Quality Assurance Project Protocols

Methods

Homer methods are slightly different, and include a summary of Homer area wetlands.

Plant Community Classification

Field

Plant cover

Most of the plant cover data were obtained from the Natural Resource Conservation Service (NRCS).  Between 1997 and 2004 NRCS collected soils and vegetation data as part of their Western Kenai Soil Survey.  Data from 22 Hydro-Geomorphic Modeling (HGM) plots collected in 1997 along the lower Kenai River watershed were also used (Hall, et. al. 2002).  The authors collected data from 100 plots to augment soil survey data, while working for the Alaska Natural Heritage Program (NHP) during the summer of 2001; these methods are described below.

 

Ocular estimates of percent cover by species are recorded using a plot-less reconnaissance method.  Because plants cover the ground at different spatial scales, a homogeneous area was sampled with attention to these different scales.  For example, tree cover is more appropriately characterized using a larger plot, while forest floor herb cover can be adequately characterized with a smaller plot.  These different scales of occurrence are taken into account when the worker chooses an area to represent plant cover.  Unlike using a fixed sized plot, where, for example, an alder may or may not occur, the sampler can record alder cover over a larger area, and use a smaller area to represent trailing raspberry cover, for example, as long as the entire area is relatively homogeneous. 

 

Between 1 and 7%, cover is recorded to the nearest 1%; values greater than 7% (up to 15%) are recorded as 10%, then values are recorded to the nearest 10%  up to 100%.  Care is taken to assure that total cover sums to at least 100%; if observation indicates that cover is obviously much greater than 100%, then the sum should reflect the plants in the plot.

Plant stratum and life form are recorded using the categories of: tall, medium, short and dwarf; and herb, grass, shrub and tree, respectively.  Tall trees are greater than 40 feet tall and medium trees greater than 15 feet.  A stunted tree category is also used for trees obviously suppressed or stunted, otherwise a regeneration category is utilized.  Shrubs are tall if greater than 10 feet tall, medium if greater than 3 feet, and low if greater than 8 inches.  Other shrubs are recorded as dwarf.  Herbs are tall if greater than 2 feet, medium if taller than 4 inches; if shorter, they are dwarf herbs.  Only two grass categories are used: tall if greater than 2 feet and medium if less.

 

These are the same protocols that NRCS biological technicians used when collecting the data we obtained from the Western Kenai Soil Survey. Plant names follow the 2000 version of the US Department of Agriculture PLANTS database.

 

Environmental data

We measured three of four environmental parameters at each site: 1) water table depth, or 2) depth to modern (versus relict) redoximorphic features; 3) pH and 4) depth of the organic horizon.  Water table, organic layer, and redox feature depths were all measured to the nearest centimeter using a metal tape.  Depth to redox features was only recorded when the water table was not encountered.  PH was measured using a YSI 63 pH/conductivity meter.  The meter was 2 step calibrated (pH 4.04 and 6.86) daily, using the methods outlined in the meter’s manual (YSI, 1998).  When measuring pH in the field, the probe was placed directly into water in a hole dug below the water table and the value recorded when the reading stabilized for 30 seconds.

Each site was pinpointed on an aerial photograph.

Data Synthesis and Summary

The largest portion of data used in this analysis originated with the Natural Resource Conservation Service (NRCS) Western Kenai Soil Survey, on which the primary author of this project was instrumental in implementing plant community data collection techniques.  Widely inclusive criteria were used to filter the entire soil survey dataset for plots that might be considered wetlands.  Wetland plots are those that meet the criteria outlined in the Army Corps of Engineers Wetland Delineation Manual (the 1987 manual; ACoE, 1987)  Soils with modern redoximorphic features or a water table closer than 31 cm to the surface; organic horizons greater than 20 cm thick; plots on soils mapped in an aquic suborder, and sites subjectively determined to flood 'commonly' to 'frequently' were retained. 

Those data were evaluated for completeness, especially during years where non-botanists/ecologists collected data unsupervised.  Unreliable data were discarded.  Reliable data were printed and error checked against raw data, and corrections re-entered into the database. 

 

We used inclusive criteria- i.e. some of the plots we included in the summary analysis do not meet the wetland criteria established in the 1987 manual, but we did exclude most upland plots.  Retention of some plots that might not be considered wetlands is useful for bracketing the classification, but can lead to misleading determinations of how well any individual plant community might indicate wetland conditions.  The best example of this pitfall is the Lutz spruce / Oakfern – Bluejoint community.  Field observations indicate that this plant community is most frequently found on uplands.  However, in this analysis, two of the three samples occupied by that community were found on marginally wet soils (with redoximorphic features within 16 cm of the surface), and all were found within a soils unit mapped as an aquic suborder.  A summary indicating that 2/3 of the samples containing this community are wet would be misleading, as the sample itself reflects only the wet end of the continuum the community spans. 

Therefore, plant community fidelity to areas considered to be wetlands using the techniques outlined in the 1987 manual is not perfect.  Some 'wetland' plant communities will be found on uplands, while some 'upland' communities will occur on jurisdictional wetlands.  As we used liberal criteria to avoid missing any communities that sometimes occur on wetlands, most of the errors should be of the first type, i.e. some of the communities described here will occur on uplands.

 

Additional data were obtained from the HGM (Hydro-Geomorphic Modeling) effort conducted in the lower Kenai River watershed in 1997 by an interdisciplinary team funded by the US Environmental Protection Agency (EPA).  The HGM data (Hall, et. al., 2002) were  evaluated for completeness and reliability, recoded to match the USDA PLANTS database (which NRCS and Heritage Program field crews used) and error checked, with corrections re-entered into the database.

 

These three plant cover data sets (NRCS, HGM and NHP) were then combined and run through the computer program TWINSPAN (Two Way INdicator SPecies ANalysis; Hill, 1979) as part of the PCORD (McCune and Mefford, 1999) software package to determine plant community dominants.  TWINSPAN is a polythetic, divisive matrix algebraic solution that arranges a matrix of items and their attributes, then divides the items into groups based on maximum differences of attribute presence and abundance.  It works well when, as in much plant ecological data, many of the matrix values are zeros (i.e. few plants occur in all plots).

 

TWINSPAN was run several times on varying subsets of the data (e.g. all the plots with spruce (Picea spp.) cover greater than 10% were run together), and iteratively, with outliers removed on successive runs.  Once the primary plants responsible for group divisions became stable, the data sheets were sorted into initial divisions defined by their occurrence (all the sitka alder plots, for example).  Data sheets from each initial division were sorted into final groups within each division using our ecological knowledge and indicator plants identified by TWINSPAN.  These final groups are defined by the occurrence of co-dominant or sub-dominant plants (all the sitka alder plots with field horsetail (Equisetum arvense) for example); or the tufted bulrush (Tricophorum caespitosum) plots with significant dwarf birch (Betula nana).  The final groups became the named plant communities.

The plant communities were described using summarized frequency of occurrence and average cover of dominant plants (greater than 10 percent cover).  Environmental data (depth to water and/or redoximorphic features; pH, and depth of organic horizon) were also summarized using average, minimum and maximum values at the soils holes dug in each plant community.  The descriptions were written using field notes and sketches, the knowledge we gained working in these ecosystems, and the plant and environmental summaries described above.  Dot maps of the sites visited in each plant community were also assembled (from NRCS and NHP aerial photo pinprick locations) and included in each plant community description.

If subsequent field visits indicated new plant communities were needed, we queried the database to find any plots satisfying group membership (e.g. the Sweetgale- Dwarf birch / Water horsetail community was created by summarizing the plots with sweetgale (Myrica gale) and dwarf birch (Betula nana) cover > 10% that also contained water horsetail (Equisetum fluviatle)).  Summaries of the remaining communities were not re-adjusted to reflect any loss of data caused by the creation of new plant communities.  This loss probably would not have changed those summaries significantly. 

No attempt was made to shoehorn every plot into a final group (plant community).  Many plots are unique and form a diverse 'unclassified plots' group.

Communities are named systematically.  The plants in the tallest layers are named first, the most frequent plants are named first, and the most abundant plants are named first.  Plants in different layers are separated by slashes, plants in the same layer by dashes.  The layer order proceeds with trees followed by shrubs followed by grasses and sedges followed by herbs.  When a taller sedge is significantly more abundant than a dwarf shrub, as in the Tufted bulrush / Sweetgale or Tufted bulrush / Dwarf birch communities, the sedge is named first.  One subset of communities with generally low vascular plant cover and high (sphagnum) moss cover is named with sphagnum moss (in the 'ground' layer) first.

Mapping

 

Mapping Classification

A mapping classification was developed in 2002.  This classification relies on the knowledge we gained while mapping soils and describing plant communities over the previous five years.  The mapping classification uses a hydro-geomorphic approach tailored to the local landscape.  The geomorphic portion of the classification involved subjectively choosing and naming common ecosystems, which were assigned based on dominant landforms.  Examples of landforms/ecosystems include: Depression, Relict Glacial Lakebed, Kettle, Headwater Fen, Discharge Slope and Riparian.  Ten ecosystems have been named. 

The mapping classification is open, that is, the number of map units is not set; they are made up of combinations of basic building blocks, the map components.  Map components are basic divisions within each ecosystem, and are used in combinations, following specific naming rules, to devise map unit names.  The number of potential map unit names, using the map components and the naming rules, is much greater than the number actually encountered on the landscape, as some component combinations are unlikely.

Within most ecosystems, map components are enumerated based on depth to the water table (the hydrological portion of the hydro-geomorphic based classification), which is usually indicated by common plant communities.  The lowest map component numbers refer to the components with the shallowest water table.  For example the 'K1' component is in the Kettle ecosystem with standing water and/or emergent vegetation (1); while the 'K4' component is in the Kettle ecosystem with a forested or woodland cover, and a deeper water table (4).

Three ecosystems use a variation on the water table depth component- naming rule.  The first, the Discharge Slope Ecosystem, uses dominant plant communities; so an 'SLS' unit is a Discharge Slope dominated by Lutz spruce (Picea X lutzii) and Barclay willow (Salix barclayii).  There are six dominant plants within this ecosystem.  Using the naming rules, 36 possible combinations could be devised;  29 were mapped.

The second, the Riparian Ecosystem, uses a modified version of Rosgen's classification (1996).  Our classification uses six basic types along with a few sub-types to describe streams.  An example of a basic type is Rosgen's "E" stream.  These are streams flowing across deposits laid down by larger processes, on the Kenai Lowlands these larger processes were Pleistocene glaciations.  Large valleys were carved by glacial meltwater rivers; these valleys are now occupied by smaller clearwater streams draining watersheds no longer occupied by glaciers.  "E" stream sub-types are: linear (Rel), sinuous (Res) and bank-full (Reb).  Another example is Rosgen's "B" stream, a moderate gradient stream composed of riffles and pools, dominated by riffles.  Six types were used, only "E" streams included sub-types.

The final ecosystem not following a component naming rule related to depth to water table is the Tidal Ecosystem.  There, degree of tidal inundation, as indicated by dominant plants, determines map component numbers.  A T1 is a Tidal Ecosystem with saltpannes, occupying low, saline soils, while a T9 is at the upper reaches of tidewater influence, occupied by manyflower sedge (Carex pluriflora) communities.  Our experience, and a study showing the relationship of intertidal plant communities to periodic tidal inundation on the nearby Susitna River estuary (Vince and Snow, 1984) is relied upon to help inform Tidal map component divisions.

 

Wetland delineation and photo interpretation

Pilot Areas

Polygons were drawn on mylar overlays using aerial photographs (NRCS 17" x 17" uncorrected 1:25,000 stereo-paired black and white photographs flown in 1996).  Homogeneous photo-signatures of areas thought to be wetlands were delineated by us and digitized (by Resource Data Incorporated, Anchorage, AK, for Soldotna Creek Watershed and Kachemak Bay Research Reserve for Daryl's fen)).

Photo-interpretation was aided by the environmental and plant data from NRCS, HGM and NHP plots and our on-the ground experience.  The points indicating soils holes sampled by NRCS were plotted digitally into a GIS (Geographic Information System; ArcView 3.3, ESRI, Redlands, CA).  As these points were obtained from hand-held Global Positioning System (GPS) readings, and since these readings have varying reliability, the digital GIS points were edited in ArcView to match the locations that soil scientists marked on their field photos (a hardcopy version of the same photos used as a backdrop while editing in ArcView).

 

Project Area

We used techniques learned from NRCS soils mapping on the Western Kenai Soil Survey.  Uncorrected stereo-paired 1996 1:25,000 black and white aerial photography was used under a stereoscope, with acetate overlays and a ultra-fine point "Sharpie" marker to delineate initial wetland polygons.  Wetland polygons are relatively homogeneous areas that fit into the mapping classification described above.  Extensive local knowledge formed by both five years of field mapping experience and the one year pilot project, informed the aerial photo interpretations and linework, done entirely by K. Noyes.  A minimum polygon size of about 3 acres was used, although many smaller polygons were delineated.

After the initial polygons were drawn on the acetate overlays, the lines were transferred, using a 0.5 mm plastic pencil, to frosted mylar overlain onto quarter-quad-centered, geo-rectified film positives of the same black and white aerial photographs.  The mylar and film positives were both pre-punched and a 7-hole register bar was used to ensure exact alignment.  Once the lines were transferred, they were re-traced onto a second pre-punched frosted mylar sheet, again using plastic pencil and register bar, creating a final clean product ready for scanning.

The clean linework was shipped to Resource Data Incorporated, in Anchorage Alaska, where it was scanned, vectorized, edge-matched and cleaned up.  The result was an ArcView 3.x Geographic Information System shapefile.

The shapefile was overlain onto the digital version of the NRCS 1996 geo-rectified black and white aerial photography in ArcView, where the polygons were assigned map units.  Map unit assignments were made while consulting the same 1996 NRCS hard-copy stereo imagery under a stereoscope, and both the published (1970) and draft versions of the soil survey still in progress.

 

Field

Pilot project 

During the summer of 2002 we attempted to visit and photograph all polygons in two pilot areas, the portion of Soldotna Creek watershed outside of Kenai National Wildlife Refuge, and Daryl’s fen, a 15,000 acre peatland complex east of Anchor Point.  Printed field images of 1996 NRCS digital aerial photography were prepared by overlaying numbered polygons on the digital NRCS 1996 black and white aerial photographs.  These field images with numbered polygons were printed on matte photo paper at 1200 dpi at two scales: 1:15,000 for navigation and interpretation and 1:25,000 for minimum polygon size decision-making. 

 

On a field visit, if the linework was found to be accurate at 1:25,000, a map unit was assigned based on visiting the entire polygon, or a representative dominant photo signature(s), if viewing the entire polygon was not practical.  If the linework was not accurate, and included more than one unique map-able unit which covered more than 10% of the polygon area, a line(s) was drawn on a printed photograph to split the polygon; if the same unique map-able unit was adjacent to another polygon of the same map unit, a note was made to join them. 

 

A map-able wetland unit is one where the vegetation pattern is relatively homogeneous on the ground and discernable on the 1:25,000 scale imagery.  Map units are frequently based on general hydrologic character (depth to ground water), which is typically reflected in vegetation type.  For example, sedge types frequently occur on areas where groundwater is at or very near the surface, and shrubby peatlands occupy areas with a deeper water table.

 

Units needed to be wetlands to be mapped.  We did not formally delineate wetlands according to US Army Corps of Engineers delineation procedures (CoE, 1987), but used our best judgment, occasionally using a soil probe to look for redoximorphic features or water table depth when deciding whether or not to include a marginal polygon or area as a mapped wetland.  When a polygon’s wetland status was uncertain, we erred in favor of calling the polygon a wetland.  Most mapped sites are obviously wetlands, with a deep peat layer and/or a water table at or very near the surface.  Uncertain sites tend to be forested, occupying slope break positions.

 

When splitting or joining polygons, new lines were drawn on the field image, and notes recorded on a datasheet.  When joining, the retained number was indicated on the datasheet, and the other number noted, and later discarded.  When splitting, the retained number was indicated on the sheet and the field image, and a new number assigned to the new polygon in the office, at the computer.  New numbers were tracked to avoid assigning duplicates.

 

In addition to evaluating and, if necessary, correcting the linework, the percent cover of each plant community (identified in the classification) present in the polygon was recorded.  All communities covering more than 10% of the polygon were recorded.  Occasionally, communities representing less than 10% of a polygon were recorded, especially if they have never been seen covering more than 10% of polygons anywhere.  If plants were present that do not fit into one of the pre-defined communities they were noted.  If these same undocumented communities were found frequently, a new community was named; otherwise, undefined community cover was left unrecorded, and the total community cover summed to less than 100%.

 

Full Project Area

Similar field protocols were followed for mapping the entire project area, but a smaller subset of polygons was selected randomly within each map unit for field visit.  The random selection process was designed so that map units were chosen in proportion to their occurrence on the landscape.  Field visits were conducted over two summers, in 2003 and 2004.  Approximately half of the project area was digitized by the summer of 2003, and the selected polygons reflect the proportion each map unit represented in that mapped subset.  During 2004 polygons were visited according to the proportion they represented in the remaining subset.

'Selected' polygons were highlighted on the printed field images.  Polygons not selected, but along the travel route, were also visited and photographed, and corrected with respect to office assigned map units.  Plant community data were only collected if the map unit assignment made in the office on selected polygons was in error.  No plant community data were collected if the map unit assignments at non-selected polygons were in error; only the assignment was changed and a photograph taken.

Data Synthesis and Summary

  

After field visits, lines drawn on the printed aerial images were used along with notes on datasheets to correct linework, heads-up in ArcView, and re-number polygons, if needed.  Digital photographs were downloaded and renamed to match polygon numbers, using lower case letter suffixes to separate multiple photographs of the same polygon. 

 

Occasionally, nearby 'selected' polygons from previous field visits needed to be re-edited (combined with another polygon on its opposite side, for example).  In these rare cases, averages of percent cover of plant communities on the old polygons were calculated and used to represent the percent cover of plant communities in the newly formed polygon. 

 

The plant community percent cover and mapping unit names were entered into a Microsoft Access database, printed and error checked against the raw data, with corrections re-entered.  The printed field images and datasheets were also used to check that all polygons had the correct assignments. 

 

For the pilot areas, the final linework was transferred onto frosted mylar overlays, using printed error checked versions of the linework, geo-rectified film positive quarter quads of the same (uncorrected) aerial photography that was used to derive the preliminary lines and a register bar.  These final lines were sent to Resource Data Incorporated to be scanned and vectorized into a final ArcView shapefile.  This file was later incorporated into the final shapefile.

 

Error-checked data were exported to tables in MS Access, and linked to the shapefile.  That way polygon editing could be done in ArcView, using the link MS Access to locate and manage polygon records.  The MS Access database file was edited to incorporate new fields linking polygon number to digital photograph filenames and ecosystem and map unit description files.  The database file with new fields was exported, and re-attached to the shapefile.  The photographs and descriptions are stored on a Kenai Peninsula Borough server and are available on the world-wide-web at www.kenaiwetlands.net.

 

Plant community frequency and percent cover were summarized by ecosystem and mapping unit.  These summaries along with field notes and observations, environmental data and photographs, were used to create the plant community and map unit descriptions.

The environmental data summaries used in the map component and map unit descriptions were generated using the corrected locations (described above, under photointerpretation) of NRCS holes.  Each hole was assigned a wetland map component based on its location in a wetland polygon, and the NRCS plant community data collected there.  The physical data collected at these holes were used to summarize wetland environmental information (depth to water table, organic layer thickness, and depth to redoximorphic features) for each map component.

Ecosystem descriptions were created from the same information.  The information was also used to generate a key to ecosystems, which incorporate a summary from the pertinent literature.  Especially important to the ecosystem descriptions are field notes taken at “type localities” where the dominant environmental gradients in each system are well defined.  At these localities, plant relationships to these gradients are described and represented by artist drawings in the descriptions.

 

The ecosystem, map unit and plant community descriptions were created in HTML and contain links to keys, soil series descriptions, literature, wetland plant indicator status and other useful information.  Some preliminary data have been successfully tested on ESRI's Internet Map Server software, housed at the Kenai Peninsula Borough's website.  There a user, with internet access and a web browser, can manipulate a map containing satellite imagery to retrieve parcel ownership information.  The wetland layer will be added soon.  Currently a user with ArcView software can download a shape- or layer file that contains URL fields linking each wetland polygon to a picture (for polygons we photographed) or a map unit description.