EQUIPMENT AND SUPPLIES--New sample splitter for water-quality samples

In Reply Refer To:                                   July 3, 1980 
EGS-Mail Stop 412


Subject: EQUIPMENT AND SUPPLIES--New sample splitter for 
                                 water-quality samples

Bob Middelburg of the Quality of Water Branch has developed a new 
sample splitter called a cone splitter. The cone splitter divides 
a water sample or sample of a water-sediment mixture into ten 
equal parts. The cone splitter was originally developed for the 
Urban Hydrology Studies Program for use in splitting samples taken 
with an automatic pumping sampler for analysis of chemical and 
physical constituents. In addition, the cone splitter will be 
quite useful in compositing proper proportions of several samples 
taken throughout a runoff event into a single discharge-weighted 
sample that represents the flow event.

Enclosed for your information is an article describing the cone 
splitter; its accuracy, application, and procedure for use. The 
differences in application between the cone splitter and the churn 
splitter are noted in the article. Cone splitters are currently 
being distributed to all projects that are part of the Urban 
Hydrology Studies Program. Further information regarding the use 
or availability of the cone splitter can be obtained by calling 
Bob Middelburg, Quality of Water Branch, FTS 928-6834.

                            R. J. Pickering 
                            Chief, Quality of Water Branch


Distribution: A, B, S, FO, PO

Key Words: Water quality, instrumentation, subsampling, 
           sample splitting

Superseded memoranda: None

                    The USGS Cone Splitter

Sampling methods have been developed that produce samples that are 
representative of flow through a cross-section. These methods 
frequently conclude with one bulk volume of water-sediment 
mixture. Unfortunately, preservation techniques and analytical 
methods do not always allow the submission of one sample in a 
single container to the laboratory for analysis. The sample must 
be subdivided, usually within a short time after collection, into 
a number of subsamples each of which must be virtually equivalent 
in concentration of suspended and dissolved constituents.

The USGS churn splitter can be used to subsample a very large 
volume (8-12 liters) sample collected for chemical analysis. It 
allows obtaining different subsample volumes from the sample while 
still maintaining the same basic chemical and physical properties 
of the original sample. The churn splitter has proven to be an 
invaluable tool for the collection and processing of composited 
cross-section samples from rivers and streams. The major 
disadvantages of the churn splitter are l) sample volumes less 
than about 6 liters cannot be split, and 2) inorganic sediments 
coarser than 62 um cannot be split with an accuracy of less than 
about +/- 10-15%.

The recent use of automatic samplers has introduced a problem that 
makes the use of the churn splitter impractical. Automatic 
samplers usually collect relatively fixed sample volumes, most of 
which are at or below the minimum volume of water required for 
proper operation of the smallest available churn splitter. Most 
automatic samplers collect only between 0.5 to 3 liters in one 
sampling cycle, which is not enough volume for proper use of a 
churn splitter.

The new cone splitter was first developed in December 1979 as a 
means to reliably subsample samples collected for the Urban 
Hydrology Studies Program conducted by the U.S. Geological Survey 
in cooperation with the U.S. Environmental Protection Agency. 
Tests have shown that the cone splitter can split samples as small 
as 250 mL volume into 10 equal subsamples, each subsample being 
with +/- 3 percent of the correct volume and sediment 

Description of Cone Splitter

The cone splitter illustrated in Figure 1 is a pour-through 
device. A funnel-shaped reservoir on the top receives the sample 
and directs it into the splitting chamber. Located in the 
reservoir funnel is a 2 mm-mesh screen which retains large debris 
such as leaves that could clog or interfere with the splitting 
process. The screen reduces the vortex action of the water leaving 
the funnel and also helps mix the sample.

Below the funnel is a short section of stand pipe. Its function is 
to direct water as a steady stream into the splitting chamber 
which contains a cone-shaped splitting head.

The cone splitter housing is machined from a solid block of Lucite 
or comparable material. Ten exit ports have been precisely drilled 
through one common point at a 45-degree angle from the vertical 
and spaced at 36-degree intervals around the circumference. The 
resultant configuration in the splitting chamber is a notched cone 
with 10 equally spaced exit ports about its base. There are no 
flat walls, benches, or surfaces inside the splitting chamber that 
can retain material or interfere with the splitting process.

The 10 exit ports direct the individual subsamples into 
distributor tubes leading to the subsample containers. The tubes 
are of sufficient size and alignment to prevent any back pressure 
or restriction of flow from the splitting chamber. They also are 
kept to a minimum length to prevent submergence of the end in the 
subsample. Any restriction of flow from an exit port will 
interfere with the rate of split-sample entry into that port, 
causing a bias in the splitting.

Evaluating the cone splitter

Two prototype cone splitters were constructed and tested for 
accuracy and bias. The tests were conducted using both clear water 
and prepared samples of water and sand-size sediment. In addition, 
tests were made to determine the effect of tilting the splitter 
and of pouring the sample into the splitter at different rates and 

To test the accuracy and bias with respect to volume, Bruce M. 
Delaney of the New Mexico District sediment laboratory prepared 
six samples of deionized water placed in l-gallon plastic 
containers, similar to commercially used milk, juice, and water 
jugs. The volumes for all observations were determined by weighing 
to the nearest 0.1 grams (essentially equivalent to 0.1 mL using 
deionized water). Samples were introduced into the splitter by 
inverting the sample bottle over the reservoir, allowing it to 
empty as rapidly as possible. The splitter was allowed to sit for 
approximately 1 minute after splitting for draindown before the 
subsamples were removed and weighed for volume tests.

The results of the six volume tests are given in Table 1. After 
weighing each subsample, it was determined that on the average 2 
mL of water was lost during a splitting process due to droplets of 
water adhering to various parts of the splitter. A small bias in 
the distribution was observed from outlet to outlet. This was 
probably due to slight variations during the fabrication process. 
Table 1 shows that the average discharge from tube No. 8 was 
consistently high by 1.5 percent, but this is considered well 
within acceptable limitations. It should be noted that the outlet 
numbers do not correspond to the sequence that the outlet ports 
were drilled.

To check the accuracy with respect to volume splits, each 
subsample was compared to the mean volume for each split. The 
maximum error observed was +1.9 percent (outlet 8, test 6) and the 
minimum error was -1.7 percent (outlet 2, test 3). The standard 
error in percent (standard deviation divided by the mean times 
100) for each test was 1.1 percent or less. These observations 
indicate that the cone splitter is capable of accurately 
subdividing a sample into 10 equal parts by volume within an 
arbitrary acceptable error limit of +/- 3 percent.

An additional series of tests were made using a water-sediment 
mixture to test the splitter capability to produce subsamples 
equivalent in physical composition to the original sample. Six 
samples were prepared. Each consisted of 1.0 grams of 62- to 125- 
um sand, 4.5 grams of 125- to 250- um sand, and 0.5 grams of 250- 
to 500- um sand plus deionized water to bring the total sample 
weight of 2500.0 grams. Using a suspended-sediment mixture of 
predominantly sands was considered to be a worst-case condition 
test because sands will not easily stay in suspension as compared 
to silts or clays. Particle sizes finer than sand (< 62 um) should 
split with an accuracy comparable to the volume-test results. If 
the cone splitter operates properly, the sediment concentrations 
of the subsamples should be virtually equivalent for each outlet 
and should not vary with the variation in volume from outlets.

The results of the water-sediment mixture tests shown in Table 2 
indicate that the splitter will subsample samples containing sand-
size sediment with a precision of 2.3 percent as calculated by 
averaging the standard deviations from each test. Test 2 produced 
both the maximum (+5.6, outlet 10) and the minimum (-4.4, outlet 
3) individual subsample errors. Figure 2 shows the plot of the 
mean, maximum, and minimum volume and concentration for each 
outlet. There does not appear to be any correlation between the 
variation in sediment-concentration means and volume means from 
outlet to outlet. Variability about the mean is greater for 
concentration than for volume, which is expected because the 
measurement of sand-size sediment concentration is less precise 
than measurement of volume alone.

The 10 subsamples obtained from test 2 were further analyzed for 
particle-size distribution. Summary results given in Table 2 show 
that the percent by weight of each subsample in the size range 125 
to 250 um is well distributed among the subsamples with a maximum 
deviation of 3 percent from the mean.

The series of sediment-concentration tests do indicate a possible 
bias in the splitter operation, although the error of the bias 
appears to be acceptable. When observing the percent variation 
from the mean concentration, the outlets having a positive 
differences are grouped together. For example, for test 1, table 
2, outlets 10, 1, 2, 3, and 4 are all greater than the 
concentration average and they all represent one side of the 
splitter outlet ports. This pattern was observed in the other 
tests, although it was not always the same group of outlets that 
contributed an above-average concentration. This phenomenon may be 
attributed to a slight vortex action associated with the flow 
through the stand-pipe. The sand leaving the funnel may tend to 
string out into a ribbon rather than mix. Further tests and 
changes of stand-pipe and screen designs will be necessary to 
determine the actual cause. It is believed, however, that such 
additional tests are not warranted considering that the observed 
errors are well within an acceptable range and the fact that the 
tests were conducted using predominantly sands, which represent 
the worst-case situation.

Testing of new cone splitters

To obtain reliable results, as observed in the series of tests 
previously discussed, a controlled operating procedure must be 
followed. Before using a new splitter, operators should 
familiarize themselves with the individual instrument by running a 
series of tests to determine any bias that could result from 
imperfection or operator procedures. The following test procedure 
should be followed:

1. Inspect the cone splitter housing and outlet ports. They should 
be smooth and symmetrical without any burrs or chips visible. Make 
sure the cone splitter is clean and place on a stable platform or 
bench in a level position. Visual leveling is sufficient.

2. Connect 10 discharge tubes to the outlet ports. All tubes must 
be approximately the same length, and the length should be as 
short as possible. The tubes need only extend into the receiving 
containers sufficiently to prevent spillage. They must not extend 
in so far that the end becomes submerged. Mark the outlets from 
1 to 10.

3. Wet the cone splitter by pouring through several liters of 
clear water. Lightly tap the system to dislodge adhering water 
drops, then discard the water. Replace an empty container under 
each outlet.

4. Accurately measure approximately 3 liters of clear water into a 
l-gallon narrow-mouth plastic bottle.

5. Rapidly invert the gallon bottle over the reservoir, letting it 
flow as fast as possible. Rest the inverted bottle on top of the 
reservoir. The rising water level in the reservoir will regulate 
the rate that water will leave the gallon bottle once the bottle 
opening becomes submerged. For proper operation, the stand-pipe 
must be discharging at its full flowing capacity.

6. After all water has passed through the splitter, tap the 
assembly several times to dislodge adhering water drops. Check for 
spills and leaks. If any are observed, discard the test, correct 
the problem, and repeat the test.

7. Accurately measure the volumes of the 10 subsample within 
+1 mL. Record the volumes for each outlet on a form similar to 
Table 1.

8. Repeat the test two more times for a total of three tests. Use 
approximately the same initial volume for each test.

Calculating Results

To determine the accuracy of the cone splitter tested, calculate 
the mean volume of each subsample (x) and standard deviation (Sx) 
for each test, by

calculate the standard deviation in percent (Ex) by the following:

Ex = Sx/x x 100

also calculate the error for each subsample (Ei) by

where Xi is the measured volume for the individual subsample. 
Finally, compute the average standard error (Ex) for the three 
tests and note the maximum and minimum errors (Ei) for all tests.

A cone splitter is considered acceptable for sample processing if 
the average standard error (Ex) for the three tests is 3.0 percent 
or less, and no individual errors (Ei) exceed +/- 5.0 percent. 
Note the error patterns for individual outlets to determine which 
outlets show consistent bias and mark them with their average 
percent bias error.

Using cone splitters

The cone splitter works best when the following procedure is 
followed. A consistent procedure such as always tapping the 
assembly at the end of a split and always wetting the system 
before a split should be practiced to help assure unbiased 

1. Set up the cone splitter on a flat open area. Check for level 
and proper tubing lengths. Visually inspect the splitter for 
broken parts, misalignment or debris.

2. Rinse through one or two liters of deionized water. Discard the 

3. Place containers under each outlet.

4. Shake the sample for 10 to 15 seconds.

5. Rapidly invert the sample container over the reservoir and rest 
it on the reservoir top.

6. After the flow has stopped, tap the assembly to dislodge 
adhering drops.

7. Remove desired subsamples, Repeat as necessary if any of the 
subsamples need splitting starting with step 3.

8. At completion of all splits for the station being processed, 
disassemble the splitter and clean before splitting another 

All subsamples do not have to be collected in separate bottles. 
Outlet tubes can be combined to collect various combinations of 
the original sample. Care must be taken, however, when combining 
outlet tubes into a single bottle to make sure there is no 
backpressure resulting from restriction of the flow.

Consider for example, the following subsamples are required from a 
3-1iter sample:

3-250 mL subsamples for chemical analyses (total).
1-500 mL subsample for chemical analysis (total).
2-250 mL subsamples for chemical analysis (dissolved). 
1-500 mL subsample for chemical analysis (dissolved).

The sample then is split by placing a 500-mL bottle under three 
outlets, two outlet tubes are combined into a l-L bottle, and the 
remaining five outlet tubes can be combined into one convenient 
container for later filtering. The resulting split of the 3-L 
sample would provide three 500-mL bottles with 300-mL each and the 
liter bottle with 600 mL. There would then be 1500 mL left for 
filtering. These volumes are close enough to the desired amounts 
for lab analysis.

If a more exact subsample volume is desired the following 
procedure is used. For example, if 440 mL is required from a 
sample of 2850 mL the first step is to compute the percentage 
needed. In this case 450 mL is 16 percent of 2850 mL. The 16-
percent split is achieved by first obtaining 10 percent from one 
tube from the first pass. The remaining 6 percent is obtained by 
pouring one of the lO-percent splits through the splitter a second 
time and drawing off 6 tubes or 60-percent. By this procedure a 
subsample of +l percent of the whole sample can be obtained by two 
passes through the splitter.

Care and maintenance of cone splitter

Cone splitters must be cleaned before being used for processing 
any samples. It is not necessary to clean it before splitting 
repetitively from one sample, but between a series of samples from 
the same station and runoff event, rinse the splitter with several 
liters of distilled water. Before using a previously cleaned 
splitter, start by pouring several liters of deionized water 
through the splitter. After using a splitter, acquiring a new 
splitter, or before starting to process a sample from a different 
station, clean the splitter by disassembling it and washing the 
parts in soap and water using a good quality laboratory detergent. 
A soft bristle test-tube brush works well for cleaning inside the 
ports. Rinse thoroughly with tap water followed with deionized 
water. Store cleaned cone splitters in plastic bags between 

The cone splitters should be visually inspected for damage 
especially the cone splitting chamber. Units that show damage or 
wear should be retested to check their serviceability. Check 
discharge tubing frequently for proper length and cleanliness. 
Replace tubes as conditions warrant.

The cone splitter is built to very close tolerances which are 
required for accurate and reliable operation. Given proper care 
and handling and operated according to the approved methods, the 
cone splitter should produce reliable results for a considerable 
number of samples.