2005 Progress Report: Assessing an HGM-based Wetland Classification and Assessment Scheme Along a 1000 km Gradient of the Appalachian Mountains: Hydrology, Soils and Wetland FunctionEPA Grant Number: R829497
Title: Assessing an HGM-based Wetland Classification and Assessment Scheme Along a 1000 km Gradient of the Appalachian Mountains: Hydrology, Soils and Wetland Function
Investigators: Cole, Charles Andrew , Brooks, Robert P. , Cirmo, Christopher P. , Wardrop, Denice Heller
Institution: Pennsylvania State University , The State University of New York at Cortland
EPA Project Officer: Packard, Benjamin H
Project Period: January 15, 2002 through January 14, 2005 (Extended to August 31, 2006)
Project Period Covered by this Report: January 15, 2005 through January 14, 2006
Project Amount: $973,301
RFA: Development of National Aquatic Ecosystem Classifications and Reference Conditions (2001) RFA Text | Recipients Lists
Research Category: Ecosystems , Water , Aquatic Ecosystems
Attributing functions to various wetland types is difficult without an organizing classification scheme. In recent years, there has been increasing focus on the development of theory and models based on the hydrogeomorphic (HGM) approach to wetland classification and assessment (Brinson, 1993; Brinson, 1995; Brinson, 1996). This approach classifies wetlands and assesses function primarily based on the location of the wetland in the landscape, the source of water for that wetland, and the dynamics of the water onsite. According to the HGM classification hierarchy, wetlands are grouped into national-level classes (Smith, et al., 1995) under which regional subclasses are defined (e.g., Cole, et al., 1997), and an analogous structure is followed for the development of functional assessment models. HGM has proven useful in organizing and understanding a variety of wetland data (e.g., Cole and Brooks, 2000a).
In Pennsylvania, a regional HGM classification and assessment protocol has been available for some time (see HGM key in Cole, et al., 1997). The primary focus to date in Pennsylvania has been on the development of data (water, soils, plants, macroinvertebrates) and functional assessment models for four HGM subclasses (riparian depression, slope, headwater floodplain, mainstem floodplain). Specifically, a long-term hydrologic data set (> 8 years for some sites) has allowed for a deeper understanding of function in each of these subclasses.
Although a wealth of data have been collected relative to these subclasses in Pennsylvania, these data have primarily been restricted to wetland sites in central Pennsylvania. These four HGM subclasses have only been described from the Ridge and Valley province of the Appalachian Mountains of central Pennsylvania. It is of great interest to ecologists and the regulatory community to determine if similar HGM subclasses occur further to the north and south of Pennsylvania also in the Appalachian Mountains. In other words, can we find similarly
classified wetlands outside of the geographic range of the original data set for which the key and functional models were developed? More to the point, do these additional wetlands function in a comparable manner to their Pennsylvania counterparts? If so, this would allow for quick application of available functional assessment models and data rather than spend considerable additional effort in their development in other areas. This project would provide valuable information on the ability to transfer developed models across geographic boundaries.
Our goal is to determine the latitudinal extent of the HGM classification and functional assessment process developed in Pennsylvania. A broad-based wetland classification and functional assessment approach would be of great utility in the development of defensible biocriteria in the Appalachian Mountain region.
Specifically, this project addresses the following objectives:
- Determine the applicability of the HGM classification key (Cole, et al., 1997) and the functional assessment models (Wardrop, et al., 1998) to regions outside of Pennsylvania.
- Develop a set of reference wetlands in New York and Virginia, using the HGM key in Cole, et al. (1997). Additional sites will thus extend the current Appalachian wetland reference data set along a much larger latitudinal gradient (1000 km) than currently exists only in Pennsylvania.
- Apply standard assessment protocols (Brooks, et al., 1996; Wardrop, et al., 1998) and models of function (Wardrop, et al., 1998) to all wetland sites to determine if both physical structure (e.g., hydrology, soils, plant communities) and inferred function are similar (within each HGM subclass) along this gradient.
This project focused on three classes of wetlands under the classification scheme presented in Cole, et al. (1997): headwater floodplain, riparian depression, and slope wetlands. In Pennsylvania, several wetlands which had been previously sampled by the Cooperative Wetlands Center were included in this project, with additional sites being located on Bald Eagle Ridge, east of Port Matilda, and adjacent to Little Fishing Creek in Bald Eagle State Forest. In Virginia, wetlands were found by exploring the area west of Shenandoah National Park in the George Washington National Forest. GIS maps of the geology of the area, streams, and road layers were used to pinpoint areas with streams where sandstone and limestone layers met layers of shale. All of our sites were found in these areas. In late spring of 2003, trips were made to the Catskills region and Adirondack regions of New York by the principal investigator to see sites selected by Chris Cirmo.
Wetlands were chosen and then classified based on a dichotomous key (Cole, et al., 1997). The classification scheme was developed using Pennsylvania wetlands and is based on three HGM principles: (1) position of the wetland in the landscape; (2) source of water; and (3) dynamics of water on-site. Headwater floodplain wetlands are positioned along first- or second-order streams, with intermittent over-bank flooding of streams providing the main source of water. Riparian depressions are wetlands located in riparian areas of streams but are typically isolated from the stream. These wetlands receive ground water and are usually bowl-shaped, therefore, retaining water. Typically there are one or several outlets for water to exit to the stream. Slope wetlands are located either along a slope (termed stratigraphic) or at the break of the slope at the bottom (termed toe of slope). The source of water to these wetlands is ground water, with the water broad-facing over the length of the slope (stratigraphic) or broad facing over the landscape at the base of the slope (toe of slope). The water is not constrained by topography as in riparian depressions.
In the Adirondack region of New York, three headwater floodplains, one riparian depression, and five slopes were chosen; in the Catskill Mountains of New York, three headwater floodplains, four riparian depressions, and four slope wetlands were chosen. In Pennsylvania, five headwater floodplain, five riparian depression, and eight slope wetlands were chosen, and in Virginia, five headwater floodplains, four riparian depressions, and six slope wetlands were chosen.
In April 2003, five headwater floodplain sites, five riparian depressions, and seven slope sites were chosen in the George Washington National Forest in Virginia and West Virginia. In May 2003, five headwater floodplain sites, five riparian depression, and six slope sites were instrumented with wells and data loggers to measure shallow ground water and inundation levels at the sites.
In July 2003, the entire field protocol was completed on nine sites, with the remainder sampled in July 2004. The field protocol includes grid setup, grid survey, survey of wells on site, survey of microtopography, debris census, vegetation protocol, soil pits, site map, GPS wells and baseline points, habitat similarity index (HSI), stressor checklist, photos, and voucher specimens. Three slope sites, two riparian depressions, and four headwater floodplain sites were completed. Denice Wardrop and Andrew Cole visited the field crews during sampling to provide guidance and quality control. In November 2004, a trip was made down to Virginia to download data from the wells and check the batteries for the coming winter. However, we encountered technical difficulties with sampling the recorders and were only able to replace the batteries and wait for resampling during the spring of 2005. Soil samples were sent to the Agricultural Analytical Lab at Penn State University (PSU). Plant voucher plant specimens have been checked by Sarah Miller at the PSU Cooperative Wetlands Center (CWC) for correct identification. Field data sheets were photocopied and the copies are stored at different locations. For all sites, unknown plant samples were identified and reconciled with the data sheets. Updated data sheets were photocopied and stored in a second location. Field and soil analysis data has been entered electronically, and data have been double checked, backed up on the website and CD.
In 2002, the entire field protocol was completed on five sites in Pennsylvania, three riparian depressions, one slope, and one headwater floodplain. Field and soil analysis data have been entered electronically and checked, as wells as data sheets archived electronically for these five sites. Hard copies of data have been stored in a second location and data backed up on the Web site and CD. In June of 2003, grid setup, vegetation protocol, soil pits, site map, GPS of wells and baseline, HSI, stressor checklist, and voucher specimens were done on eight sites (four slopes, two headwater floodplains, and two riparian depression. The remainder (n = 5) were completed during summer of 2004. Unknown plant specimens were identified and reconciled with the data sheets. Data sheets have been photocopied and stored in a second location. Soil samples were sent to the Agricultural Analytical Lab at PSU. Voucher plant specimens have been checked by Sarah Miller from the CWC. These field and soil data have been entered electronically and double checked and backed up on the Web site and by CD.
Over the summer of 2003, the entire field protocol was run on 11 sites in the Catskills. Grid setup, vegetation protocol, soil pits, site maps, HSI, stressor checklist, photos, and vouchers were completed at all sites. Four of these 11 sites in the Catskills group were surveyed in the spring of 2004. This led to 11 sites completed in the Catskills area of New York. Eleven sites in the Adirondacks were completed during 2004. Soil samples and voucher specimens were sent to the Agricultural Analytical Lab at PSU and the plant voucher specimens sent to the CWC at PSU for identification. All 11 sites have been instrumented with wells. Jessica Peterson made two separate visits in June to the Catskills in New York to run the vegetation protocol with the New York team. During these trips she checked for consistency in the application of the protocol and consistency in plant identification between the New York and PSU teams. Well data from Catskill sites were downloaded in January 2004 and the Adirondacks in October of 2004, with batteries replaced.
To compare hydrologic information across sites over a common time period, only data from June 1, 2003 - June 10, 2004, were considered. Data within this time period were not available for all sites because of bear damage or failure from cold weather. Data within each state/class combination were compiled into summary files. If more than one well was present at a site, data were averaged together. Excel was used to generate the percentages of time water was in the root zone (greater than -30 cm), time site was saturated (water level between -30 and 0), and the time the site was inundated (water level greater than 0).
Data of adjusted water depth were compiled with date/time, site, state, class, region (NY), and adjusted depth as variables, which was then used in SAS to generated median, n, maximum, and minimum statistics for water levels for each state and within each state/class/New York region combination. SAS was also used to run Friedman’s two-way non parametric analysis of variance using ranked data by class with state and state (site) as factors. Additionally, a test for significant differences between medians of states within a class were tested with the Kruskal-Wallis Test with SAS; for tests between more than two groups, SAS uses a x2 test statistic to approximate the H test statistic.
Functional Models. Field data and GIS data were used to score a set of variables, which were then plugged into functional equation models for 11 functions to yield a functional capacity index (FCI) score (Brooks, et al., 2004). Variables and FCI scores ranged between 0 and 1. Variables and FCI scores were generated for all wetlands sampled in 2002 and 2003 (results for 2004 are still being tabulated). For each function, mean values of FCI within each subclass and state combination were generated along with standard error, maximum, and minimum values using SAS. Nonparametric Kruskal-Wallis tests were also conducted to look for significant differences in FCI scores between states within subclass. Additionally, an anthropogenic disturbance score was generated for each wetland based on the percent of natural cover surrounding the wetland in a 1-km radius circle (natural cover being the addition of forest, water and wetland), the number of classes of anthropogenic stressors marked on the checklist, the size and type of the buffer surrounding the wetland, and any penetrations to the buffer. The anthropogenic disturbance score for each wetland sampled in 2002 and 2003 was represented graphically using Sigma Plot along with mean values of FCI scores.
Ordinations. Several ordinations were run using CANOCO, using only those sites that were completed in 2002 and 2003. We first analyzed wetlands by all the HGM variables (used to score models) which were scored for all three subclasses (i.e., we discarded the variables such as Vgrad as it was only scored for Huntington Wildlife Forest and SLP). Wetlands appeared to separate out better with linear versus unimodal method of ordinations; we used a principal component analysis, focusing scaling on inter-sample distances, with species scores divided by standard deviations, and centered and standardized species. Additionally we ran a clustered component analysis and a redundancy analysis (RDA) with these variables and the median depth to water table as the environmental constraint. The RDA analysis showed better separation between samples; we ran this analysis focusing on intersample distances, dividing species scores by standard deviation, centered and standardized by species, and we did not use forward selection or evaluate with the Monte Carlo permutation test.
We fell behind somewhat in our schedule when Jessica Peterson left to return to graduate school. As such, we still need to complete the calculations inherent in the functional assessment models and conduct the appropriate analyses. We are in the process of hiring additional expertise to assist us and the no-cost extension requested to gather water data will also allow us to get back on schedule for our analyses and writing. As stated, we will continue to collect hydrologic data until early 2006 such that we have 3 years worth of data for the New York and Virginia sites.
We will collect data from the water level recorders in early May and in late October. We have resolved the technical issues that hindered data collection in November 2004.
Water level data will be downloaded from all sites in May 2005 and October 2005.
Water level data will be downloaded from all sites in May 2005 and October 2005.
Journal Articles on this Report : 2 Displayed | Download in RIS Format
|Other project views:||All 18 publications||3 publications in selected types||All 3 journal articles|
||Brooks RP, Wardrop DH, Cole CA, Campbell DA. Are we purveyors of wetland homogeneity? A model of degradation and restoration to improve mitigation performance. Ecological Engineering 2005;24(4):331-340.||
||Brooks RP, Wardrop DH, Cole CA. Inventorying and monitoring wetland condition and restoration potential on a watershed basis with examples from the Spring Creek Watershed, Pennsylvania, USA. Environmental Management 2006;38(4):673-687.||