and it is a simple and fast 
way of determining volumes of samples needed for the composite 

The following information is needed before compositing can be 
done. It should be obtained as soon after the runoff event as 
possible so that the compositing can be completed and the 
sample(s) split and preserved as required.

l. Begin time of each sampling cycle.

~. Gage height associated with each sampling cycle.

3. Begin time and end time for the event.

4. Composite volume required by laboratory.

                        R. J. Pickering
                        Chief, Quality of Water Branch

Distribution: A, B, S, FO, PO

Key words: water-quality, sampling, sediment, sample compositing, 
           urban hydrology.

This does not supersede any previous QW Technical Memorandum.