USGS Home
Contact USGS
Search USGS

Field

We studied 40 basins in 1992 and 36 in 1993. We studied 32 of these basin one of the two years only and 21 in both years. We selected July and the first week of August for field surveys because peak standing crops occur during these months in the north-temperate United States (Bernard, 1974). A soils and sediment research team accompanied us in the field. We visited study basins in a general south-to-north route to help compensate for the approximately two-week difference in phenology between the southernmost and northernmost plots. We used county road maps, NWI maps, and high-level, black-and-white aerial photographs of the 10.4-km² plots to locate the study basins in the field.

We delineated wetland zones on low-level, color aerial photographs of each basin taken in mid-June of the same year. Although vegetation usually forms a virtual continuum around prairie wetlands, zonation is usually evident (Johnson et al., 1985). Zones are areas closely related to degree of water permanence and have characteristic assemblages of plants (Steward & Kantrud, 1971). Peripheries of wet-meadow zones were usually considered the boundaries of the study wetlands. In cropland, these boundaries were often difficult to determine and were estimated, based on basin overflow (spill) elevations, soil characteristics, and vegetation. On the low-level photographs, we also delineated plant communities within zones. Plant communities are defined by Westhoff & Van der Maarel (1973) as vegetation in relatively uniform environments with floristic composition and structure relatively homogeneous and distinct from surrounding vegetation. Many zones had a single plant community. In a few instances, drastic water-level increases, cultivation, or rapid crop growth in late June or early July made portions of the low-level photograph obselete. In those cases, we delineated zone and community boundaries using whatever reference points (boulders, haystacks, fencelines) were available.

If a wetland was subject to different land uses, as is often the case where basins are fenced or in multiple ownership, we restricted sampling to communities in the portion of the wetland with predominant land use. Within this area, we further restricted sampling to plant communities occupying at least 10 percent of the area. We numbered all plant communities, noted their land use, and assigned them to wetland zones of Stewart & Kantrud (1971).

We used a modified method of Barker & Fulton (1979) for vegetation sampling. We sampled vegetation along the long axis through the center of each plant community to avoid edge effects. We paced the long axis and threw a marker buoy overhead at each of five roughly equidistant points along the axis. At the point of impact, we centered a 1-m² quadrat frame (Figure 1).

Figure 1. Prairie pothole region of Minnesota and the Dakotas showing strata based on wetland density and location of 10.4-km2(4mi2) plots containing study wetlands.

We assigned a Daubenmire (1959) cover class to each macrophyte taxon in the quadrat, but here we report only information on species richness. We noted plant taxa not encountered in the quadrats while walking between quadrats. Total taxa recorded inside and outside the quadrats were summed to provide a measure of taxon richness for each community.

We identified nearly all plants to species, but identified a few only to genus or family. We were unable to identify four plants. We classified all pteridophytes and spermatophytes as to life history (annual or biennial, perennial, native, or introduced) from standard floras. After we completed the plant sampling, the soils and sediment researchers cored the bottom substrate at the center of each quadrat with a hand auger and measured litter depth to the nearest cm. The fresh litter cores were recognizable as undecomposed or partially decomposed fallen vegetation. We considered the bottom of the litter layer the point where decomposing material changed from fibric (peat) to hemic (muck) or where plant remains became unrecognizable as such when observed through a 10 x hand lens. We measured water depth to the nearest cm at the centers of quadrats and marked their locations on the field photographs. Within each quadrat, we also estimated the percentages of unvegetated bottom, open water, and percentage of standing dead vegetation.

Analytical

At the end of each field season, we scanned and georeferenced the low-level aerial photos of each basin with a map and image processing system (Skrdla, 1992) to determine areas of the sampled plant communities. On each image, we classified all polygons within the basin as to wetland zone and community and marked the locations of quadrats. We also classified artificial wetland types within the basins (e.g., excavated dugouts for livestock watering) and uplands within the basins (e.g., spoils or rockpiles). Data were averaged across the five quadrats within each community prior to analysis.

We used analysis of variance (ANOVA) techniques to assess the effects of watershed quality and year on total zone area. Because approximately half of the sample wetland basins were measured in both 1992 and 1993, the design was a repeated measures with year serving as the repeated-measures factor. ANOVAs were done separately for each of the five zones: low-prairie, wet-meadow, shallow-marsh, deep-marsh, and fen.

We also used ANOVA techniques to assess the effects of watershed quality, zone, and year on all response variables measured at communities within zones (Table 1).

Table 1. Response variables

1. Variable measured at zone within wetland basin:
   1. Area of zone (ha)
2. Variable measured at quadrats within communities within zones (all analyses were done by first averaging across quadrats):
   1. Water depth (cm)
   2. Percent standing dead vegetation
   3. Length (cm) of litter core
   4. Percent unvegetated (bare) bottom
   5. Percent open water
   6. Percent vegetation
3. Variables measured at communities within zones:
   1. Area of community (ha)
   2. Total plant species
   3. Total perennial introduced plant species
   4. Total perennial native plant species
   5. Total annual introduced plant species
   6. Total annual native plant species

The sampling design was a split-plot with repeated measures. Each basin was assumed to be the independent whole-unit, with zone and community combination being the sub-unit (see Figure 2).

Figure 2. Study wetland showing sampled and unsampled hydrophyte communities, location of quadrats, and land-use of uplands.

The fen and low-prairie areas were combined with wet-meadows. Because approximately half of the sample wetland basins were measured in both 1992 and 1993, year served as the repeated-measures factor. However, because of the highly unbalanced design structure (Table 2), the three-way interaction effect and least-squares means of wetland quality by zone by year were not fully estimable for the repeated-measures design. Therefore, we randomly deleted one year's data on basins that were used in both 1992 and 1993 (Table 2 footnote). This allowed basin to become "nested" within year and wetland quality and thus made the three-way interaction testable, albeit with slightly less power, and all least-squares means estimable. This procedure eliminated the need to consider the repeated-measures aspect as part of the analysis. We report the least-squares means from this "balancing" approach as all combinations among year, water quality, and zone. Multiple passes were made through the data with a different random selection each pass. In all passes, ANOVAs yielded similar conclusions. We used Fisher's protected least significant difference (LSD) to isolate differences in least-squares means following significant effects in the ANOVAs (Milliken & Johnson, 1984) where applicable. We considered α = 0.05 to be significant.

Table 2. Sampling design lay-out. Single (X or x) or multiple communities (number) were sampled within deep-marsh (DM), shallow-marsh (SM), or wet meadow (WM) zones in basins in good- and poor-quality watersheds, 1992-1993

Identification number		1992			1993
Plot	Basin	DM	SM	WM	DM	SM	WM
Good-quality watersheds
73	29	X	X	X	x	x	xa
374	100	x	x	x	X	X	X
374	225	X	X	X	x	2	x
442	301	x	x	x	X	X	X
156	22		X	X		x	x
363	22	x		x	X		X
363	58		X	X		x	x
442	93		x	x		X	X
442	295		X	X	x	x	x
73	86			X			x
156	24			x			X
156	26			X			x
156	42			x			X
374	272
		2			2
374	65			x			X
60	58		X	X
60	128			X
249	50		X	X
249	86		X	X
59	111		X	X
396	106			X
396	107			X
396	130		X	X
407	67						X
407	109						X
498	146				X	X	X
498	227
498	277				X	X	X
133	386					X	X
407	168				X	X	X
Total communities		5	13	24	9	13	23
Grand Total							87
Poor-quality watersheds
134	140		2			X	X
134	165		X			x	x
134	270			X			x
134	406		x	x	X	X	X
134	432			X			x
442	260			x			X
442	261			X			x
442	281		x	x		X	X
38	62			X
54	39			X
59	42	X	X	2
134	272			X
241	3	X	X	2
241	48			2
246	34			X
246	37			X
246	53		X	X
133	370						2
133	380					X	X
134	158						X
327	72					X	X
327	117					X	2
327	147					X	X
Total communities		3	8	17	1	8	16
Grand Total							53

^a Data from communities designated by lower case x's and underlined numbers were not used in ANOVAs or for estimating least-squares means for response variables measured at community level (see methods).

Footnote:

Effects	Type	No. levels	Levels
Basin quality	Fixed	2	Good, poor
Wetland zone	Fixed	3	Deep-marsh,
			shallow-marsh, wet-meadow
			(includes fen and
			low prairie)
Year of study	Fixed	2	1992,1993
Wetland basin	Random	-	-
Communitya	Random	-	-

^a Community by zone was considered the sampling unit and randomness assumed.

All ANOVAs were conducted using the general linear model procedure (PROC GLM) of SAS (SAS 1989). Least-squares means (SAS 1989) were computed and reported when adequate data were available for ANOVAs. Otherwise, arithmetic means are reported. Effects considered fixed and random are listed in Table 2 footnote. For most of the response variables, we conducted the ANOVAs both in the original unit of measurement and using a ln(Y+ 1) transformation (Steel & Torrie, 1980). However, we do not report the results of the transformation because ANOVA results were similar for both transformed and untransformed data; this indicates no gross departures from ANOVA assumptions for untransformed data (see Conover (1980):337). Least-squares means in tables are at the highest-order interaction for reporting purposes, with standard errors based on ANOVA mean-squared error terms.