SUBSAMPLING AND TAXONOMIC PROCEDURES EMPLOYED BY

STREAMKEEPERS OF CLALLAM COUNTY

(revision of 6 December 2006)

 

Arthur J. Frost

 

INTRODUCTION

 

                This document attempts to show the subsampling and taxonomic procedures that are used in dealing with benthic macroinvertebrate samples collected by Streamkeepers of Clallam County.  Some of the procedures given are tentative (not yet subjected to review by others); all may be considered subject to change (when such changes have occurred over the course of Streamkeepers’ existence I have noted them; note also the revision date above).  The sources of non-tentative procedures will be given immediately following the discussions of them; most of these sources will be electronic mail documents received by Streamkeepers.

                It amazes me that a well-established taxonomic laboratory has not already prepared a document such as this for general distribution (granted, I have been doing this sort of thing professionally for about ten years but I do not consider myself to be well-established).  The hope is that I am not “reinventing the wheel” by preparing this document.

                Copies of published work which answer questions raised below (or which tell me how much “wheel reinventing” I am doing) would be appreciated.

 

SUBSAMPLING

 

                The main purpose of subsampling is to keep specimen counts to a reasonable number; Streamkeepers has set 500±10% as the target.  Samples which do not reach 450 are thus not subjected to subsampling.  If subsampling is done records must be maintained showing the procedures used; records showing running counts might also be useful.

                Most of the subsampling that Streamkeepers has done to date has used a grid of the type described by Larry Caton in the Fall 1991 Bulletin of the North American Benthological Society.  This grid consists of a solid outer tray, a mesh-bottomed inner tray, a square “cookie cutter” and a scoop.  Deciding which samples to subsample involved first “floating” the sample and estimating the number of macroinvertebrates that floated out; if this number seemed to be in excess of 1000, the sample was subsampled.

                When subsampling, the first task is to let the floating macroinvertebrates sink; this is accomplished by keeping the sample in water for a couple of hours before doing anything more to it.  Once this is done, the sample may be spread out onto the Caton grid, attempting to distribute the sample solids as evenly as possible.  For small samples, a portion of the area may be used; a sample that does not take up at least five grid squares should be subsampled in some other manner.  A sample that contains sand or gravel (this should not be the case for Streamkeepers, but it does happen) should have such materials removed: evenly distributing sand and gravel on the grid consumes more time than the subsampling would save, and cleaning sand out of the grid afterwards can be particularly difficult.  Sand and gravel (if present) would be examined, and any macroinvertebrates found would be retained for possible inclusion in the additional taxon search (which is more conmpletely described below).

                Once the organic portion of the sample is spread out on however much of the Caton grid is needed, grid squares are selected in some random manner (I usually use polyhedral dice: a twenty-sided die for the “down” coordinate, a six-sided die for the “across” coordinate; a result that points to an unoccupied or already-used square is ignored.  This procedure is best carried out well ahead of time, with resulting grid[s] used as needed).  The entire solid contents of a selected square are removed from the grid in the following manner: the “cookie cutter” is placed in the grid square to outline its edges; the scoop is placed along one edge of the cookie cutter and the contents are routed to the scoop using a paintbrush or some other tool (I prefer to use a narrow plastic scraper; cleaning out a paintbrush so as to not miss anything is not exactly easy).  The contents of the scoop are then transferred to a petri dish or other container (I use a “sorting dish” made from two PVC flat end caps: a gap is cut into each cap that is about as wide as my field of view at 10X, and the two caps glued together at the gap) and sorted using a binocular stereoscope.  Nematode and annelid worms are counted and recorded as they are sorted (specimens of these taxa often break; fragments including one natural end point count as one-half, intact specimens count as one); other taxa are simply counted (the full identification phase can wait).  Terrestrial and semi-aquatic macroinvertebrates are not included in the count.  A running count is maintained as each square taken is dealt with; the count target is 500 (a count up to 50 over the target is acceptable).  Each square taken is completely dealt with (final counts will almost always be some figure over 500); at least two squares must be taken (it is not impossible for a single grid square to hold 500 or more specimens).

                In the event that a specimen or large piece of debris overlaps two grid squares.  Large pieces of debris should be cut along the grid square border once that is determined.  Specimens should be shifted such that they completely occupy whichever grid square initially contained the head (assuming that can be determined); if that is impossible specimens may be cut.  If a specimen is cut, only the head is counted (for nematodes and annelids see the procedure given above).

                The number of grid squares actually used and the number of those that were taken are recorded; these numbers are used to determine the percentage of the sample actually analyzed with the following equation:

(grid squares taken/grid squares used) X 100

There is a field in the Streamkeepers database that accepts this information; the information is used to estimate the total number of organisms in the sample.

                It is possible for subsampling to fail (the penultimate square undershooting the target, the last square overshooting the target by more than 50); if this happens, either the sorted organisms are placed in a smaller tray (such as that described below) and a subsample taken from that, or all sorted macroinvertebrates are identified and an electronic subsample generated (see further comments below).  In either case, it must be remembered that the final percentage is in fact a product of two initial percentages (either two iterations of the formula above, or one run of the above formula plus a run of (specimens identified/specimens subsampled) X 100).

                Once the count target has been reached the non-worms of the subsample would be identified and tallied, with the results recorded and the specimens stored as usual.  The used grid squares that were not taken are then subjected to an additional taxon search.  The standard for this type of search is to examine what remains in the grid using some magnification for a set period of time, removing one specimen of any taxon not found in the subsample.  I consider this method all too likely to miss taxa, and so the procedure I use for this search varies from the standard (and is sometimes modified under exceptional circumstances).  My usual procedure is as follows:

1) Clean the grid, saving all debris still therein, preserving the debris in 90+% alcohol; label this container as apopropriate.

2) Wait 48 hours for the remaining specimens to reabsorb alcohol (don’t waste this time; go on to the next sample), then re-float the sample remainder.

3) Examine the float under magnification, extracting one specimen of each taxon not found in the subsample.  To make the task of a Quality Assurance tech a little easier, these specimens are stored separately (labelling the container with the site, date, and the phrase “Uncounted Sorted Specimens”).

4) Examine the remaining organic debris, extracting one specimen of each taxon not found in either the subsample or the float discussed above.  I usually move specimens still in this debris that represent taxa that had been found in a previous step into the “Uncounted Sorted Specimens” container; this presents a QA tech with organic debris that is free or nearly free of specimens (thus reducing the amount of time such a tech would need to look through the debris).  The one exception to the preceding sentence involves a sample that is largely matted filamentous algae; removing specimens from such is never easy, and so is done only when required.

5) If sand or gravel were in the initial sample, such would be sorted at this time, extracting one specimen of any taxon not found in a previous step (the remaining specimens would be referred to the “Uncounted Sorted Specimens” container, for the same reasons given under 4) above).

This procedure does take a fair amount of time (at least an hour after the initial wait), but I believe that so doing has a better chance of finding most (if not all) of a sample’s additional taxa than does the standard method described above.  In the event that a number of samples are in progress at the same time, adequate labelling of all containers in use is essential!!!

                When a Caton grid is too big for a sample that nonetheless is known to hold more than 1000 specimens, an alternative method uses a small-area tray (we have a transparent plexiglass tray with a grid area of eighteen square inches; the underside of the tray is marked–and the markings protected by a sealant–for the grid).  This tray is small enough to fit under a boom stand stereoscope; once the sample is evenly spread out here some water would either be allowed to evaporate or would be siphoned off (this to prevent drift as the tray is shifted under the stereoscope).  Most of the remaining procedure as outlined above would be followed (the dice used to determine selection of grid squares are in this case two six-sided dice of different colors).  The testing I have done of this method suggests that it is best to use a 24-cell culture tray as an intermediate destination, completely emptying the gridded tray (this helps prevent specimen dessication).

                If a sample after being completely sorted and identified is found to hold more than 550 specimens, a subsampling computer program is used (one such may be found at www.cnr.usu.edu/wmc; Streamkeepers has a copy of this program).  The target count for such programs is set at 500; if a random number is required, the same number should be used throughout (this reduces the number of variables that would have to be considered).

                My handwritten bench sheets for each reach all follow nearly the same format, whether or not subsampling is done; one clue that subsampling has been done to a sample is the letter “A” rather than a number entered for a given taxon (indicating that taxon was found in the additional taxon search rather than the subsample).

                The Streamkeepers database records the taxa (and counts) found in the subsample and the results of the additional taxon search in separate fields; the first is numeric, the second boolean.

 

“TAXONOMY”

 

                As can be seen from the heading, I consider the word “taxonomy” to be misused for what I am about to describe; a better word would be “identification.”  I would restrict the word “taxonomy” primarily to what Stribling, Moulton & Lester (2003) called “research taxonomy” and that part of “production taxonomy” that is concerned with preparing identification keys; the remainder of “production taxonomy” would be called “identification.”  That being said, the remainder of this discussion will focus on the following: the kinds of organisms to be counted, and circumstances under which an otherwise eligible organism would not be counted.

                The samples as collected include not only benthic macroinvertebrates, but terrestrial organisms that for some reason had fallen in, semiaquatic macroinvertebrates, and fish; of these only the benthic macroinvertebrates are to be counted.  Streamkeepers applies macroinvertebrate identifications and counts to a ten-metric Benthic Index of Biotic Integrity.  With this tool the taxonomic rank considered the final identification varies by group, as follows:

                Nematoda.  Stops at phylum level.  A parasitic nematode must be completely free from its host to be counted.  A damaged specimen must have at least one natural end point (those with only one are counted as one-half).

                Cnidaria, Platyhelminthes, Annelida.  Stops at class level.  For budding Hydrozoa, clumps count as one.  Damaged Turbellaria are counted only if the fragment includes the head (the usual standard) or the pharynx (the standard I use); counts may be different depending on the standard used, which therefore must be specified by the initial identifier.  Damaged Annelida are treated in the same way as damaged nematodes (see above).

                Mollusca.  Stops at family level.  N.B.: originally it was believed that Gastropoda were not taken that deeply; this has resulted in some entries for “Gastropoda 1” in the Streamkeepers database.  Empty shells are not counted.

                Arthropoda general comment.  Cast-off exoskeletons are not counted; damaged specimens must include the animal’s head.

                Arthropoda other than Insects.  Generally stops at order level.  For some subgroups of the subphylum Crustacea, the “order level” is not necessarily that currently accepted (for example, Copepoda, Ostracoda and Cladocera remain undivided; specialists consider each of these to hold more than one order.  A further note on Cladocera and Copepoda: at least one Streamkeepers correspondent states that these should not be considered; as ecological variable values are available, we are continuing consideration for the moment; we need to get more opinions on this subject).

                Insects.  Generally stops at genus level.  Exceptions are as follows: Plecoptera families Capniidae and Leuctridae; Coleoptera family Dytiscidae; Diptera families Chironomidae, Syrphidae, Sciomyzidae, Ephydridae (?) and Muscidae; all of these stop at family level.  For taxa with a pupal stage: pupae (which are generally identifiable only to the family leve1) are included in the count only if one of the following conditions is met: a) the taxon is customarily identified to the family level; b) larvae identifiable to genus are not also present in the sample; or c) all larvae in a given sample are identified as the same genus (in which case the pupae are assumed to also represent this genus).  Condition c) may be deleted in the future.

 

LITERATURE CITED

 

Stribling, James B., Stephen R. Moulton II, and Gary T. Lester.  2003.  Determining the quality of taxonomic data.  Journal of the North American Benthological Society, 22(4):621-631.  December.