Abstract
Physical factors, including both in-stream and riparian habitat characteristics
that limit biomass or otherwise regulate aquatic biological condition,
have been identified by previous studies. However, linking the ecological
significance of nutrient enrichment to habitat or landscape factors
that could allow for improved management of streams has proved to be
a challenge in many regions, including agricultural landscapes, where
many ecological stressors are strong and the variability among watersheds
typically is large. Riparian and associated habitat characteristics
were sampled once during 2003–04 for an intensive ecological and nutrients
study of small perennial streams in five contrasting agricultural landscapes
across the United States to determine how biological communities and
ecosystem processes respond to varying levels of nutrient enrichment.
Nutrient concentrations were determined in stream water at two different
sampling times per site and biological samples were collected once per
site near the time of habitat characterization. Data for 141 sampling
sites were compiled, representing five study areas, located in parts
of the Delmarva Peninsula (Delaware and Maryland), Georgia, Indiana,
Ohio, Nebraska, and Washington. This report examines the available data
for riparian and associated habitat characteristics to address questions
related to study-unit contrasts, spatial scale-related differences,
multivariate correlation structure, and bivariate relations between
selected habitat characteristics and either stream nutrient conditions
or biological responses.
Riparian and associated habitat characteristics were summarized and
categorized into 22 groups of habitat variables, with 11 groups representing
land-use and land-cover characteristics and 11 groups representing other
riparian or in-stream habitat characteristics. Principal components
analysis was used to identify a reduced set of habitat variables that
describe most of the variability among the sampled sites. The habitat
characteristics sampled within the five study units were compared statistically.
Bivariate correlations between riparian habitat variables and either
nutrient-chemistry or biological-response variables were examined for
all sites combined, and for sites within each study area.
Nutrient concentrations were correlated with the extent of riparian
cropland. For nitrogen species, these correlations were more frequently
at the basin scale, whereas for phosphorus, they were about equally
frequent at the segment and basin scales. Basin-level extents of riparian
cropland and reach-level bank vegetative cover were correlated strongly
with both total nitrogen and dissolved inorganic nitrogen (DIN)
among multiple study areas, reflecting the importance of agricultural
land-management and conservation practices for reducing nitrogen delivery
from near-stream sources. When sites lacking segment-level wetlands
were excluded, the negative correlation of riparian wetland extent with
DIN among 49 sites was strong at the reach and segment levels.
Riparian wetland vegetation thus may be removing dissolved nutrients
from soil water and shallow groundwater passing through riparian zones.
Other habitat variables that correlated strongly with nitrogen and phosphorus
species included suspended sediment, light availability, and antecedent
water temperature.
Chlorophyll concentrations in seston were positively correlated with
phosphorus concentrations for all sites combined. Benthic chlorophyll
was correlated strongly with nutrient concentrations in only the Delmarva
study area and only in fine-grained habitats. Current velocity or hydraulic
scour could explain correlation patterns for benthic chlorophyll among
Georgia sites, whereas chlorophyll in seston was correlated with antecedent
water temperature among Washington and Delmarva sites. The lack of any
consistent correlation pattern between habitat characteristics and organic
material density (ash-free dry mass) within study areas may indicate
that the density of organic matter is not generally sensitive to nutrient
enrichment in small agricultural streams. For all sites, and for the
Nebraska, Delmarva, and Georgia subsets of sites, the reach-mean areal
coverage of aquatic macrophytes and macroalgae was strongly related
to channel shading.
Data reduction techniques were applied to select
a subset of 29 variables, representing 20 categories of habitat characteristics,
for multivariate analysis. Factor analysis was used to identify and
interpret three leading modes of variation (principal factors) in two
data subsets—one for the Georgia sites and one for all other sites combined.
The factor analysis for Georgia sites indicated that riparian land use
and land cover (LULC) (wetland extent in particular) and channel shading
correspond to dominant modes of variability in the habitat data set.
The variables that best characterize variation in riparian habitat for
the other four study areas included mid-channel measures of canopy shading,
riparian cropland extent in the 15-meter buffer and 150-meter buffer,
and measures of the patchiness of woodland cover in the 15-meter buffer
(patch length and gap frequency). LULC metrics calculated for riparian
buffers, particularly at the segment scale, were more correlated with
the principal modes of variation in the overall habitat data set than
was LULC extent for the total basin drained by each site.
Correlations of woodland extent within 15 to 50 meters of the channel
(reach- and segment-level data) with woodland extent in a series of
longitudinal bands of the riparian buffer that were located at increasing
distance from the channel showed decreased strength as the compared
band shifted beyond the first 50 meters from the channel, becoming negligible
for areas beyond 100 meters from the channel. For many of the studied
agricultural streams, the riparian buffer includes a heterogeneous mix
of riparian and upland land covers when the summarized buffer area extends
more than about 50 to 100 meters from the streambank, depending upon
basin (or stream) size. Comparisons between the extent of reach- and
segment-level median values of woodland and other cover types within
the riparian buffer extending 50 meters from the stream suggest that
the reach length used for this study generally is not long enough to
accurately represent both the overall composition and patch structure
that characterizes the riparian areas along small, agricultural streams.
The mean extent of forest plus woody wetland ranged from 5.4 to 76
percent of the riparian buffer area. For the Georgia sites, where riparian
woody wetlands were more extensive than for any other study area, canopy
closure over the channel was greatest, whereas it was least for sites
in Washington and Nebraska.
To the extent that riparian woodland is the most important LULC type
affecting algal-nutrient relations, correlations indicated that basin
characteristics might be effective surrogate predictors of riparian
effects at the drainage-network scale. But the results also indicated
that basin-level cropland was not an accurate surrogate for riparian
cropland extent