Full Text :COPYRIGHT 2002 Soil & Water Conservation Society

Land-use planning for environmentally sustainable agricultural production is a process that demands the availability and processing of large amounts of spatial information in support of decision-making. Assessing land qualities specific to the location of agricultural production systems and identifying the need for conservation practices are essential steps in this process. When planning goals include the reduction of pollution potential, comparison of relative pollution potentials across land areas can produce valuable information for decision- and policy-making. Such decision-making is related to implementing conservation practices or selecting lands for inclusion in conservation programs such as the Conservation Reserve Program (CRP).

As conservation programs cannot enroll entire geographic areas, only the most environmentally sensitive lands, yielding the greatest environmental benefits, are accepted into the programs. For lands with surficial water-quality problems, nonpoint sources of phosphorus (P) to water bodies need to be identified and characterized to aid in improving management plans by focusing on land areas where the problem is greater. As the direct assessment of P content in surface water is expensive, indirect methods are frequently used to obtain a spatial representation of P contribution to streams. These indirect methods are used to identify critical regions and are less expensive and time-consuming than direct monitoring of contributing areas to surface-water bodies (Daniel et al. 1994).

The use of available spatial information and modeling of biophysical processes can be a cost-effective way to identify land areas that may contribute to water-quality problems. However, although many sophisticated process-based hydrological models are available for use in predicting P loss at the watershed scale, their use usually requires a large amount of data for calibration and validation of the results. Additionally, the current state of science does not allow the use of simple models and general inputs to produce estimates of P export to waters without a substantial degree of uncertainty. Therefore, the use of simpler information analysis approaches can help screen the study area and recognize priority areas for implementing conservation practices before results of those more complex modeling analyses are available. Many geographic land characteristics are available in map form, both analog and digital, and can be used to assess the potential of a site to export P.

The objective of this study was to develop and apply spatial indices at the watershed scale for identifying and ranking land areas with respect to potential P export to water bodies, with the goal of prioritizing such areas for inclusion in conservation programs. Although the indices are process-based, they are simple; they do not supplant the need for more-advanced modeling techniques. The indices are not intended to provide quantitative estimates of P exports to streams, but are only intended for use in making relative comparisons among geographic areas.

Methods and Materials

Study Area. The New York City (NYC) watershed is the source of drinking water for NYC residents. Land use in this watershed aggregates natural, agronomic, and economic conditions that have resulted in P transport to water bodies in amounts that may be in excess of current water-quality standards, as in the Cannonsville Reservoir Basin. There, total P concentrations in stream inflow range from 5 to 80 mg [m.sup.-3] with a median of 30 mg [m.sup.-3] over a six-year period (1993 to 1999) (NYC DEP 2001), and nonpoint runoff is a major source (Longabucco and Rafferty 1998). Although New York state does not have a numerical P water-quality standard, it has a P guidance value for lakes and reservoirs of 20 mg [m.sup.-3] (NYC DEP 2001). Actions for keeping the high quality of the drinking water in this watershed include the improvement of farm management practices, the implementation of conservation practices, and the reduction of certain farming activity in the watershed by inclusion in the CRP.

The Cannonsville Reservoir Basin is located in Delaware County, New York, and is part of this water-supply system. This basin has a catchment area of 1,178 [km.sup.2] and comprises 31 subbasins (Figure 1).The major land use in the basin is low-density family farming, mainly characterized by dairy farms. Dairy farming in the basin has contributed to a reduction in water quality resulting from elevated P content in surface water and eutrophication of aquatic environments caused by cropland drainage, manure and fertilizer applications on fields, soil erosion, and barnyard runoff (Porter and Beckhardt 1997). Large amounts of P are excreted in the manure of dairy cows and pose an environmental concern, especially where livestock are concentrated near water bodies.

The model. The Animal Manure Pollution Potential Index (AMPPI) provides a relative ranking of the capacity of a particular site for supporting an animal-based agriculture with low impact on water quality. The index uses spatial data to estimate the potential transport of P to surface waters, considering land use and biophysical characteristics of selected land areas. The AMPPI was adapted from the Animal Waste Pollution Index developed by Heatwole and Shanholtz (1991) and estimates the potential delivery of some fraction of P derived from average annual additions of manure from an area to the nearest receiving water body as follows:

AMPPI = [L.sup.*] [A.sup.*] MDR, (1)

where: AMPPI = Animal Manure Pollution Potential Index (weight [area.sup.-1] [time.sup.-1]), L [P.sub.2][O.sub.5] load or application rate (weight [area.sup.-1] [time.sup.-1]), A = availability factor (A = 0 to 1), MDR = modified delivery ratio (MDR = 0 to 1).

The availability factor (A) is an estimate of the proportion of P from annual additions that moves into runoff (that may or may not reach the stream) from surface-applied manure and could be estimated from values in the literature. The modified delivery ratio (MDR) is the fraction of the estimated P in runoff that reaches the streams and is calculated as follows: MDR = [Sd.sup.*] DR, (2)

MDR = [Sd.sup.*] DR, (2)

where: MDR = modified delivery ratio, Sd = soil drainage factor, DR = delivery ratio (Heatwole and Shanholtz 1991).

The soil drainage factor was included to account for differences in water infiltration and resultant runoff as they vary with respect to soil wetness and water-storage capacity. The values for the soil drainage factor are 0.1 for excessively and well-drained soils, 0.3 for moderately well-drained soils, 0.7 for somewhat-poorly drained soils, and 1.0 for poorly and very poorly drained soils (Bryant et al. 2000).

The delivery ratio was developed by Heatwole and Shanholtz (1991) based on the work of Draper et al. (1979) and is calculated as follows:

DR = [e.sup.[-k.sub.1][DS.sub.f]], (3)

where: DR = delivery ratio, D = distance that water has to travel until reaching a water body (m), [S.sub.f] = slope factor.

The slope factor ([S.sub.f]) is calculated as (Heatwole and Shanholtz 1991):

[S.sub.f] = [S.sub.fmin] + [e.sup.[-k.sub.2](S+[S.sub.o])], (4)

where: [S.sub.fmin] = minimum value for [S.sub.f], S = slope gradient along D (m [m.sup.-1), and [k.sub.1], [k.sub.2], and [S.sub.o] are parameters for fitting the equations to the field observations by Draper et al. (1979).

As described in Heatwole and Shanholtz (1991), [S.sub.f] is a slope factor that increases the delivery ratio as slope gradient increases. While [S.sub.f] could potentially vary between 0 and 1 (for 100% and 0% slopes, respectively), a lower limit of [S.sub.f] = [S.sub.fmin] was used for steep slopes. Therefore, the effect of the slope gradient on the delivery ratio will reach some maximum value at minimum [S.sub.f], and the greater importance of the slope gradient in the delivery ratio will be maintained. As in Heatwole and Shanholtz (1991), a value [S.sub.fmin] = 0.6 and parameters [k.sub.1] = 0.0161 m [m.sup.-1], [k.sub.2] = 16.1, and [S.sub.o] = 0.057 were used to provide a good representation of the delivery ratio estimated by Draper et al. (1979). The curvilinear relationships between distance and DR and between distance and the MDR are shown in Figures 2 and 3, respectively.

The Total P Export Index (TP) for an area is calculated by the multiplication of the AMPPI by the total area. The TP for each subbasin may be divided by the number of animal units and by the total area of each subbasin, generating the P Export per Unit of Area Index (PA) and the P Export per Animal Unit Index (PAU). Geographic areas or fields can be characterized by the mean value of these indices.

Although the indices are physically based, calculations are based on current land use, animal population densities, and modern estimates of P content in manure. Without taking into account the history of land use, they should not be used to infer average annual P export to surficial water bodies. The lack of data on fractions of annual additions of P that are lost in runoff (apart from losses attributable to historical P loading) does not allow validating the numerical values. The AMPPI is used in this study without defined units, because the calculated values cannot be validated. However, given that the indices reflect current land use and management, they do identify existing landscapes as currently managed that pose the greatest future risk to water quality, and these are the areas that should be targeted by conservation programs.

Data collection and model application. The indices were calculated for all subbasins with dairy farming in the Cannonsville Reservoir Basin. Digital data layers used in this analysis were USGS 7.5' Digital Elevation Model (DEM), SSURGO digital soil survey of Delaware County (NY), 1992 land use (NYC DEP 1999), location of dairy barns, watershed and subbasins boundaries (NYC DEP 1999), farm locations and tabulated characteristics (Lamont 1999). For calculating the DR, the DEM was pre-processed and used to calculate flow direction and flow accumulation. In addition, the stream network was redefined to represent stream extent similar to that as mapped in the SSURGO soil survey.

Unlike the procedure used by Heatwole and Shanholtz (1991), this study used the actual distance that water would have to travel to teach a water body. This was calculated using a flow-direction data layer derived from DEM. That layer was modified so that all stream cells received a value of zero. The actual distance that water had to travel from each nonstream cell to reach a stream cell was obtained by calculating a modified flow-direction data layer using standard hydrologic functions in ArcView 3.2 (ESRI 2000). The resulting flow-length data layer had, in each nonstream cell, the value of the distance (D) that water in each nonstream cell has to travel until reaching the nearest stream.

The mean slope gradient (S, in m [m.sup.-1]) of the water pathway from each cell to the stream was calculated using this distance (D) and the difference in elevation between each nonstream cell and the stream cell where the water teaches the stream. The elevation of the stream cells where the water reached the stream was defined using an automatic routine for calculating the weighted-flow length, using as weight a data layer with elevation values for each stream cell and "no data" in nonstream cells. The result is a data layer where nonstream cells assume the value of the elevation of the stream cell where the water reaches the stream network. The mean slope gradient (S) for each nonstream cell was calculated by dividing the difference in elevation between the DEM data layer and the data layer with the elevation of the cell where water enters the stream by D. These derived data layers, D and S, were used for deriving the DR.A drainage-factor layer was obtained by reclassifying the SSURGO Delaware County Soil Survey. Cell values from this data layer were multiplied by cell values in the DR layer for obtaining the MDR.

Manure loads to the fields were calculated based on animal distribution. The animal number in each subbasin was calculated based on the 1997 Census of Agriculture (USDA 1999) and data provided by Lamont (1999). Numbers of animals of each type are given in ranges. For every farm, the total number of animals was calculated using the midpoint of each range. The total number of animals in each subbasin was added to obtain the total number of animals in the watershed. This total was scaled by area to obtain the same number of animals expressed in the 1997 Census of Agriculture (USDA 1999), which represented an increase of 11% over the number calculated using the farm data. The total manure produced was estimated assuming a manure production of 0.079 [m.sup.3] of manure per animal per day for milking cows producing 11,000 kg of milk per year. For other dairy animal types (e.g., dry cows, heifers, and calves), manure production was assumed to be 0.087 [m.sup.3] of manure per animal unit (AU) per day, where one AU is 454 kg of live weight (Midwest Plan Service 1985, NRAES 1999). The estimated total amount of manure produced was used for estimating the total amount of [P.sub.2][O.sub.5] produced in each subbasin, assuming a [P.sub.2][O.sub.5] content in manure of 1.8 kg [P.sub.2][O.sub.5] [m.sup.-3] (Midwest Plan Service 1985).

The 1992 land use and subbasin boundaries layers were used for determining area of croplands and pasture in each subbasin. Crops, alfalfa, and bare soils were grouped together and are referred to as cropland. Grass/shrub and grass were grouped together and are referred to as pastures. It was assumed that manure application rate on croplands is four times larger than the application rates on pastures, based on what is known about the actual farming practices in the basin. These application rates or loads were spatially located, generating a [P.sub.2][O.sub.5] load data layer with different manure application rates for each land-use type and subbasin. Given that farm boundary information is not available, it was assumed that all manure produced in a subbasin is applied within the same subbasin. Subbasins having no dairy barns located within their boundaries were assumed to have no dairy farm activity and no import of manure from other subbasins. Therefore, they were excluded from this analysis.

MDR and manure load data layer were multiplied by the availability factor for obtaining the AMPPI. Several literature sources were analyzed to obtain an estimate of the value of A (Coote and Hore 1976, Draper at al 1979, Heatwole and Shanholtz 1991, Swanson et al. 1971). As most studies indicated less than 0.10, a mean value of 0.05 was adopted, following the procedure adopted by Draper at al. (1979) and Heatwole and Shanholtz (1991). Resulting data layers have a value for AMPPI for each grid cell in each subbasin. Subbasins were characterized by mean values of AMPPI for crop areas and for pasture areas. The AMPPI was used to calculate the TP in each subbasin, which was further divided by subbasin areas and by number of animal units in the subbasins, calculating the PA and PAU indices. As the objective of this study is to differentiate geographic areas regarding potential P transport to water bodies rather than forecast the amounts of P exported from each area, and the numerical results are very uncertain giv en the lack of data needed to validate them, no units were used for these indices, and they were evaluated in a comparative way.

Results and Discussion

The MDR estimated from terrain analysis is an indicator of the inherent potential for P loss from land areas if P is applied. Under similar conditions of management and P loading, land areas with high MDR values will potentially deliver more nutrients to stream waters. The estimated MDR values for those subbasins having dairy farms located within their boundaries and estimated MDR values for areas of croplands and pastures within those subbasins are summarized in Figure 4. In most of the subbasins, the mean MDR for pastures is higher than the mean MDR for crops because of pastures being located on moderately well-drained and somewhat-poorly drained fragipan soils that occupy steeper slopes and to crops being grown on well-drained, valley-bottom soils. This fits the general land-use pattern in the Cannonsville Reservoir Basin. The best agricultural soils are well-drained and located in the valley bottoms, pastures are located on the mid-slopes where runoff is sufficiently concentrated to form concentrated flow s, and the steepest upper slopes are forested.

Highest mean MDR values for croplands are located in subbasins 26, 28, and 30. The total areas of croplands in these subbasins are small (especially in subbasins 28 and 30), and [P.sub.2][O.sub.5] export from these subbasins should not make significant contributions to the total P pollution potential. The highest mean MDR value for pastures is located in subbasin 33, but the acreage in pasture is relatively low. As in the case of croplands, land areas least-suited for pasture are not being used for pasture. By comparison, those subbasins (14 20, 24, 27, 28,30, and 41) having the next highest MDR values also have higher proportional acreages of pasture and are important sources of nutrients. There is no relationship between the mean MDR values for entire subbasins and mean MDR values for croplands and pastures. This indicates that mean MDR value for the entire subbasin is not a good indicator of the potential delivery of nutrients from the areas of pasture and crops where manure is applied.

The AMPPI accounts for manure application for each land-use type within subbasins in combination with the inherent potential for P transport represented by the MDR. Mean values of AMPPI calculated for croplands and pastures are shown in Table 2. As previously explained, absolute values are based on current land use and management and should not be interpreted as representing average annual P loading to streams. Because the fraction of P loss derived from current annual additions has not been quantified, the AMPPI and the indices derived from the AMPPI are reported as unitless values and are indicative only of relative differences between subbasins.

In most subbasins, the mean AMPPI for croplands is larger than the mean AMPPI for pastures. Croplands are generally located on lands having MDRs lower than pastures; the higher AMPPI values are due to 1) the much higher manure application rates (4X) on croplands compared with pastures and 2) croplands being located closer to streams. Higher AMPPI values were found on croplands in four subbasins (16, 18, 26, and 28) that have high MDR (as in the case of subbasins 26 and 28) and/or high [P.sub.2][O.sub.5] load caused by high animal density compared with the area of crops and pastures (16 and 18).

For each subbasin, the mean AMPPI for each land-use type was multiplied by the areas of the respective land-use type for determining the TP for croplands and pastures. The sum of these products for croplands and pastures within each subbasin provides an estimate of TP for the Subbasins (Table 2). Subbasins with higher TP are large subbasins (17,24, and 41) or subbasins with high animal density (12). The estimate of TP alone is not an indicator of environmental problems, but this estimate can be such an indicator when related to variables such as subbasin area or number of AU in the subbasin.

PA and PAU indices are shown in Table 2 and allow comparison of these environmental indicators in map form showing ranges of values for animal density, [P.sub.2][O.sub.5] load, MDR, AMPPI, TP, PA, and PAU (Figures 5 and 6). The PA of the entire subbasin accounts for the dilution of runoff from pastures and croplands caused by addition of stream flow from forested lands and runoff from nonagricultural areas. These values are indicators of the concentration of P in water that reaches the streams. In the study area, most P enters the stream by overland flow; therefore, these values are indicative of overall stream-water quality. Subbasins with higher PA values have higher animal densities. The six subbasins (12, 13, 16, 18, 21, and 41) having the higher PA also have high animal density. If stream quality in individual subbasins is of concern, then these subbasins should be prioritized for enrollment in conservation programs such as the CRP, because they are likely to produce the highest P transport to stream wat ers.

The PAU compares subbasins by normalizing TP by the number of animal units and is therefore representative of how well the manure is managed in each subbasin. Subbasins with higher PAU (e.g., 3