The effects of forest on stream water quality in two coastal plain watersheds of the Chesapeake Bay

The effects of forest on stream water quality in two coastal plain watersheds of the Chesapeake Bay

Ecological Engineering 14 (2000) 337 – 362 www.elsevier.com/locate/ecoleng The effects of forest on stream water quality in two coastal plain waters...

1MB Sizes 1 Downloads 43 Views

Ecological Engineering 14 (2000) 337 – 362

www.elsevier.com/locate/ecoleng

The effects of forest on stream water quality in two coastal plain watersheds of the Chesapeake Bay M.M. Norton *, T.R. Fisher Horn Point Laboratory, Center for En6ironmental Science, Uni6ersity of Maryland, P.O. Box 775, Cambridge, MD 21613, USA

Abstract Forest had varying effects on stream nutrients in two coastal plain basins of the Delmarva Peninsula, USA. In the Choptank basin, forest was strongly associated with low stream total nitrogen (TN) and nitrate (NO− 3 ) concentrations (r 2 0.70), and forest placement along first order streams was important in maintaining low stream nitrogen (N) concentrations (r 2 0.35). In addition, a multiple regression model explained  40% of the stream total phosphorus (TP) variance and indicated that forest directly adjacent to streams (0 – 100 m) acted as a TP source and forest further away (100–300 m) from streams acted as a TP sink. In contrast, stream nutrients in the nearby Chester basin demonstrated a strong relationship with soil hydrologic properties. Forest had no significant effect on stream N and P because the finer-textured soils, higher stream slopes, and higher runoff potential of the Chester basin appeared to result in less baseflow compared to that in the Choptank basin. This reduced the opportunity for forest to intercept N via plant uptake and denitrification in the high runoff potential soils of the Chester basin. The high percentage of stormflow (40%) coupled with high stream slopes resulted in high soil erosion potential, which may explain the higher TP stream concentrations measured in the Chester compared to that in the Choptank. Differences in the hydrologic pathway appear to explain the different effects of forest on water quality in these two basins. © 2000 Elsevier Science B.V. All rights reserved. Keywords: Riparian forest; Nonpoint source pollution; Landscape analysis; Hydrologic characteristics; Soil texture; Nutrient export; N; P

1. Introduction Chesapeake Bay is the largest estuary in North America. Studies conducted since the 1970s indicate that increases in the intensity of agriculture, population growth, and sewage discharge are * Corresponding author. Tel.: +1-410-2288200; fax: +1410-2218490. E-mail addresses: [email protected] (M.M. Norton), [email protected] (T.R. Fisher)

causing the Bay to become eutrophic (EPA, 1983). Nutrient enrichment of streams and other anthropogenic disturbances within the Bay’s watershed have triggered a series of undesirable effects. Decline of water quality in Chesapeake Bay is indicated by increased phytoplankton biomass (Malone et al., 1988; Harding, 1994), increased sediment loads (Cooper and Brush, 1991), decline of submerged aquatic vegetation (Kemp et al., 1983; Dennison et al., 1993; Stevenson et al., 1993), and loss of commercially valuable fish and

0925-8574/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 5 - 8 5 7 4 ( 9 9 ) 0 0 0 6 0 - 9

338

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

waterfowl due to habitat loss and over-harvesting (Officer et al., 1984; Lubbers et al., 1990). A regional, multi-state program has been created to reduce nutrient inputs to Chesapeake Bay by 40% by the year 2000 (EPA, 1987). While improvements in wastewater treatment and a ban on P-containing detergents has resulted in reduction of point-source loads, non-point source N and P from agriculture continue to be the largest contributor of nutrients to the Bay: 39 and 49% of total N and P nutrient sources, respectively (Chesapeake Bay Program, 1995). Reduction of nonpoint source N and P from agricultural activities will require reduction strategies which are tailored to local environmental conditions, including the three major physiographic regions within the basin (coastal plain, piedmont, ridge and valley). The difficulty of addressing diffuse N and P pollution from agricultural activities is compounded by the long residence time of groundwater. Studies conducted by the US Geological Survey (USGS) (Bohlke and Denver, 1995) on the Delmarva peninsula have demonstrated that current stream discharge originated from groundwater recharge more than 20 years ago. Since large increases in fertilizer application occurred during 1960 –1980 (Bohlke and Denver, 1995), nutrient input to Delmarva streams has increased (Fisher et al., 1998) and is expected to continue to increase during the next two decades. Economic pressure on farmers to provide an inexpensive food supply for the growing Bay population will increase the intensity of agricultural land use, making the 40% nutrient reduction goal signed by the Chesapeake Bay Commission increasingly difficult to achieve by the year 2000. The ability of riparian vegetation, particularly forest, to mediate non-point source pollution in agricultural landscapes has received much attention (Hill, 1996). Riparian forests are those which flank stream banks and usually intercept a significant fraction of N and P moving towards the stream from adjacent uplands. Riparian forests which have saturated soils for extended periods during the growing season are also referred to as riparian wetlands. Riparian forests positively impact water quality by: (1) acting as effective sedi-

ment traps; (2) consuming and storing nutrients by accreting biomass; (3) stimulating microbial assimilation of nutrients in forest soils and (4) providing an environment conducive to anaerobic microbial dissimilation of nitrate to nitrogen gas (denitrification) or ammonia. The ability of riparian vegetation to intercept nutrients and eroded sediment depends not only on the presence of vegetation but also on a microbial energy source, (electron donor), adequate temperatures, redox conditions, groundwater aquifer characteristics, and position of vegetation within the landscape. Reviews by Haycock et al. (1993), Osborne and Kovacic (1993), Gilliam (1994) and Hill (1996) offer excellent discussions on the water quality functions of riparian vegetation. Water quality functions of riparian forests have been studied throughout the Atlantic coastal plain in landscapes on or similar to those of the Delmarva peninsula. A riparian forest studied by Jordan et al. (1993) effectively removed nitrate from an adjacent corn field. The groundwater contained nitrate concentrations of  8 mg l − 1 at the edge of a corn field which was reduced to 0.4 mg l − 1 halfway (20 m) into the forest. A riparian forest studied by Peterjohn and Correll (1984) demonstrated a 90% reduction in annual sediment load, an 80% reduction of nitrate in overland flow and an 85% reduction in groundwater nitrate originating from an adjacent agricultural field. Lowrance et al. (1984) calculated that riparian forest in Georgia’s Little River Watershed retained 68% of N and 30% of P received from adjacent cropland and precipitation. In contrast to the large number of detailed studies on the role of nutrient interception by riparian forests, little attention has been given to the spatial distribution of these forests at the large watershed area. While some studies have stressed the importance of landscape position (Whigham et al., 1988; Brinson, 1993; Haycock et al., 1993), few have included this variable in water quality studies. In a nationwide study conducted by Omernik et al. (1981), positional effects of forest and agriculture did not improve a regression model’s ability to predict stream nutrient levels over the use of land cover’s total extent within a watershed. The lack of forest positional effects on

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

stream nutrients may have resulted from the high variation attributed to regional watershed characteristics which influence hydrologic flow and stream chemistry (soil type, precipitation, geology). Stream nutrient concentrations in the Salt Fork basin, a central Illinois watershed with : 90% of its land in agriculture, was found to be more influenced by proximity of urban areas than by extent or position of cropland (Osborne and Wiley, 1988). However, with only 10% forest in the basin, most of which was in the lower portion, it is not surprising that positional effects were small in this highly disturbed basin. Johnston et al. (1990) determined that proximity of herbaceous wetlands to streams in a nine-county region surrounding Minneapolis significantly influenced nitrate and dissolved P during base flow as well as ammonium, nitrate, and TP during storm flow. In Chesapeake Bay’s Rhode River estuary, Houlahan et al. (1992) showed that an implementation of best management practices on cropland directly adjacent to streams could reduce agricultural nutrient loadings by 90%. We have also attempted to understand the role of forest in retention of agricultural nutrients. Water quality studies were conducted in two adjacent basins dominated by agriculture on the Atlantic coastal plain in the Chesapeake Bay watershed. In addition to the extent of forest cover in these watersheds, this study attempted to characterize the effects of spatial distribution of forest throughout these basins. We hypothesize that riparian forests in coastal plain watersheds effectively remove and/or retain anthropogenic nutrients from agricultural activities. Furthermore, we hypothesize that the spatial distribution of forest within a watershed has an important effect on local stream nutrients. We analyze forest and other landscape variables and their effects on N and P concentrations in streams throughout the Choptank and Chester watersheds.

2. Study site description The Choptank and Chester basins are located on the Atlantic coastal plain on the western side of the Delmarva peninsula within the Chesapeake

339

watershed (Figs. 1 and 2). The basins span the border between the states of Maryland and Delaware, and land use is dominated by agriculture and forest. Most of the agriculture is cropland, with small amounts of pasture (  5%), nurseries ( 2%) and feedlots (  0.5%). The majority of the cropland in these two watersheds is occupied by a corn–soybean rotation (Maryland Agricultural Statistics Service, 1986, 1992), and forest consists of a wide variety of maple, oak and pine species with  40% temporarily or seasonally flooded during the growing season (Cowardin et al., 1979; Tiner and Burke, 1995). The Chester and Choptank River basins have shallow unconfined aquifers. The majority of both watersheds are underlain by the Columbia surficial aquifer with the lower portion of the watersheds underlain by the confined Chesapeake Group aquifer. Recharge to the Columbia aquifer is from local precipitation, and the groundwater flow paths are short, generally B2 km (Bachman and Wilson, 1984; Hamilton et al., 1993). The water table in most of the region lies between 0 and 3 m below the land surface, but in the northcentral portion of the Chester basin, the water table ranges from 3 to 9 m below the surface (Bachman and Wilson, 1984; Hamilton et al., 1993). Surface sediments are mostly Talbot and Wicomico formations that are reworked sands, clays and gravels of the Pleistocene era (Carpenter, 1983; Hamilton et al., 1993). Three soil series comprise approximately half of both watersheds; Sassafras ( 29%), Woodstown ( 7%), and Fallsington ( 12%). These soil series are all formed from the same sandy coastal plain sediments which contain moderate amounts of silt and clay. These series form a catena, a sequence of soils formed from similar parent material and subjected to similar climatic forces but with different physical properties due to their landscape position. Sassafras soils are well-drained and are found on level uplands and rolling hills with slopes ranging from 2 to 15%. Woodstown soils are moderately well-drained and found in flatter areas with slope ranging from 0 to 5%. Fallsington soils are poorly drained and found on upland flats, low-lying depressions, and at the heads of

340

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

streams with slopes ranging from 0 to 2%. All three soils belong to the ultisol soil order which is characterized by their argillic or kandic horizons. Long-term precipitation and temperature patterns are similar within the two basins (Fig. 3). Long-term average annual precipitation for these two watersheds is 110 cm year − 1 and is evenly distributed throughout the year (10 – 15 cm per month). Average daily air temperatures peak in July and August ( 25°C) and are at a minimum during January and February (2°C). In contrast to the similarity of long-term precipitation and

temperature patterns, long-term stream discharge is considerably different between the two watersheds (Fig. 3). The Choptank watershed has a pronounced seasonal maximum in March ( 7 cm per month) and a minimum in Sept. (  1 cm per month), whereas the Chester has a more evenly distributed discharge throughout the year (2–3 cm per month). Furthermore, water yields in the Choptank are 10 cm more per year compared to the Chester (Table 1). These important differences are due to the larger proportion of soils with low runoff potential in the Choptank basin.

Fig. 1. 1990 land use/land cover in the Choptank basin. Data compiled from Maryland Office of Planning (1990) and Delaware Dept. of Agriculture (1987).

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

341

Fig. 2. 1990 land use/land cover in the Chester basin. Data compiled from Maryland Office of Planning (1990) and Delaware Dept. of Agriculture (1987).

3. Methods In a cooperative effort with the US Fish and Wildlife Service, we sampled at 59 non-tidal subwatersheds within these two basins (Fig. 4). Avoiding tidally influenced areas was necessary to estimate local hydrology and restricted the sampling to portions of the basin far from estuarine areas. Monthly base- and stormflow for each subwatershed were estimated using GWLF (Haith and Shoemaker, 1987). The sampled subwatersheds of both the Choptank and Chester basins were sparsely populated and devoid of pointsource nutrient inputs.

Five to six samples per month were manually collected for 16 months in 1985–1986 (Choptank) and for 12 months in 1992–1993 (Chester). For each month, TN, TP and NO− concentrations 3 were averaged to provide a single monthly value associated with the GWLF-estimated stream flow. Nitrate (NO− 3 ) and nitrite were measured colorimetrically by autoanalyzer in filtered (G FF) samples, and TN and TP were measured by persulfate oxidation of whole samples, followed by colorimetric analysis (Valderrama, 1981). Concentrations in both watersheds displayed a log-normal frequency distribution which was confirmed using a chi-square goodness of fit test

342

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

Fig. 3. Seasonal variations in precipitation, stream discharge and air temperature for the Choptank and Chester River basins. Bars represent standard error of the mean.

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

343

Table 1 Landscape and water quality characteristics of the Choptank and Chester basinsa Landscape component

Units

Atmosphere Precipitation N deposition P deposition

cm year−1 kg ha−1 year−1 kg ha−1 year−1

Land Total basin area Land (drainage area) Agricultural input of N N retention

ha ha % basin kg ha−1 year−1 %

Land co6er (1990) Total forest Upland forest Riparian forest Cropland Feedlots/agr. facilities Urban/developed Water

% % % % % % %

Hydrochemistry Ave. annual TN concentration Ave. annual TP concentration Ave. annual NO− 3 concentration Water yields (1946–1992) Ave. TN yield Range TN yield Ave.TP yield Range TP yield Estimated yields for 100% forested system in Chesapeake Bay

N:P ratio (WY 1980–1990) Sewage N Sewage P Geomorphology 1st order stream length

mg l−1 mg l−1 mg l−1 cm year−1 % of rain kg ha−1 year−1 kg ha−1 year−1 kg ha−1 year−1 kg ha−1 year−1 kg ha−1 year−1 kg ha−1 year−1 molar rayio kg person−1 year−1 kg person−1 year−1

2nd order stream length Stream slope Stream density

km % of total length km m km−1 m ha−1

Soil hydrologic classes Low runoff potential Moderately-low runoff potential Moderately-high runoff potential High runoff potential

% % % %

Choptank

111.0 13.0 0.8 205 690 177 661 86 84 85

26.4 15.7 10.7 52.0 0.5 4.6 14.6 3.42 (1.58) 0.058 (0.026) 2.52 (0.59) 39.5 35.5 10.0 3−20 0.2 0.06–0.34 4.64 (TN) 0.16 (TP)

Chester

109.0 13.0 0.8 128 476 114 745 89 84 81

26.1 15.1 11.0 55.8 0.4 3.7 12.2 3.69 (0.12) 0.100 (0.008) 2.87 (0.12) 29.7 27.2 11.0 5–22 0.3 0.08–0.57 4.64 (TN) 0.16 (TP)

65.3 5.1 1.1

58.3 na na

1053.7 45.4 582.5 3.2 14

604.1 48.8 330.1 5.1 9

37.7 18.0 24.5 19.8

3.8 43.2 24.9 28.1

a Precipitation represents multi-station, multi-year averages within and near each basin from NOAA weather stations. N and P deposition are derived from (a) wet DIN and PO4-P deposition at NADP site MD13 at the Wye Research and Education Center of the University of Maryland; (b) DON and DOP data reported for other regions (Haines, 1976; Lewis et al., 1985; Cornell et al., 1995) and (c) dry NH+ 4 and NO− 3 deposition at the Blackwater National Wildlife Refuge NDDN site BWR139. Land area is based on basin boundaries interpreted from topographic maps, and aerial extent based on GIS digital database coverages for each basin. N agricultural inputs are based on estimates developed by Jordan and Weller (1996) and reported by Jordan et al. (1997). Nutrient concentrations represent those sampled throughout the Chester and Choptank basins with standard error of the mean shown in parentheses. N outputs based on measured stream TN concentrations and long-term stream discharge (Fisher, 1992; Victoria, unpublished). TN and TP yields estimated for 100% forested basin were based on those published by EPA (1992). Water yields are from USGS Water Resources Reports. Nutrient concentrations are from our water chemistry sampling and USGS discharge data (Fisher et al., 1998; Victoria, unpublished). Sewage discharge is estimated from Fisher et al. (1998). Soil hydrologic categories are based on SCS county soil surveys.

344

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

Fig. 4. Subwatersheds sampled for water chemistry within the Choptank and Chester basins. USGS stream discharge gauges (shaded) and NOAA precipitation gauges (points) shown.

(P B 0.05). Log-transformed nutrient averages were therefore used in a forward, stepwise multiple regression analysis using the PC-based software SigmaStat. The standard convention used throughout this text to display statistically significant relationships were those associated with error probabilities B0.05 and P \ 0.01 (indicated by a ‘*’), strong significant relationships are those with PB0.01 and P \0.001 (indicated by a ‘**’), and highly significant relationships

are those with PB 0.001 (indicated by ‘***’). The significance of the correlation coefficient was based on (1) the sample size and (2) the statistical test used and (3) alpha. Alpha was the acceptable probability of incorrectly rejecting the null hypothesis (Type I error). The data sets used in this study had large sample sizes (n= 25 for the Chester, n= 34 for the Choptank), and some of the correlation coefficients were low but significant.

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

A vector-based, geographic information system (GIS) database was compiled for the Choptank and Chester watersheds. Each database included layers on streams, land cover, soils, wetlands, and subwatershed boundaries. PC-based ARC/INFO was used to manipulate and manage the GIS data layers. Subwatershed boundaries were super-imposed on data layers in order to tabulate the extent of watershed characteristics. Digital land use/land cover data for the watersheds was interpreted from aerial photography. The Maryland portion of the watersheds was obtained from Maryland Office of Planning (1990) and the Delaware portion from Delaware Dept. of Agriculture (1987). The land use/land cover classification systems, a modified Anderson level II (Anderson et al., 1976), was generalized to a modified Anderson level I within ARC/INFO. The state portions of the land cover data were transformed to a common map projection and datum. The boundaries of adjacent land cover parcels were then edge-matched. Streams were obtained from Maryland State Planning Office and GIS Corp. (Austin, TX, USA). Stream order was visually interpreted and added to the digital database using the methodology described by Strahler (1952). Watershed and subwatershed boundaries were manually interpreted from USGS 7.5 minute topographic map contours. Boundaries were then digitized using ARC/INFO. It was assumed that surface and groundwater flow followed topographic divides. Digital data on wetlands were obtained from US Fish and Wildlife Service’s National Wetland Inventory (NWI) (Cowardin et al., 1979). All palustrine forested wetlands not associated with ponds or lakes were assumed to be riparian and are herein referred to as ‘riparian wetlands’. Forested wetlands were also obtained by intersecting hydric soils (from the soils data layer) with forest (from the land cover data layer). This second wetland distribution is referred to as ‘forest on hydric soils’. Soil hydrologic groups used for this project were determined from the county soil surveys. The SCS hydrologic symbols A, B, C and D represented water transmission rates of \ 0.75, 0.40 –0.75, 0.15–0.40 and 0 – 0.15 cm h − 1, respec-

345

tively. Hydrologic symbols A, B, C and D represent the following soil runoff potentials; low, moderately-low, moderately-high and high, respectively (Soil Conservation Service, 1990). In some counties, soil series were not classified by hydrologic categories. In these instances, infiltration was based on a depth-averaged permeability obtained from the soil survey. The soil survey maps for the Chester and Choptank basins were digitized by Earth Satellite Corporation in Rockville Maryland. The series attributes were entered by Earth Satellite, University of Maryland, and Fish and Wildlife Service personnel. Hydric soil characteristics (Soil Conservation Service, 1991) were compared to those described in individual county soil surveys of the Chester and Choptank watersheds. Soils that met the SCS criteria outlined in the manual were selected as ‘hydric’. Criteria included temperature regime, soil hydrologic class, high water table depth, months of flooding, permeability within top 50 cm, and flood frequency and duration. A ‘hydric’ attribute was added to the GIS soils database for all series that met the proper criteria. Landscape proximity analysis was conducted by delineating concentric zones around the streams within each subwatershed (Fig. 5). Zone widths 0–100, 100–300, 300–500, and \ 500 m from streams were selected based on the resolution of the vector data and the size of the subwatersheds. The analysis used for generating basin zones and calculation of the amount of land cover within each zone was similar to that used by Omernik et al. (1981) and Osborne and Wiley (1988). Spatial distribution of forest was quantified in three different ways. First, the total extent of forest within each subwatershed was estimated by summing the area of total forest, riparian forest, and forest on hydric soils within the subwatershed boundary. Second, the amount of total forest within each basin zone was calculated as percent cover. Third, the total amount of forested stream length was estimated by intersecting the stream vectors with the land cover polygons. Forested stream length was calculated as the percent of total stream length that intersected forest land use/cover. Where streams and/or rivers were rep-

346

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

Fig. 5. Example of how basin zones were constructed.

resented by two lines (one for each bank), land use/land cover on each bank was averaged by the total stream length of both banks All forest area estimates were conducted using GIS analyses. Long-term data on precipitation and air temperature for the two basins were obtained from four National Oceanic and Atmospheric Administration (NOAA) climatological stations. Longterm stream discharge information was obtained from USGS gauging stations at Morgan Creek for the Chester and at Greensboro for the Choptank. Names and locations of the stations used for long-term estimates are given in Table 2.

4. Results

4.1. Climate A comparison of air temperature, precipitation and stream discharge during the sampling periods and long-term averages was made in order to determine if sampling conditions reflected ‘typical’ conditions. Monthly air temperatures for the Chester and Choptank watersheds during the sampling periods followed long-term trends with little deviation (Fig. 3). During the sampling period in the Choptank, February was significantly

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

347

Table 2 NOAA climatological and USGS stream discharge data used to estimate long-term air temperature, precipitation and stream discharge for the Choptank and Chester basinsa Climatic parameter

Choptank Station, lat., long.

Chester Station, lat., long.

Hydrologic parameter

Choptank (cm year−1)

Chester (cm year−1)

Temp., precip.

Greenwood 38° 50% 00%%, 75° 35% 00%% Cambridge 38° 34% 00%%, 76° 04% 00%% Denton 38° 53% 00%%, 75° 48% 00%% Royal Oak 38o 43% 00%%, 76o 11% 00%% Greensboro 38° 59% 50%%, 75° 47% 09%%

Millington 39° 16% 00%%, 75° 52% 00%% Chestertown 39° 13% 00%%, 76° 04% 00%% Dover 39° 09%00%%, 75° 31% 00%% Royal Oak 38° 43% 00%%, 76° 11% 00%% Morgan Creek 39° 16% 48%%, 76° 00% 54%%

Long-term precip.

111.6 (1.4)

110.8 (1.12)

Sample precip.

100.4 (4.5)

108.5 (3.8)

Long-term discharge

39.6 (1.8)

29.7 (0.5)

Long-term stormflow

10.7(7.0)

12.2

Long-term baseflow

27.7 (9.5)

17.5 (5.0)

Sample period discharge (gauge) Modeled discharge for basin during sample period

30.9 (2.0)

27.1 (1.6)

30.9 (2.0)

30.3 (0.8)

Temp., precip. Temp., precip. Temp., precip. Discharge

a Standard deviation is given in parentheses. Precipitation and discharge values represent annual totals. Long-term stormflow and baseflow for Choptank from Lee (2000) based on an average of 10 water years (1980–1990). Long-term baseflow for Chester from Bohlke and Denver (1995). Stormflow for Chester by subtraction of long-term base flow from long-term average annual total. Average modeled discharge for Choptank based on water yields from Greensboro gauging station. Average modeled discharge for Chester based on GWLF (Haith and Shoemaker, 1987).

cooler and March and November were significantly warmer compared to long-term air temperatures (Fig. 3). In the Chester, air temperatures during sampling were significantly warmer in January and cooler in February, March, August and October compared to long-term averages (Fig. 3). While these departures are statistically significant, they are relatively small and the trends in the sampling period air temperatures reflect long-term averages. There were many seasonal deviations in both basins. In the Choptank, the sample period precipitation was lower in the spring and much higher in September compared to long-term averages (Fig. 3). In the Chester, sample period precipitation was much higher in the Spring and lower in the late fall compared to long-term averages (Fig. 3). While the Choptank watershed experienced a slightly drier year (100 cm year − 1) compared to the long-term average total (111 cm year − 1), total annual precipitation in the Chester watershed (108 cm year − 1) was similar to longterm totals (Fig. 3, Table 3). Total annual precip-

itation during the study periods for the Chester and Choptank represented 95 and 90% of the long-term average, respectively. Monthly discharge patterns followed those of precipitation but with smaller deviations from long-term averages (Fig. 3). Discharge during the sampling period at the Morgan Creek gauge was 91% of the long term total, while discharge at Greensboro was only 78% of the long term total. Small short-falls in precipitation during the study periods resulted in much larger short-falls in discharge for the Greensboro gauging station compared to Morgan Creek. Discharge at Greensboro during the sampling period was greater during September, October and February and less during the remaining months compared to long-term averages. During the sampling period, discharge at Morgan Creek was higher during March and April, and slightly less than the long-term average for the remainder of the year (Fig. 3). Long-term total annual discharge was significantly lower (29 cm) in Morgan Creek compared to that in Greensboro (39 cm) (Table 3).

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

348

Table 3 Summary of precipitation and discharge for Choptank and Chester basinsa Hydrologic parameter

Choptank (cm year−1)

Chester (cm year−1)

Long-term precip. Sample precip. Long-term annual discharge Long-term stormflow Long-term baseflow Discharge at gauge during sampling Ave. modeled discharge for study area during sampling Discharge/precip. (%) long-term Discharge/precip. (%) sample

111.63 (1.43) 100.39 (4.50) 39.59 (1.77)

110.79 (1.12) 108.50 (3.75) 29.69 (0.52)

10.7 (7.0)

12.2

27.7 (9.5) 30.85 (1.97)

17.5 (5) 27.12 (1.57)

30.85 (1.97)

30.32 (0.77)

35.47

30.73

26.80

28.32

a Precipitation and discharge values represent annual totals. Standard deviation given in parentheses. Long-term stormflow and baseflow for Choptank from Lee (2000) based on average of 10 water years (1980–1990). Long-term stormflow and baseflow for Chester from Bohlke and Denver (1995). Stormflow for Chester by subtraction of long-term base flow from long-term ave. annual total. Average modeled discharge for Choptank based on water yields from Greensboro gauging station. Average modeled discharge for Chester based on GWLF hydrologic model for each subwatershed.

4.2. Stream chemistry Average monthly stream nutrient concentrations for both the Choptank and Chester subwatersheds were highly variable throughout the year (Fig. 6). Less variability was demonstrated by stream TN concentrations compared to stream concentrations (Fig. 6), and the monthly NO− 3 variation was similar to that observed by Fisher et al. (1998) in a more detailed data set for the Choptank basin which included 10 years of sampling. Nitrate comprised most of the TN; 74% in the Choptank and 78% in the Chester. No significant difference between average annual TN and NO− 3 concentrations in the Choptank and Chester was observed (Table 1). Average TP concentrations were significantly higher (**) in Chester

Fig. 6. Average monthly stream N and P concentrations for the Chester subwatershed 20B and Choptank subwatershed 29. Bars represent standard error of the mean.

streams; almost double what was measured in the Choptank basin (Table 1). N and P export coefficients (kg ha − 1 year − 1) were estimated from the monthly stream flows and concentration data. While N export rates (3− 22 kg ha − 1 year − 1) were comparable with those calculated for other coastal plain watersheds dominated by agriculture (Beaulac and Reckhow, 1982), P export rates (0.06–0.57 kg ha − 1 year − 1)

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362 Table 4 Simplified land use/land cover characteristicsa

4.3. Land co6er extent

Land cover

Chester study area (%)

Choptank study area (%)

Cropland Developed/urban Feedlots Total forest Upland forest Forested wetlands Temp. flooded (NWI)* Seasonally flooded (NWI)* Water*

65.3 1.9 0.5 31.9 19.7 12.2 6.8

62.9 2.8 0.6 33.6 19.3 14.3 9.2

(2.7) (0.4) (0.1) (2.7) (1.8) (1.1) (0.7)

349

(2.0) (0.4) (0.1) (2.0) (1.4) (1.0) (0.7)

4.2 (0.5)

2.7 (0.3)

0.4 (0.2)

0.0 (0.1)

a

Water is present in the Choptank study area, but occupies less than one tenth the size of the sampled area. Cropland contains crops as well as small amounts of pasture and orchards. Standard error of the mean is shown in parentheses.

were relatively low (Table 1). Since most of the stream P loading occurs during only a few storm events (Edwards and Owens, 1991; Pionke et al., 1996; Fisher et al., 1998), the Choptank and Chester were probably under-sampled with respect to TP. In addition, soils in these watersheds have a high mineral content and P-adsorption capability (Walbridge and Struthers, 1993).

The extent and types of land cover found in the sampled areas of the two watersheds were very similar (Table 4, Figs. 1 and 2). No significant differences were found between the land cover categories except for the larger amount of water (*) in the Chester. Cropland and forest together occupy : 96% of the sampled subwatersheds (Table 4), resulting in a high negative correlation between the two land cover types (r 2 = 0.97). Riparian forested wetlands comprise 40% of the total forest found within the sampled areas (Table 4), and most of the forested riparian wetland ( 62%) is temporarily flooded for brief periods during the growing season with a water table which lies well below the ground surface for much of the year. Only a small amount of the riparian wetland is seasonally flooded ( 27%) for extended periods during the growing season with a water table close to the ground surface during the rest of the year (Cowardin et al., 1979). In both basins, no significant relationships were identified between stream TP concentrations and land cover variables. However, the models for predicting stream N concentrations based on land cover for the Chester and Choptank basins exhibited very different results (Table 5). In the

Table 5 Land cover regression models for predicting stream nutrientsa Nutrient Chester ln TN ln TP ln NO− 3 Choptank ln TN ln TP ln NO− 3

a

Model

Formula

Land Land Land Land Land Land

cover cover buffer cover cover buffer cover cover buffer

NSV y= −0.293+0.022*(cropland 100–300 m) NSV NSV y= 1.690–0.045*(upland forest) y= −1.640+0.036*(cropland 100–300 m)

Land Land Land Land Land Land

cover cover buffer cover cover buffer cover cover buffer

y= −1.006+0.031*(cropland)+0.232*(feedlots) y= 0.172+0.021*(cropland 100–300 m)−0.012*(forest \500 m) NSV y= −3.326+0.027*(forest 0−100 m)−0.020*(forest 100–300 m) y= −1.867+0.039*(cropland)+0.259 * (feedlots) y= −0.362+0.025*(cropland 100–300 m)−0.015*(forest \500 m)

r2

0.27*

0.27* 0.31* 0.73*** 0.76*** 0.41** 0.74*** 0.74***

The dependent variable (y) is indicated on the left hand side of the table under ‘Nutrient’. No significant relationships were found for developed land cover, unforested wetlands or water as well as for buffers not listed below. NSV represents ‘no significant variable’

350

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

Fig. 7. Relationship between forest and stream nitrate in the Choptank and Chester watersheds. Riparian forested wetlands are based on the distribution of the National Wetlands Inventory (NWI) palustrine forested wetlands. Forest on hydric soil is based on the intersection of total forest land cover and potentially hydric soils. Upland forest is that which is not considered a wetland by NWI definitions. Non-significant statistical relationships noted on graphs by ‘NS’.

Choptank, cropland was highly correlated with annual average stream TN and NO− (r 2 = 3 0.69*** and 0.70***, respectively). Regression models for predicting N in Choptank streams which included both cropland and feedlot variables significantly increased the models’ prediction ability by 4% (Table 5) beyond that o f using cropland alone. Riparian forest, forest on hydric soil, and upland forest all showed strong negative correlations with stream TN and NO− 3 concentrations (Fig. 7) in the Choptank basin. Thus, stream TN and NO− concentrations in the Choptank 3 basin were strongly related to both forest and cropland in the surrounding area. In contrast, land cover was a poor predictor of stream N concentrations in the Chester basin (Table 5). Regression analysis of the Chester land cover variables yielded only a weak relationship between stream NO− 3 concentrations and upland forest (r 2 =0.27*). Furthermore, total forest, riparian forested wetland and forest on hydric soil

was not significantly correlated with stream TN concentrations in the Chester basin and NO− 3 (Fig. 7). We found these results to be surprisingly different since the Chester and Choptank basins were geographically contiguous and had similar amounts and types of land cover.

4.4. Land co6er distributions Since forest type and total extent did not differ significantly between the two watersheds, a landscape analysis was conducted in order to examine the potential differences in the spatial distribution of forest within basin zones around streams. The amount of total forest in each basin zone is shown in Table 6. The only significant difference between the two watersheds was in the zone directly adjacent to streams (0–100 m). The Chester basin contained significantly more forest (52%*) and less cropland (44%*) compared to the Choptank (37% forest* and 60% cropland*). There was no significant differences between feedlots in any

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362 Table 6 Position and extent of land cover in relation to streams for sampled subwatershedsa Zone (m from stream)

Forest (%)

Cropland (%) Feedlots (%)

Chester 0–100 100–300 300–500 \500

52.2 29.0 27.3 31.6

(9.7)b (12.1) (15.8) (20.5)

44.2 67.5 69.6 64.3

(10.6)c (11.7) (15.6) (20.4)

0.5 0.7 0.5 0.5

(1.0) (1.6) (0.7) (0.6)

Choptank 0–100 100–300 300–500 \500

36.8 30.1 33.4 37.6

(11.2) (12.5) (16.2) (19.0)

59.8 66.3 63.1 59.0

(11.4) (12.7) (15.7) (18.6)

0.8 0.7 0.5 0.3

(1.3) (0.8) (0.8) (0.6)

a Land cover is presented as % cover within basin zone. Standard error of the mean is shown in parentheses. b Significantly greater (*) than in corresponding buffer in Choptank and in all other Chester buffers. c Significantly less (*) than in corresponding buffer in Choptank and in all other Chester buffers.

basin zone nor in any landuse \100 m from the streams. The lack of a relationship between forest and stream N concentration in the Chester was not due to a lack of forest in close proximity to streams. The Choptank had approximately equal proportions of the total forest distributed through the watershed. The Chester, however, had more forest directly adjacent to streams and significantly less in the other three basin zones (Table 6). If anything, this spatial distribution would be expected to lower TN and NO− 3 stream concentrations, in contrast to the similar concentrations observed in these basins. Thus, landscape position analysis did indicate that there is a difference between the distribution of forest in these two basins; however, this difference did not add to our understanding of forest water quality effects. A multiple regression analysis was run on the landscape proximity variables and stream nutrient concentration averages (Table 5). Cropland in the second zone (100 – 300 m) significantly contributed to stream N concentrations in both watersheds. In the Chester, the land cover distribution significantly increased the model’s TN

351

prediction ability by 6% compared to the upland forest model, and the land cover distribution significantly increased the model’s NO− 3 prediction ability by 4% (Table 5). While the land cover distribution variables did not explain much more of the stream N variance compared to the total extent of upland forest in the Chester basin, these models did indicate that a small portion of the subwatersheds ( 30% of the area) in the Chester basin explained about 30% of the stream N concentration variance (Table 5). The landscape regression model significantly increased the correlation between Choptank stream TN concentration and land cover by 3%. Although this is again a small increase, the landscape model was useful in identifying those portions of the watershed which significantly influenced stream nutrient concentrations. In addition to cropland 100–300 m from streams, forest far from local streams ( \500 m) was important in reducing TN and NO3– concentrations (Table 5). A much larger portion of the Choptank watershed (50–85%) influenced local stream concentrations compared to that in the Chester. Finally, stream TP was found to be positively correlated with forest next to streams (0–100 m) and negatively correlated with forest 100–300 m from Choptank streams. This result indicated that 30% of the Choptank subwatershed area increased stream TP chemistry and 30% decreased concentrations (Table 5).

4.5. Effect of stream order The landscape distribution analysis explained more of the stream N variance in the Chester compared to the total extent of land cover. However, the relationship between forest and stream nutrients in the this basin was not as strong as those observed in the Choptank (Fig. 7). In an attempt to better understand these differences, an analysis of forest distribution with respect to stream order was conducted. The landscape analysis examined the lateral position of forest in relation to local streams, while the stream order analysis determined where the forest was longitudinally along the hydrologic gradient.

352

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

Fig. 8. Riparian forested stream banks and stream N concentrations in the Choptank River basin. Bars represent standard error of the mean. There were no parallel significant relationships in the Chester. Non-significant statistical relationships noted on graphs by ‘NS’.

The Chester contained more total forested stream length (62%***) and more forested first order streams (55%***) compared to the Choptank (40 and 33%, respectively, Fig. 8). Despite the presence of more forest along streams in the Chester, stream N concentrations were similar to those in the Choptank. Furthermore, forested stream length displayed no significant relationship to stream TN and NO− concentrations in the 3 Chester basin. Total amount of forested stream length was not significantly correlated to stream TN and NO− 3 in the Choptank; however, forested first order streams were associated with low 2 stream TN and NO− 3 (r =0.33 and 0.38, respectively) (Fig. 8).

4.6. NWI flood regime Most ( 95%) of the NWI riparian wetlands were classified as temporarily (a) or seasonally flooded (c) (Cowardin et al., 1979). No significant relationship was found between the flood regime (a or c) and stream N concentrations in the Chester samples. In contrast, the ‘wetter’ seasonally flooded forest (c) had a higher correlation with stream TN and NO− (r 2 = 0.38** and 3 0.43**, respectively) compared to the temporarily flooded forest (r 2 = 0.27* and 0.24*, respectively, Fig. 9). Thus, the ‘wetter’ riparian forests in the Choptank were associated with lower stream N concentrations while the degree of saturation of

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

riparian forest in the Chester had no significant effect on stream nutrients.

4.7. Soil properties Landscape patterns do not explain the different effects forest has on stream nutrients in the Choptank and Chester basins. In an attempt to find other watershed variables which might be important in influencing stream N and P, several soil properties were examined. The amount of surface texture, subsurface texture (particle size) and soil hydrologic categories were measured in each sampled subwatershed. Soil hydrologic characteristics reported on Table 1 reflect the entire Chester and Choptank basins while those on Table 7 reflect the subwatersheds sampled for stream chemistry.

353

The Chester and Choptank soils had very similar particle size compositions (Table 7). The only significant difference between the two basins was the slightly higher amount of coarse-loamy soil in the Choptank (19%*) compared to that in the Chester (8%*). This probably contributed to the better-drained nature of the Choptank compared to the Chester. Most soils in both watersheds had a fine- loamy particle size ( 70%). The similarity of particle sizes in these two watersheds was attributed to the similarity of the parent material in the control section which lay in or near the argillic horizon. None of the eight particle size classes were significantly correlated with stream N and P concentrations for either watershed. Soils found in the Chester subwatersheds had lighter, siltier surface textures compared to the

Fig. 9. Riparian wetland (NWI) flood regime and stream N concentrations in the Choptank basin. Bars represent standard error of the mean. There were no parallel significant relationships in the Chester.

354

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

Fig. 10. Soil with low runoff potential (A+ B)% and stream nutrient concentrations in the Choptank and Chester basins. Bars represent standard error of the mean. The alternate pathways N and P take through the watershed may result in P limitation in watersheds with low runoff potentials and N limitation in subwatersheds with high runoff potentials.

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362 Table 7 Average soil parameters for sampled subwatershedsa Soil parameter

Chester (%)

Choptank (%)

Surface texture Loamy sand* Gravely loam* Sandy loam* Fine sandy loam* Loam* Silt loam*

1.6 2.4 31.6 1.1 31.6 31.4

(0.3) (1.1) (3.7) (0.5) (3.7) (5.4)

5.0 0.0 44.2 0.0 41.4 9.1

(0.6) (0.0) (2.8) (0.0) (2.3) (1.9)

Particle size Sandy Coarse-loamy* Loamy Fine-loamy Fine-silty Clayey

0.7 8.4 2.2 74.4 7.4 6.9

(0.2) (1.7) (0.5) (2.1) (1.5) (1.4)

0.7 18.6 1.8 68.1 5.7 5.0

(0.2) (3.2) (0.3) (3.6) (2.5) (1.1)

Hydrologic group A* B* C D

1.4 48.2 18.0 32.4

(0.3) (3.4) (3.6) (3.9)

28.9 18.2 17.4 35.5

(4.0) (1.6) (1.2) (4.4)

Hydric* 30.6 (3.3) a

45.5 (3.1)

Standard error of the mean is shown in parentheses.

sandier surface textures of the Choptank soils (Table 7). This texture reflects that of the surface, or mineral layer of the soil. Gravely loam soil texture is only represented by the north central Chester subwatersheds and is completely lacking in any of the Choptank watersheds. The general surface texture of Chester soils was represented by sandy loam, loam, and silt loam, while that of the Choptank was largely characterized by sandy loam and loam textures. Pair-wise correlation analysis between all individual surface textures and stream nutrients resulted in only three significant associations. There was a negative correlation between sandy loam and TP concentration in the Chester (r 2 =0.20*) and a negative correlation between loam and stream TN and NO− 3 concentrations in the Choptank (r 2 =0.41** and 0.37**, respectively). The negative association between fine-textured loam soils and stream nitrogen concentrations in the

355

Choptank emphasized the impedance of fine soil textures to infiltration and leaching. In the Chester, greater amounts of sandy loam emphasized greater infiltration and thus, less potential soil erosion and transport of P from surface soils to streams. Soil hydrologic characteristics contrasted markedly between the two watersheds. The major difference in hydrologic categories between the Chester and Choptank watersheds was the amount of soils with low (A) and moderately low (B) runoff potentials. The Choptank study area contains considerably more (*) soil with low runoff potential compared to that in the Chester, while the Chester has more (*) soil with moderately-low runoff potential compared to that in the Choptank (Table 7). The standard deviation for all hydrologic classes was high, which reflected the high variation found between individual subwatersheds. There were significant effects of hydrologic class on stream N and P concentrations. The four hydrologic classes were combined into two categories (A +B and C + D), and were found to explain the most variation in multiple regression analysis for predicting stream nutrient concentrations (Fig. 10). Correlation coefficients determined for soils with low runoff potential and nutrients in the Chester (r 2 = 0.40** for TN and NO− 3 , 0.24* for TP) were much higher compared to those computed for forest (upland forest and 2 NO− 3 , r = 0.27*). The correlation coefficients computed for Choptank soils with low runoff potential and stream N (r 2 = 0.63** for TN and 0.58** for NO− 3 ) were lower than those calculated from forest (r 2 = 0.75*** for TN and 0.71*** for NO− 3 and forest on hydric soil).

4.8. Forest 6ariables Several of the forest variables presented here were inter-correlated. The more equal distribution of forest in the Choptank basin resulted in high correlations between forested first order stream length, total forest, and upland forest (Table 8). Forest on hydric soil was more closely related to seasonally flooded wetlands (r 2 = 0.59) than temporarily flooded wetlands (r 2 = 0.23). The amount

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

356

Table 8 Correlation coefficients (r 2) for forest variables in the Choptank and Chester basinsa wetfor

Uplfor

totfor

Forbuf

for1buf

tempfld

seasfld

hydfor

for01m

for13m

Choptank wetfor uplfor totfor forbuf for1buf tempfld seasfld hydfor for01m for13m

1.00 – – – – – – – – –

0.20 1.00 – – – – – – – –

0.62 0.81 1.00 – – – – – – –

0.00 0.43 0.23 1.00 – – – – – –

0.27 0.71 0.69 0.59 1.00 – – – – –

0.73 0.08 0.37 0.02 0.19 1.00 – – – –

0.53 0.07 0.28 0.03 0.07 0.13 1.00 – – –

0.66 0.37 0.66 0.01 0.33 0.23 0.59 1.00 – –

0.07 0.59 0.43 0.91 0.73 0.06 0.00 0.11 1.00 –

0.62 0.65 0.87 0.19 0.65 0.36 0.24 0.69 0.41 1.00

Chester wetfor uplfor totfor forbuf for1buf tempfld seasfld hydfor for01m for13m

1.00 – – – – – – – – –

0.40 1.00 – – – – – – – –

0.73 0.89 1.00 – – – – – – –

0.05 0.09 0.01 1.00 – – – – – –

0.02 0.20 0.06 0.74 1.00 – – – – –

0.66 0.32 0.53 0.00 0.00 1.00 – – – –

0.60 0.22 0.41 0.05 0.02 0.21 1.00 – – –

0.67 0.34 0.55 0.02 0.00 0.62 0.51 1.00 – –

0.00 0.23 0.11 0.70 0.73 0.01 0.01 0.02 1.00 –

0.48 0.61 0.81 0.05 0.19 0.38 0.27 0.42 0.19 1.00

a

The following abbreviations are used to represent forest variables: wetfor, NWI forested wetlands; uplfor, upland forest; totfor, total forest; forbuf, forested stream length; for1buf, forested first order stream length; tempfld, temporarily flooded wetlands (a), seasfld, seasonally flooded wetlands (c), hydfor, forest on hydric soil, for01m, forest extent in 0–100 m zone, for13m, forest extent in 100–300 m zone. All correlation coefficients were based on %cover for each sampled subwatershed.

of forest in the 0–100 m zone was highly correlated to the amount of total forested stream length (r 2 = 0.91). In the Chester basin, more of the forest was located in close proximity to local streams. Thus, high correlations existed between forested first order stream length and total forested stream length but not with total forest and/or upland forest (Table 8). Forest on hydric soil was more equally related to temporarily flooded (r 2 =0.62) and seasonally flooded (r 2 =0.51) wetlands compared to those in the Choptank. The amount of forest in the 0–100 m zone was related to the amount of forested first order streams (r 2 =0.73) while the amount of forest in the 100 – 300 m zone is related to upland forest (r 2 =0.81). Thus, the identification of individual predictor variables in multiple regression analysis does not exclude the importance of other basin variables in explaining stream N and P concentrations.

5. Discussion Many studies of coastal plain riparian forests have demonstrated N and P retention/removal capabilities (Hill, 1996). However, the ability of these forests to retain nutrients from agricultural uplands requires an efficient delivery of nutrients. When coastal plain watersheds are characterized by sandy soils, there appears to be a strong relationship between riparian forest cover and low stream nutrient export (Peterjohn and Correll, 1984; Jordan et al., 1993; Gilliam, 1994). However, a watershed characterized by fine-textured soil (the Chester) did not appear to efficiently transport nutrients originating from agricultural activities in uplands to riparian forests. In this watershed, we could not demonstrate any association between low N and P and the presence of riparian forest. Ritter (1986) demonstrated that

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

357

Fig. 11. A and B: Hydrologic budget for the Choptank (A) and Chester (B) basins. Values are given in cm year − 1 based on long-term averages from National Oceanic and Atmospheric Administration (NOAA) weather stations and US Geological Survey (USGS) gauging stations. Percentages are based on proportion of total annual precipitation. The biggest difference between the two basins lies in baseflow and evapotranspiration. The finer-textured soils in the Chester result in less infiltration, less baseflow and more evapotranspiration compared to the Choptank basin which has coarser-textured soils. Losses to deeper aquifers are probably small due to the continuous aquiclude but are unknown. Evapotranspiration was computed by difference, assuming no deep recharge. Average stream slope was computed using USGS topographic maps with the procedure described by Carpenter (1983). Unsaturated zone depths were visually estimated using ground elevations from USGS topographic maps and surficial aquifer depths as mapped in Bachman and Wilson (1984). Drawings not to scale and represent the authors’ conception of possible conditions in these basins.

358

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

stream nutrient export was more highly correlated with hydrologic characteristics than with land use and livestock density in southern Delaware watersheds, some of which were within the Choptank and Chester basins. Sonzogni et al. (1980) stated that soil texture was the single most important watershed characteristic which effected water quality of the Great Lakes. We have shown that the soil characteristics can completely over-ride land cover effects. Although the Choptank and Chester watersheds had very similar land cover extent and spatial patterns (Figs. 1 and 2), they differed in their topographic and soil characteristics (Table 7) which undoubtedly influenced the hydrologic flow path of nutrients from cropland to streams. We ignored differences in forest and crop species which may effect basin N and P export and assumed these plants were homogeneously distributed throughout each sampled basin zone. Analysis of crop production estimates for counties within the Chester and Choptank basins (Norton, 1998) and visual inspection of the forests throughout these basins (data not shown) revealed similar crop as well as upland and wetland forest species. Thus, the differences in N and P export from uplands to streams caused by differences in forest and/or crop species appeared to be minimal. Differences in surface soil texture and soil hydrologic class may be the key to understanding the different effect of forest on stream N and P concentrations in these basins. Soils with moderately-high or high runoff potential in the Chester basin clearly damped the long-term annual hydrograph (Fig. 3), as well as that of individual storms (data not shown). Furthermore, water yields were 25% lower in the Chester (29 cm year − 1) compared to the Choptank (39 cm year − 1) despite similar precipitation totals (Table 3) This suggested that a much larger fraction of rainfall was lost via evapotranspiration in the Chester and did not contribute to discharge. While subsurface texture in the Choptank and Chester watersheds was found to be similar (fineloamy particle size), surface soil textures and hydrologic characteristics differed greatly (Table 7). The coarse textured, low runoff potential soils of the Choptank may have provided a direct path-

way between nitrogen inputs (fertilizers) at the soil surface and the riparian forest. The fine textured, high runoff potential soils of the Chester do not provide this link, possibly due to the greater amount of overland flow and evapotranspiration in this basin as a result of these soil properties (Table 3, Fig. 11). Other investigations have indicated that the hydrologic flow path differs between these two watersheds (Fig. 11, Bohlke and Denver, 1995; Fisher et al., 1998; Lee, 2000). The Chester appears to have a deeper unsaturated zone; the result of steeper topography, finer surface textures and less baseflow compared to that in the Choptank basin. If stormflow comprises a large amount of total stream discharge ( 40%) in the Chester, then N reduction via plant uptake and denitrification is minimized and the potential for transport of P via eroded sediment is increased. Riparian forests may have minimal impact on reduction of N in overland runoff (Verchot et al., 1997) and long-term sediment-bound P (Whigham et al., 1988). The larger component of stormflow as well as higher stream slopes may explain the higher stream TP concentrations in streams of the Chester as well as the inability of forest to effect stream N concentrations (Fig. 11). The basin zone analysis indicated that riparian forest in the Choptank may act as both a source and sink of TP. Forest directly adjacent to streams (0–100 m) was positively correlated to stream TP and forest in the 100–300 m zone was negatively correlated to stream TP in a multiple regression model (r 2 = 0.41**). Forest in the 100– 300 m zone with high redox conditions may have trapped and retained sediment-bound P while the riparian forest zone (0–100 m) with potentially low redox conditions may have contributed dissolved P from sediment as well as from the forest itself (organic P). Others have observed that forest acts not only as a sink for particulate-bound P but a source of dissolved, organic P (Peterjohn and Correll, 1984; Cooper et al., 1995) The riparian zone may have also contributed dissolved, inorganic P from trapped sediment if wet conditions in the forest created low redox conditions (Whigham et al., 1988).

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

The relatively large amount of soil with high runoff potential and hydric conditions present within these two basins was not strongly related to stream TP concentrations. Apparently, the low redox conditions required for soils to become hydric did not encourage a P flux from these watersheds to local streams. Since the spatial distribution of hydric soils was highly variable in both watersheds (data not shown), P de-sorbed from soil in low redox conditions may have been re-adsorbed in adjacent pockets of well- drained soil. The importance of the spatial distribution of soil properties with respect to TP export from cropland needs more detailed field-based observations to validate this hypothesis. Despite the significantly higher stream TP concentrations observed in Chester streams compared to those in the Choptank, both basins exhibited low TP export relative to other Atlantic coastal plain watersheds (Beaulac and Reckhow, 1982; Jordan et al., 1997). Soils on Maryland’s eastern shore have demonstrated a high potential for P adsorption (Richardson, 1985; Walbridge and Struthers, 1993). Although groundwater pH was not measured in this study, Jordan et al. (1993) measured a pH range of 5 – 7 in a small watershed within the Chester basin. Interception of P from cropland under acidic conditions may occur by adsorption of phosphate by Fe and Al oxides and clay minerals at low phosphate concentrations and by precipitation of insoluble Fe and Al phosphates at higher concentrations (Anderson, 1981; Stevenson, 1986). More stream P sampling needs to be conducted, particularly during storm events, in order to compare P export from the Choptank and Chester basins with others in the Chesapeake Bay and coastal plain regions. Riparian forest may have minimal impacts on reduction of N in overland runoff (Verchot et al., 1997) and long-term sediment-bound P (Whigham et al., 1988). However, restoration and conservation of coastal plain forest should be encouraged for several reasons. Forest provides valuable habitat for wildlife, stabilizes stream banks, lessens the impact of storm discharge associated with flooding and provides recreational and scenic resources. An additional benefit is that land which contains forest is not occupied by cropland, and is

359

not being fertilized. While forest in the Chester may not have the same water quality functions as those observed in the Choptank, it does not exclude them from having value.

6. Conclusions Forest had a different effect on stream nutrients in the Choptank and Chester coastal plain basins. In the Choptank, forest cover was strongly associconcentrations. ated with low TN and NO− 3 Within first order streams, the conduits of water from terrestrial to aquatic systems, the presence of forested stream banks also had a strong relationship with low stream N. In addition, the amount of riparian wetlands and degree of ‘wetness’ was inversely correlated with stream N in the Choptank basin. In contrast, forest cover in the Chester basin did not have a strong impact on stream nutrients regardless of landscape position and/or flooding regime. Hydrologic characteristics, rather than land cover, had the strongest effect on predicting Chester stream nutrient concentrations. Stream TP concentrations in both watersheds were relatively low compared to other basins. A regression model which explained 40% of the Choptank TP variance indicates that forest adjacent to streams (0–100 m) acted as a TP source while forest 100–300 m from streams acted as a TP sink. These effects may be a result of lower redox conditions closer to streams and/or contributions of dissolved organic-P by riparian forest. Although stream TP concentrations were higher in the Chester basin compared to the Choptank, no significant relationships with land cover or land cover distributions were found. The influence that forest has on stream N and P must be viewed in context of a wide variety of basin characteristics. While stream N and P in the Choptank were strongly linked to land cover, stream nutrients in the Chester were more closely associated with soil properties. Analysis of soil texture and hydrologic properties indicates that the Choptank had coarser-textured surface soils and was better drained compared to soil in the Chester. The fine-textured soils with higher runoff potentials in the Chester appeared to limit the

360

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

transport of leached N from cropland to riparian forest. In addition to soil properties, higher stream slopes resulted in higher soil erosion potential and thus, P transport potential in overland flow. Results of this investigation reinforce the need for integration of hydrological, geological, land cover and soil properties. To understand the anthropogenic effects on water quality, we needed to consider all four properties. The Choptank and Chester River watersheds are geographically contiguous, have similar land cover characteristics, and receive similar amounts of precipitation. However, differences in soil surface texture, hydrologic characteristics, and stream discharge indicated that the path N and P take through the landscape differ. In the Choptank, N appeared to be primarily transported via groundwater flow and the low overland flow resulted in low stream P concentrations. In the Chester, with less groundwater flow, there appeared to be more P mobilized and N apparently by-passed riparian and other landscape sinks. Riparian forest has tremendous value as ecological habitat and for recreation, but its ability to improve water quality is a function of its interaction with the hydrologic flow path.

Acknowledgements This work was supported under US Fish and Wildlife Service cooperative agreement 14-48-000592-9036 with University of Maryland, Center for Environmental Science. Computer support was provided by the US Fish and Wildlife Service, Chesapeake Bay Field Office. Special thanks are due to Chris Victoria and Anne Gustafson for their efforts in stream sampling and GIS database compilation. We thank Ray Weil, Dick Weismiller, Karen Prestagaard and Mike Kearney for ideas and suggestions. We also thank three anonymous reviewers for their comments and suggestions concerning this manuscript.

References Anderson, J.R., Hardy, E.E., Roach, J.T., Witmer, R.E., 1976. A land use and land cover classification system for use with

remote sensor data. U.S. Geological Survey Professional Paper 964. U.S. Government Printing Office, Washington, D.C. Anderson, M.A., 1981. Expectations and limitations for aqueous adsorption chemistry. . In: Anderson, M.A., Rubin, A.J. (Eds.) Adsorption of Inorganics at Solid – Liquid Interfaces. Ann Arbor Scientific, Ann Arbor, MI, USA, pp. 327 – 349. Bachman, J., Wilson, J.M., 1984. The Columbia Aquifer of the Eastern Shore of Maryland. Report of investigations no. 40. Maryland Geological Survey. Dept. of Natural Resources. Baltimore, MD. Beaulac, M.N., Reckhow, K.H., 1982. An examination of land use-nutrient export relationships. Water Res. Bull. 18 (6), 1013 – 1024. Bohlke, J.K., Denver, J.M., 1995. Combined use of groundwater dating, chemical, and isotopic analyses to resolve the history and fate of nitrate contamination in two agricultural watersheds, Atlantic coastal plain, Maryland. Water Res. Res. 31 (9), 2319 – 2339. Brinson, M.M, 1993. Changes in the functioning of wetlands along environmental gradients. Wetlands 13 (2), 65 – 74. Carpenter, D.H., 1983. Characteristics of streamflow in Maryland. Report of investigations no. 35. Maryland Geological Survey. Dept. of Natural Resources. Baltimore, MD. Chesapeake Bay Program, 1995. The state of the Chesapeake Bay. Printed by the U.S. Environmental Protection Agency. U.S. Government Printing Office, Washington, D.C. Cooper, A.B., Smith, C.M., Smith, M.J., 1995. Effects of riparian set-aside on soil characteristics in an agricultural landscape: Implications for nutrient transport and retention. Agric. Ecosys. Envir. 55, 61 – 67. Cooper, S.R., Brush, G.S., 1991. Long-term history of Chesapeake Bay anoxia. Science 254, 992 – 996. Cowardin, L.M., Carter, V., Golet, F.C., LaRoe, E.T., 1979. Classification of wetlands and deepwater habitats of the United States. U.S. Government Printing Office, Washington, D.C. Available from: U.S. Fish and Wildlife Service, Washington, D.C. Cornell, S., Rendell, A., Jickells, T., 1995. Atmospheric imputs of dissolved organic nitrogen to the oceans. Nature 376, 243 – 246. Dennison, W.C., Orth, R.J., Moore, K.A., Stevenson, J.C., Carter, V., Kollar, S., Bergstrom, P.W., Batiuk, R.A., 1993. Assessing water quality with submersed aquatic vegetation. BioScience 43 (2), 86 – 94. Edwards, W.M., Owens, L.B., 1991. Large storm effects on total soil erosion. J. Soil Wat. Conserv. 46 (1), 75 – 78. EPA, 1983. The Chesapeake Bay Agreement of 1983. Signatory states of Virginia, Maryland, Pennsylvania and District of Columbia. EPA, 1987. The Chesapeake Bay Agreement of 1987. Signatory states of Virginia, Maryland, Pennsylvania and District of Columbia.

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362 EPA, 1992. Progress report of the Baywide nutrient reduction reevaluation. Document no. CSC.LRIB. 12/19. Printed by PEA for the Chesapeake Bay Program, 68 p. Fisher, T.R., Lee, K.Y., Berndt, H., Benittz, J.A., Norton, M.M., 1998. Hydrology and chemistry of the Choptank River Basin in the Chesapeake Bay drainage. Water Air Soil Pollut. 105, 387 –397. Fisher, T.R., 1992. Nutrient Inputs to the Choptank River. Final report to U.S. Fish and Wildlife Service under cooperative agreement with University of Maryland. Chesapeake Bay Field Office, Annapolis, Maryland. Gilliam, J.W., 1994. Riparian wetlands and water quality. J. Envir. Qual. 23, 896 –900. Haines, E.B., 1976. Nitrogen content and acidity of rain on the Georgia coast. Water Res. Bull. 12, 1223–1231. Haith, D.A., Shoemaker, L.L., 1987. Generalized watershed loading functions for stream flow nutrients. Water Res. Bull. 23, 471 – 478. Hamilton, P.A., Denver, J.M., Phillips, P.J., Shedlock, R.J., 1993. Water-quality assessment of the Delmarva peninsula, Delaware, Maryland, and Virginia-effects of agricultural activities on, and distribution of, nitrate and other inorganic constituents in the surficial aquifer. U.S. Geological Survey open-file report 93-40, Towson, MD. Harding, L.W., 1994. Long-term trends in the distribution of phytoplankton in Chesapeake Bay: roles of light, nutrients and streamflow. Mar. Ecol. Prog. Ser. 104, 267–291. Haycock, N.E., Pinay, G., Walker, C., 1993. Nitrogen retention in river corridors: European perspective. Ambio 22 (6), 340 – 346. Hill, A.R., 1996. Nitrate removal in stream riparian zones. J. Envir. Qual. 25, 743 –755. Houlahan, J.W., Marcus, A., Shirmohammadi, A., 1992. Estimating Maryland critical area act’s impact on future nonpoint source pollution along the Rhode River estuary. Water Res. Bull. 28 (3), 553–567. Johnston, C.A., Detenbeck, N.E., Niemi, G.J., 1990. The cumulative effect of wetlands on stream water quality and quantity. A landscape approach. Biogeochemistry 10, 105– 141. Jordan, T.E., Correll, D.L., Weller, D.E., 1997. Effects of agriculture on discharges of nutrients from coastal plain watersheds of Chesapeake Bay. J. Envir. Qual. 26, 836– 848. Jordan, T.E., Weller, D.E., 1996. Human contributions to terrestrial nitrogen flux. BioScience 46, 655–664. Jordan, T.E., Correll, D.L., Weller, D.E., 1993. Nutrient interception by a riparian forest receiving inputs from adjacent cropland. J. Envir. Qual. 22, 467–473. Kemp, W.M., Boynton, W.R., Means, J.C., 1983. The decline of submerged vascular plants in upper Chesapeake Bay: summary of results concerning possible causes. Mar. Tech. Soc. J. 17 (2), 78 – 89. Lee, K.Y., 2000. The integration of GIS and GWLF to estimate nutrient export from Maryland coastal plain basins. Ph.D. dissertation, University of Maryland.

361

Lewis, W.M. Jr., Grant, M.C., Hamilton, S.K., 1985. Evidence that filterable phosphorus is a significant atmospheric link in the phosphorus cycle. Oikos 45, 428 – 432. Lowrance, R., Todd, R., Fail, J. Jr., Hendrickson, O. Jr., Leonard, R., Asmussen, L., 1984. Riparian forests as nutrient filters in agricultural watersheds. BioScience 34 (6), 374 – 377. Lubbers, L., Boynton, W.R., Kemp, W.M., 1990. Variations in structure of estuarine fish communities in relation to abundance of submersed vascular plants. Mar. Ecol. Prog. Ser. 65, 1 – 14. Malone, T.C., Crocker, L.H., Pike, S.E., Wendler, B.W., 1988. Influences of river flow on the dynammics of phytoplankton production in a partially stratified estuary. Mar. Ecol. Prog. Ser. 48, 235 – 249. Maryland Agricultural Statistics Service, 1992. Maryland agricultural statistics. Summary for 1992. Issued cooperatively by agricultural statistics service and Dept. of Natural Resources. Annapolis, Maryland. Maryland Agricultural Statistics Service, 1986. Ibid. Norton, M.N., Ph.D. dissertation, 1998. Landscape effects of land cover, soil properties and geomorphology on stream water quality draining two coastal plain watersheds in the Chesapeake Bay. University of Maryland. Officer, C.B., Biggs, R.B., Taft, J.L., Cronin, L.E., Tyler, M.A., Boynton, W.R., 1984. Chesapeake Bay anoxia: origin, development, and significance. Science 223, 22 – 27. Omernik, J.M., Abernathy, A.R., Male, L.M., 1981. Stream nutrient levels and proximity of agricultural and forest land to streams: some relationships. J. Soil Water Conserv. 36 (4), 227 – 231. Osborne, L.L., Kovacic, D.A., 1993. Riparian vegetated buffer strips in water-quality restoration and stream management. Freshwat. Biol. 29, 243 – 258. Osborne, L.L., Wiley, M.J., 1988. Empirical relationships between land use/cover and stream water quality in an agricultural watershed. J. Envir. Manag. 26, 9 – 27. Peterjohn, W.T., Correll, D.L., 1984. Nutrient dynamics in an agricultural watershed: Observations on the role of a riparian forest. Ecology 65 (5), 1466 – 1475. Pionke, H.B., Gburek, W.J., Sharpley, A.N., Schnabel, R.R., 1996. Flow and nutrient export patterns for an agricultural hill-land watershed. Water Res. Res. 32 (6), 1795 – 1804. Ritter, W.F., 1986. Water quality of agricultural coastal plain watersheds. Agricult. Wastes 16, 201 – 216. Richardson, C.J., 1985. Mechanisms controlling phosphorous retention capacity in freshwater wetlands. Science 228 (4706), 1424 – 1427. Soil Conservation Service, 1990. Soil series of the United States, including Puerto Rico and the U.S. Virgin Islands; their taxonomic classification. Soil Conservation Service miscellaneous publication number 1483. U.S. Government Printing Office, Washington, D.C. 723-472/20319. Soil Conservation Service, 1991. Hydric soils of the United States. U.S. Department of Agriculture miscellaneous publication number 1491. In cooperation with the National

362

M.M. Norton, T.R. Fisher / Ecological Engineering 14 (2000) 337–362

Technical Committee for hydric soils. NTCHS, SCS, Room 152, Federal Building, 100 Centennial Mall North, Lincoln, NE. Sonzogni, W.C., Chester, G., Coote, D.R., Jeffs, D.N., Konrad, J.C., Ostry, R.C., Robinson, J.B., 1980. Pollution from land runoff. Envir. Sci. Tech. 14 (2), 148–153. Stevenson, F.J., 1986. Cycles of Soil: C, N, P, S, Micronutrients. John Wiley and Sons, New York, NY, USA. Stevenson, J.C., Staver, L.W., Staver, K.W., 1993. Water quality associated with survival of submersed aquatic vegetation along an estuarine gradient. Estuaries 16 (2), 336–346. Strahler, A.N., 1952. Hypsometric (area-altitude) analysis of erosional topography. Bull. Geol. Soc. Am. 63, 1117–1142. Tiner, R., Burke, D.G., 1995. Wetlands of Maryland. U.S. Fish and Wildlife Service. Ecological Services, Region 5, Hadley, MA, and Maryland Dept. of Natural Resources, Annapolis, MD.

.

Valderrama, J.C., 1981. The simultaneous analysis of total nitrogen and total phosphorus in natural waters. Mar. Chem. 10, 109 – 122. Verchot, L.V., Franklin, E.C., Gilliam, J.W., 1997. Nitrogen cycling in piedmont vegetated filter zones: I. Surface soil processes. J. Environ. Qual. 26, 327 – 336. Victoria, C., unpublished. Assessment of Water Quality, Basin Characteristics and Biological Condition of the Chester River Watershed. U.S. Fish and Wildlife Service Report, Chesapeake Bay Field Office, Annapolis, Maryland. Walbridge, M.R., Struthers, J.P., 1993. Phosphorus retention in non-tidal palustrine forested wetlands of the mid-atlantic region. Wetlands 13 (2), 84 – 94. Whigham, D.F., Chitterling, C., Palmer, B., 1988. Impacts of freshwater wetlands on water quality: a landscape perspective. Environ. Manag. 12 (5), 663 – 671.