Maps, Images, and Modeling in the Assessment of Wetlands
Wetlands are sometimes delineated by use of aerial photography, satellite imaging, maps, or modeling, as an alternative to collecting field data. Offsite methods are recommended by the 1989 interagency manual for use in areas where "information on hydrology, hydric soils, and hydrophytic vegetation is known, or an inspection is not possible due to time constraints or other reasons." They are also the primary methods used by Natural Resources Conservation Service to delineate wetlands under the terms of the Food Security Act and by the Fish and Wildlife Service to map wetlands for the National Wetlands Inventory (NWI). Delineation of wetlands by offsite methods is subject to errors that do not affect delineation by use of data collected directly from the field. Offsite methods should be used only when their inherent limitations are recognized, as described here.
AERIAL PHOTOGRAPHY AND SATELLITE IMAGING
Aerial photography has been used to map wetlands for at least three decades (Olson, 1964; Anderson and Wobber, 1973; Seher and Tueller, 1973; Cowardin and Myers, 1974; Hardy and Johnston, 1975; Gammon and Carter, 1979; Johnston, 1984). Because the photographs provide a synoptic view of wetlands and their surrounding terrain, they facilitate rapid boundary determination. Interpretation of photographs is difficult, however, in areas where changes, vegetation, soils, or hydrology are indistinct or variable through time. Such areas are generally difficult to delineate by field methods as well.
Satellite remote sensing also holds promise for wetland delineation (FGDC, 1992), but it is not used routinely. When methods for the NWI were being evaluated in the late 1970s, it was determined that the images provided by satellites did not have sufficient spatial or spectral resolution to map wetlands reliably, and their use was rejected in favor of aerial photographs (Tiner, 1990). Recent research, however, suggests that satellite images could be superior for delineating wetland hydrology in some cases, particularly for agricultural areas. A review of the use of satellite data for mapping and monitoring wetlands, based on the experience of several private and public agencies, has been published by the Federal Geographic Data Committee (FGDC, 1992).
Detection of Standing Water
Aerial photographs are recommended by the 1987, 1989, and FSA manuals as suitable indicators of hydrology because they can provide direct evidence of inundation if they are taken when there is standing water on the soil surface and there is no obscuring vegetation. Interpretation of this type of photo requires little training; one merely looks for the dark areas associated with surface water. If photographs are available for several dates, they can provide a history of inundation. As with any hydrologic interpretation, soil moisture and antecedent precipitation must be considered, however.
Satellite images also can be used to detect standing water. The Landsat Thematic Mapper (TM) satellite detects energy returned from surface features in several wavelength bands. Band 5, which detects infrared wavelengths between 1.55 and 1.75 mm, is particularly useful for detecting soil moisture and standing water (Lillesand and Kiefer, 1979). The National Aeronautics and Space Administration (NASA) Space Remote Sensing Center in Huntsville, Alabama, used this band to map natural and farmed wetlands in the Yazoo River Basin in Mississippi and Arkansas as part of a pilot project for NRCS (R. Pearson, presentation to NRC Wetlands Characterization Committee, Nov. 23, 1993). A hydrologic model was used to determine the stage height corresponding to a 2-year flood, and TM images coinciding with those conditions were used to map fields that were inundated for 15 consecutive days, the FSA threshold for wetlands. Unlike aerial photography, which is done infrequently because of its high cost, TM images are acquired every 16 days as the satellite passes over the Earth. This allows NASA analysts to find a scene that coincides with flood conditions. This methodology also was tested for delineation of wetlands in the prairie pothole region of North Dakota (FGDC, 1992). Maps produced from TM images were adopted by NRCS for the Yazoo River Basin in Mississippi, but not for North Dakota, because their accuracy was considered too poor in the drier western half of the state (North Dakota State Conservationist, presentation at the National Interagency Memorandum of Agreement Meeting, St. Paul, Minnesota, May 18, 1994).
In the most sophisticated applications, the photo interpreter uses not only the tone or color in the photograph, but also landscape position, land slope, and the appearance of vegetation, to distinguish wetlands from uplands (Hardy and Johnston, 1975). Ancillary data, from soil surveys or topographic maps, also can be useful. The accuracy of interpretation is greatest where wetland boundaries coincide with changes in the density and structure of the dominant vegetation; wetlands that are easily discernible on the ground are generally also easiest to see in aerial photographs. Because an experienced interpreter uses a combination of clues to make a determination, keys and descriptions of methods are rare. Given the importance of aerial photography in the preparation of NWI and FSA surveys, however, there is a need for more explicit documentation on this delineation method.
Stereoscopic viewing of photographs greatly increases the accuracy of interpretation (Lillesand and Kiefer, 1979; Soil Survey Staff, 1993) because it shows topographic breaks, which help show wetland boundaries, and it reveals changes in vegetation height and shape that can indicate changes in soil moisture. Stereoscopic viewing also helps the interpreter locate wetlands because wetlands are more likely to occur in some topographic positions than in others. Stereoscopic interpretation is a standard procedure in preparation of the NWI (1990), but not in FSA inventories conducted by NRCS.
A boundary delineated on an aerial photograph corresponds to a zone on the ground with a width that equals the width of the line divided by the scale (the representative fraction) of the photograph. For example, a 0.0197-inch (0.05 cm) pen line on a 1:24,000 photo would represent 39 ft (11.85 m) on the ground. Although a boundary of this width could depict ecological reality, it would be insufficiently resolved for jurisdictional purposes.
People who work on the ground often have difficulty determining their location from a map or aerial photograph. The unfamiliarity of most nonspecialists with maps is a major disadvantage of delineating wetlands on aerial photographs: Even if an area is perfectly depicted, it can have little meaning if a bulldozer operator or tractor driver cannot relate the mapped boundary to a field location. The increasing accuracy and decreasing cost of global positioning systems are improving the georeferencing of ground locations, but this technology is not yet widely available at the fineness of resolution needed for wetland delineation.
WETLAND DELINEATION UNDER THE FOOD SECURITY ACT
In contrast to the field-intensive methods used to identify wetland boundaries for Clean Water Act Section 404 permits, wetland delineations required by FSA are done primarily by offsite methods. The 1990 FSA amendments direct NRCS to conduct a field wetland determination if possible whenever requested to
do so by an owner or operator. In practice, field determinations are done only when an owner or operator questions the validity of the offsite determination.
Determinations Before 1994
After enactment of FSA, each state developed mapping conventions for indirect wetland determinations on agricultural lands, with technical guidance from the four NRCS National Technical Centers (Midwest, 1988; Northeast, 1989; South, 1989; West, 1988). Mapping conventions vary slightly by state because of regional differences in wetland characteristics and in the availability of data, but the general methodology is the same except where satellite remote sensing has been used. In all states, five land classes are differentiated: wetland (W), farmed wetland (FW), converted wetland (CW), prior converted cropland (PC), and artificial wetland (AW) (Chapter 6).
The primary source of data for FSA determinations is 35 mm aerial color slides that have been taken each year since the early 1980s by the Agricultural Stabilization and Conservation Service (ASCS). The slides, which are acquired primarily for confirmation of cropping claims under ASCS subsidy programs, are taken at a time of the year optimal for distinguishing crops, usually late June or July in the Midwest and West. Their use by NRCS for wetland determinations began later, and has not influenced the timing of their acquisition.
The mapping conventions from the four NRCS regions specify the following as photographic indicators of wetland: hydrophytic vegetation, water or drowned crop (mud), crop stressed by water (yellow leaves), lush crop in dry years, and differences in crop color caused by delay of planting. Steps in the delineation procedure described by the Midwest Center (1988), which are similar to those from the other three regions, are as follows:
Review NWI maps. All areas mapped by NWI are considered wetlands unless review of the ASCS slides indicates the contrary. Some wetlands shown by NWI might have been drained since the NWI photos were taken, or there can be errors on NWI maps.
Review the soil survey for evidence of wetlands (such as hydric soils).
Review ASCS color slides for the previous 5-7 years in conjunction with precipitation data collected 2 to 3 months before the date of the slide to determine prevailing climatological conditions at the time of photography. The slides are projected sequentially, and evidence of wetness over the 5 to 7 year interval is used to make decisions about the appropriate classification of the area.
The boundaries of areas determined to be wetlands are transferred to a 1:7,920 (8 in. = 1 mile) (20.32 cm = 1.61 km.) ASCS enlarged black-and-white aerial photograph that is used as a base map.
The major disadvantage of the FSA approach is that, in many parts of the
country, the slides are taken after surface waters have receded from wetlands. A further disadvantage is that air photo interpretation is monoscopic, rather than stereoscopic. The indicators of wetland are the color differences on the slides. There is no supplementary information about topography that stereoscopic interpretation or field inspection would provide. Thus, the appearance of stressed crops might be similar for an eroded hilltop and for a wetland depression.
The NRCS delineations are sent to the landowner for review. If the landowner contests the delineation, it is reviewed in the field by methods described in NFSAM, and corrected as necessary. There are four stages in the appeal process, corresponding to the four levels within NRCS: district, area, state, and national (Chapter 7). At the national level, there have been only about 200 appeals out of 2 million determinations (M. Fritz, U.S. EPA, presentation to NRC Wetlands Characterization Committee, Feb. 2, 1994); all other disputes either were resolved at a lower level or were not appealed to a higher level. FSA delineations are not subject to public review, and the wetland maps are not published. In addition to making wetland determinations, NRCS also prepares wetland inventory maps by use of ASCS 35 mm slides, aerial photographs, soil surveys, and NWI maps.
The Food, Agricultural, Conservation, and Trade Act of 1990 amended FSA to include a certification requirement for all wetland determinations. Determinations done before Nov. 28, 1990, become certified if the decision is appealed at least one level (to the district conservationist), or if the state conservationist determines that the inventories and mapping conventions are adequate and finds a sample of determinations in the field office to be accurate. Determinations done between Nov. 29, 1990, and Jan. 6, 1994, become certified if the decision is appealed at least one level (to the district conservationist), or if the decision is not appealed and the state conservationist determines that the inventories and mapping conventions are adequate, and if the affected persons are notified and given rights of appeal. Wetland determinations had been completed by these conventions for 60% of U.S. Department of Agriculture (USDA) program participants when, in 1991, NRCS offices were directed to discontinue mapping until a review of the FSA delineation manual could be completed (B. Teels, presentation to NRC Wetlands Characterization Committee, Sept. 15, 1993). From 1991 until the signing of the 1994 interagency Memorandum of Agreement (MOA), determinations were made by NRCS only as needed for evaluation of easements under the Wetlands Reserve Program (16 U.S.C. § 3837) or the Environmental Conservation Acreage Reserve Program (16 U.S.C. § 1230), and for cost-sharing under the Agricultural Water Quality Incentives Program (16 U.S.C. § 3838) (P.L. 101-624).
Determinations After 1994
On Jan. 6, 1994, the interagency MOA gave NRCS responsibility for making
delineations for the swampbuster provisions of FSA and Section 404 of CWA (Chapter 4). Under the new MOA, representatives of U.S. Army Corps of Engineers (USACE), Environmental Protection Agency (EPA), FWS, and NRCS must concur in writing on the mapping conventions to be used in each state. Mapping conventions were discussed by representatives of all four agencies at an interagency meeting convened May 16-20, 1994, in St. Paul, Minnesota. Mapping conventions approved by the agencies for use at the state level will be reviewed by the headquarters of the signatory agencies to ensure national consistency. All wetland determinations done after Jan. 6, 1994, become certified on agricultural lands if approved mapping conventions were used. Determinations done on nonagricultural lands after Jan. 6, 1994, become certified if: USACE, under EPA review, makes the final wetland determination; if NRCS makes the determination on wetlands included within agricultural land; or if NRCS makes the determination by request on lands owned or operated by a USDA program participant (NFSAM Part 514.52).
The memorandum also establishes a monitoring and review process that is intended to improve wetland delineation. EPA will lead the signatory agencies in establishing interagency state oversight teams in periodic reviews of wetland delineations. Each team will attempt to reach agreement on wetland delineation issues that arise during these reviews, which will be conducted quarterly for the first year, semiannually for the second year, and annually thereafter. NRCS also will provide information to FWS about wetlands on NWI maps that are not verified as wetlands by NRCS mapping conventions or field investigation (NFSAM Part 513.31).
The third edition of NFSAM manual (Parts 513.30 and 527.4) contains general guidance for developing wetland mapping conventions that are consistent with mapping conventions previously used for FSA. In addition to the five wetland categories (W, FW, CW, PC, AW), Part 526.100 lists 19 new delineation categories related to various types and uses of agricultural wetlands.
The purpose of the NWI is to map the wetlands of the united States for resource assessment rather than for regulation. NWI was begun by FWS in the mid-1970s and is still in progress. Maps are prepared by interpretation of photographic transparencies under stereoscopic magnification. Boundaries are drawn directly on the photos with a 0000 drafting pen, which on 1:58,000 photo represents approximately 40 ft (12 m) on the ground. Photo interpretation and field checking are done by contractors from consulting firms, universities, and state and federal agencies, with quality control by NWI personnel. Interpreters are usually expected to spend 20 hours (2 1/2 work days) in the field for each 1:100,000 map (the equivalent of 32 1:24,000 topographic maps) (National Wet-
lands Inventory, 1990). Final map preparation is done by transfer of the wetland boundaries from the photographs to 1:24,000 maps (Figure 8.1).
Most of the aerial photographs used for NWI mapping were obtained by the U.S. Geological Survey (USGS) for general use and therefore are not always optimum for wetland mapping. Until the early 1980s, 1:80,000 (0.39 in. = 2,624 ft; 1 cm = 800 m) black and white panchromatic photos acquired by USGS were used. Each photo covers the equivalent of a 1:24,000 topographic quadrangle. After 1980, 1:58,000 (0.39 in. = 1,902.4 ft; 1 cm = 580 m) color infrared photos taken by the National High Altitude Photography program were used. This program was replaced in 1992 by the National Aerial Photography Program, which acquires 1:40,000 (0.39 in. = 1,312 ft; 1 cm = 400 m) color infrared aerial photography.
Because NWI maps depict wetlands that were present on the date of photography, wetland extent can be estimated incorrectly if atypical expansion or contraction of vegetation was occurring at the time of photography. Interannual variation causes fewer errors where ground water maintains wetland hydrology or where vegetation is resistant to interannual variation (such as in forested wetlands). Even though NWI is incomplete for much of the country (Figure 8.2), some of the earliest maps are 20 years old. The age of NWI maps is a particular problem in areas where agricultural and urban development have altered wetlands.
Wetland delineation on NWI maps is generally accurate areas where there is an abrupt change in hydrology, soil, or vegetation at the wetland boundary. In the prairie pothole region, for example, wetlands smaller than 1.24 acres (0.5 ha) are mapped routinely by NWI (Tiner, 1990). Mapping of wetlands in level landscapes, such as coastal or glaciolacustrine plains, is less precise because boundaries are not as evident. Forested wetlands are particularly difficult to map because foliage obscures the ground. Temporarily flooded, forested wetland is one of the most difficult types to map because, for most of the year, the water table usually lies below the surface.
NWI maps tend to be less inclusive of wetlands than are other wetland maps. Farmed wetlands usually are not included on NWI maps (NWI, 1990; NFSAM, 1994), and areas mapped as hydric soils on USDA soil surveys are generally much more extensive than are areas mapped as wetland on NWI maps (Street, 1993). In Washington and Tyrrell counties of coastal North Carolina, for example, only 19% of the hydric mineral soils were mapped as wetland by NWI, even though 82% of the hydric organic soils were mapped as wetlands (Moorhead and Cook, 1992). In the same area, Lukin and Mauger (1983) mapped as wetland nearly 35,000 acres (14,000 ha) that NWI showed as upland; the addition of these sites to the NWI maps would have increased the total wetland area by 16%. These three sites are clearly not a comprehensive sample, however, the NWI maps should be evaluated broadly in relation to field-delineated wetlands.
NWI was not designed to be used for regulatory delineation. It is a useful
source of background information for wetland delineations, and it is recommended as an ancillary data source by all federal delineation manuals (1987; 1989; NFSAM). Its utility as an ancillary data source varies regionally; it is least useful in areas where broad expanses of mineral soils with facultative vegetation and little topographic variation complicate wetland delineation. Given NWI's
utility for delineation in many parts of the country, however, as well as its importance in providing synoptic information about the nation's wetlands, it should be completed.
GEOGRAPHIC INFORMATION SYSTEMS
Geographic information systems (GISs), computerized systems for the analysis and display of spatially distributed data, have great potential for use in management, regulation, and study of wetlands. A GIS can be used to store, retrieve, and edit data; it can be used to create new data bases; and it can be used for tabular, graphic, and digital presentation of information. A GIS offers numerous advantages in extraction and analysis of data, and in the revision of data files (Johnston et al., 1988a).
The primary disadvantages of GIS use are the time and expense required to digitize maps, and the expensive equipment and trained personnel that are needed. Also, a GIS data base is only as good as the source from which it was derived.
NWI maps are being digitized in a GIS-compatible format, which should greatly increase their utility (Tiner and Pywell, 1983). This effort is progressing even more slowly, however, than is map production. Digitizing is complete or nearly complete for only 10 states: Delaware, Florida, Indiana, Illinois, Maryland, Minnesota, New Jersey, Rhode Island, Virginia, and Washington (Figure 8.3). Most of these states paid part of the cost of the work.
A GIS can be used as part of a spatial decision support system (SDSS) for wetland delineation. An SDSS is a computerized system for data interpretation, manipulation, and analysis that is used for support of complex decisions based on spatially distributed information (Djokic, in press). It is designed to be interactive and easy to use. The solution procedure is developed interactively by the user, who creates a series of alternative solutions and then selects the most viable. This approach could be used with a wetland delineation decision tree and digital information on soils, vegetation, and topography. An SDSS also could facilitate management decisions about wetlands by putting them in a landscape or historical framework. For example, an SDSS could be used to evaluate the effect of additional wetland loss relative to past wetland losses within a region.
Protection of wetland functions requires that wetlands be considered in context with the surrounding landscape. A GIS can be useful for this purpose. A GIS can place individual wetlands in appropriate spatial context (such as in a watershed or in a waterfowl flyway) and can combine information about wetlands with information about their surrounding environment. Empirical relationships between resource loss and measures of environmental degradation can be developed with a GIS (such as degradation of water quality, or loss of biodiversity). Rates of change in number or extent of wetlands can be quantified with a GIS that contains wetland maps for two periods. Transition probabilities derived from
such analyses can then be used for predicting wetland trends (Pastor and Johnston, 1992).
A GIS can be used in assessments of cumulative environmental change (Johnston et al., 1988b; Johnston, 1994b). Disturbances that affect wetlands directly (such as the location of logged areas within wetlands) and indirectly (such as upstream sources of water pollution) also can be analyzed with a GIS.
A major barrier to the use of GISs has been the lack of suitable digital data on wetlands. Fortunately, this situation is changing. NWI maps are being digitized, and NRCS plans to digitize county soil survey maps under its Soil Survey Geographic Data Base program, which should facilitate identification of hydric soils (Reybold and TeSelle, 1989). EPA's North American Landscape Characterization program, which was developed in collaboration with the USGS EROS Data Center and the NASA Landsat Pathfinder program, will produce digital land cover maps from 1991 Landsat Multispectral Scanner images. It also will generate image-derived digital land cover change maps for the 1970s, 1980s, and 1990s (EPA, 1993). EPA has digitized all streams in the nation, coded by reach number, associated with its STORET water quality data base. USGS is producing digital elevation models (topographic data) for 7 l/2-minute areas corresponding to its 7 1/2-minute quadrangle series, and has digitized the watersheds it uses for its hydrologic units. These data bases could be useful for delineation, but their utility in a landscape context will rest on their accuracy and on the ability of wetland delineators to interpret them correctly.
Hydrologic models describe natural processes by representations that are either conceptual or mathematical (NRC, 1990). Conceptual models deal with interactions of hydrologic processes by the use of simplifying approximations and assumptions. For example, the accuracy of water budget estimates for most wetlands is limited by measurement difficulties, but this does not prevent the development of a conceptual model that shows how the budget components interact (Winter, 1988). A mathematical model can be developed from a conceptual model. The mathematical model makes specific quantitative estimates of the hydrologic characteristics of a watershed or wetland. The success of mathematical models is largely predicated on the validity of the underlying conceptual model. For example, the results of a mathematical model of water flow in a raised bog can be quite different if it is assumed that the bog is separated from or hydraulically connected to the water table. In general, it is difficult to describe with models both the flow system and the boundary conditions for natural systems. In such cases, models are best used to determine empirically what is possible or probable, even though they might not produce accurate predictions (Oreskes et al., 1994). The effects of modifications such as those caused by drainage ditches are generally easiest to describe because the boundary condi-
tions and principal flow directions are better defined than they are for natural systems.
Mathematical Models to Assess Wetland Hydrology
There are two major applications of mathematical models in wetland hydrology. The first involves assessment of the hydrology of a site in relation to hydrologic thresholds for wetlands. Modifications by drainage or other activities can make it impossible to determine wetland status from soils and vegetation, both of which could be characteristic of the site's former hydrology but not of its altered condition. Models can be especially useful in assessing wetlands under such conditions. The second important application of models is in estimating or projecting the effects of cumulative wetland loss on regional hydrologic characteristics, particularly flood flows. Several of watershed-response simulation models, such as HEC2 (Bedient and Huber, 1988), have been developed to project flood flow in response to precipitation, soil moisture and depression storage, infiltration, ground water flux, evapotranspiration, and natural or constructed drainage systems.
Types of Mathematical Models
Mathematical models can be either analytical or numerical. Analytical models solve the fundamental equations for conservation of mass and fluid flow for steady-state processes and for transient events (Kirkham, 1957; van Schilfgaarde, 1974; Luthin, 1978). They are used to calculate water table response to drainage by open ditches and drainage pipes. Analytical models usually assume constant values for hydraulic properties of the soil (such as hydraulic conductivity and drainage porosity) and simple, well-defined boundary and initial conditions of the wetland. Although analytical models are mathematically exact solutions to the equations governing water flow, their application usually involves the use of approximations that can introduce errors.
Water in soil moves in three dimensions in the general case, but most analytical models treat it in one or two dimensions process. These approximations can be used in some situations to estimate the most probable location of the water table over a long period. Where the physical properties of soil and the sources of water vary widely in a wetland, however, analytical models are less likely to be successful. The wide range of models suitable for determining the steady-state and transient water table response to drainage is given by Kessler et al. (1973), van Schilfgaarde (1974), and Cohen and Miller (1983). Although analytical models can describe drainage processes and water table responses for steady rainfall and short-term drawdown, they are not generally applicable to describing water table fluctuations caused by the combined effects of rainfall, drainage, and evapotranspiration over long periods.
Numerical models can be functional (lumped parameter) or discritized (distributed parameter). Both solve the equations for fluid flow and conservation of mass. Functional models usually assume that the wetland behaves as a relatively uniform soil system with well-defined boundary conditions, although layers with different hydraulic properties can be considered. The water budget for the wetland is reduced to algebraic equations, and analytical algorithms are used to calculate flux as a function of water table position, depth, and spacing of natural or constructed drains, potential evapotranspiration, and the like. Unlike purely analytical models, which are usually used for steady-state conditions or short transient events, functional models can be used to project the rise and fall of water levels with time over a long period of record. Three well-known functional models that simulate wetland hydrology in drained areas are DRAINMOD (Skaggs, 1978; 1991), PHIM (Guertin and Brooks, 1986; Guertin et al., 1987), and SWATRE (Feddes et al., 1978). Functional models are effective tools for evaluating the hydrologic features of an entire wetland or certain specific points within a wetland. They are generally less useful for determining spatial differences in the water table regime for a wetland and adjacent upland.
Distributed-parameter numerical models are used for estimating detailed subsurface ground water flow and flood flows in watersheds. They are powerful, but they are difficult to construct and use (NRC, 1990). The study area (the flow domain) is partitioned into sections, each of which can be assigned values for the physical or hydraulic properties of the soil, rates of recharge and water loss, initial water levels, slope, storage capacity, and other variables that affect water movement and distribution. The numerical routines calculate water flow into and out of each section and for the domain as a whole. As with the analytical and functional models, distributed-parameter models originated primarily from two sources: they come from models that describe ground water processes and models that predict the performance of drainage systems. The most theoretically rigorous of these methods, the so-called exact approach, is difficult and expensive and can be applied only by an expert in ground water modeling. A basic limitation is the requirement of detailed descriptions of the properties of unsaturated soil and boundary conditions throughout the flow domain. These barriers will likely confine the use of distributed-parameter models to wetland research rather than to regulation.
Distributed-parameter ground water models have been used to determine the major controls over wetland ground water hydrology in many settings (Siegel, 1983; Winter, 1988; McNamara et al., 1992). A useful and easily applied distributed-parameter approach for modeling wetlands is based on numerical solutions to the Boussinesq equation. This approach is appropriate where subsurface water movement is primarily horizontal in the saturated zone. Models such as those developed by de Laat et al. (1981) and WATRCOM, developed by Parsons et al. (1991a,b), can be used to predict water table fluctuations continuously throughout a wetland on a continuous basis. By this means, the effect of a drainage ditch
on the water table can be evaluated as a function of distance from the ditch. Boundaries that satisfy hydrologic conditions for wetlands can be determined from simulations of water table positions over long periods. Differences in soil properties and land uses from point to point in the wetland and adjacent upland also can be considered. These models can be applied to large, heterogeneous areas.
Model Selection and Application
The simplest model that will provide the required accuracy and resolution is the most desirable because the reliability of predictions often decreases with complexity. Hydrologic models require information on the physical attributes of the site, soils, vegetation, and climate. All data to be used in models are subject to uncertainty, so it is essential that the user consider the effects of uncertainty through the use of sensitivity analysis. Field data can increase the reliability of modeling. Often several months or even years of field hydrologic data on a wetland are insufficient for classification of a marginal site because of the inherent variability of hydrology (Chapter 5). In such cases, the field data can be used to test and calibrate a model, which can then be used for simulating water table conditions over a long period of record, providing a good basis for determining whether wetland hydrology exists on the site.
Advantages and Disadvantages of Hydrologic Modeling
Models offer several advantages. They can be used to analyze the effects of alterations, and thus provide a means of assessing wetland hydrologic conditions without reference to vegetation and soils, which might not be reliable indicators under these conditions. In addition, the assessment of the hydrology by use of models is not limited by short-term weather conditions. Models also can be used to predict the effects of agricultural drainage or other activities on the hydrology of adjacent wetlands, and the effects of wetland alterations on regional flood flows.
Models also have disadvantages. Because they are data intensive, they can be expensive and time-consuming. Their application also requires specialized training; if they are improperly applied, models can lead to erroneous conclusions. Overall, however, models should prove increasingly useful, particularly for quantifying hydrologic features of ecosystems for which direct hydrologic information is unavailable or inadequate.
QUANTITATIVE ANALYSIS OF BOUNDARIES
Boundary detection is one aspect of the analysis of spatial change. Although the position of an ecological boundary might be evident with little analysis,
boundary detection can be difficult if the change is gradual (Hansen et al., 1988). The clarity of a boundary varies with the amount of change over distance, as well as with the overall magnitude of change. The most distinct boundaries are those for which there is a large change over a short distance.
Threshold Values are often used to set ecological boundaries (Chapter 5). An example is the use of the prevalence index for vegetation to distinguish between wetland and upland. Although there can be statistically significant differences between the wetland and upland, it does not follow that the boundaries between them will be distinct.
Transect Data for Boundary Determination
The use of transects for determination of wetland boundaries is recommended by the 1987 and 1989 manuals. Biological, physical, or chemical data are collected along a perceived gradient and are analyzed for ecological discontinuities. Quantitative techniques can be used to locate and characterize ecological discontinuities along transects (Webster and Wong, 1969; Webster, 1973; Ludwig and Cornelius, 1987; Wierenga et al., 1987; Brunt and Conley, 1990), including transects across boundaries between wetland and upland. Extensive use of quantitative methods might be beneficial for application to gradients in redox potential, particularly in areas where vegetation and soils have been disturbed.
When suitably located and sampled, transects can provide a large amount of data with minimal effort. Because a transect locates only one or two points on a boundary, however, identification of an entire wetland perimeter requires multiple transects. Also, because transects are often placed perpendicularly to perceived boundaries, the data are biased toward visible gradients, and therefore could be unrepresentative of gradients that are not visible.
Detection of Boundaries with Image Analysis
Images provide information about the entire landscape, including boundaries. Information about boundaries can be extracted from an image by use of the moving-window technique, which involves a scan of the image with a two-dimensional window. The moving window technique can be applied to any 2-dimensional digital data, including aerial photography scanned with a video digitizer or scanning camera. For example, Johnston and Bonde (1989) have used ''textural analysis,'' a moving-window technique that measures boundary contrast as the relative difference between the reflectance values of picture elements (Musick and Grover, 1991), to analyze boundaries within a Landsat satellite image of a portion of northern Minnesota. They applied this technique to a map of normalized difference vegetation index, which is a measure of the spectral properties of vegetation. Nellis and Briggs (1989) performed a similar analysis on the Konza prairie in Kansas.
Scientific vs. Legal Boundaries
Just as the interior of a wetland can be classified by soil type, vegetation, or other variables, its boundary can also be classified. A vector-based GIS, which depicts features as a series of connected points and lines, inherently classifies boundaries based on the features that are being separated and can be used to classify boundaries by other attributes. For example, a land cover map of Scotland produced by the Macaulay Land Use Research Institute uses different line widths to indicate the precision with which ecological boundaries are located (Aspinall et al., 1993). Similarly, boundaries could be classified by their strength (magnitude of change), width, or permanence. From a scientific perspective, wetland boundaries must be shown in some cases as broadly placed within a transition zone.
Land ownership and regulation in the United States are based upon discrete lines separating one piece of property from another. The mathematical analogue of this approach is classical set theory, in which space is discretely subdivided by use of threshold values rather than by probabilities. Herein lies a basic problem of wetland delineation: ecological properties often change gradually, rather than sharply, whereas legal boundaries are lines without width. The authors of the 1987 and 1989 manuals have done a credible job of establishing thresholds that compact a wetland boundary into widthless line, but legal boundaries will never be fully reconciled with ecological reality.
Aerial photography can be useful for wetland delineation and mapping if its timing, frequency, and scale are suited for making wetland determinations. Aerial photographs should be acquired specifically for wetland delineation in areas where these requirements are not met by existing photographs.
The interpretation of aerial photography should be done by personnel trained in this method of wetland determination and who have field knowledge of wetlands in the area being interpreted.
Monoscopic interpretation of aerial photography should be supplemented with information on topography and soils and should be validated by periodic field reconnaissance and regional assessments of accuracy.
The accuracy of offsite wetland determinations for agricultural lands should be evaluated comprehensively in the field before mapping conventions are adopted and wetland determinations are certified.
Remote sensing by satellite and high-altitude aircraft has promise for wetland delineation and should be evaluated further as a potential technique for wetland delineation where large areas of land are flooded seasonally.
Models, if verified in the field, should be accepted for analysis of the hydrology of some wetlands, including altered wetlands.
Documentation of wetland boundaries by use of global positioning systems should be encouraged as GPS technology is refined.