The transition and forest zones of both flat transects occupied significantly lower elevations and were more hummocky as compared with the steep transects, while the high marsh occupied similar elevations. The flat transition zone had less drainage and a higher mean annual water table, higher and more variable mean groundwater and pore water salinities, and similar redox potentials for the 10 cm depths in the transition and forest zones. However, the steep slope showed significantly lower redox potential at 20 cm depth for all zones which is hypothesized to be related to the discharge of anaerobic groundwater. Hydrogen sulfide was very low throughout, but was significantly higher for the flat transect high marsh. Results of univariate analysis were supported by multivariate factor analysis which identified three factors accounting for 69.4% of the variation in the data: brackish water, redox potential, and sulfide variables. Factor scores for brackish water and redox potential for the flat transition and forest zones were significantly higher than for corresponding steep zones, but similar in the high marsh zone. All of these trends suggest that the dominant source of water is atmospheric for the flat transition and drainage and groundwater discharge for the steep transition.
Degree of slope affects transition zone size and position, such that the steep transition zone is narrower and changes more abruptly than the flat transition. Hydroperiod, salinity, and redox potential are believed to directly influence the distribution of plant species and their relative dominance. I could not determine whether hydroperiod or salinity controlled the position of the transition zone. However, the position of the transition zone roughly corresponds to the distance at which mean salinity decreases most rapidly. This occurs at the base of the forest area for the steep slope due to drainage or groundwater discharge, but at a greater distance from the tidal creek and at lower elevation for the flat slope. Marsh vegetation may restrict overland flow over long distances from the tidal creek during flood events, thus allowing the glycophytic vegetation to inhabit lower elevations, further away from the creek. Hummocks appear to allow glycophytic vegetation to colonize closer to the tidal creek thus increasing the width of transition zones.
This work benefitted enormously from the help and guidance of my committee: Drs. Mark M. Brinson, Robert R. Christian, Claudia L. Jolls, and Kevin F. O'Brien. I would like to especially acknowledge my thesis advisor, Dr. Mark Brinson, for his patience, advice, energy, insight, and his ability to always make sure I was fed (funded). I have many fond memories of personally guided tours through wetlands from Edmonton, Canada, to New Orleans, Louisiana, of which I will not soon forget. I have yet to meet a person who can keep up with him in the marsh. Special thanks also go to Dr. Robert Christian for his advice and help in the field; to Dr. Claudia Jolls who continually reminded me of the vegetation aspects of my work; and to my statistician guru, Dr. Kevin O'Brien.
So many people helped me along the way that I am sure that I will never be able to thank them all. The hard working field helpers include: James Taylor, who helped greatly in field collection and installation of sampling instruments, John J. Russell, and Paul Farley. The research at the VCR/LTER went most smoothly with the help of Dr. Linda Blum, Dr. John Porter, Marcio Santos (the Sipper King), Dave Osgood, Randy Carlson, Jimmy Spitler, and many UVA REUs. I would also like to thank Dr. Tom Chenier for helping me through the panics of factor analysis. Binita Kwankin provided a much needed review of an early draft, and helped me through the turmoil of life on Interstate 95. And lastly, I would like to thank my family for emotional and fiscal support, who thought I was crazy for leaving a great job.
This work was funded in part by a National Science Foundation Grant DEB-9211772 and BSR-8702333-06 through the University of Virginia, VCR/LTER, and East Carolina University.
Page
LIST OF FIGURES
vi
LIST OF TABLES
ix
1. INTRODUCTION
1
1.1 Factors Affecting Salt Marsh Vegetation
2
1.1.1 Oxygen Availability
2
1.1.2 Hydrogen Sulfide
3
1.1.3 Salinity
7
1.2 Effects of Slope
9
1.2.1 Effects on Hydrology
9
1.2.2 Effects on Vegetation
13
1.3 Objectives and Hypotheses
15
2. SITE DESCRIPTION AND METHODS
18
2.1 Site Description
18
2.2 Study Transects
22
2.3 Methods
24
2.3.1 Vegetation Sampling and Analysis
24
2.3.1.1 Vegetation Sampling
24
2.3.1.2 Vegetation Analysis
27
2.3.2 Surface Elevation
29
2.3.3 Soil Characteristics
31
2.3.4 Water Table Fluctuations
33
2.3.5 Groundwater Salinity
34
2.3.6 Pore Water
35
2.4 Data Analysis
36
3. RESULTS
40
3.1 Vegetation
40
3.1.1 Zone Detection and Delineation
40
3.1.1.1 Flat Near Transect
40
3.1.1.2 Flat Far Transect
43
3.1.1.3 Steep Near Transect
47
3.1.1.4 Steep Far Transect
48
3.1.2 Plant Species Presence
49
3.2 Elevation
52
3.2.1 Variation in Elevation by Transect
52
3.2.2
Variation in Elevation by Zone 55
3.2.3 Surface Macrotopography
58
3.3 Soil structure
58
3.3.1 Depth to mineral soil
58
3.3.2 Bulk Density and Organic Carbon
61
3.3.2.1 Bulk Density and
Organic Carbon by Slope 61
3.3.2.2 Bulk Density and
Organic Carbon by Zone 63
3.3.3 Chlorinity
65
3.3.3.1 Chlorinity by Slope
65
3.3.3.2 Chlorinity by Zone
65
3.4 Hydroperiod
66
3.4.1 Water Table Fluctuations
69
3.4.2 Precipitation, PET, and Water Table
Drawdown 72
3.4.3 Groundwater Salinity Patterns
74
3.5 Pore Water Chemistry
77
3.5.1 Pore Water Salinity
77
3.5.2 Redox Potential, Hydrogen Sulfide,
and pH 83
3.6 Factor Analysis
88
3.6.1 Analysis of Entire Data Set
88
3.6.2 Analysis of Transition Data Set
94
4. DISCUSSION
100
4.1 Factors Influencing Position of Transition Zone
101
4.1.1 Brackish Water Factor
101
4.1.1.1 Effect of Slope on
Hydroperiod 102
4.1.1.2 Effect of Slope on
Salinity 104
4.1.1.3 Effect of Slope on pH
106
4.1.2 Redox Factor
107
4.1.3 Sulfide Factor
108
4.2 Slope Effects on Vegetation
110
4.2.1 Position of Transition Zone
110
4.2.1.1 Steep Transitions
111
4.2.1.2 Flat Transitions
114
5. CONCLUSIONS
118
6. LITERATURE CITED
121
LIST OF FIGURES
Figure 1.Elements of a water budget in a tidal salt marsh as
they are
affected by slope.
10
Figure 2.Simple conceptual model of the effects of slope on the
hydrology
and pore water and vegetation used to formulate hypotheses.
14
Figure 3.Location and arrangement of study transects at the
Brownsville,
VA, marsh representing Steep Near (SN), Steep Far (SF),
Flat
Near (FN), and Flat Far (FF) transects in relation to
Phillips
Creek.
19
Figure 4.Frequency of occurrence for shoreline to upland soil
assemblages
for Northampton County, VA, using soil maps (USDA 1989).
20
Figure 5.Possible combinations of landscape position in relation
to proximity
to tidal creek and gradients of slope used to choose study
transects. 23
Figure 6.Layout of the Flat Near and Steep Near study transects
indicating
locations of the surficial groundwater wells, lysimeters,
and water
level recorders.
25
Figure 7.Sampling frame for vegetation analysis. (A) Frame size
and
geometry used for vegetation sampling. Frame size for
measuring
vegetation were: herbaceous (H) 1 m2, shrub (S) 4 m2, and
trees
(T) 8 m2.
26
Figure 8.Squared Euclidean distance computed using moving
split-window
analysis.
41
Figure 9.Peaks in the squared Euclidean distance for the Flat
Far transect
using window sizes of 10, 8, and 6 sampling points.
42
Figure 10.Percent of the total cover for each species grouped by
zone and
vegetation type.
44
Figure 11.Elevation contour relative to the creekbank benchmark
(HAYD). 53
Figure 12. Mean elevation for the high marsh zone, transition zone,
and forest
zone.
57
Figure 13.Topographic variation in
marsh surface and underlying mineral soil
from high marsh to terrestrial forest.
59
Figure 14.Mean soil structure data from core samples at 10 and
20 cm depths
for both Flat Far and Near (Flat) and Steep Far and Near
(Steep),
q standard error.
64
Figure 15.Mean monthly water table levels for the high marsh,
transition, and
forest zones determined by shallow (1.5 m) groundwater
wells. 67
Figure 16.Hydrograph during September 1992 showing the effects
of rainfall
and a storm surge due to Tropical Storm Diane for the Flat
Near
transition.
70
Figure 17.Monthly precipitation totals from 1990 to 1993
(Brownsville
Marsh) and 1949 to 1981 (Cheriton, VA).
73
Figure 18.Hydrograph during June and July 1993 comparing the
diurnal pulse
of water table drawdown due to evapotranspiration for flat
and
steep slopes.
75
Figure 19.Mean monthly salinity for the high marsh, transition,
and forest
zones determined by shallow (1.5 m) groundwater wells.
76
Figure 20.Box plot of the groundwater salinities along the Flat
Near and
Steep Near transects for the period beginning 4 November
1991 for
the Flat Near and 5 August 1992 for the Steep Near
transect. 78
Figure 21.Pore water salinity collected for the Flat Near and
Steep Near
transects from lysimeters at 10 and 20 cm depths for the
period 14
Jan to 29 July 1993.
80
Figure 22.Mean pore water data collected for the Flat Near and
Steep Near
transects from lysimeters at 10 and 20 cm depths for the
period 14
Jan to 29 July 1993.
81
Figure 23.Comparison between salinities measured using shallow
groundwater
and pore water salinities collected using lysimeter at 10
cm for the
Flat Near transect for the period January to July 1993.
84
Figure 24. Pore water redox potential collected for the
Flat Near and Steep Near
transects from lysimeters at 10 and 20 cm depths for the
period 14 Jan
to 29 July 1993.
85
Figure 25.Sulfide concentration collected from the 0 m lysimeter
at the 10 cm
depth and from the 150 m lysimeter at the 20 cm depth for
the Flat
Near transect.
87
Figure 26.Mean factor scores for the high marsh, transition, and
forest
vegetation zones for the period 14 January to 29 June 1993,
q
standard error.
91
LIST OF TABLES
Table 1.Algorithm used to convert dominance rankings to percent
cover. 28
Table 2.Vascular plant species in the Phillips Creek marsh.
50
Table 3.Slope summary table for study transects determined by
linear
regression.
54
Table 4.Summary of leveling transects separated by vegetation
zone. 56
Table 5.Mean depth (m) to organic layer for the Flat Near and
Steep Near
study transects, q standard error.
60
Table 6.Soil properties for study transects obtained from
sediment cores. 62
Table 7.Characteristics of hydroperiod at the transition zone
water level
recorders for the Steep Near and Flat Near transects.
68
Table 8.Mean pore water data collected from low-tension
lysimeters for the
period 14 January to 20 May 1993, q standard error.
82
Table 9.Correlations of physical and chemical variables.
89
Table 10.Correlations of physical and chemical variables with
factors. 90
Table 11.Two-way analysis of variance comparing factor scores
for
lysimeter data among slopes and sampling dates for the
entire data
set.
92
Table 12.Mean factor scores for the Flat Near and Steep Near
transects
grouped by vegetation zone, q standard error.
93
Table 13.Correlations of physical and chemical variables with
factors using
data from transition zone only.
96
Table 14.Two-way analysis of variance comparing factor scores
for
lysimeter data among slopes and sampling dates for the
transition
data set only.
97
Table 15.Mean factor scores for
Transition zone dataset for the Flat Near
and Steep Near transects, q standard error.
99
Salt and brackish marshes are characterized by sharp zonation patterns of vascular plants which have been the object of intensive study by many researchers. The vegetation patterns have been shown to be determined by the frequency and duration of tidal flooding along with its related affects on substrate oxygen levels, pore water salinity, pore water sulfides, and nutrient availability. These soil and pore water characteristics have been shown to limit growth, reproduction, and establishment of marsh vegetation.
Much of the work which has been done on the factors influencing zonation and growth of salt marsh vegetation has been conducted on the regularly flooded low and irregularly flooded high marsh vegetation zones. However, the transition from marsh to forested and upland areas at the landward edge of salt marshes has received little attention despite the fact that these areas are particularly susceptible to the effects of sea level rise. The transition from marsh to upland represents a continuum of soil and pore water characteristics which influence vegetation types and abundances.
The existing paradigms for coastal wetlands suggest that abiotic variables associated with inundation by brackish water and their influence on vegetation should decrease along the continuum from low marsh to upland as a function of elevation. Elevation has been shown to exert its influence mainly through the extent of tidal inundation. However, areas with little or no change in elevation from the marsh landward may be influenced more by the distance to a tidal source than elevation. The effect of landscape slope on the marsh-forest soil characteristics, and ultimately the position of vegetation zones, has not been well documented.
The purposes of this section are to review the literature on the factors known to affect marsh and upland vegetation and to formulate hypotheses about the patterns of these factors along the continuum from marsh to upland along gradients of slope. The introduction is organized into three sections which will discuss: (1) the physical and chemical factors affecting salt marsh and upland vegetation; (2) the hydrologic processes underlying the continuum from marsh to forested areas; and (3) suggested hypotheses for the physicochemical patterns as they are affected by slope.
The inundation by brackish water has been associated with changes in sediment physical and chemical characteristics which, in turn, affect the types of vegetation which can tolerate the potential stresses. The major physical and chemical stresses from flooding and its related effects include sediment oxygen availability, hydrogen sulfide, and salinity.
The saturation of soil with water reduces the rate of oxygen diffusion (Gambrell and Patrick 1978) and can limit root and microbial oxygen availability. Plants growing in submerged soils have two major adaptations to growth in anoxic sediments: oxygen transport from the aerial parts to the roots and fermentation.
Oxygen transport from aerial parts to the roots in some marsh plants occurs mainly through gas spaces or aerenchyma (Conway 1940; Armstrong 1975). Marsh plants have well developed aerenchyma and their root cells no longer depend on diffusion of oxygen from the surrounding soil, which is the main source of root oxygen in terrestrial plants. However, Bartlett (1961) found that terrestrial plants vary widely in their resistance to waterlogging and that this resistance was linked with the capacity of the root to oxidize the rhizosphere. Terrestrial plants respond to oxygen stress in the roots by forming intracellular gas spaces in the cortex (Bryant 1934). Bryant (1934) suggests that limited amounts of oxygen may be transferred from the shoot to the root to enable terrestrial plants to survive short periods of waterlogging.
Reduced soil environments have been associated with limited productivity and growth in Spartina alterniflora (Teal and Kanwisher 1965; Linthurst 1979; Mendelssohn and Seneca 1980; Howes et al. 1981; Mendelssohn and McKee 1988) and two European marsh species Spartina townsendii and Molinia caerulea (Goodman and Williams 1961). 1.1.2 Hydrogen Sulfide
When oxygen is limiting, microbial communities use other terminal electron acceptors to break down organic matter (Turner and Patrick 1968), converting the soil to a biochemically reduced state and producing end products which are potentially toxic to wetland vegetation. Hydrogen sulfide (H2S), an end product of sulfate reduction, has been shown to be a phytotoxin in wetland plants (Goodman and Williams 1961; Havill et al. 1985; Koch and Mendelssohn 1989; Koch et al. 1990). Linthurst (1979) found that growth of S. alterniflora in a greenhouse experiment was negatively correlated with hydrogen sulfide concentration of the sediment. Also in a greenhouse experiment, Koch and Mendelssohn (1989) directly added sulfide to intact cores of a salt marsh species, S. alterniflora, and a fresh marsh species, Panicum hemitomon, to determine the effects of sulfides on growth. Sulfide significantly reduced culm, root, and rhizome biomass in P. hemitomon, but only root biomass in S. alterniflora, suggesting interspecific difference in tolerance levels.
Although some of the effects of sulfide on the growth of some marsh plants are known, the precise physiological mechanism whereby sulfide exerts its effect is not clear. Recent investigations have attempted to clarify the growth limitation mechanisms of sulfide toxicity. In a greenhouse experiment, Koch et al. (1990) found that the effect of sulfide results from the inhibition of the anoxic generation of energy via alcoholic fermentation and a concomitant reduction in plant nitrogen uptake. They conclude that sulfide has a significant negative effect on the anoxic production of energy in roots via alcohol fermentation, leading to inhibition of plant nitrogen uptake.
Sulfate reduction to sulfide occurs mainly in the top 30 cm of New England salt marsh soils (Howarth and Teal 1979), Georgia salt marshes (King et al. 1982; Howarth and Giblin 1983), and Long Island Sound sediments (Martens and Berner 1977). Howarth and Teal (1979) found that the rates of sulfate reduction tended to be slightly lower in the top few centimeters than between 4 and 18 cm, even though all the sediments below the top few millimeters were reducing.
Spatial variability in the rates of sulfate reduction is thought to be due to the amount of organic matter available to fuel sulfate reduction (Martens and Berner 1977; Howarth and Teal 1979) and gradients in tidal water movement (King et al. 1982). Howarth and Teal (1979) suggest that the growth and decomposition of the belowground biomass provide a large annual input of organic carbon throughout the top 20 cm of the peat resulting in high sulfate reduction rates over depth. King et al. (1982) found higher sulfide concentrations in sites which had the least water movement, mainly areas away from the creekbank and at a higher elevation.
Sulfate reduction has also been shown to be temporally variable. Sulfate reduction rates in marsh soils have been shown to vary on a daily and monthly basis (Howarth and Teal 1979; Peterson et al. 1983). In a New England salt marsh, Howarth and Teal (1979) showed that for three years the average sulfate reduction for all sediment depths had maxima during the late summer and fall, and minima in the winter. However, the seasonal trend was not entirely explainable by temperature alone. The authors concluded that sulfate reduction in the peat is controlled by substrate availability (i.e., energy source), not by the amount of SO42-.
Almost no information exists in the literature describing the effect of sulfate reduction in forested areas fringing salt marshes. This is probably because forest sediments rarely experience reducing conditions conducive to sulfate reduction despite occasional sulfate influxes and flooded conditions from seawater during extreme flooding events. Unlike strictly terrestrial systems where sulfate can be but is rarely limiting (Coleman 1966), forested areas surrounding marshes probably are never sulfate limited due to these influxes, especially areas which experience flooding more frequently. No information exists on whether these flooding events contribute to the sulfur cycling in forests which fringe salt and brackish marshes and whether the cycling is significant from an ecological perspective.
Very limited information exists on the effects of sulfate reduction on the growth of forest species. During photosynthesis, SO42- and HSO3- are reduced and many species can emit the end product as sulfide gas (Winner et al. 1981). The majority of studies which have been conducted on sulfate reduction are primarily concerned with sulfide emission rates from the leaves of green plants (Winner et al. 1981), not soil sulfide toxicity. However, Spaleny (1977) added K2SO4 to spruce seedlings to determine the effect of increased levels of sulfate in the soil on sulfide production. He found that the addition of sulfate caused a metabolic inhibition resulting in a brown necrosis of the needles, similar to field observation in sulfate rich, polluted areas. Spaleny (1977) concludes that increased levels of soil sulfate due to pollution would considerably damage spruce seedlings. 1.1.3 Salinity
High soil salinities (e.g., 35-50 ppt) are hostile to most vascular plant species thus excluding all but the most tolerant species known as halophytes. The physiological mechanisms of halophytes differ from other plants since salt accumulation in the cells can lead to toxic ionic imbalances. Halophytes can withstand salinity by a combination of mechanisms of quantitative or qualitative salt exclusion at the roots (Smart and Barko 1978; Antlfinger and Dunn 1983), and the leaves (Anderson 1974), and by a compartmentalization of salts within cells (Jefferies 1981).
The dominant factors which control the salinity of salt marsh soils are frequency and duration of tidal flooding (Wells 1928; Chapman 1940; Hinde 1954; Chapman 1960; Adams 1963), incidence and amount of rainfall, and evaporation (Chapman 1960; Ranwell et al. 1964; Mahall and Park 1976). Other factors which have been shown to influence soil salinity are soil texture (Chapman 1960; Smart and Barko 1978), depth to the water table (Penfound and Hathaway 1938), fresh groundwater inflow (Chapman 1960; Lindberg and Harriss 1973; Harvey et al. 1979), and proximity to a tidal creek (Chapman 1960). It is generally thought that soil salinity concentration at any one point in the marsh is influenced by a combination of one or more of these factors.
Research into the spatial patterns of soil salinity from the marsh landward in tidal salt marshes has shown that soil salinity increases landward to a maximum at or about mean high water and then gradually decreases (Gilham 1957a; 1957b; Kurz and Wagner 1957; Adams 1963). In contrast, other researchers indicate that salinity decreases from the edge of S. alterniflora zone landward without an increase at mean high water (Penfound and Hathaway 1938; Reed 1947). However, Reed (1947) studied marshes with sediments that had a thin layer (2-25 cm) of fine sand on the surface, and found that drainage increased landward. Reed (1947) suggests that the improved infiltration and drainage landward lead to decreased salinities.
The mechanisms behind the spatial patterns of sediment salinity are thought to be related to the extent of tidal flooding which gradually decreases from the marsh landward up to mean high water. Because the presence of salt in the soil above mean high water is transported by only the higher high tides, soil salinity must decline with increasing distance landward. However, the lower frequency of flooding above mean high water increases the influence of evapotranspiration and rainfall as controls on salinity.
These controls on salinity are highly variable temporally, and can lead to extremely high salinity in the summer when rainfall is minimal and evapotranspiration is high (Ranwell et al. 1964; Mahall and Park 1976). Nestler (1977) found that interstitial salinities dropped in the Juncus portion of the high marsh following a rainfall, but increased later in drought situations. He concluded that high interstitial salinities indicated less frequent tidal flooding.
Studies of the soil salinity patterns in tidal marshes have also been shown to vary with depth of sediment (Christian et al. 1978). In a tidal salt marsh along the northeastern Gulf coast of Florida, Lindberg and Harriss (1973) found that the sediment surface had the highest concentration of pore water salinity. 1.2 Effects of Slope 1.2.1 Effects on Hydrology
Slope has the potential to control the magnitude and direction of surface and subsurface water movement. The effect of slope on hydrology characteristics is paramount since the transport of all nutrients, metabolic toxins, and solutes is hydrologically mediated. In general, salt marsh sediments may receive water by runoff or regional groundwater aquifer discharge from upland areas, direct precipitation, or from tidal flooding (Figure 1). Water is lost in the sediment by evaporation, transpiration, or drainage through the sediment and out into the tidal creek. The degree of slope has the potential to influence water movement in the form of runoff or tidal flooding, drainage, and groundwater discharge.
Slope, which has both an elevation component and distance component, influences the extent of flooding by tidal water since surficial water is physically restricted by friction caused by vegetation as it travels over longer distances from the tidal creek. Figure 1. Elements of a water budget in a tidal salt marsh as they are affected by slope. Slope has the potential to control water movement in the form of runoff, drainage, or discharge, and is indicated with an asterisk (modified from Nuttle 1988). This restriction is analogous to the non-channelized overland sheetflow situation for flooded riparian systems and can be calculated using the Manning equation (Mitsch and Gosselink 1986). Because the slope component in the Manning equation is represented by a square root, this suggests that very low gradients of slope would result in proportionately higher restrictions to sheet flow, all other variables being equal. The restriction due to vegetation may explain why Jackson (1952) found that the extent of land inundated by spring tides was less than based on elevations alone. Conversely, if the degree of slope is high, upland runoff would be expected to increase.
The amount of water movement in sediments depends on the sediment hydraulic conductivity and the hydraulic gradient. In sediments with relatively high hydraulic conductivity, slope is positively correlated with subsurface drainage. For example, Harvey (1990) found that drainage had a strong positive correlation with slope for a tidal freshwater marsh in Virginia. These findings supported modeling results which determined that horizontal hydraulic gradients highly correspond to topographic gradients in tidal marshes (Harvey et al. 1987). However, these studies were conducted in marshes which had relatively high soil conductivities (7.4 x 10-4 cm s-1; Harvey et al. 1987). Sediments with lower hydraulic conductivities (0.05 x 10-4 cm s-1, Harvey 1990) restrict water movement and require a larger hydraulic gradient for drainage to overcome this restriction (Harvey and Odum 1990).
Drainage in tidal salt marshes may be significant for sediment in areas which are near a tidal creek. For example, Nuttle (1988) found that there is essentially no horizontal water movement greater than 15 m from the creekbank in a New England salt marsh. Without substantial drainage, the sediment at the interior of the marsh is mainly influenced by evapotranspiration (Hemond and Fifield 1982) and precipitation (Ranwell et al. 1964). Although drainage often constitutes a small proportion of water loss from tidal marsh soils when compared to evapotranspiration, solutes (e.g., salt) are exported in drainage water and not evapotranspiration. This suggests that areas which experience less infiltration will have higher solute concentrations.
At the upland border, the influence of groundwater becomes increasingly important. Groundwater from regional aquifers is shown to have a large upward hydraulic gradient beneath the hillslope and marsh (Harvey et al. 1988; Harvey and Odum 1990). However, the interaction between groundwater from hillslopes and pore water from adjacent tidal marsh soils is highly dependent on sediment composition. For example, highly conductive aquifers of glacial till deposits in Sippewissett, MA, which are composed of low organic coastal sands, had large groundwater discharge (4397 L m-1 d-1) (Valiela et al. 1978).
In contrast, groundwater discharge is much less (6-10 L m-1 d-1) (Harvey and Odum 1990), and constitutes a small component (<10%) of the subsurface water budget in sediments where clay or organic muds are present in substantial proportions, as in the coastal plain of Virginia (Harvey et al. 1988). Because of this low sediment hydraulic conductivity, vertical gradients of groundwater discharge were not detected within 2 m of the surface of the marsh and the upward hydraulic gradient diminished exponentially as a function of distance into the marsh (Harvey et al. 1988). 1.2.2 Effects on Vegetation
Ultimately, landscape slope has the potential to affect the distribution and abundance of marsh and upland plant species by its effect on physicochemical variables. Based on previous studies in low and high marshes, it is expected that pore water salinity, sulfides, and redox potential should play a significant role in determining plant species composition in the high marsh-upland transition zone. Since many of these abiotic variables are interrelated, a simple hierarchical conceptual model was created to illustrate the connections and relationships between the abiotic variables and vegetation as influenced by slope (Figure 2). Each of these soil physicochemical variables is heavily influenced by differences in hydroperiod. It should be noted that the conceptual model was created to illustrate the potential effects that slope has on vegetation and is not meant to be complete since it lacks obvious feedback mechanisms.
Along the continuum from marsh landward, vegetation becomes less adapted to flooding by brackish water and its related effects. For example, halophytic vegetation (e.g., S. alterniflora) can withstand high pore water salinities, shrubs (e.g., Iva frutescens L.) can tolerate intermediate salinities, and trees (e.g., Pinus taeda L.) Figure 2.
Simple conceptual model of the effects of slope on the hydrology and pore water and vegetation used to formulate hypotheses. have a limited capacity to tolerate saline water. Thus, the degree of slope may ultimately affect which types of vegetation can survive in a given area due to the effects on soil properties.
The degree of slope may also have a profound impact on the response of wetland and forested areas to a rising sea level. Brinson et al. (in preparation) suggest marsh migration over land (i.e., transgression) may be influenced by the slope of the antecedent landscape. Marsh areas which abut steep sloped uplands are expected to stall the transgression process whereas marshes which abut gradual or flat sloped uplands would undergo more rapid transgression. Thus, the effects of slope are likely to play an integral role in the dynamics of marsh response to sea level fluctuations. 1.3 Objectives and Hypotheses
The overall objective of this research is to determine the effects of slope on abiotic and biotic variables along a brackish marsh-upland continuum. Specifically, the objectives are to assess the effects of slope on sediment physicochemical variables (i.e., hydroperiod, pore water salinity, redox potential, and sulfides) along gradients of steep and flat slopes; to assess the distribution of vegetation communities along the same continuum; and to correlate the soil variables with the vegetation distribution.
Because vegetation distribution responds to abiotic variables, and these variables are affected by slope, slope effects are manifest in the position of vegetation zones. In particular, the position of the transition zone between marsh and upland is indicative of the most landward extent of salt water influence and the most seaward extent of terrestrial influence. The position of the transition zone integrates these influences and thus represents a common point of reference to compare the effects of slope.
I used the position of the marsh-upland transition zone as a frame of reference to formulate testable hypotheses on the effects of slope on the following variables: (1) elevation of the transition zone, (2) hydroperiod, (3) salinity, (4) redox potential, and (3) sulfide concentration. Care was taken in constructing the hypotheses to address variables separately since many of the abiotic variables were expected to be related. The null hypotheses in each case were that each variable would be the same in steep and flat transition zones.
The transition zone of the flat transect is hypothesized to occupy a lower elevation, but at a greater distance from the source of tidal flooding for a flat slope because the resistance to tidal flow increases as distance increases. The distance to the tidal source is expected to ultimately affect many of the pore water variables because of its affect on hydrologic controls. The flat transition zone is expected to have a lower hydraulic gradient resulting in decreased potential for drainage and thus is hypothesized to exhibit a greater hydroperiod. Although not directly related to drainage, the flat transition is hypothesized to exhibit a higher flood frequency from extreme high water events because of its lower elevation in relation to mean high water. The effects of hydraulic gradient on drainage is expected to play a major role in influencing pore water characteristics. It is hypothesized that pore water salinities of a steep slope will be lower than those of a gradual slope because of lower potential for drainage and higher potential for evapotranspiration. The effect of slope on drainage potential is also hypothesized to result in lower pore water redox potentials and higher concentrations of hydrogen sulfide for a flat slope.
The study of the factors which affect vegetation dynamics is usually conducted at the ecological (i.e., less than a decade) or paleoecological (i.e., centuries to millennia) time scales. The vegetation dynamics between these time scales (i.e., decades to centuries) is difficult to obtain since it is limited by methodology. Long term records by pollen analysis usually lack the spatial resolution of pattern, and vegetation analyses address the spatial scale at which individuals interact, but lack the historical perspective. Thus, this work establishes the initial effort of characterizing the abiotic and biotic patterns along transitions from marsh to upland within the framework of long term ecological research. 2. SITE DESCRIPTION AND METHODS 2.1 Site Description
The study area, Phillips Creek Marsh (37x 27', 75x 50'W), is located at the Virginia Coast Reserve (VCR-LTER) in Brownsville, VA, and is owned by The Nature Conservancy (Figure 3). The site is an expansive brackish marsh on the mainland side of the coastal lagoon complex on the Eastern Shore of Virginia. Flooding of the marsh is from the southwest by the tidal creek, Phillips Creek, which has a tidal range of about 145 cm. The average monthly rainfall is 86 mm, with annual maximum rainfall in July and August due to conductive storms (USDA 1989).
The stratigraphy of the coastal plain of Virginia is composed of sands and clays of Cretaceous age or younger that were deposited atop bedrock in fluvial, estuarine, or marine environments (Powars et al. 1988). The water table aquifer is typically 3 to 20 m thick and is underlain by the semi-confining layers of clayey-sand (Pliocene age) that contain fresh interstitial water or groundwater (Cederstrom 1943, as cited in Harvey 1990). The surficial pore water of marsh soil is composed of deposits of Holocene age or younger.
The soil types along the shoreline range from very poorly to poorly drained and consist of Chincoteague (low marsh), Magotha (high marsh), and Nimmo, Munden and Bojac (upland) (USDA 1989). The dominant soil assemblage along the coast of Northampton County, VA, consists of Chincoteague-Magotha-Nimmo and Chincoteague-Magotha-Munden soil types (Figure 4). These soil assemblages account Figure 3.
Location and arrangement of study transects at the Brownsville, VA, marsh representing Steep Near (SN), Steep Far (SF), Flat Near (FN), and Flat Far (FF) transects in relation to Phillips Creek. Figure 4. Frequency of occurrence for shoreline to upland soil assemblages for Northampton County, VA, using soil maps (USDA 1989). Capital letters designate abbreviations of soil types from shore to upland. Soil abbreviations are: tidal marsh soil (low marsh) Chincoteague (C), and Fisherman (F); intertidal (high marsh) Magotha (M), Camocca (Ca), and Udorthents (U); and upland (forest) Nimmo (N), Munden (Mu), Bojac sandy (Bk), Bojac fine sandy (Bo), Dragston (D), Seabrook (S), Molena (Mo), and Polawanna (P). Soil assemblage frequencies were determined by drawing a perpendicular line from the marsh edge to upland for every 300 m of shoreline and recording soil types. for over 60% of the shoreline soils. The principal soil types for the study site consist of Chincoteague, Magotha, Nimmo, and Munden.
The Magotha soil series is very deep, fine sandy loam, and poorly drained. Magotha soils are more loamy in the upper part of the soil than Chincoteague soils, have higher levels of sodium than Nimmo soils, and are grayer than Munden soils. The Nimmo series are also very deep, poorly drained and are grayer than Munden soils. Munden soils are moderately well drained and have moderately coarse textured sediments.
The brackish marsh is surrounded by forested uplands to the north and west, and by an agricultural field to the south (Figure 3). The marsh grades gradually into the forested areas to the north and steeply in the south.
Well developed low and high marsh plant communities occur in the study area with gradients of high marsh-upland transition zones. In general, the high marsh is dominated by Spartina patens, J. roemerianus, Panicum virgatum, and Distichlis spicata. In the transition zone, these species are joined occasionally by the shrubs I. frutescens, Baccharis halimifolia, Myrica cerifera, and small (< 3 m) tree species Juniperus virginiana. The transition is also characterized by many dead red cedar trees and stumps, especially along gradual slopes, similar to high marsh forest transitions from Long Island, New York (Clark 1986a; 1986b), to Florida (Kurz and Wagner 1957). The forest species are predominantly pines such as P. taeda, and large J. virginiana. 2.2 Study Transects
Four linear transects perpendicular to the high marsh-upland transition zone were established to represent extremes of slope (i.e., flat or steep), and distances from a tidal source (i.e., far or near) (Figure 3). Each combination of slope and landscape position: Steep Near, Steep Far, Flat Near, and Flat Far was represented by a different transect (Figure 5). However, this designation does not represent a complete array of slope and landscape positions since some arrangements (e.g., flat transition immediately adjacent to a tidal creek) are improbable.
Transect lengths were established initially by choosing common end members (i.e., predominance of high marsh vegetation and distance to tidal creek for the beginning of the transect, and lack of marsh vegetation cover for the end of the transect) to represent starting and ending points. This designation inherently may misrepresent the actual length of the transect since starting and ending points are arbitrary. However, I felt that capturing the transition between marsh and forested vegetation types was more important than the actual length of the transect. As such, it should be noted that the actual length of the study transects could differ considerably (q 20-50 m) if others had chosen it. The starting point was at approximately 50 m away from Phillips Creek for Steep Near, 200 m for Steep Far, 150 m for Flat Near, and 250 m for Flat Far. Transect distances were numbered to begin at 0 and increased sequentially toward the forest area.
Vegetation cover, elevation, sediment bulk density, percent carbon, percent Figure 5. Possible combinations of landscape position in relation to proximity to tidal creek and gradients of slope used to choose study transects. water, chlorinity, and depth to mineral sediment were assessed once for all transects and are described in section 2.3. Two of the four transects, Steep Near and Flat Near, were instrumented at sampling stations to measure hydrologic and pore water variables. Water table position, pore water salinity, redox potential, acidity, and sulfide concentration are described in section 2.3. Ten sampling stations per transect were installed and the distance between sampling stations was arbitrarily chosen to reflect the width of the transition zone and the total length of the transect (Figure 6). In general, sampling was conducted monthly during the cool season and twice monthly for the warm season for up to 2 years. 2.3 Methods 2.3.1 Vegetation Sampling and Analysis 2.3.1.1 Vegetation Sampling
The percent cover of grasses, shrubs, and trees was assessed continuously along study transects using a plot method. A quadrat was created by placing meter tapes on the ground to create a 2 x 4 m frame. The frame was sub-divided into a 1 x 1 m frame to measure herbaceous ground cover, 2 x 2 m frame to measure shrub cover, and a 2 x 4 m frame to measure tree cover (Figure 7a) (after Mueller-Dombois and Ellenberg 1974). Percent cover for trees and shrubs was directly estimated within the sampling frame by calculating the total area occupied for each species divided by the total area. For herbaceous species, cover was difficult to judge for Figure 6. Layout of the Flat Near and Steep Near study transects indicating locations of the surficial groundwater wells, lysimeters, and water level recorders. (A) Flat Near transect. (B) Steep Near transect. Figure 7. Sampling frame for vegetation analysis. (A) Frame size and geometry used for vegetation sampling. Frame size for measuring vegetation were: herbaceous (H) 1 m2, shrub (S) 4 m2, and trees (T) 8 m2. (B) The moving split-window for analysis of vegetation using a six sample window (asterisks designate sample points). The mean vegetation cover for each species is computed for each window half and the dissimilarity is calculated between windows. The window is then moved along the transect one sampling point at a time (after Johnston et al. 1992). individual species and dominance was assessed instead. An algorithm converting dominance to cover was then used. The algorithm was based on field observations between cover and dominance (Table 1). Vegetation layers were chosen (after Mueller-Dombois and Ellenburg 1974) to fall into 3 categories: (1) herbaceous layer (<30 cm to 1 m); (2) shrub layer (>1 m to 3 m); and (3) tree layer (>3 m). Species were identified using Radford et al. (1968). 2.3.1.2 Vegetation Analysis
The moving split-window analysis was used to reveal vegetation discontinuities which were then used as a basis for partitioning study transects into three zones: (1) high marsh, (2) transition, and (3) forested/upland. The moving split-window technique identifies boundaries by peaks in squared euclidean distance (SED) (Wierenga et al. 1987; Ludwig and Cornelius 1987).
The SED is calculated by placing a double window over equally spaced sampling points (a sampling point is the total cover of all species types in one 2 x 4 m frame), and the dissimilarity (distance) between attribute values in each window half is statistically compared (Figure 7b). The window is moved sequentially along the transect until all comparisons are made for the entire length of the transect. The mean attribute value is calculated for each window half and the SED is computed as the square of the difference between the means of each variable (i.e., individual species cover) in adjacent windows, summed across all variables measured (i.e., all
Table 1. Algorithm used to convert dominance rankings to
percent cover. Algorithm was based on field observations between cover
and dominance.
Species Dominance Rank Multiply total
cover by:
Example:
total cover=80%
AA= 11.0A= 80% cover
A > BA= 1
B= 2 0.75
0.25 A=
60% cover
B=
20% cover
A > B > CA= 1
B= 2
C= 3 0.60
0.30
0.10 A=
48% cover
B=
24% cover
C=
08% cover
A = BA= 1
B= 1 0.50
0.50 A=
40% cover
B=
40% cover
species) (Brunt and Conley 1990; Johnston et al. 1992). Peaks in the SED are
then
used to indicate boundary locations (Wierenga et al. 1987).
Although the data collection methodology was identical for all study transects, it became apparent that the analytical window size would have to be sensitive with regard to the overall transect length. Thus, the analysis of the steep gradient transects necessitated a smaller analytical window width (e.g., 4) than the flat transect (e.g., 8 or 10) by nature of its shorter overall length. A smaller window width is more sensitive to small changes in species composition (e.g., for one life form) and can cause greater noise (i.e., many peaks which do not correspond to vegetation zones). A larger window may include two or more boundaries and may obscure boundary locations. Thus, after trying a range of window widths, I chose the analytical window sizes to best represent field observations. 2.3.2 Surface Elevation
Elevations along study transects were determined using a laser level (Pentax Total Survey Station III, Model #5, Asahi Precision Co., Tokyo, Japan), data logger (Pentax, SC-5, Pentax Instruments, Englewood, CA) and reflecting prism. Optical leveling used to augment the laser leveling in areas where views were obstructed using an auto level (Topcon Model AT-F2, Tokyo Optical Company, Tokyo, Japan) and stadia rod. All elevations were referenced to the Hayden Benchmark (Virginia Coast Reserve benchmark HAYD, N 372732.021 W 754958.036) located adjacent to Phillips Creek creekbank water level recorder (Figure 6). This standardized elevations for all transects relative to the benchmark and facilitated elevation comparisons between transects.
Slope for the entire transect was calculated using linear regression and represents the change in elevation (m) divided by change in distance (m). Slope is reported as m m-1, and thus is unitless. The conversion to percent slope can be calculated by multiplying the slope number by 100, and conversion to degree by multiplying the percent slope by 0.9x.
The humocky nature of the surface elevations were assessed by measuring surface macrotopography (i.e., at the scale of 5-10 m). Macrotopography along the transects was determined by the departure from linearity of the linear regression line. The regression line would be expected to have a better fit (i.e., higher r2 value) to the surface elevations if the transect profile departed little from the regression line. However, the r2 value from regressions of data having zero slope (e.g., the flat slope) is zero, so the root mean square error term was used instead. The regression line would be expected to have a better fit to the surface elevations (i.e., a lower root mean square error) if the transect departed little from the regression line. A hummocky area would be expected to have a higher root mean square error. Thus, the root mean square error value was used as a surrogate for actual small scale macrotopographic measurement. It should be noted that this is a crude estimation of the surface macrotopography and is only used as a comparison between transects. 2.3.3 Soil Characteristics
Core samples were obtained for quantitative analysis using a thin walled cork borer (size #12, 2 cm o.d.) in August 1993. Samples were obtained by excavating a small soil pit (approximately 30x30x30 cm) by shovel and extracting cores from the vertical faces of the pit. Core samples were obtained with minimal compression by rotating the corer slowly into the sides of the pit at the appropriate depth (i.e., 10 and 20 cm). Core samples were extruded and immediately placed in Ziploc bags. Depth to mineral soil was measured directly from the sides of the of the pit by visual approximation and meter stick. Cores were taken every 20 m for the Flat Near transect, every 10 m for the Flat Far, and every 5 m for the Steep Near and Steep Far transects.
Soil samples for bulk density and percent water were dried to a constant mass (105xC) for 48 h. Soil samples were allowed to cool inside a desiccator prior to weighing. Soil bulk density was calculated as the sample dry mass divided by the volume (SSSA 1986). Percent water was calculated by the mass of the wet soil divided by the mass of the dry soil less 1, multiplied by 100 (SSSA 1986).
Soil samples for total carbon and chlorinity were air dried for 72 h, milled by mortar and pestle, and sieved through 2 mm copper screen to remove macroorganic matter (i.e., large roots and rhizomes). Total carbon was determined using a H-C-N elemental analyzer (Leeman Labs, Model CE 440, Lowell, MA) and was divided by the sample mass to obtain percent total carbon. Percent total carbon was chosen over organic matter as a measure of organic carbon in the soil because: (1) the amount of inorganic carbon in the soil (i.e., mainly dolomite) is very low or non-existent (as determined by the Test for Presence of Inorganic Carbon using concentrated HCl, SSSA 1986, p. 563); and, (2) the conversion value from organic carbon to organic matter is inexact. If conversion to organic matter is desired, a rough estimation would be to multiply the total carbon figures by a factor of two (SSSA 1986). An attempt was made to mill all material of the soil sample, but large organic clumps (i.e., large rhizomal pieces and woody fragments) were unable to be used by the HCN analyzer. It should be noted that the actual total carbon may be underestimated for soil samples which contained mostly large rhizomal material. This underestimation would be most likely affect the 10 cm depth soil samples and at locations closer to the transect beginning (e.g., 0 m).
Soil chlorinity was determined by placing a known volume of milled sediment into large (50 mL) centrifuge tubes. A known volume of distilled water (10-15 mL) was added and then shaken for 48 h to assure mixture. The resulting mixture had a volumetric soil-water ratio of 1:2. The samples were then centrifuged for 2 h at 3,000 rpm and allowed to settle. The supernatant was then automatically titrated using an automatic/amperometric Cl- titrator (Buchler Chloridometer Model 4-2500, Fort Lee, NJ), calibrated with known standards. Each sample was titrated twice. 2.3.4 Water Table Fluctuations
The Flat Near and Steep Near transects were instrumented with shallow wells to monitor the position of the water table during sampling dates and with water level recorders to monitor fluctuations in the water table continuously. The wells were evenly spaced along the transect at 5 m intervals for the steep transect (n=11) and at approximately 25 m intervals (n=9) for the flat transect (Figure 6). Water level recorders were installed in the transition zone for both study transects (Figure 6).
The position of the water table was monitored using wells constructed of PVC pipe (4.8 cm diameter) slotted from top to bottom and inserted 1.5 m into the substrate (after Bouma et al. 1980). Wells were screened with either multiple wraps of fiberglass window screening, or nylon stockings, and were augered to depth, backfilled, and sealed with Bentonite to prevent water from entering from the top. Distance from the top of the tube to the water table was measured to the nearest mm by meter tape. The water table was rarely more than 1.2 m and could be easily seen.
For continuous monitoring of the water table, one water level recorder (Stevens Type F) was installed in the transition zone for each of the Flat Near and Steep Near transects. The water level recorder was installed on 11 July 92 for the Flat Near transect, and 18 June 93 for the Steep Near transect. A stilling well (20.5 cm o.d.) slotted and screened with fiberglass window screening from top to bottom was inserted 2 m into the substrate to allow the float to travel beneath the surface undisturbed. Water table position was determined at the start and end of each chart period (i.e., 8-32 d) by measuring the distance from the water table to a fixed point on the water level recorder stand. The charts were then digitized and converted to date/times and water levels using a video recorder and JAVA software (JAVA, Jandel Scientific, San Rafael, CA).
Seasonal patterns of drawdown were estimated for the entire study area using simple water balance estimates. Periods with a high potential for water table drawdown were identified on the basis of precipitation totals and mean monthly potential evapotranspiration (PET) for two periods: (1) long term (1949-1981) data set for a town nearby the study site, Cheriton, VA (USDA 1989); and (2) a short term (1990-1993) data set collected on-site by a meteorological station (Krovetz and Porter 1993). PET was calculated using the Thornthwaite method (Thornthwaite and Mather 1957). Differences between monthly precipitation totals, mean monthly PET and the frequency of deficits of precipitation relative to PET were calculated for all months for the study period (1990-1993). 2.3.5 Groundwater Salinity
The salinity of the surficial groundwater was assessed using the same wells for determining water table characteristics. Groundwater salinity was determined using a S-C-T meter (Yellow Springs Instrument, Model 33, Yellow Springs, OH) which was calibrated with known standards. Salinities were taken at the water table surface and at the bottom. Values reported are the mean of top and bottom measurements. 2.3.6 Pore Water
Soil pore water was collected for chemical analysis using low-tension lysimeters (after Harvey 1990). The lysimeter allowed collection of pore water at a single level per instrument, at depths ranging between 10 and 60 cm. The body of the lysimeter was constructed of PVC pipe (5 cm o.d.) with the inlet capped with 70 fm porous nylon fritware. Lysimeters were installed vertically into the soil with the top stoppered and fitted with a sampling port and a gas port. Prior to sampling, a hand vacuum pump (Nalgene) was used to remove all water from the instrument and discarded. The instrument was then flushed with nitrogen gas to replace the headspace gas and evacuated. Pore water was collected about 12 h later with a 60 mL syringe. Shorter intervals between evacuation and collection were attempted, but failed to collect appreciable amounts of pore water due to very low soil hydraulic conductivity. Pore water samples were used to determine salinity, redox potential, hydrogen sulfide, and pH.
Pore water samples for salinity and pH determinations were collected in 20 ml glass scintillation vials, stoppered, and brought back to the lab and measured immediately. Salinity was measured directly using a temperature compensated refractometer (Reichert, Model 10419). The pore water pH was measured using an automatically temperature compensated pH meter (Fisher Model 1002, Springfield, NJ) and gel filled combination pH/ATC probe (Fisher 74613, Springfield, NJ). A two point calibration method was used and the electrode was recalibrated for each new sample.
Redox potential was measured in the field using two kinds of electrodes: (1) a commercially produced redox combination electrode (Orion H4402, Cambridge, MA), and (2) a brightened platinum electrode and a calomel reference probe (Fisher 74613). Sampler headspaces were kept purged with N2 prior to sampling. Pore water was removed from the sampler by syringe. The syringe was attached to a modified syringe chamber which contained the electrode (after Marcio Santos, University of Virginia, personal communication). The measurement was recorded after stabilization of the reading which occurred within about one minute after immersion of the electrode. Readings were standardized using Zobell's solution and referenced to the standard hydrogen electrode (SHE) by adding 244mV. The redox values were not normalized with respect to pH.
Hydrogen sulfide was determined colorimetrically (after Cline 1969). Pore water used for redox determinations was carefully added to stoppered test tubes which contained Cline's reagent and kept on ice. After collecting, samples were brought back to the lab and read immediately on the spectrophotometer. 2.4 Data Analysis
In order to evaluate the hypotheses for the physical and chemical variables as stated, statistical tests were performed on the data. Prior to the use of any statistical test, the data for all variables were tested for the necessary assumptions of normal distributions and homogeneity of variances; in all cases, assumptions were violated. Transformations were attempted using a Box-Cox power transformation test, but none adequately met these assumptions. Thus, the data were rank-transformed (PROC RANK, SAS User's Guide 1985) and all subsequent statistical tests were performed on the ranked data. The RANK procedure produces rank scores across observations which creates a normal distribution because the sapling distribution of ranks are equal.
Univariate analyses of variances (ANOVA) were performed separately on the variables that were measured once (i.e., elevation, carbon content, chlorinity, percent water, and bulk density) to test slope (i.e., flat, steep) and zone effects (i.e, high marsh, transition, and forest). Because of the small sample size for individual transects, data were pooled for vegetation zone by slope (i.e., flat, steep) and treated as replicates within zone. The assumption that data from separate transects could be pooled by zone was supported by a randomized block design using the variable elevation. Individual contrasts showed that no significant difference occurred between the transition zones for each type of slope. Significant differences occurred between the forest zone of the flat transects and the high marsh zone of the steep transects, but these areas had very small sample sizes (e.g., 2-4 sampling points) making comparisons difficult. It should be noted that this assumption was tested for elevation only, but is probably tenable for other variables because they are heavily influenced by elevation.
Differences in the variables that were measured repeatedly over time (i.e., pore water salinity, acidity, redox potential, hydrogen sulfide, and water table position) were assessed using a randomized block design blocked on date. I chose to block on date because many of these variables would be expected to respond similarly to date effects. It should be noted that the assumption of repeated measures as replicates is somewhat problematic since the argument can be made that the replicates are not strictly independent from each other. However, changes in these variables occur at time scales much less than the two week sampling interval and can, for the most part, be considered independent.
Many of the pore water variables and physical variables were thought to be interdependent so a factor analysis was used to determine common factors underlying the composition of multiple variables. Unlike the univariate analysis which used ranked data, the factor analysis was performed on actual scores. However, the factor analysis was used primarily for data reduction and distribution assumptions were not completely necessary. An orthogonal rotation method, Varimax factor rotation, was used to clarify factor interpretation (Duntemann 1984). Factor scores were then used as variables in ANOVA to determine if factor scores differed across zones and sampling dates. Tukey's (HSD) test was used to isolate which means were different, if any. Unless specified, significant refers to p < 0.05, and highly significant refers to p < 0.01.
Two subsets of the data were analyzed using factor analysis: data set for all zones, and a subset consisting of the transition zone only. The subset of the entire data were analyzed to determine if properties unique to the transition zone could be found using a smaller scale. 3. RESULTS 3.1 Vegetation 3.1.1 Zone Detection and Delineation
The moving split-window analysis revealed vegetation discontinuities which were used as a basis for partitioning study transects into three zones: (1) high marsh, (2) transition, and (3) forested/upland. In general, the flat transects have considerably more peaks than the steep transects (Figure 8), making boundary determination more difficult. Thus, it became necessary to use field notes and observations to "ground truth" and augment the window analysis for the flat transects. A window width of 4 sampling points was used for the steep transects and 8 sampling points for the flat transects. The effect of window width on peak identification is illustrated by data for the Flat Far transect. Varying the window width from 6 to 10 sampling units did not appreciably affect the interpretation of the boundary locations, only the emphasis of certain peaks (Figure 9). 3.1.1.1 Flat Near Transect
For the Flat Near transect, the analysis revealed 10 peaks between 0 m and 120 m (Figure 8a), but these were considered to represent mostly shifts in the types of herbaceous ground cover. For this transect, the zones were delimited as high marsh (0-120 m), transition (120-140 m), and forested (140-190 m). Figure 8. Squared Euclidean distance computed using moving split-window analysis. Open circles represent sampling points. Window size for analysis was 8 for the Flat transects, and 4 for the Steep transects. (A) Flat Near transect: high marsh (0-120 m), transition zone (120-140 m), and forest (140-190 m). (B) Flat Far transect: high marsh (0-62 m), transition zone (62-116 m), and forest (116-130 m). (C) Steep Near transect: high marsh (0-12 m), transition zone (12-22 m), and forest (22-50 m). (D) Steep Far transect: high marsh (0-20 m), transition zone (20-28 m), and forest (30-50 m). Figure 9. Peaks in the squared Euclidean distance for the Flat Far transect using window sizes of 10, 8, and 6 sampling points. The high marsh zone was predominantly comprised of D. spicata, S. patens, and J. roemerianus which accounted for the majority of the total cover (Figure 10a). In this zone, the shrubs I. frutescens and B. halimifolia, occasionally appeared, but do not make up a large part of the total cover. The landward extent of the high marsh and beginning of the transition zone was harder to detect over such a gradual change. A peak at approximately 120-135 m corresponded with field notes as the beginning of tree cover, as well as continued shrub cover and was felt to represent the beginning of the transition zone. The transition zone was comprised of the tree species P. taeda and J. virginiana, and the shrubs M. cerifera, and B. halimifolia (Figure 10a). However, the herbs S. patens and D. spicata continued to be dominant along with P. virgatum especially in higher, hummocky areas.
The landward extent of the transition zone and the beginning of the forested area was delimited by a sharp peak around 140 m (Figure 8a). The forested species was predominantly P. taeda, although an occasional J. virginiana was also found mostly in the marshward extent of the forested area. The shrub species M. cerifera was present along the edge of the forested area and was the only shrub species found in this zone. The ground cover was sparse and consisted predominantly Rhus radicans and Uniola laxa. 3.1.1.2 Flat Far Transect
The Flat Far transect had similar numerous, steep peaks as the Flat Near Figure 10. Percent of the total cover for each species grouped by zone and vegetation type. Species abbreviations are as follows: herbaceous: A=Spartina alterniflora, D=Distichlis spicata, P=Spartina patens, J=Juncus roemerianus, V=Panicum virgatum, and O= Other; shrubs: I=Iva frutescens, B=Baccharis halimifolia, J=Juniperus virginiana (< 3 m), and M=Myrica cerifera; and forest: P=Pinus taeda, and J=Juniperus virginiana (> 3 m). (A) Flat Near transect. (B) Flat Far transect. (C) Steep Near transect. (D) Steep Far transect. Figure 10. Concluded. transect (Figure 8b). Thus for the Flat Far transect, the zones were delimited as high marsh (0-62 m), transition (62-116 m), and forested (116-130 m). The steep peak at the 70-75 m location was primarily due to a large bare patch and was not considered a transition area. For the high marsh zone, the predominant ground cover was similar to the Flat Near transect and comprised of D. spicata, S. patens, and J. roemerianus (Figure 10b). Also present were P. virgatum, S. viridis, and Juncus spp., which were not found in the Flat Near transect. The shrubs present in the high marsh zone were B. halimifolia, M. cerifera, and small (<3 m) J. virginiana. The landward extent of the high marsh zone and the beginning of the transition zone was delimited by a sharp peak at approximately 58 m (Figure 8b). The peak corresponds to the emergence and dominance of tree and shrub species. In the transition zone, the dominant tree species were P. taeda and large J. virginiana (Figure 10b). The shrub M. cerifera and small (<3 m) P. taeda and J. virginiana comprised all of the shrub species in the transition zone. The herbaceous ground cover was predominantly P. virgatum, S. patens, and Setaria viridis, although D. spicata and J. roemerianus were present.
The large, steep peak at 116 m marked the end of the transition zone and the beginning of the forested zone (Figure 8b). The tree P. taeda and the shrub M. cerifera, comprised the majority of cover in the forested zone (Figure 8b). The ground cover was mainly R. radicans, and occasionally Campsis radicans. 3.1.1.3 Steep Near Transect
The peaks in the SED for the Steep Near transect were less numerous and broader than both of the flat transects (Figure 8c), making zone boundaries easier to define. The zones were delimited as high marsh (0-12 m), transition (12-22 m), and forested (22-50 m). The large peak at 44 m was primarily due to patchy bare areas and very dense pockets of vine species (e.g., R. radicans), and was not considered to be a separate zone.
For the Steep Near transect, the high marsh cover consisted of mostly D. spicata and S. patens with S. alterniflora appearing at the marsh end of the transect (Figure 10c). The herbaceous ground cover was joined by the emergence of the shrub species I. frutescens and B. halimifolia at 10 m.
The landward extent of the high marsh and the start of the transition zone was signalled by a large peak around 12 m (Figure 8c). The peak was primarily due to a change in the ground cover species, but also signalled the beginning of cover for forested species. S. patens and P. virgatum became the dominant herbaceous species (Figure 10c) in the transition zone. Shrub cover was predominantly B. halimifolia and M. cerifera. Tree cover was predominantly comprised of P. taeda and J. virginiana, although these accounted for only a small amount of the overall cover.
The edge of the forested ecotone and the marsh transition was delimited by another sharp peak at approximately 22 m (Figure 8c). The peak was mostly due to the shift of cover from shrub species to forest species, with all shrub species terminating at 32 m. The dominant tree cover species was P. taeda, although few very large (>3 m) J. virginiana were present (Figure 10c). The shrubs B. halimifolia, M. cerifera and small J. virginana accounted for very little of the total cover. The ground cover was mostly R. radicans, Sagittaria latifolia, and P. virgatum. The percent cover of these ground species was very sparse, with large bare patches and occasional dense pockets. 3.1.1.4 Steep Far Transect
The peaks in the SED for the Steep Far transect were very distinct and less numerous than any of the study transects. Peaks were delimited at 22 m and at 28 m, and correspond to high marsh (0-20 m), transition (20-28 m), and forested (30-50 m) vegetation zones (Figure 8d).
The high marsh consisted of mostly herbaceous ground cover with S. patens and D. spicata as the sole species (Figure 10d). The shrub I. frutescens was also present, but contributed little to the total cover.
The peak delimiting the landward extent of the high marsh and beginning of the transition zone was mostly due to the shifting of the total cover from herbaceous species to the shrub I. frutescens, which was the sole shrub member (Figure 10d). In the transition zone, the herbaceous cover was dominated by S. patens and D. spicata with traces of P. virgatum.
The tree cover of the forested zone which began at 30 m was comprised of exclusively P. taeda (Figure 10d). The shrub species cover was comprised of M. cerifera and small (<3 m) J. virginiana. The herbaceous ground cover was similar to the Steep Near, with P. virgatum, R. radicans, and S. latifolia. A vine, C. radicans, and the grasses, U. laxa, and S. viridis, were also present and together comprised most of the ground cover. 3.1.2 Plant Species Presence
In total, twenty-four species were identified along the four transects (Table 2). Seven species could be found in all three zones and included the graminoids D. spicata, P. virgatum, and S. patens, the shrubs I. frutescens, B. halimifolia, M. cerifera, and a tree species J. virginana. One species was unique to the high marsh (S. virginica) and one was unique to the transition zone (S. bona-nox). Eight species were unique to the forested areas and generally represent grasses and trees which cannot tolerate saline water. In general, the high marsh were the least rich, the transition were the next rich, and the forested areas were the most rich.
Comparisons of species presence and diversity between the gradients of slope, the flat slopes were slightly less rich than the steep slopes. J. roemerianus was found in the high marsh and transition on both flat slopes, but was noticeably absent on the steep slopes. Similarly, Limonium carolinianum was found along the steep slopes, but was absent on the flat transects.
Table 2.3.2 Elevation 3.2.1 Variation in Elevation by TransectVascular plant species in the Phillips Creek marsh. Zones are designated as HF: High marsh, flat transect; HS: High marsh, steep transect; TF: Transition zone, flat transect; TS: Transition zone, steep transect; FF: forest, flat transect; FS: forest, steep transect. Presence of a species in a zone is indicated by a "1" in the zone columns. Species nomenclature follows Radford et al. (1968). Family Species Common Name Zone HFHSTFTSFFFS Poaceae Distichlis spicata Greene Panicum virgatum L. Uniola laxa (L.) BSP. Setaria viridis (L.) Beauv. Spartina alterniflora Loisel Spartina patens (Aiton) Muhl Juncaceae Juncus roemerianus Scheele Juncus spp. Cyperaceae Fimbristylis spadicea Vahl Scirpus olneyi Gray Anacardiaceae Rhus radicans L. Plumbaginaceae Limonium carolinianum (Walter) Britton Alismataceae Sagittaria latifolia Willd. Salt grass Switch grass Foxtail grass Saltmarsh cordgrass Saltmeadow cordgrass Black needlerush Marsh fimbristylis Three-square sedge Poison ivy Sea lavender Broad leaved arrowhead 1 1 1 1 0 1 1 0 1 1 1 1 0 0 0 0 1 0 1 0 0 0 1 0 0 1 0 1 0 0 1 1 1 1 1 1 1 0 1 0 0 0 1 1 0 0 1 0 1 0 1 1 0 0 0 0 1 1 0 1 0 0 1 0 1 1 0 1 0 1 0 0 0 0 0 0 1 1 Bignoniaceae Campsis radicans (L.) Seeman Chenopodiaceae Salicornia virginica L. Liliaceae Asparagus officinalis Smilax bona-nox L. Asteracea Iva frutescens L. Baccharis halimiflora L. Myricaceae Myrica cerifera L. Cupressaceae Juniperus virginiana L. Pinaceae Pinus taeda L. Rosaceae Prunus serotina Ehrh. Aquifoliaceae Ilex opaca Aiton Trumpet vine Perennial saltwort Asparagus Greenbriar Marsh elder Groundsel tree Wax myrtle Red cedar Loblolly pine Black cherry American holly 0 0 0 0 1 1 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 1 1 1 1 0 1 1 0 1 1 0 1 1 0 1 0 1 1 1 0 1 0 1 1 0 0 1 0 1 1 0 0 0 0 0 1 0 0 0 0 0 1 TOTALS24 Species11712101115
The study transects were chosen to represent flat and steep slope gradients. In general, the flat transects had negligible slopes and more topographic variation than the transects with steep slopes (Figure 11).
The Flat Near transect had a very negligible slope (0.29 x 10-3, Table 3) with numerous hummocks along the entire length of the transect (Figure 11a). The general profile of the transect was flat with a large depression between 150-160 m. The topographic high at 60 m was represented by a spoil probably from the construction of ditches for mosquito control. Other than the ditch, the topographic high was around 100 m which was at the edge of a small pond (radius approximately 5 m). The topographic low along the transect was located at 160 m which is well into the forested area. This low spot represented the lowest elevation measured for all transects.
The Flat Far transect also had a negligible slope (0.81 x 10-3 Table 3), but a smoother surface and was less hummocky than the Flat Near transect (Figure 11b). The general profile of the transect was flat with a slight high elevation between 25-50 m followed by a depression around 60 m. The topographic high was at 85 m and was located on a large hummock in the transition zone. Unlike the Flat Near transect, there was only a slight depression immediately after the beginning of the forested area. Figure 11. Elevation contour relative to the creekbank benchmark (HAYD). Open circles represent sampling points; a linear regression line used to determine slope is indicated. (A) Flat Near transect. (B) Flat Far transect. (C) Steep Near transect. (D) Steep Far transect. Table 3. Slope summary table for study transects determined by linear regression. Regression analysis was computed from elevation measurements. Transect N Slope (x 10-3) r2 MSEa P Flat Near Flat Far 69 20 0.29 0.81 0.042 0.135 44.5 39.3 0.0932 0.1103 Steep Near Steep Far 33 12 13.56 7.28 0.892 0.668 5.0 8.3 0.0001 0.0010 aDesignates root mean square error.
In contrast to the flat transects, the Steep Near transect had an appreciable slope (13.6 x 10-3, Table 3) and contained fewer hummocks than either of the flat transects (Figure 11c). The change in elevation was continuous, with the topographic low at the start of the transect (0 m) and the high near the end of the transect (40 m) in the forested zone. There was a slight depression between 42-50 m which occurs well into forest cover.
The Steep Far transect also had an appreciable slope (7.3 x 10-3, Table 3), but the change in elevation was not as continuous as the Steep Near transect. The profile of the transect was relatively flat until 32 m where the change in elevation was the greatest (Figure 11d). Similar to the Steep Near transect, the Steep Far transect had fewer hummocks than the flat transects. 3.2.2 Variation in Elevation by Zone
The study transects were grouped by gradient of slope (i.e., flat, steep) and divided into zones (i.e., high marsh, transition, and forested) for comparison of mean elevations between zones (Table 4).
The elevations for the high marsh zone were similar for both flat and steep transects. For this zone, the flat transects had a slightly higher elevation than the steep transects, but the difference (2.5 cm) was not significantly different (Figure 12).
However, for the transition and forested zones, the difference between mean elevations for the flat and the steep transect was highly significant (Table 4). The Table 4. Summary of leveling transects separated by vegetation zone. Transect N Elevation (m) Standard Deviation Mean Maximum Minimum High marsh Flat Near Flat Far Steep Near Steep Far Flat Steep 52 7 7 5 59 12 1.034 1.044 0.963 1.080 1.037 1.012 1.310 1.110 1.040 1.110 1.310 1.110 0.880 0.940 0.910 1.050 0.880 0.910 0.074 0.057 0.053 0.025 0.071 0.073 Transition Flat Near Flat Far Steep Near Steep Far Flat Steep* 12 10 7 2 22 9 1.045 1.065 1.153 1.120 1.054 1.146 1.150 1.280 1.220 1.140 1.280 1.220 0.967 0.970 1.080 1.100 0.967 1.080 0.065 0.107 0.055 0.028 0.085 0.051 Forest Flat Near Flat Far Steep Near Steep Far Flat Steep* 5 3 19 3 8 22 1.015 1.160 1.453 1.360 1.070 1.440 1.140 1.220 1.594 1.440 1.220 1.594 0.809 1.130 1.280 1.280 0.809 1.280 0.142 0.052 0.072 0.080 0.133 0.078 * p < 0.001 Figure 12. Mean elevation for the high marsh zone, transition zone, and forest zone. Data were pooled data for both Flat Far and Near (Flat), and Steep Far and Near (Steep). steep transect transition zone was 9.0 cm higher than that for the flat transect. The forested zone for the steep transect was 37.0 cm higher than that for the flat transect. 3.2.3 Surface Macrotopography In general, the flat transects had more topographic variation than the steep transects which corroborated field notes and observations. Using linear regressions, the Flat Near and Flat Far transects had very high root mean square errors compared with the steep transects (Table 3). The Flat Near transect had the most topographic variation and highest root mean square error (44.5), whereas the Steep Near transect had the least topographic variation and lowest root mean square error (5.0). 3.3 Soil structure 3.3.1 Depth to mineral soil The depth to mineral soil along both flat and steep transects was highly variable. In general, the organic layer became thinner along the transect from marsh landward (Figure 13). The mean organic thickness for the Flat Near and Steep Near transect was similar when compared over the entire transect (Table 5). The overall mean for the Steep Near transect was probably overestimated due to the thick organic layer at 40 m in the forested zone (Figure 13b). The organic layer in the forested zone is comprised of leaf litter of terrestrial origin and is not strictly depositional peat material. When only the depositional peat material was considered (i.e., high marsh Figure 13. Topographic variation in marsh surface and underlying mineral soil from high marsh to terrestrial forest. (A) Flat Near transect. (B) Steep Near transect. Table 5. Mean depth (m) to organic layer for the Flat Near and Steep Near study transects, q standard error. Zone N Mean Depth to Organic Layer (m) Flat TransectSteep Transect High and Transition120.087 q 0.180.071 q 0.14 Total140.078 q 0.180.079 q 0.19 and transition zones), the Flat Near transect had a mean 1.6 cm thicker organic layer than the Steep Near transect (Table 5). The grand mean (8 q 2 cm) for both transects suggests that the 10 cm depth is mostly organic and the 20 cm depth is mostly mineral. The thickness of the organic layer smoothed the hummocky nature of the underlying mineral layer and created less variable surface elevations (Figure 13). The organic matter accumulation on the antecedent landform for the flat slope resulted in a flattening of surface elevations. If considered separately from the surface elevations, the slope of the underlying mineral layer for the Flat Near transect was 1.4 x 10-3 and the slope of the surface was slightly flatter at 0.29 x 10-3. The difference is small, but acts to flatten an already flat slope. However, organic matter accumulation for the Steep Near transect resulted in a steeper surface slope (13.6 x 10-3) than the underlying mineral layer (12.8 x 10-3). The thickness of the organic layer was weakly negatively correlated with elevation (r= -0.259, n= 50), but not significant (p = 0.07), suggesting that as elevation increased, the organic layer became thinner. 3.3.2 Bulk Density and Organic Carbon 3.3.2.1 Bulk Density and Organic Carbon by Slope
The mean bulk density of soil samples from the flat slopes were slightly less and the organic carbon content slightly more than for the steep slopes (Table 6). For the 10 cm depth, the mean bulk density of the flat slope was significantly less than for Table 6. Soil properties for study transects obtained from sediment cores. Values
reported are means of pooled values: Flat represents both Flat Far and Flat Near, and Steep represents both Steep Near and Steep Far, q standard error. Variable Zones Transect Mean High MarshTransitionForest Bulk Density (g cm-3) Flat 10 cm Steep 10 cm 0.29 q0.03 0.38 q0.03 0.35 q0.04 0.52 q0.03 0.42 q0.11 0.51 q0.02 0.35 q0.251 0.47 q0.20 Flat 20 cm Steep 20 cm 0.55 q0.03 0.59 q0.02 0.62 q0.03 0.52 q0.03 0.60 q0.02 0.52 q0.05 0.59 q0.20 0.55 q0.20 % Carbon Flat 10 cm Steep 10 cm 13.0 q2.7 2.1 q0.8 8.5 q1.8 2.1 q0.4 2.6 q0.1 1.9 q0.7 8.3 q2.33 4.4 q2.0 Flat 20 cm Steep 20 cm 2.1 q0.8 1.4 q0.6 1.0 q0.2 1.8 q0.7 2.0 q0.6 1.0 q0.2 1.7 q0.78 1.4 q0.62 % Water Flat 10 cm Steep 10 cm 55.4 q8.0 33.5 q4.0 37.6 q7.7 7.3 q1.5 17.4 q1.9 3.6 q1.4 36.8 q4.43 14.8 q4.0 Flat 20 cm Steep 20 cm 17.0 q4.0 5.8 q1.0 10.3 q2.1 9.1 q4.5 9.8 q3.0 3.3 q1.1 12.4 q2.03 6.1 q1.7 Chlorinity (mM) Flat 10 cm Steep 10 cm 151.4 q22.0 188.6 q24.2 107.6 q27.0 56.8 q14.2 26.0 q12.8 10.4 q 4.0 95.0 q11.23 85.2 q13.6 Flat 20 cm Steep 20 cm 78.6 q 9.2 98.2 q10.2 43.2 q8.2 72.0 q6.2 23.6 q5.6 6.6 q1.8 48.4 q7.43 59.0 q9.8 1 p < 0.05 2 p < 0.01 3 p < 0.001 the steep slope. The organic carbon content was also significantly higher for the flat slope. However, the 20 cm samples were essentially the same for both bulk density and organic carbon content for both steep and flat slopes (Table 6). 3.3.2.2 Bulk Density and Organic Carbon by Zone Soil samples at 10 cm in all zones of the flat transects had lower bulk densities and higher carbon content than for the steep transects (Figure 14a, c). None of the differences in bulk density were significant, although the transition zone of the flat transect was marginally significantly lower (p= 0.06) than that for the steep transect. The organic carbon content of the high marsh and transition zone for the flat transects were significantly higher than that for the steep transects (Table 6). The forested zone of the flat transects was only slightly higher in organic content than for the steep transects.
The mean bulk densities at 20 cm for the flat transect were higher for the transition and forested zones, but lower for the high marsh (Figure 14b). Again, none of the differences were found to be significant, although the transition zone of the flat transect was marginally significantly higher (p= 0.06) than for the steep transect. The organic content at 20 cm was low for both the steep and flat transects (Figure 14d) with no significant differences. Figure 14. Mean soil structure data from core samples at 10 and 20 cm depths for both Flat Far and Near (Flat) and Steep Far and Near (Steep), q standard error. (A) Bulk density at 10 cm. (B) Bulk density at 20 cm. (C) Total carbon at 10 cm. (D) Total carbon at 20 cm. (E) Chlorinity at 10 cm. (F) Chlorinity at 20 cm. 3.3.3 Chlorinity 3.3.3.1 Chlorinity by Slope The sediment chlorinity ranged from a high of 310 mM for the steep high marsh at 10 cm depth to 0 mM for the steep forest at 60 cm depth. In general, the chlorinty was highest for the high marsh and decreased toward the forest for steep and flat slopes. For both steep and flat slopes, chlorinity also decreased with depth. The chlorinity of soil samples for the 10 cm depth was higher than that for the 20 cm depth for both steep and flat transects. 3.3.3.2 Chlorinity by Zone
The chlorinity of the samples from the high marsh, transition, and forested zones at 10 cm was generally higher for the flat transect than for the steep transect, but the differences were not significant (Figure 14e).
The soil samples at the 20 cm depth of the flat transect were highest for the high marsh, and essentially the same for transition and forested zones. The chlorinity was highest for the high marsh, intermediate for the transition zone, and lowest for the forested zone of both flat and steep transects (Figure 14f). The chlorinity of the high marsh and forested zones of the flat transects was higher than that for the steep transects; in the transition zone, it was lower in the flat transect. Despite the large apparent differences in mean chlorinity, none were not significantly different between zones. 3.4 Hydroperiod
Hydroperiod evaluations were conducted for the Steep Near and Flat Near transects. The water table levels showed differences between months, seasons, and years of record (Figure 15a,b). The spring and summer of 1993 had the lowest water tables for all zones during the study period. Based on the long-term data set (USDA 1989), lower than seasonal rainfall occurred during the late spring and early summer which resulted in lower water tables for both transects. The lowest water table levels of the study period were recorded in July and August 1993. Higher than usual rainfall for August may have prevented the summertime drawdown for 1992. The water table levels were generally highest in January and February.
Differences between seasons are reinforced by summary statistics of hydroperiod from continuous water level recorders placed in the transition zone of both the near steep and flat transects (Table 7). The percent of time flooded and mean water level were considerably lower during the warm period than the cool period for the flat slope. The warm period showed variation between 1992 and 1993, with higher water levels and percent of time flooded in 1992. This may be the result of the higher rainfall in the summer of 1992. For the warm period, the transition zone for the flat transect experienced more flooding and a higher mean water level than the steep transition zone. Because of the extremely dry summer, no flooding occurred for the period of record for the steep transition water level recorder. However, data collection for the steep transition zone water level recorder Figure 15. Mean monthly water table levels for the high marsh, transition, and forest zones determined by shallow (1.5 m) groundwater wells. Sample points represent the mean of three wells for each zone, q standard error. The water table position was measured twice in August 1992 to October 1992, and from March to August 1993. For these dates, the mean represents six wells. (A) Flat Near transect. (B) Steep Near. (C) Mean water table for the period October 1992 to August 1993 for Flat Near (Flat), and Steep Near (Steep) transects. Table 7. Characteristics of hydroperiod at the transition zone water level recorders for the Steep Near and Flat Near transects. Characteristics Flat1 (1991-1993) Steep2 (1992-1993) WarmaCoolaWarm % of time flooded Mean depth (cm)b (S.D.) Total wks in period # wks precipitation was received # floods from storm events 13.4 - 18.83 (5.43) 27 21 2 31.0 0.88 (2.13) 18 17 3 0 - 81.84 (3.73) 11 7 0 1 Total period from 11 July 92 to 31 Aug 93. 2 Total period from 18 June 93 to 31 Aug 93. a Warm season = April through September; Cool season = October through March. b Calculated from hourly value. commenced 18 June 93, which was already into the warm period. When the identical period was examined, the flat transition zone still experienced higher mean water levels than the steep (e.g., -45.3 cm for the flat, and -81.8 cm for the steep), but the percent of time flooded decreased to 1% from 10.7%. Comparison among slopes for the cool period was not possible because data were collected only during the summer 1993 for the steep transition zone. Flooding from the estuary occurred more often during the cool period and mainly during extratropical storm events. The number of storm events during the warm period was higher than would be expected and due to events which occurred during the beginning or end of the warm period (e.g., a storm on September 30). Storm events are characterized by very large, steep rises in the water level which are greater than expected due to rainfall alone (Figure 16). Storm surges were often influenced by astronomical tides once water levels are above the surface and result in a quick succession of peaks in water level about 12 h apart. 3.4.1 Water Table Fluctuations Overall, the high marsh zone experienced water level fluctuations at or near the surface for most of the study period. The transition zone had a somewhat lower water table with slightly more seasonal variation. The forested zone experienced the largest seasonal and yearly variation. The overall trends in seasonal water table fluctuations were generally the same Figure 16. Hydrograph during September 1992 showing the effects of rainfall and a storm surge due to Tropical Storm Diane for the Flat Near transition. Position of the marsh surface is indicated by the 0 cm water level. Asterisk indicates beginning of storm. for both steep and flat transects for the period of record. However, the drawdown during the summer 1993 was more pronounced for the steep transect, with lowest water levels recorded in August (Figure 15b). The drop in water levels was more severe for the steep slope, especially between May and June. Rainfall from convection storms in August resulted in a rise in water levels for all zones of the flat slope, but the steep slope only showed slight increases in water levels. Although the general patterns of seasonal water table fluctuations were similar for both steep and flat slopes, the extent of these characteristics was different among slopes. For both steep and flat slopes, the mean water table was lowest in the forest (Figure 15c). However the mean water table was highest for the steep transect in the high marsh, but highest in the transition zone for the flat transect. The monthly water table levels for the flat slope were actually higher in the transition zone than the high marsh for the most of the spring and summer of 1993, even during periods of drawdown.
The mean water table level for the high marsh showed little difference between flat and steep slopes and was less than 10 cm from the surface (Figure 15c). However, highly significant differences were observed between the transition and forested zones of flat and steep slopes. The mean water table levels of both transition and forested zones were closer to the surface for the flat slope than for the steep slope. The mean water table in the transition zone for the flat transect ranged from 5 cm above the surface to 50 below the surface. For the steep transition zone, the mean water table ranged from at the surface to 75 cm below the surface. For the forested zone, the mean water table of flat transect ranged from 10 to 80 cm below the surface whereas the range of the steep transect forested zone ranged from 20 to 100 cm below the surface. 3.4.2 Precipitation, PET, and Water Table Drawdown
From 1949-1981, total precipitation averaged 103 cm y-1, virtually all as rain. Monthly means were relatively constant throughout the year and were generally highest in July through October and lowest from April through June (Figure 17). >From 1990 to May 1993, the mean monthly precipitation was highest in August (21.8 cm mo-1) and lowest in June (2.93 cm mo-1). For the three year period, the total precipitation was highest in 1992 (113.4 cm y-1) and lowest in 1990 (94.7 cm y-1).
Based on data for 1949-1981, mean monthly potential evapotranspiration (PET) ranged from 1 cm in January to 14 cm in July and was less than mean monthly precipitation in all months except for May through September (Figure 17). Thus, these months are predicted to experience drawdown due to precipitation deficits. However, over the period January 1990 to August 1993, only June, July and October had precipitation deficits for all three years. Due to the unusually high rainfall during the study period for August, rainfall exceeded the predicted PET between 1990 and 1992, but not in 1993.
Despite PET being less than precipitation in November 1992, the water table Figure 17. Monthly precipitation totals from 1990 to 1993 (Brownsville Marsh) and 1949 to 1981 (Cheriton, VA). The dotted line connects monthly means for 1949 to 1981. The small open and solid circles connect mean Thornthwaite evapotranspiration for the same period. Points below the line connected with small circles represent months with precipitation deficits. decreased from October to November for the steep slope for all zones and increased during the same period for the flat slope. The water table did increase substantially in December for the steep slope. Drawdown during the summer of 1993 began in May for both transects. A typical drawdown for flat and steep slopes shows that the recharge period and extent for the steep slope is more pronounced than for the flat slope (Figure 18). 3.4.3 Groundwater Salinity Patterns Salinity patterns show variation seasonally, yearly, and among vegetation zones and gradients of slope (Figure 19). Average salinities ranged from a high of 30 ppt in the high marsh to a low of 2 ppt in the forested zone.
In general, seasonal and yearly variations in salinity can be attributed to specific environmental conditions (e.g., flooding by the estuary, rainfall). For example, the Halloween Storm (31 October 1991), and a Northeaster (4 January 92) flooded the marsh and forested areas, and resulted in increased salinity which remained elevated until March 1992 (Figure 19a). Unusually high rainfall during the spring and summer of 1992 lowered salinities throughout the summer and fall. Another storm on 25 September 92 (Tropical Storm Diane) flooded the marsh and forest with saline water (salinity of the flood water at the forest zone was 10-20 ppt, measured during the event) and resulted in increased salinity of the transition and forest zone and decreased salinities for the high marsh. Salinities increased and Figure 18. Hydrograph during June and July 1993 comparing the diurnal pulse of water table drawdown due to evapotranspiration for flat and steep slopes. Depth is shown relative to the marsh surface. (A) Hydrograph for the Flat Near transition water level recorder. (B) Hydrograph for the Steep Near transition water level recorder. Figure 19. Mean monthly salinity for the high marsh, transition, and forest zones determined by shallow (1.5 m) groundwater wells. Sample points represent the mean of three wells for each zone, q standard error. The salinity was measured twice in August 1992 to October 1992 (for the Flat Near transect), and from March to August 1993. For these dates, the mean represents six wells. (A) Flat Near transect. (B) Steep Near. remained high following another flood by a strong Northeaster in December 1992.Salinity was found to be highest in the high marsh, declining both toward the forest and creekbank (Figure 20). Although the general patterns of salinity with distance appear to be similar for both flat and steep slopes, the nature of the change is different. The flat transect showed a very gradual decrease in salinity with increasing distance from the marsh, with much variation (Figure 20a). In contrast, the steep transect had a very abrupt change in salinity with increasing distance, and comparatively little variation (Figure 20b).
The area with the largest drop in salinity with distance corresponds with the transition zone for both flat and steep transects. The largest drop in salinity for the flat slope occurred between 123 to 150 m which is the location of the transition zone. Similarly, for the steep slope, the greatest drop in salinity with distance occurred between 20 to 25 m for the steep slope which is also the location of the transition zone. 3.5 Pore Water Chemistry
The lack of precipitation during the spring and summer of 1993 resulted in periods of severe drawdown. Consequently, pore water was unable to be collected after June and only data collected from January to June 1993 were used for analysis. 3.5.1 Pore Water Salinity
The salinity of the pore water ranged from a high of 34 ppt for the high marsh Figure 20. Box plot of the groundwater salinities along the Flat Near and Steep Near transects for the period beginning 4 November 1991 for the Flat Near and 5 August 1992 for the Steep Near transect. Dotted bars represent the mean, solid bars represent the median, and the box and error bars encompass the range. (A) Flat Near transect. (B) Steep Near transect. zone and a low of 1 ppt for the forested zone. Pore water salinities were stable for all zones of both steep and flat slopes, and for 10 and 20 cm during the period of study (Figure 21). As the summer began, salinity increased for the high marsh and transition zones for the flat slope, but only for the high marsh zone of the steep slope. Variation was the greatest for the forest zone of the flat slope and the high marsh zone for the steep slope. For the lysimeter data collection period mean salinity decreased with distance landward for both steep and flat slopes and for 10 and 20 cm (Figure 22a, b). Mean salinity increased with depth for all zones and slopes. The 20 cm depth was about 3 ppt more saline than the 10 cm depth for the high marsh zone.
The flat transect experienced higher mean pore water salinities at 10 and 20 cm than the steep transect for all vegetation zones (Figure 22a, b). The differences between the transects were highly significant for the transition and forest zones. The mean salinity for these zones was more than double the mean salinity for the same zones of the steep transect (Table 8).
A comparison between the salinities obtained from lysimeters and shallow groundwater wells shows that the wells measure slightly higher salinities (Figure 23). The wells average about 3 ppt higher than the lysimeters for the high marsh and transition zones, and about 1 ppt higher than the lysimeters for the forest. This discrepancy may be due to increased salinity with depth as was found between 10 and 20 cm lysimeters. Figure 21. Pore water salinity collected for the Flat Near and Steep Near transects from lysimeters at 10 and 20 cm depths for the period 14 Jan to 29 July 1993. Points represent the mean of three lysimeters for the Flat transect and six lysimeters for the Steep transect, q standard error. (A) Flat Near transect at 10 cm. (B) Flat Near transect at 20 cm. (C) Steep Near transect at 10 cm. (D) Steep Near transect at 20 cm. Figure 22. Mean pore water data collected for the Flat Near and Steep Near transects from lysimeters at 10 and 20 cm depths for the period 14 Jan to 29 July 1993. Bars represent the mean of three lysimeters for the Flat transect and six lysimeters for the Steep transect, q standard error. (A) Mean pore water salinity at 10 cm. (B) Mean pore water salinity at 20 cm. (C) Mean pore water redox potential at 10 cm. (D) Mean pore water redox potential at 20 cm. (E) Mean pore water sulfide concentration at 10 cm. (F) Mean pore water sulfide concentration at 20 cm. (G) Mean pore water pH at 10 cm. (H) Mean pore water pH at 20 cm. Table 8. Mean pore water data collected from low-tension lysimeters for the period 14 January to 20 May 1993, q standard error. Data were used as a basis for Figure 21. Variable Depth1 (cm)Zone2 Flat3 Steep4 Salinity (ppt) 10 H T* F* 16.7 q 0.6 14.4 q 0.4 7.5 q 0.7 16.3 q 0.8 7.0 q 0.3 3.0 q 0.2 20 H T* F* 19.8 q 0.5 14.1 q 0.3 8.4 q 0.6 19.5 q 0.9 7.1 q 0.3 4.4 q 0.2 Redox potential (mV) 10 H T F 195 q 11 211 q 6 219 q 7 182 q 12 195 q 10 199 q 9 20 H* T* F* 206 q 9 210 q 6 218 q 7 157 q 11 166 q 11 170 q 8 Hydrogen Sulfide (fM) 10 H* T F 105.6 q 44.9 0.2 q 0.1 3.8 q 2.3 34.1 q 9.6 0.4 q 0.1 4.0 q 0.5 20 H T F 2.4 q 1.0 0.4 q 0.2 0.1 q 0.1 1.6 q 0.4 0.3 q 0.2 1.5 q 0.4 pH 10 H T* F* 6.5 q 0.2 5.8 q 0.2 3.5 q 0.1 6.4 q 0.1 4.4 q 0.1 4.2 q 0.1 20 H T* F 5.8 q 0.3 5.2 q 0.2 3.4 q 0.1 5.9 q 0.1 4.0 q 0.8 3.6 q 0.6 * p< 0.01 1 Depths were 10 and 20 cm for High marsh and Transition zones, 20 and 60 cm for Forest zone. 2H= high marsh zone; T= transition zone; F= forest zone. 3n=24 for all zones; 4n=48 for H and T; n=64 for F 3.5.2 Redox Potential, Hydrogen Sulfide, and pH Mean redox potential increased over time during the study period for all vegetation zones and depths (Figure 24). Values were lowest in January, then rose and remained high in February through June. For both slopes, redox potentials ranged from a low of -50 mV for the high marsh to a high of +260 mV for the forest. Very little variation was seen among zones and depths within type of slope. It should be noted that the low readings on 14 January 93 may have been due to instrument malfunction, since it was the first measurement. However, sediment redox potentials measured for two months prior to the lysimeter sampling period, which are not reported here, were similarly low (e.g., approximately 100 mV). In general, the redox potential was higher for the flat transect than the steep transect for all zones and depths (Figure 22c, d). The mean redox potential for the flat transect was significantly higher than the steep transect at 20 cm for all zones. The flat transect was roughly 50 mV more positive than the steep transect for all zones at 20 cm. Although higher for the flat transect, the differences between flat and steep transects at 10 cm were not significant. The range of redox potentials during the study period was often low enough to create conditions favorable to hydrogen sulfide production. However, with the exception of the high marsh lysimeter at 10 cm, very little sulfide was measured during the study period (Figure 22e, f). Sulfide ranged from a high of 1200 fM for the high marsh to below the level of detection. Figure 23. Comparison between salinities measured using shallow groundwater and pore water salinities collected using lysimeter at 10 cm for the Flat Near transect for the period January to July 1993. Figure 24. Pore water redox potential collected for the Flat Near and Steep Near transects from lysimeters at 10 and 20 cm depths for the period 14 Jan to 29 July 1993. Points represent the mean of three lysimeters for the Flat transect and six lysimeters for the Steep transect, q standard error. (A) Flat Near transect at 10 cm. (B) Flat Near transect at 20 cm. (C) Steep Near transect at 10 cm. (D) Steep Near transect at 20 cm. Mean pore water sulfide concentration was significantly higher for the flat transect at 10 cm in the high marsh zone. The high marsh zone had a mean sulfide concentration three times that of the same zone and depth of the steep transect. The mean concentrations for other depths and zones were highly variable making comparisons difficult. An examination of two lysimeters (0 m at 10 cm depth and 150 m at 20 cm depth) of the flat slope shows that a time lag exists between peaks in sulfide production with distance landward (Figure 25). The 0 m lysimeter showed peak sulfide production between 20 May and 17 June, whereas production peaked around 17 June for the 150 m lysimeter. Although the differences in concentrations are about two orders of magnitude, the sulfide production shows a time lag of about two weeks. The decline in sulfide concentrations coincided with the beginning of the summertime drawdown.
Sulfide detection by the lysimeters was very patchy in nature; at times, one lysimeter would have high concentrations of hydrogen sulfide and the replicate lysimeters would not. This resulted in very large standard errors and made it difficult, if not impossible, to compare means.
Mean pH ranged from a high of 8.1 for the high marsh zone to a low of 2.5 for the forest zone. The mean pH decreased from the marsh landward at both depths (Figure 22g,h). The flat transect had significantly higher pH for the transition zone at 10 and 20 cm than the steep transect (Table 8). The forest zone at 10 cm had a Figure 25. Sulfide concentration collected from the 0 m lysimeter at the 10 cm depth and from the 150 m lysimeter at the 20 cm depth for the Flat Near transect. Bars represent sampling dates from 8 April to 13 August 1993. significantly lower pH for the flat transect, but was not significant at 20 cm. The high marsh zone was similar for both transects and depths. 3.6 Factor Analysis 3.6.1 Analysis of Entire Data Set Six of the 11 physical and chemical variables had correlations greater than 0.48 (Table 9) suggesting factor analysis would be appropriate. The factor analysis constructed 11 factors of which three had eigenvalues greater than 1.0, a criterion used in deciding the number of factors to extract (Dunteman 1984). The fourth factor had an eigenvalue of 0.81 indicating the three factor solution was satisfactory. The three factors accounted for 69.4% of the variation in the data (Table 10). After rotation, all factors had at least two variables with a loading score whose absolute value was 0.5 or greater.
The first factor was interpreted as a brackish water factor. Salinity and pH at 10, and 20 cm, and water table had high positive loadings on this factor, while distance and elevation had large negative loadings (Table 10). From the analysis of the univariate data, it follows that as distance and elevation increased (from high marsh toward forest), salinity, pH, and water table levels decreased (Figure 26a).
Analysis of variance for factor one showed significant zone differences, but not significant date differences (Table 11). The transition and forest zones of the flat transect had higher mean factor scores for the study period (Table 12), reflecting Table 9. Correlations of physical and chemical variables1. Correlations with absolute values greater than 0.48 are significant (p < 0.05). P10 P20 E10 E20 W S10 S20 H10 H20 D E P10 1.00 P20 0.72 1.00 E10 -0.27 -0.21 1.00 E20 -0.25 -0.18 0.86 1.00 W 0.41 0.38 -0.23 -0.13 1.00 S10 0.52 0.46 -0.10 -0.11 0.36 1.00 S20 0.62 0.48 -0.10 -0.11 0.35 0.86 1.00 H10 0.22 0.28 -0.06 -0.01 0.10 0.19 0.23 1.00 H20 0.25 0.33 0.01 -0.06 0.06 0.06 0.20 0.25 1.00 D -0.75 -0.67 0.18 0.23 -0.42 -0.58 -0.66 -0.30 -0.29 1.00 E -0.47 -0.50 0.12 0.04 -0.41 -0.65 -0.63 -0.19 -0.10 0.53 1.00 1Variable abbreviations are: P = pH; E = Eh (Redox potential); W = water table; S = pore water salinity; H = hydrogen sulfide; D = distance from creekbank to lysimeter; E = elevation at lysimeter; 10 = 10 cm depth; 20 = 20 cm depth. Table 10. Correlations of physical and chemical variables1 with factors. Variable Factor 1 Factor 2 Factor 3 S10 S20 P10 P20 W D E E10 E20 H20 H10 0.88 0.87 0.71 0.62 0.58 - 0.74 - 0.81 - 0.11 - 0.07 0.01 0.14 0.02 0.01 - 0.24 - 0.18 - 0.19 0.16 - 0.03 0.95 0.95 0.00 0.03 0.00 0.14 0.38 0.50 - 0.01 - 0.42 - 0.04 0.01 - 0.03 0.82 0.65 % variance explained by each factor 42.10 16.26 11.06 1Variable abbreviations are: P = pH; E = Eh (Redox potential); W = water table; S = pore water salinity; H = hydrogen sulfide; D = distance from creekbank to lysimeter; E = elevation at lysimeter; 10 = 10 cm depth; 20 = 20 cm depth. Figure 26. Mean factor scores for the high marsh, transition, and forest vegetation zones for the period 14 January to 29 June 1993, q standard error. (A) Brackish water factor. (B) Redox factor. (C) Sulfide factor. Table 11. Two-way analysis of variance comparing factor scores for lysimeter data among slopes and sampling dates for the entire data set. Factor Variation df SS F P Factor 1Model Date Slope x Date Error 5 13 38 90 96.6 4.2 3.2 31.1 55.97 0.93 0.25 0.0001 0.5273 1.0000 Factor 2Model Date Slope x Date Error 5 13 38 90 2.9 76.9 23.9 5.4 9.50 98.15 10.45 0.0001 0.0001 0.0001 Factor 3Model Date Slope x Date Error 5 13 38 90 30.8 9.2 13.6 92.8 5.98 0.69 0.35 0.0001 0.7729 0.9998 Table 12. Mean factor scores for the Flat Near and Steep Near transects grouped by vegetation zone, q standard error. Factor Zone1 Flat Transect Steep Transect 1 H 0.49 q0.07 0.72 q0.10 T* 0.07 q0.05-1.01 q0.10 F*-1.07 q0.12-2.08 q0.07 2 H- 0.04 q0.16- 0.15 q0.14 T* 0.07 q0.22-0.17 q0.16 F* 0.57 q0.28-0.01 q0.35 3 H 0.61 q0.31 0.24 q0.13 T-0.41 q0.09-0.29 q0.06 F-0.77 q0.08 0.12 q0.06 * p< 0.05 (Tukey's) 1H = high marsh; T = transition zone; and F = forest area higher salinities, pH, and water tables. The high marsh was similar for both transects. The differences in factor one over time were not significant and probably reflect the stable nature of salinity and pH during the study period (see, for example,Figure 21). The second factor was interpreted as the redox factor because the redox at 10 and 20 cm was the dominant loading variable (Table 10). No other variable had high loadings on this factor. The redox factor generally increased with distance landward for both slopes (Figure 26b), although variability was very high. The redox factor showed significant zone and date differences over the period of study (Table 11). The transition and forest zones of the flat transect had significantly higher mean factor scores (Table 12) which is in agreement with the univariate data for redox potential. As in the univariate analysis of redox potentials, the redox factor increased from January to February and remained high for the rest of the study period.
The third factor was interpreted as the sulfide factor since sulfide at 10 and 20 cm had the only dominant loading value for the factor (Table 10). The sulfide factor scores decreased with distance (Figure 26c), although no significant differences between zones or dates existed (Table 11). The variability of this factor was very high and reflects the high variability of the sulfide variable (Table 12). 3.6.2 Analysis of Transition Data Set
A factor analysis was conducted using the data set for the transition zone to determine if this zone contained any unique properties. Similar to the analysis for the entire data set, the factor analysis constructed 11 factors of which three had eigenvalues above 1.0. The factor loadings were very similar to the loading for the entire data set with differences mainly between the loading scores of distance and water table (Table 13). Distance remained as a loading variable on the first factor, but changed its value from a positive to negative loading. The water table variable remained a positive, but changed position to load on the third factor. The factors were interpreted as being identical to the factors described for the entire data set and are brackish water, redox potential, and sulfide.
For factor one, the brackish water factor, it follows that as distance increased landward, the salinity of the pore water also increased. The 2 y well data for salinity support this trend for the transition zone (see, for example, Figure 20a). The finding that as distance increases for the transition, salinity also increases is the reverse of what was found using the entire data set. Although pH contributed to the loading on this factor, the values were much smaller, suggesting that the influence due to pH may play a smaller role.
For factor three, the sulfide factor, water table was found to be positively associated with sulfide production (Table 13). It follows that as the water table increases, sulfide concentrations would also increase, especially during the period of study (i.e., spring and summer). Table 13. Correlations of physical and chemical variables1 with factors using data from transition zone only. Variable Factor 1 Factor 2 Factor 3 S10 S20 D P20 P10 E E20 E10 H10 W H20 0.90 0.79 0.78 0.73 0.68 - 0.86 0.10 - 0.07 - 0.13 0.41 0.23 0.05 0.20 0.00 - 0.23 - 0.21 - 0.12 0.95 0.94 0.05 0.05 - 0.40 0.26 0.27 - 0.23 0.12 - 0.12 - 0.35 - 0.02 0.06 0.78 0.57 0.48 % variance explained by each factor 39.00 19.15 11.59 1Variable abbreviations are: P = pH; E = Eh (Redox potential); W = water table; S = pore water salinity; H = hydrogen sulfide; D = distance from creekbank to lysimeter; E = elevation at lysimeter; 10 = 10 cm depth; 20 = 20 cm depth. Table 14. Two-way analysis of variance comparing factor scores for lysimeter data among slopes and sampling dates for the transition data set only. Factor Variation df SS F P Factor 1Model Date Slope x Date Error 1 11 6 23 28.8 1.1 0.3 1.6 414.96 1.39 0.61 0.0001 0.2445 0.7170 Factor 2Model Date Slope x Date Error 1 11 6 23 0.2 28.9 2.2 1.7 2.15 35.59 4.99 0.1563 0.0001 0.0021 Factor 3Model Date Slope x Date Error 1 11 6 23 2.1 13.5 6.4 20.0 2.45 1.41 1.23 0.1313 0.2352 0.3279