Relationships Among Hydrology, Vegetation and Soils in Transition Zones of Rhode Island Red Maple Swamps

Hydrologic data were gathered weekly for three growing seasons along a soil drainage toposequence running from very poorly drained to moderately well drained soils at three forested sites in southern Rhode Island. Cluster analysis was performed using eighteen hydrologic variables for each sampling station. Stations within the cluster diagram were subjectively designated as wetland, transitional, or upland. Discriminant analysis was used to classify the transitional stations as wetland or upland. The wetland/upland break fell most frequently between very poorly drained and poorly drained soils, or within the poorly drained soil zone. These locations were quite low on the moisture gradient and were probably due to the mesic nature of the upland end of the transects. Using stepwise discriminant analysis, the percentage of the growing season during which air-filled porosities within 30 cm of the ground surface were 15 % or less was selected as the most important hydrologic feature distinguishing between wetland and upland. When hydrologic analyses for individual years were compared, the location of the wetland/upland break was found to vary in accordance with the timing and magnitude of annual and seasonal precipitation levels. In years of high precipitation, much of the poorly drained soil zone might be classified wetland using hydrologic data. While hydrologic data are vital to our understanding of wetland systems, their use as a wetland boundary identification tool is limited by high annual and seasonal variability, and by the intensity of data collection required to adequately describe a site.

Reaching a consensus on wetland boundary determination criteria has been hampered by imprecise definitions of the distinguishing features of wetland and upland. Wetland transition zone research, of which this thesis project was a part, was undertaken in red maple swamps in Rhode Island to examine the relationships among hydrology, vegetation, and soils, and to develop field criteria for locating wetland boundaries using these parameters.
In this thesis, three years of hydrologic data were used in cluster and discriminant analysis to classify sampling stations as wetland or upland.
In most cases, the wetland/upland hydrologic break fell on the border between very poorly drained and poorly drained soils. Variables describing the percent of the three growing seasons during which high soil moisture levels occurred within 30 cm of the ground surface were most useful in distinguishing between wetland and upland. The location of the hydrologic break varied between years, and suggested that the location of the break may move upslope to include more of the poorly drained soil zone in years of high precipitation.
Wetland/upland breakpoints based on _ hydrology, hydric soil status, and herb-layer vegetation were compared. The hydrologic break was lowest on the moisture gradient, and the vegetation-based break was highest. The break based on hydric soil status appeared to be the most reasonable location for the wetland boundary from both ecological and management perspectives. These results suggest that poorly drained soils should be considered wetland, and that hydric soil status is a    Using stepwise discriminant analysis, the percentage of the growing season during which air-filled porosities within 30 cm of the ground surface were 15 % or less was selected as the most important hydrologic feature distinguishing between wetland and upland. When hydrologic analyses for individual years were compared, the location of the wetland/upland break was found to vary in accordance with the timing and magnitude of annual and seasonal precipitation levels. In years of high precipitation, much of the poorly drained soil zone might be classified wetland using hydrologic data.
While hydrologic data are vital to our understanding of wetland systems, their use as a wetland boundary identification tool is limited by high annual and seasonal variability, and by the intensity of data collection required to adequately describe a site.

INTRODUCTION
The U.S. Fish and Wildlife Service (FWS) defines wetlands as "lands transitional between terrestrial and aquatic systems where the water table is usually at or near the surface or the land is covered by shallow water" (Cowardin et al. 1979:3). As this definition suggests, hydrology is the driving force maintaining wetland conditions Turner 1978, Carter et al. 1979). Wetland functions such as flood reduction, wildlife habitat, and pollution abatement are largely influenced by hydrologic characteristics such as the extent and duration of surface flooding and degree of soil saturation (Van der Valk et al. 1978.
Wetland hydrology is difficult to describe due to seasonal, annual, and longer-term fluctuations . As a result, few thorough· studies of wetland hydrology have been performed, especially in freshwater wetlands of the Northeastern United States. Most research addressing wetland identification and delineation has concentrated on vegetation and soils, since these two parameters are more easily quantified than hydrology. Although wetland vegetation and soil are generally considered to reflect hydrologic conditions, the precise nature of the relationships among these three parameters is poorly understood.
In 1985, the University of Rhode Island initiated a study of hydrology, vegetation, and soils along _ a moisture gradient extending from wetland to upland in southern Rhode Island deciduous forests. The goal of that study was to describe the relationships among the three 3 parameters as they changed along the moisture gradient.
This paper presents the hydrologic results from that study. The objectives of this paper are: 1) to characterize the hydrologic gradient in the transition zone between forested wetland and the adjoining upland; 2) to determine a wetland/upland boundary based on hydrologic features; and 3) to identify which hydrologic parameters are most useful in distinguishing wetland from upland.
The Great Swamp site was located at the edge of a large wetland system 4 including a 400-hectare shallow lake and over 800 hectares of forested wetland. The two Laurel Lane sites were located approximately 200 m apart in a small watershed drained by a perennial stream. All sites were within 10 km of the University of Rhode Island in Kingston.
All three sites were dominated by red maple (Acer rubrum) at the lower end of the moisture continuum, and by white oak (Quercus alba) at the upper end . A well-developed shrub layer occurred underneath much of the canopy at each site; sweet pepperbush (Clethra alnifolia) was the most abundant shrub species overall. Soils in the PD, SPD, and MWD zones were predominantly weakly developed Entisols composed of loamy sands and sands.
In the VPD zone, Inceptisols with histic epipedons were common, and Histosols occurred in the wettest areas of the Laurel Lane sites ).

Data Collection
At each study site, three transect lines were established perpendicular to the slope contours and spaced 15 m apart. Along each transect line, six sampling stations were located according to soil drainage class. Drainage classes were determined from auger samples, using criteria specified by   . The cores were then oven-dried at 105 °c to drive off all free water and reweighed. Soil water content was expressed as volumetric water content. Since soil aeration, not soil moisture, is frequently restricted in wetlands, air-filled porosity at each soil moisture potential was calculated as the difference between the total soil porosity and the volumetric moisture content. Percent air-filled porosity estimates the percent of the soil core volume filled with air .

Statistical Analyses
Exploratory statistics were employed to examine hydrologic relationships among the sampling stations.
To achieve an adequate sample-to-variable ratio of at least 3-to-l    and .

RESULTS AND DISCUSSION
The most gradual topographic slopes occurred at GSW; slopes were intermediate at LLl, and steepest at LL2 (Figure 1). At all sites, there was relatively little change in elevation between Stations 1 and 3; most of the rise occurred between Stations 3 and 6.
Inspection of soil morphology in the soil pits corroborated the initial soil drainage class determinations at most of the stations.
Station 3, placed on the VPD/PD border, was classified PD on the three transects at LLl, and VPD on all transects at GSW and LL2. On two transects at GSW, where Station 5 initially had been classified SPD, the designation was changed to PD after closer inspection.
alculated from weekly measurements over three growing seasons Cn=91-93). ater level. cPercent of time water level was within 30 cm of ground surface.

Water Levels
In every case, the water table was closest to the surface at Station 1, and its depth increased steadily between Stations 3 and 6 (Table l, Figure 1). Mean water levels at GSW were consistently shallower than at the Laurel Lane sites throughout the transects; the differences were particularily noticable above Station 3. On all transects, seasonal fluctuations in water levels were greatest at Station 6, and decreased down the transects; the standard deviation of the water level mean at Station 6 was typically two to three times as high as at Station 1 ( Table 1).
The duration of soil saturation near the surface directly influences plant species distribution Forsythe 1981, Paratley and. Examination of soil profiles at the Rhode Island sites indicated that most of the tree, shrub, and herb roots were within 30 cm of the ground surface. For that reason, the percentage of the growing season during which the water level was within 30 cm of the ground surface was determined. The water table was within that zone less than 2% of the growing season at Station 6 on all transects; the percentage increased to 87-95% of the growing season at Station 1 (Table 1).
Monthly precipitation levels for the study period are presented in   Figure 2 illustrates the mean soil moisture potentials for the six stations on Transect 2 at LLl as an example. Consistent with water level trends, soil moisture potentials were highest at Station 1, indicating near-saturated conditions at all depths, and lowest at Station 6. Mean soil moisture potentials for the 54 sampling stations are presented in Appendix C.
The air-filled porosities of the mineral soil samples were significantly higher than those of organic soil samples based on two-sample t-tests (p<0.05) at each pressure increment ( Figure 3) .
These findings concur with other studies which have shown that, because of small pore sizes and colloidal proper~ies, well-decomposed organic materials retain more water compared to coarser soils under the same environmental conditions (Taylor 1949, Boelter 1964).
Air-filled porosities were not significantly different within the mineral or organic soil classes.
Using the separate air-filled porosity curves derived for organic and mineral soils (  Because most of the organic horizons occurred at the lower ends of the transects, this conversion further enhanced the soil moisture differences between the wetland and upland ends of the transects.
The air-filled porosity curves showed a sharp change in slope at 25 cm water of pressure ( Figure 3). Several authors (Flocker et al. 1959 have reported that restricted aeration results when air-filled porosities drop below 10%-20% in wetted soils, leading to oxygen defi~iencies, and that above 10%-20% air-filled porosity, sufficient oxygen diffusion can occur to inhibit anaerobiosis. Based on these studies and the break in slope observed in the laboratory-derived air-filled porosity curves, oxygen deficiencies were assumed to occur in soils with air-filled porosities at or below 15% in this study. Since the duration of anaerobiosis affects soil morphology (Zobeck andRitchie 1984, Evans andFranzmeier 1986) and plant species distribution Forsythe 1981, Paratley and, the percentage of each growing season during which the air-filled porosity was 15% or less was calculated for the four soil depths at each station.

25
GREAT SWAMP

Key Hydrologic Characteristics
Stepwise discriminant analysis was used to determine which hydrologic variables contributed most to the separation of wetland and upland stations. Five of the 18 variables contributed 93% of the discriminatory power of the model; four of these were air-filled porosity variables (Table 3). The first variable to enter the model (the 1986 30-cm air-filled porosity variable) was responsible for 87% of the discriminatory power. These results suggest that the air-filled porosities of the soils were more useful than groundwater levels in distinguishing wetland from upland.
Two of the five variables selected as significant were air-filled porosity variables at 30 cm. This finding supports the hydric soil criteria (SCS 1987) which calls-for a high water level within 30 cm of the surface in highly permeable very poorly and poorly drained soils.
At the Rhode Island study sites, moisture levels in the 30-cm zone have additional ecological significance in that they are likely to influence vegetation distribution since the bulk of the roots occurred within this zone.

WETLAND UPLAND
Ov er three times the range of 26 cm observed between 1985 and 1987 was (Table 2). This suggests that changes in station classification between years of high and low rainfall could potentially be much greater than those observed in this study.

correlation with Hydric Soil Status
The hydric status for the soil series at the 54 stations was compared to the wetland/upland classifications resulting from the . three-year model hydrologic analysis (Table 4). All 15 of the stations with nonhydric soils were classified upland in the hydrologic analysis, but 10 of the 39 stations with hydric soils also were classified upland. One transect (GSW, Transect 1) showed perfect correlation between hydrologic and soil-bas£d classifications.
On six of the transects, the hydrologic break was one station lower on the transect than the hydric/nonhydric soils break and on two transects, the hydrologic break was two stations lower than the soils break. Of the ten stations where discrepancies occurred, nine were PD and one was SPD.
For highly permeable (>15 cm/hr) soi~s to be considered hydric, water levels must be within 30 cm of the surface for a week or more during the growing season (SCS 1987). When this criterion was applied to the water level data collected during the three growing seasons of this study, six of the stations with hydric soils and upland hydrologic classifications did not meet the water level criterion in any year; four met the criterion during two of the three growing seasons (Table   4). However, the soil morphology at these stations supported the hydric  (1987) water level criteria and weekly water level measurements over three growing seasons. NH -Nonhydric; H -Hydric. *Disagreement between hydric soil status and hydrologic-based classifications of station.
status , which suggests that the hydrologic classifications and the water level criteria were incorrect at these stations.
Since reducing conditions can occur in nearly-saturated as well as as fully-saturated soils, wetland morphology can develop in soils lacking an observable water table Bouma 1976, Pickering and . Therefore, soil moisture potential may be a more precise characteristic than groundwater level to correlate with hydric soil features.

SUMMARY AND CONCLUSIONS
Through cluster and discriminant analysis of hydrologic data Soil moisture data, expressed as air-filled porosity, contributed more to the discrimination between wetland and upland stations than did water In the Northeastern United States, few comprehensive studies of freshwater wetlands have been performed.  correlated soil coloration with soil moisture regimes, temperature, and redox potential in a soil drainage toposequence in Massachusetts. In a Rhode Island study of 12 forested wetlands, 7 years of water level data were correlated with tree growth rates, plant community composition, and microrelief . Damman and Kershner (1977) and Messier

(1980) described wetland plant communities in Connecticut and
demonstrated that floristic dif~erences could be explained by water regimes and nutrient levels.
A few studies have correlated wetland hydrology, vegetation, and soils, but with varying success. In a study of forested wetland transition zones in Connecticut,  found an inverse relationship between soil moisture content and soil acidity, and that vegetation could be grouped accor~ing to soil moisture content and relative elevation.  determined that soil moisture was closely correlated with both soil morphology and the distribution of vegetation in a forested wetland in upstate New York.
In 1985, the University of Rhode Island initiated a three-year study of hydrology, vegetation, and soils along a moisture gradient at three deciduous forested sites in southern Rhode Island. The goals of that study were to determine the relationships among the three parameters as they changed along the moisture gradient and to identify key aspects of each that could be used as field criteria for wetland delineation.
Each parameter was analyzed independently and a separate wetland/upland break was identified for each. This paper examines the extent of agreement among hydrologic, vegetative, and soil criteria by comparing the relative locations of the three breaks along the moisture gradient. Possible causes of disagreement among the three breaks are discussed.

Study Sites
Three study sites which shared the following features were selected: a continuous deciduous forested canopy; freedom from recent disturbance; a gradual slope from wetland to upland; stratified glacial deposits; and a drainage toposequence including very poorly drained  increasing pressure . The cores were then oven-dried at 105 °c to drive off all ·free water. Soil water content was expressed as volumetric water content. Since soil aeration, not soil moisture, is frequently restricted in wetlands, air-filled porosity at each soil 57 moisture potential was calculated as the difference between total soil porosity and volwnetric moisture content. Percent air-filled porosity estimates the percent of the soil core volwne filled with air ).

Data Analysis
Exploratory statistics were employed to determine hydrologic relationships among the sampling stations. Data from all 54 stations from the three study sites were pooled to achieve a minimwn sample-to-variable ratio of at least 3-to-l . All statistical analyses were performed using SAS software (Statistical Analysis Systems 1985). Cluster analysis (PROC CLUSTER, CENTROID) was used to group the sampling stations according to similarities in their hydrologic characteristics. Wetland, upland, and transitional clusters of stations were subjectively identified from cluster dendrograms.
Using discriminant analysis (PROC DISCRIM), wetland and upland linear discriminant functions were derived from the hydrologic data for stations within the wetland and upland clusters, respectively. Each station within the transitional cluster was then classified wetland or upland. The hydrologic variables contributing most to the separation of wetland and upland stations were identified using stepwise discriminant analysis (PROC STEPDISC). A significance level of 5% was required for a variable to enter the stepwise analysis. Comprehensive reviews of cluster and discriminant analysis may be found in  and . Importance measures used were basal area in the tree layer, stem density in the shrub layer, and percent cover in the herb layer.
Weighted averages equal to or below 3.0 were considered wetland, while scores above 3.0 were considered upland.

RESULTS
The most gradual slopes occurred at GSW, where the mean slope for the three transects was 1.43% . Slopes were intermediate at LLl (mean-2.21%), and steepest at LL2 (mean=3.67%). All transect lines showed relatively little change in elevation between Stations 1 and 3, with most of the rise occurring between Stations 3 and 6 (see Figure 1 in Manuscript 1).
Inspection of soil morphology in the soil pits corroborated most of the initial soil drainage class determinations. Station 3, placed on the VPD/PD border, was classified PD on the three transects at LLl, and VPD on all transects at GSW and LL2.
On two transects at GSW, where Station 5 initially had been classified SPD, the designation was changed to PD after closer inspection.

Hydrology
Hydrologic analyses were performed using six characteristics for each station: mean growing-season water level, percentage of the growing season that the water level was within 30 cm of the ground surface, and percentage of the growing season during which soil air-filled porosity was 15% or less at each of the four depths monitored (15, 30, 45, and 60cm). Values for each characteristic were calculated for each of the three growing seasons, so that a total of 18 hydrologic variables were entered into the analyses.

Discriminant analyses of hydrologic variables by individual years
showed that the location of the wetland/upland breakpoint moved by one or more stations between years (see Manuscript 1). For 1985, in which annual precipitation was near the 30-year mean, but growing season precipitation was 41% above average (primarily due to two storm events in late August), the wetland/upland breakpoint moved upslope on two transects compared to the three-year breakpoint location. For 1986, with annual precipitation 10% higher than average and near-average growing season precipitation, the breakpoint moved downslope on one transect. For 1987, with annual precipitation 10% below average and a  (Table 1, Figure 3). Transect 2 at GSW was classified entirely as wetland.
A moisture-related vegetation gradient also was observed in the tree layer, but wetland/upland classification using the 3.0 breakpoint was unreasonable because of the overall high values of the scores (Appendix F); on only one transect did the weighted average drop below 3.0 (to 2.9) in the VPD soil zone. The high scores were due to the dominance of Acer rubrum (FAC) in the VPD and PD soil zones, and ~uercus alba (FACU) in the SPD and MWD soil zones. Together these two species composed 96% of the total basal area at GSW, 90% at LLl, and 72% at LL2.
There was no observable moisture-related gradients in the tall or low shrub layers. Weighted averages in these layers were narrow in range, largely between 2.0 and 3.0, and showed no consistent trends along the transect lines.  (Table 1). Organic soils were classified into the Adrian and Carlisle series. Sudbury (MWD) and Walpole (PD) series were found at individual stations at GSW and LLl, respectively. Based on the hydric soils list (SCS 1987), all of the VPD and PD soils were hydric, while all of the SPD and MWD soils --with the exception of a SPD Wareham soil at LL2 --were nonhydric ( Table 1).
The hydric soils list also provides the criteria used to determine which soils should be designated hydric (SCS 1987  . The hydric status of the series included in the national list was supported by observed water levels in all 3 years of the study at 42 of the 54 sampling stations (Table 1) In this study, the low position of the hydrologic break was largely a result of the mesic nature of the stations used to create the upland endpoint function for discriminant analysis.  and  state that the composition of, and the degree of separation between, the endpoints affects the classification of unknowns.
Had it been possible to include well drained or excessively drained soils in this study, the separation between the wet and dry endpoints would have been greater, and some of the stations classified upland in the present analysis most likely would have been 71 classified wetland.
Conversely, had the transects been extended into habitats as wet as the red maple swamps described by , the wetland/upland break might have been pulled further downslope.
Discriminant analysis of hydrologic data from separate years indicated that the location of the wetland/upland break moved one or more stations between years. During the three years of this study, annual precipitation levels were within 10% of the 30-year mean (see Manuscript 1) --although considerable variation occurred in seasonal precipitation levels.
Since wetland water levels roughly reflect precipitation patterns , it is likely that the wetland/upland hydrologic break would move upslope to include more of the PD soil zone in years with appreciably greater precipitation.
Since engineering constraints and wetland functions such as flood control are of more critical concern during periods of high water levels, the wetland/upland breaks identified here using discriminant analysis should be viewed as conservative . However, the large data requirements and the relative, and somewhat arbitrary, nature of the wetland and upland classifications that result from discriminant analysis limits the value of this method in wetland delineation for regulatory purposes.
Herb-layer weighted averaging results indicated that all of the VPD, PD, and SPD stations, as well as one MWD station supported wetland vegetation. Thus, the wetland boundary based on herbs was one station farther upslope than the soils boundary and two or more stations higher than the hydrologic break on 7 of the 9 transects. The majority of herb-layer species at the PD and SPD stations were classified FACW or FAC, so that weighted averages at those stations were just below or 72 equal to 3.0. The abundance of facultative species throughout the transition zone caused the wetland boundary based on weighted averages to lie too far upslope in most cases.  suggested that weighted averages falling between 2.5 and 3.5 represented a "gray zone" and hence were inconclusive; in such cases, they recommended that another parameter such as soils be used to confirm wetland status.
In this study, this gray zone was very wide, including 3 VPD stations, all of the PD and SPD stations, and all but one of the MWD stations (Table 1)  At most stations, the hydric status assigned in based on soil series appeared reasonable. Even at the twelve "hydric" stations where observed water levels indicated nonhydric conditions in one or more years, a hydric status was probably appropriate in most cases.
Although the number of years that a soil must satisfy SGS (1987) water level criteria in order to qualify as hydric are not specified, it seems reasonable to assume that the five PD stations where hydric water level criteria were met at least once during the three growing seasons are hydric. At the six PD stations which did not meet hydric criteria in any of the years, morphologic features in the soil profiles indicated that a hydric status was appropriate ).
At all six stations, water levels were within 39 cm (three were within 35 cm) for at least a week during one or more growing seasons; in years of higher rainfall, groundwater levels might rise enough to meet 30-cm criterion. Alternatively, soil moisture potential may better correlate with hydric soil features than observable water level, since reducing conditions, which drive the development of wetland soil morphology, can occur in soils that are not fully saturated (Vepraskas and Bouma 1976;. A nonhydric status appears to be appropriate for the SPD Wareham station at LL2. At this station, the highest water level observed for at least a week in any growing season was -44 cm. According to the water level criteria (SCS 1987), high-permeability SPD soils must have a water table within 15 cm of the surface for at least a week to be designated hydric (SCS 1987); it is unlikely that water levels would rise an additional 30 cm, even in years of exceptionally high rainfall.
Cluster analysis of hydrologic data also indicated that the SPD Wareham station was more similar hydrologically to the nonhydric stations than to the hydric stations. This station was included in the upland cluster and was closely linked with most of the other SPD stations ( Figure 2). The single SPD station in the transitional cluster is from the Great Swamp site. The eleven PD stations with disagreement between hydric status based on observed water levels and by series all were contained in the transitional cluster.   x x x x % Organic matter 5-6.9 7-8.9 9-10.9 11-12.9 13-19 .9 x x x x x