Understanding Plant-Soil-Livestock Interactions: A Key to Enhanced Sustainability in Southern-Pine Silvopasture Systems

NOTE: All Tables, Figures, Drawings, Publications and other Materials referenced in the following sections are found in the separate appendix.

Objectives 1 and 2
1. Determine the impacts of N supply (fertilization versus clover) on above- and below-ground forage productivity, forage quality and plant diversity in young longleaf pine (Pinus palustris) silvopasture.
2. Determine the impacts of N supply on pasture soil structural stability and relationships to soil compaction in young longleaf pine silvopasture.

Field Research Site
This research was conducted from 2005 to 2007 in a young 4-ha longleaf pine (Pinus palustris Mill.)-bahiagrass (Paspalum notatum Flugge) silvopasture and adjoining 4-ha bahiagrass pasture without trees (open-pasture) at the USDA-NRCS Jimmy Carter Plant Materials Center, Americus, Georgia (32° 3' N, 84° 14' W). The bahiagrass pasture to be converted to silvopasture was prepared in summer 2000 by in-row subsoiling and application of glyphosate in a double-row set configuration: 1.82-m tree-to-tree-in-row spacing and 3.04-m spacing between the double-row sets of trees; alleys for forage production between double-row tree sets were 12.2-m wide. Longleaf pine seedlings were planted in the double-row set configuration in December 2000. All trees had emerged from the grass-stage by April 2005 and had reached an average height of 5.9 ± 0.05 m and diameter at breast height (DBH) of 11.5 ± 0.11 cm by the end of the study in Fall 2007; tree height and DBH were not different between plots (multivariate ANOVA, Wilk’s Lambda, F probability = 0.4674). Trees were not pruned at any time during the study. Soil at the site was an Orangeburg loamy sand (fine-loamy, kaolinitic, thermic Typic Kandiudults), a series of large extent on the Southern Coastal Plain. At this site, the soil particle size distribution was 850 g/kg sand, 125 g/kg silt, and 25 g/kg clay, 22 g/kg organic matter, and an estimated ion exchange capacity of 6.23 cmol/kg. Using annual Auburn University soil test recommendations, levels of plant available P and K were adjusted as needed with mixed commercial fertilizer in late spring, and soil pH was maintained at 6.0 with addition of dolomitic limestone in the fall.

Climatic Conditions
March to August 2005 precipitation was consistently higher than the 47-y average except in May, but was consistently lower than the 47-y average in September and October. Precipitation in 2006 was consistently lower than the 47-y average from January to June, except in May, and in September. Precipitation also remained below the 47-y average from January to May, July, and from September to November 2007. With few exceptions, monthly average minimum and maximum temperature mostly remained similar for all three years.

Treatments
Silvopasture and open-pasture were each divided into three blocks (replications) and within each block, two 0.2-ha plots were randomly assigned one of two N-source treatments: commercial N fertilizer (ammonium nitrate) or crimson clover (Trifolium incarnatum L. ‘Dixie’). Nitrogen fertilizer was applied annually as a single application of 67 kg/ha N in late spring; this rate was based on current Auburn University soil test recommendations for bahiagrass pasture. Crimson clover was overseeded with a Truax FLEXII (Truax Co., Inc., New Hope MN) grass drill with no-till attachment in October 2004 at a rate of 11.2 kg/ha. Crimson clover was overseeded again in October 2006 because drought conditions in September and October 2005 inhibited clover germination resulting in an almost non-existent stand of crimson clover in the treatment plots in spring 2006 (Table 1).

Forage and Soil Sample Collection and Analysis
Silvopasture plots included four double-row tree sets (average 449 ± 63 trees/ha) and three 12.2-m x 36-m alleys. To locate permanent points for sample collection, randomization was performed at three levels: 1) distance in 1-m intervals (2-35 m) along the length of the alley, 2) alley (right, middle, left) within the plot, 3) left or right side of the selected alley for location of the sample positions relative to the tree base. At each randomly-selected distance along the selected alley, points representing the alley center position were located 6.1 m from the center of the tree base; the alley side position was located 1.0 m from the center of the tree base. Meter tapes and pin flags were used to locate and mark all sampling points. For example, in one plot, sampling points were located in the middle alley at 15 m along the length of the alley and then 6.1 m (alley center) and 1.0 m (alley side) from the center of the tree base at the right side of the alley. The result was five sub-samples from both the alley-center position and the alley-side position within each plot. A similar sampling scheme was established in the open-pasture. In 2006, an additional sample point for shoot biomass collection was added at 3.5 m from the center of the tree base (equidistant between the 1.0-m and 6.1-m sample points) for all sample collection locations in the silvopasture.
To estimate shoot biomass and quality, forage within a 0.25 sq m quadrat was clipped to 5 cm from the ground. Pine straw included within the quadrat at a height of 5 cm or more was collected separately. Immediately after sample collection, plots were mowed (2005, 2007) or grazed (2006) to 5 cm then allowed to re-grow. Shoot samples were collected three times a year: April or May, June or July, and August or September. Shoot sample collection dates were based on maturity stage of crimson clover (full-bloom) in the spring and bahiagrass (one-third full-bloom, approximately 25 cm) during the summer months. For the April or May (cool-season) sample collections, crimson clover, other legumes, grasses, and forbs in the shoot sample were separated, and weighed separately. The category ‘other legumes’ included small amounts of legume species {white clover (Trifolium repens L.), small hop clover (Trifolium dubium L., common vetch (Vicia sativa L.)} present in all silvopasture and open-pasture plots when the study began and constituted an equal proportion of the cool-season biomass in all silvopasture and open-pasture plots (Table 1). Forage tissue samples were dried at 60ºC for 72 h. All components of oven-dried shoot biomass samples including pine straw (when present) were combined then ground to pass a 1-mm sieve. Ground tissue samples were composited by alley position within a plot to estimate Kjeldahl-N and acid detergent fiber (ADF) (Goering and Van Soest, 1970).
Root samples were collected in August 2005 and October 2007 with a 5-cm (diameter) x 10-cm depth) core sampler and kept cool (4ºC) until processing was completed within 14 days of collection. Soil was washed from root cores over a 500-micrometer mesh sieve. After debris was removed, the root tissue was dried at 60ºC for 72 h.
When soil sample collections were scheduled, they were coordinated with forage biomass sampling. Throughout the study, five sub-samples were collected from both the alley-center position and the alley-side position within each plot at positions around (WSA) or below (DFH) the quadrat area where shoot biomass had been collected. A similar sampling scheme was established in the open-pasture. Soil samples for water stable aggregates (WSA) were collected to 7.6 cm in May and August of 2005 and 2006, and April and September 2007. Samples were sieved (2-mm) in a field-moist condition, allowed to air dry, then analyzed using an Eijkelkamp wet-sieving apparatus (Soil Moisture Equipment Corp., Goleta CA). To determine the percentage of unstable aggregates, 4 g of air-dried 2-mm aggregates were weighed into each of 8 sieves (0.250-mm), pre-moistened with deionized water and then moved up and down a set distance for 3 min in pre-weighed cans filled with 80 ml deionized water (water-filled). After 3 min, the water-filled cans that contained the unstable soil aggregates were removed and replaced by pre-weighed cans filled with 80 ml 2.0 g/L NaOH (NaOH-filled). The remaining water-stable aggregates were then moved up and down an additional 11 min in the NaOH-filled cans to disperse all remaining aggregates. All cans were evaporated in a convection oven at 110°C for 24 hr then reweighed. Percent water-stable aggregates was calculated as equal to the weight of soil obtained in the NaOH-filled can divided by the sum of the weights obtained in the NaOH-filled can + the water-filled can. In this calculation, the actual weight of soil obtained in the NaOH-filled can was adjusted for the weight of the NaOH by subtracting 0.2 g.
Soil samples for density of fungal hyphae (DFH) were collected in August 2005, May and August 2006, and April and September 2007; samples were kept cool (4ºC) until processing was completed within 14 days of collection. DFH was estimated using the membrane filter technique (Bardgett, 1991) to prepare two membrane filtrate slides for each sample. These slides were examined at 200x magnification by observing five random fields of view for each slide; total hyphal length for each slide was estimated following method four of Olson (1950). Average hyphal length from two slides prepared for each sample was used to estimate DFH in m/g of wet soil, which was then converted to m/g of oven-dried soil based on the gravimetric water content of a subsample of the initial DFH sample.
In June 2005 and October 2007, soil compaction was measured in terms of penetration resistance (PR) in-situ at four depths from the soil surface: 0-5 cm, 5-10 cm, 10-15 cm, and 15-20 cm using a dynamic cone penetrometer (Herrick and Jones, 2002). Soil samples (0-5 cm) were taken at the same time from points nearby the PR measurement locations, oven dried at 100ºC for 72 h, and weighed to determine soil moisture content.

Data Analysis
The mixed procedure (SAS 9.1) was used to analyze the data with block as a random factor and sampling date as a repeated factor with spatial power law as a covariance structure (Littell et al., 2006). Main sources of variation included pasture type, N source, and sampling date. For soil PR, data from 2005 and 2007 were analyzed separately with depth as a repeated factor and first-order auto-regressive as a covariance structure (Littell et al., 2006); all possible interaction effects were also assessed. Data from silvopasture were also analyzed separately to assess the alley position effect as a result of proximity to trees. Probability level for rejection of the null hypothesis (Ho) was set at 0.05.

Measurement of Forage Species and Ground Cover Composition
Composition of ground cover and understory vegetation was measured by the point intercept method (USDA-FS 1996) using an optimal point projection device (Buckner 1985) during the early growing season (mid-April to mid-May) and late growing season (late-August to early-November) each year from 2005-2007, except in 2006 when observations were not made during the late growing season. From September 2005 to September 2007, baselines were established in the 0.16-ha silvopasture blocks across alleys perpendicular to tree rows on either side of the alley at five permanent points. Measurements were made at five alley positions directly on the baseline relative to the tree base by starting at one meter from the center of the tree base on left side of the alley and ending at one meter from the center of the tree base on the right side of the alley; three middle points between the two one-meter positions were flagged equidistant to one another. The point projection device was placed at each measurement position and cover categories recorded at 0º, 45º, 90º, 135º, and 180º relative to each baseline by moving the projection device in a semi-circle (125 readings per block); live vegetation was recorded by species. Measurements were taken similarly in the open-pasture blocks. Measurement of overstory coverage in silvopasture began in September 2005, and understory cover composition by alley position along the baseline was recorded during the 2006-2007 observation periods.

Data Analysis
Species diversity (species richness weighted by species eveness) and evenness (distribution of individuals among species) indices were calculated using Shannon’s method (Magurran 1988) and the similarity index (a comparison of species composition of two or more plant communities) was calculated using the method described by Cook and Stubbendieck (1986). Occurrence of different understory and overstory plants as well as litter and pine needles as land covers was tabulated. The mixed procedure (SAS 9.1) was used to analyze the data with block as a random factor and sampling date as a repeated factor with spatial power law as a covariance structure for unequally spaced sampling dates (Littell et al. 2006). Main sources of variation included pasture type and sampling date. For samples from silvopasture, alley position relative to the tree base was an additional source of variation. All possible interaction effects were also assessed. Alpha probability level for rejection of the Ho (null hypotheses) in favor of Ha (alternative hypotheses) was set at 0.05.

Greenhouse Studies
Three 12-week experiments were conducted Sept.-Dec. 2005, and May-Aug. and Sept.-Dec. 2006 at the Plant Science Research Center (PSRC), Auburn University, Auburn, Alabama (AL), USA. For the Sept.-Dec. 2005 experiment, both cool-season and warm-season forage species were sown in field-state soil (pH 5.0) on September 14. The study was repeated during May-Aug. 2006 with the sowing of warm-season forage species on May 25 and cool-season forage species during Sept.-Dec. 2006 with sowing on September 15. Forages were grown at both field-state and adjusted soil pH levels during both 2006 experimental periods. Eleven cool-season forage species or mixtures and nine warm-season forage species (Table 2) commonly grown in the Southeast USA for livestock forage, wildlife habitat enhancement or bioenergy were designated as treatments; a control treatment (no plants present) was included with each set of plants during each experimental period at each pH level. The experimental design was a randomized complete block with five replications of each treatment combination (forage species or mixture + pH level) and the control.

Soil Microcosm Preparation and Seeding
An Orangeburg loamy sand (fine-loamy, kaolinitic, thermic Typic Kandiudults) was collected (0-15 cm) from the Jimmy Carter Plant Materials Center, Americus, Georgia (GA), USA, sieved through a 15-cm hardware cloth on site, and then transported to the PSRC greenhouse. The field-state soil was then sieved through a 2-mm sieve in the field-moist state and microcosms were constructed by filling plastic pots (8.5x8.5 sq cm bottom area, 10.5x10.5 sq cm top area, and 12.5 cm depth) with 1 kg soil each. For the May-Aug. and Sept.-Dec. 2006 studies, each forage species or mixture treatment was sown in both field-state and adjusted-pH soil treatments. The field-state pH treatment consisted of 2-mm soil with no lime addition. For the adjusted-pH treatment, a composite sample of 2-mm soil was tested for its original pH (1:1 soil:water). Pulverized dolomitic limestone was added to the 2-mm soil and mixed thoroughly to raise the soil pH by approximately 1.5 units, and then reanalyzed. Other soil characteristics of interest for the field-state soil prior to 2-mm sieving and pH adjustment were: particle size distribution of 850 g/kg sand, 125 g/kg silt, 25 g/kg clay; 22 g/kg organic matter; cation exchange capacity 6.23 cmol/kg; water stable aggregates 472 g/kg, and extractable concentrations (mg/kg) of 31.5 P, 50.0 K, 92.7 Mg and 40.0 Ca. All of these extractable ion levels are considered in the high range for this soil.
Before sowing, the soil in each microcosm was wetted with tap water. Field-state and adjusted-pH soil treatments were each randomly allocated to five replications within each pH level, and each species treatment was allocated randomly to each replication. An equal amount of seeds were sown for each replicate within each treatment to uniformly cover the soil surface area; the sown seeds were then covered with a thin layer of soil with appropriate pH level. Sown microcosms of each replication were randomly allocated to the designated greenhouse bench (cool-temperature or warm-temperature zone) according to the experimental design. The day/night temperature settings for the cool-temperature zone were 24ºC/21ºC, and 28ºC/21ºC for the warm-temperature zone.

Care and Management of Plants
Soil in each microcosm was watered daily to approximately 85% of field capacity. When seedlings were well-established, unwanted seedlings were thinned to leave six uniform, healthy seedlings per microcosm. For the mixtures, three seedlings of each species were maintained in each microcosm. Pesticides were sprayed according to the regular insect management routine of the greenhouse as well as whenever insects appeared; weeds were removed manually. Based on development of an apparent nutrient deficiency symptom (chlorosis) during the Sept.-Dec. 2005 experimental period, a complete fertilizer with the elemental analysis 20 N, 4.4 P, 16.6 P was applied as a 0.25 g/kg solution (based on N) to approximately 85% field capacity of the field-state (pH 5.0) soil in each microcosm on alternate days for the last 25 days of the experimental period; water was not applied separately on the day of fertilizer application. No fertilizer was applied during the May-Aug. and Sept.-Dec. 2006 experimental periods because no nutrient deficiency symptoms were observed. Treatments within each replication were re-randomized weekly to minimize possible variation among the treatments caused by the greenhouse environment.

Soil and Plant Collection and Analyses
Plant shoots were harvested at soil level after 12 weeks of growth, dried at 60ºC for 72 hours and then weighed. Soil samples were designated for determination of root biomass, water stable aggregates (WSA), and density of fungal hyphae (DFH) by dividing the entire soil volume of each microcosm lengthwise into two equal parts with a sharp knife. One-half of the soil volume was designated for determination of root biomass. The remaining one-half of the soil volume was sub-divided in half lengthwise resulting in two fourths: one-fourth of the soil volume was designated for determination of WSA, and the top 2.5 cm of the remaining one-fourth of the soil volume was used to estimate DFH. The portion of the soil volume designated for WSA determination was sieved through a 2-mm sieve, allowed to air dry, and then analyzed following the method of Nimmo and Perkins (2002) using an Eijkelkamp wet-sieving apparatus (Soil Moisture Equipment Corp., Goleta CA) equipped with 0.250-mm sieves; 2.0 g/L NaOH was used as the dispersing agent. The portion of the soil volume designated for root dry matter and DFH determinations was kept cool (4ºC) until analyses were completed within 14 days of collection. To recover root dry matter, the designated soil volume was washed over a 500-micrometer mesh sieve, debris removed, and then the root tissue dried at 60ºC for 72 hours before being weighed. DFH was estimated using the membrane filter technique described by Bardgett (1991). Prior to processing the 2.5-cm soil volume designated for DFH determination, one-half of the soil volume (approx. 40-50 g) was separated and dried at 105ºC for 24 h for estimation of gravimetric water content. The remaining one-half of the soil volume was used to prepare two 13-mm diameter membrane filters for each sample. These filters were examined under a microscope at 200x magnification by observing five fields of view for each filter; total hyphal length for each filter was estimated following the method 4 of Olson (1950). Average hyphal length determined from the two filters prepared for each sample was used to estimate DFH in m/g of wet soil. This value was then converted to m/g of oven-dried soil based on the gravimetric water content.

Data Analysis
The mixed procedure (SAS 9.1) was used to analyze the data with block as a random factor (Littell et al. 2006). Multiple comparisons among the means of forage species treatments were performed by using the Tukey-Kramer method. Analyses were also performed for Pearson product-moment correlation coefficients (r) to quantify the association among the plant and soil variables. Probability level of alpha was set at 0.05 for determining significant treatment effect or significant correlations among the variables measured.

Objective 3. Compare the impact of goat stocking rates for control of invasive perennial broadleaf plants within developing silvopasture systems.
This objective was accomplished at the Fort Valley State University Technology Development Transfer Center. The field site is a 10-acre, volunteer, woodlot currently dominated by volunteer pines. The 10-acre site was divided such that 3 treatments were replicated within the site.
Treatment 1: Goats at low stocking rate – four, 0.3-ha plots
Treatment 2: Goats at high stocking rate - four, 0.3-ha plots
Treatment 3: No grazing - two, 0.3-ha plots

Measurement of Plant Species and Ground Cover Composition
Understory and overstory plant species composition was measured before goats were stocked onto the site and after two years of browsing by goats.
Composition of ground cover and understory vegetation was measured by the point intercept method (USDA-FS 1996) using an optimal point projection device (Buckner 1985) during the early summer (late May-June) in 2006 and 2008. Permanent locations for species composition measurements were located at ten randomly selected points along a baseline that ran diagonally across each 0.3-ha plot. Transects were drawn perpendicularly to the baseline at each of the ten random points. Whether the transect was drawn to the right or left at each point along the baseline was also randomly determined. Measurements of both understory and overstory were made at ten positions spaced one meter apart on the transect by starting at one meter from the baseline and ending at ten meters from the baseline. The point projection device was placed at each measurement position and cover categories recorded at 0º, 45º, 90º, 135, andº 180º relative to each transect by moving the projection device in a semi-circle. A total of 500 understory and 500 overstory point readings were made per plot; live vegetation was recorded by species. Disturbance of vegetation on the side of the transect where understory cover was being recorded was avoided until all readings for that transect were complete.

Data Analysis
Occurrence of different understory and overstory plants as well as litter and pine needles as land covers was tabulated. Species diversity and evenness indices were calculated for perennial plant species using Shannon’s method (Magurran 1988) and the eneness index was calculated using the method described by Cook and Stubbendieck (1986). The t-test was performed to find out whether species diversity was changed due to the various levels of treatments in the research area as described by Magurran (1988). Probability level for rejection of the null hypothesis (Ho: species diversity index would not differ before and after the application of treatment) was set at 0.05.

Clearing and Staining Intact Leaves of Plant Species
A collection was made of intact leaves and stems of understory plant species that would potentially be browsed by goats at the Fort Valley site. These plant species were identified and then cleared with 10% NaOH, dehrdrated with an ethanol series and stained with safranin O (Berlyn and Miksche 1976) to obtain drawings of the microscopic characteristics of the leaf and stem epidermal cells. Since epidermal cells are heavily cutinized, many fragments escape degredation in the ruminant GI tract and are found in fecal material and thus can be used to identify the particular diet being selected. There are currently atlases of epidermal plant fragments ingested by grazing animals for the Great Plains (Howard and Samuel, 1979) and one for 50 selected plants common on longleaf-slash pine-bluestem range in the southeastern United States (Johnson et al., 1983). The goal for this project was to expand the southeastern atlas to include drawings of the micro-anatomical features of epidermal fragments of plant species found at the Fort Valley site that had not been described previously. This would ultimately allow estimation of the botanical composition of goat diets (Croker 1958; Sparks and Malechek 1968) over time and in response to stocking density at this and similar sites.

Goat Performance
In 2006, 45 mature Spanish does weighing about 45kg were stocked at two rates. The rates were the equivalent of 15.4 animals, and 30.9 animals per hectare, respectively. Animal weight per unit of land was estimated to be 693 kg per hectare and 1,390 kg per hectare respectively.
In 2007 and 2008 the same number of does was used but two different rates were used because the necessary changes in configuration of the paddocks to better balance vegetation and level of light penetration. Stocking rates in those two years were the equivalent of 11.6 does in the lighter rate, and 23.1 does per hectare in the heavier stocking rate. This calculated to be approximately 522 kg per hectare and 1,040 kg per hectare. Although the goats in this study were not expected to be particularly productive since that was not their function in this setting, industry recommended stocking rates under most production conditions is 11 to 13 mature goats per hectare.
Does were weighed only at the beginning of the study each year in order to balance experimental groups, and to obtain an estimate of stocking weight per hectare. Because stocking rates all exceeded the carrying capacity of this wooded area, all groups had to be removed two to three times during the growing season because of the disappearance of vegetation. During that time they were held in existing conventional pastures on surrounding farmland. They were returned to the experimental paddocks when vegetation regrowth was observed to be sufficient to support grazing for at least four weeks during the growing season. Because the goats were forced to consume all vegetation they would willingly consume, it is likely that for 7 to 10 days prior to each removal they were in a weight-loss situation. Bermuda grass hay was provided as a supplement as the observed supply of palatable fresh vegetation became depleted. The goats were removed from the experimental site when vegetation was no longer available based on human ocular judgment. Animals were returned when adequate regrowth was available to support grazing for at least 30 days.
Goats were maintained in the experimental areas with commercially-available electric netting with a 32-inch vertical height. A solar energizer was used in this remote location. The fencing generally worked well following initial off-site training of the animals to this type of fence. There were occasional goat break-outs, general during times when vegetation within paddocks was depleted or when a dead short occurred for various reasons including personnel vagaries. Known “jumpers” were not selected to be part of this study. A recently available product of 38-inch vertical height would probably work better for goats under occasional hunger stress.

Objective 4. Examine the economic feasibility and level of landowner acceptance of management practices being proposed.
Forage production data on silvopasture plots were used to compare the nitrogen cost relationship of forage produced when utilizing a legume (crimson clover) as the sole nitrogen source for growth of bahiagrass versus application of commercial nitrogen.
Landowners were exposed to the project results at a total of eight field days, tours, workshops or conferences. Three of the field days were held at the producer-cooperator properties and included participant evaluation and feedback on management practices. Two additional field days are upcoming for 2009 on 15 April (Owens Farm, Chipley FL) and 2 June (USDA-NRCS-Jimmy Carter Plant Materials Center).

Objective 5. Estimate effects of silvopasture management practices on watershed-level hydrology using the Hydrology Simulation Program-Fortran (HSPF).
The purpose of this research was to evaluate the relative impacts of agroforestry practices on hydrology and water quality processes with a calibrated St. Louis Bay watershed model.

Introduction of St. Louis Bay Watershed Model
The St. Louis Bay watershed is located in the Gulf Coast region of Mississippi. The study area, depicted in Figure 1, drains approximately 500,000 acres (781 sq km), covering nearly half of the Mississippi Coastal watershed (USGS Cataloging Unit 03170009). The watershed is comprised of two main river systems, the Jourdan and Wolf Rivers, as well as numerous bayous that drain into the St. Louis Bay. The St. Louis Bay is a shallow estuarine system of approximately 9900 acres that empties into the Mississippi Sound.
The majority of the study area, especially the northern and middle portions of the watershed, remains fairly undeveloped. Figure 2 displays the land use distribution. Over half of the land area is covered in forest. Scattered throughout the forest and scrubland are areas of agricultural land. Pasture and grasslands account for nine percent of the land within the study area. The majority of the urban land use is located in the southern portion of the watershed, along the perimeter of the Bay.
The St. Louis Bay watershed hydrology and water quality model was developed for the Mississippi Department of Environmental Quality to develop total maximum daily loads (TMDLs) (Huddleston et al., 2001). The selected model was Hydrological Simulation Program Fortran (HSPF), one of the most widely used watershed model for environmental studies (Bicknell et al., 2001).
St. Louis Bay watershed has been delineated into 37 sub-watersheds to simulate the hydrological processes (Figure 3). The simulated constituents by St. Louis Bay model included flow, bio-chemical oxygen demand, dissolved oxygen, nitrogen species (ammonia, nitrate, and organic nitrogen), and phosphorus species (ortho-phsophate and organic phosphorus). The complicated AGCHEM module was applied to simulate the complex nutrient processes occurring in the cropland. The modeled nutrient processes included atmospheric deposition, fertilization practices, manure application, and nutrient transformation processes. The PUQAL/IQUAL module was used to simulate the water quality processes in the non-crop lands.
The St. Louis Bay watershed model has been calibrated with the observed data by Mississippi Department of Environmental Quality. In addition, some of the developed model inputs have been substantiated by soil sampling and edge-of-field data. The results of soil sampling and edge-of-field experiment were presented separately.

Soil Sampling
The soil samples from St. Louis Bay study area were collected and analyzed. The sampling results in a similar agricultural watershed at Mississippi State University Pontotoc Ridge/Flatwoods Branch Experiment Station (Evans, 2005) were also used to confirm the model inputs.
The sampling locations in St. Louis Bay watershed were selected in Pearl River, Hancock, and Harrison counties so as to represent the widest possible array of soil types and agricultural land practices. These sites span the study area (Figure 4). The samples represent these series and the associated soil series. Taken together they represent between 40 to 60 percent of the soils in the Bay St. Louis Study area (USDA-NRCS, 2006). The soil samples were taken from 0 to 1 in., 1 to 6 in, 6 to 18 in., and 18 to 30 in. depths. These depths were chosen to provide the basis for parameter estimation for the various soil depths in the AGCHEM module. Soil nutrient quantities were determined using procedures described by Sparks et al. (1996).
Figure 5 and Figure 6 represent the typical spatial distribution of nitrogen and phosphorus in the measured soil samples from the study area. It can be observed that most of the nutrients in the soil samples are found in the soil depth of 0-6 inches (Figure 5 and Figure 6). In addition, the amount of nutrients in the soil depth of 0-1 inch is much greater than that in the soil depth of 1-6 inch.
The development of nitrogen and phosphorus loadings from fertilization, for the St. Louis Bay watershed model, was documented by Huddleston et al. (2003). The nutrient loadings from hay cropland dominate the nutrient contributions from croplands due to the comparatively larger area and higher unit loading rates. For the developed model, it was assumed that the typical application method is broadcast, which only applies the nutrients to the surface soil layer, with the prescribed soil depth of 0 – 0.5 inch.
The results of nutrient distribution in the soil samples proved the validity of assumed nutrient distribution in the developed model. For the model, all the nutrients from fertilization were assumed to be applied in the top soil layers, which could cause the majority of nutrients to be retained in the soil surface layer, which was demonstrated by the nutrient distribution in the soil samples. In addition, the phosphorus transportation with vertical water flow could cause the stepwise-deceasing of phosphorus level in the vertical soil profile, which was also demonstrated by the nutrient distribution in the soil samples.
Evans (2005) investigated the relationships among phosphorus concentrations and soil properties and land use were explored in a 259-acre agriculture watershed on the Mississippi State University Pontotoc Ridge/Flatwoods Branch Experiment Station. Over the entire study area, 400 soil samples were collected and analyzed for PH, Mehlich III-extractable P (M3P), Olsen-extractable P, and total P.
The spatial distribution of mean M3P in the soil was shown in Figure 7. The soil samples were collected from four layers: thatch, 0-3 inch, 3-9 inch, and 9-18 inch. In order to be compared with model inputs, the spatial distribution of M3P was converted to be compatible with soil layers specified in our model: 0-0.5 inch, 0.5-6.5 inch, and 6.5-47.5 inch. In order to do this, several assumptions have to be made. The measured M3P in each soil layer is assumed to be uniformly distributed over the entire soil layer. The measured M3P in the Thatch is assumed to be in the soil surface layer specified by the model since Thatch is the interface between vegetation and soil. The adjusted spatial distribution of M3P was shown in Figure 8. It can be observed that the result also proved the validity of assumed nutrient distribution in the developed model, which assumed that all the phosphorus from fertilization was assumed to be applied in the top soil layers.

Edge-of-field Experiment
Beavers (2005) initiated the edge-of-field experiment to evaluate the dynamics and forms of phosphorus in sediment and runoff, determine phosphorus losses under two tillage and two planting treatments, and examine phosphorus concentrations resulting from rainfall and runoff influence. This study was conducted at the North Mississippi Branch of the Mississippi Agriculture and Forestry Experiment Station (MAFES) in Holly Springs, Mississippi, in conjunction with the USDA-ARS National Sedimentation Laboratory in Oxford, Mississippi. The applied phosphorus fertilizer is chicken litter. Edge-of-field data also confirmed the nutrient inputs for the developed model. For the developed St. Louis Bay watershed model, the phosphorus input from fertilization practice was assumed to be 100% in the form of inorganic phosphorus. This assumption has been substantiated by the edge-of filed data. The majority of phosphorus in the surface runoff was in the form of inorganic phosphorus (Figure 9) (Beaver, 2005). The nature of phosphorus in poultry litter is like phosphorus fertilizers.
The developed phosphorus mass balance in the hay cropland for the St. Louis Bay watershed model was also substantiated by the edge-of-field experiment. The results from edge-of-field experiment indicated that the ratio of phosphorus uptake to phosphorus input ranged from 6.15% to 9.82%, whereas the ratio in the St. Louis Bay model was 8.69% (Table 3). It can be observed that the phosphorus input for the St. Louis Bay model is lower than that by Beavers (2005). This is because the phosphorus rate used in the St. Louis By model reflects the average fertilization condition over the entire simulation period from 1965 to 2001, whereas the loading rate from Beavers (2005) represents the loading rate of recent decade. In both cases, the phosphorus application rates were developed based on nitrogen application rate, which could cause the higher levels of phosphorus in the surface runoff and in-stream.

Experiment Design
In this research, the calibrated St. Louis Bay watershed model was used to assess the impacts of agroforestry practices on flow and water quality processes. The land use distribution in sub-watershed 019 was re-configured, but the hydraulic properties including channel length, cross section, and slope, were kept unchanged (Figure 3). For the designed base scenario, it was assumed that sub-watershed 019 was a pure pasture land with a total area of 2,301 acres (Table 4). Scenario 1 is a designed agroforestry ecosystem with 50% of pasture land and 50% of forest (Table 4).
Generally, forest land has higher infiltration rate, interception capacity, roughness, and evaportranspriation than pastureland. The differences in the hydrological properties between pasture land and forest were described by 4 parameters of HSPF. These four parameters are INFILT, NSUR, CEPSC, and LZETP (Table 5). INFILT is an index representing soil infiltration capacity. It controls the overall distribution of the available moisture among surface runoff, interflow, upper zone storage, lower zone storage, and groundwater. High values of INFILT will produce more water in the lower zone and groundwater, and result in higher base flow to streams, whereas lower values produce more upper zone and interflow storage water, and result in higher direct surface flow and interflow (Huddleston et al., 2003). Parameter of NSUR is the empirical Manning’s coefficient used to estimate the overland flow plane. Value of CEPSC determines how much of rainfall has been intercepted by the vegetation, and eventually evaporated. As defined by USEPA (2000), LZETP is a coefficient to define the evapotranspiration opportunity; it affects evapotranspiration from the lower zone which represents the primary soil moisture storage and root zone of the soil profile.

Additional Objective 6. Characterization of microclimate and evapotranspiration within young and mature silvopasture versus open pasture landscapes.

HOBO© (Onset Computer Corp., Bourne MA 02532) weather stations were established in silvopasture and open pasture landscapes at two locations in Fall 2005 to monitor microclimatic conditions. The first location was the young longleaf pine silvopasture at the USDA-NRCS-Jimmy Carter Plant Materials Center, Americus GA (32°3'N, 84°14'W) and the second site was the mature loblolly pine silvopasture at Owens’ Farm, Chipley, Florida (30º46’46.53” N, 85º32’18.51” W). Within each pasture type at each location, total solar radiation, air temperature, relative humidity, wind and gust speeds, soil temperatures at 5-cm and 10-cm depths, and dew point have been sampled every five minutes for a two-minute period since Fall 2005.
Weather data collected from both silvopasture and open-pasture landscapes at each research site for the whole study period were classified on monthly or seasonal basis (winter: Dec. 21 – March 20, spring: March 21 – June 20, summer: June 21 – Sept. 20, fall: Sept. 21 – Dec. 20). Data in each season were further sub-classified according to diurnal periods; morning (winter 600-1000 h, other seasons 500-1000 h), midday (1000-1400h), afternoon (winter 1400-1700 h, other seasons 1400-1800h), evening (winter 1700-2200 h, other seasons 1800-2200 h), and night (winter 2200-600 h, other seasons 2200-500 H) to find out whether pasture type had a significant effect on weather variables depending on the time of a day.
Data were arranged as a split plot in time when time of a day was not considered with pasture type as a main plot and season or month as a sub-plot, or split-split plot in time design with pasture type as a main plot, season as a sub-plot and time of a day as a sub-sub-plot. Mutivariate Analysis of Variance (MANOVA) test was performed using GLM procedure (Johnson and Wichern, 2002; Littell et al., 2005) in SAS 9.1 to analyze the data. Probability level for rejection of the null hypothesis (Ho: weather variables would not differ between silvopasture and open-pasture regardless of season, month, or time of a day) was set at 0.05.
Evapotranspiration is also being calculated for both study sites using the FAO Penman-Monteith equation (http://www.fao.org/docrep/X0490E/x0490e06.htm).

Additional Objective led by PhD graduate student (directed by Dr. Mary Goodman):
Objective 7. Quantify diurnal distribution and behavior of cattle in loblolly-pine (Pinus taeda) silvopasture versus open-pasture landscapes and relate forage production and quality and microclimatic differences to possible differences in cattle distribution and behavior between the landscapes.

Study Site and Design
This study was conducted during three portions (March, June, September) of the warm portion of the 2007 grazing season at Owens’ Farm, Chipley, Florida (30º46’46.53” N, 85º32’18.51” W). Two 5-ha pastures were used: one within a 20-yr old loblolly-pine (Pinus taeda)-bahiagrass (Paspalum notatum) silvopasture with a tree density of 247 ha-1 and a nearby open bahiagrass pasture with unlimited access to a 1-ha wooded area (Fig. 10). Crimson clover (Trifolium incarnatum) was managed to reseed in both pastures but was only present in March.
To assess the distribution patterns and behavior of cattle, the whole area of silvopasture under study was delineated into four (March) or five zones (June, September: one zone open); one zone contained the water source (Fig. 10a). The open-pasture was delineated into six zones with the area around the water source and the wooded (tree) habitat designated as separate zones (Fig. 10b). Cattle had free access to every zone in each pasture. The experimental design was a split-split-plot in time with pasture type as the main plot, observation date as the split-plot, and portions of a diurnal period as the split-split plot.

Observation of Cattle Distribution and Behavior
Six to eight (March: 6 each pasture; June: 6 silvopasture, 7 open-pasture; September: 8 each pasture) mature dry beef cows (Bos taurus) were stocked onto each pasture two days prior to each observation day. Distribution patterns and behavior of each animal were monitored simultaneously (one observer per landscape) in each pasture from 6-m high tree stands located such that grazing animals would not be distracted as a result of the observer’s activities (Fig. 10). Observations were made with binoculars every 15 minutes and recorded from dawn-to-dusk (diurnal) for each observation date in March, June, and September 2007. The diurnal observation period was 13 hours in March for both pastures, 15 hours for silvopasture and 15.25 hours for open-pasture in June, and 12.75 hours for silvopasture and 12 hours for open-pasture in September. Behavior categories recorded included grazing, lying, and loafing; loafing represented activities other than grazing or lying, such as moving, standing, scratching, or playing.

Forage Sample Collection
To estimate available forage biomass, ten random 0.25 sq m quadrats were clipped to 5 cm within each pasture on the day prior to each observation day. Forage tissue samples were dried at 60ºC for 72 hrs then ground to pass a 1-mm sieve. All tissue samples were analyzed for acid detergent fiber (ADF) following the method of Goering and Van Soest (1970) and for nitrogen (N) using the Kjeldahl method to estimate total digestible nutrients (TDN) and crude protein (CP).

Weather Data Collection
HOBO© (Onset Computer Corp., Bourne MA 02532) weather stations were established in each landscape to monitor microclimatic conditions. Within each landscape, total solar radiation, air temperature, relative humidity, wind and gust speeds, soil temperatures at 5-cm and 10-cm depths, and dew point were sampled every five minutes for a two-minute period during the observation periods.

Data Analysis
Distribution patterns of cattle were quantified using the Distribution Evenness Index (DEI) developed by Zuo and Miller-Goodman (2003). For analysis, DEI and behavior data as well as weather data for each observation day and landscape type were divided into three diurnal periods: morning (dawn-1100h), midday (1100h-1400h), and post-midday (1400h-dusk).
Because of a serious non-normality, DEI and behavior data were analyzed using the Wilcoxon rank-sum test (Gibbons and Chakraborti 2003) in SAS package 9.1. Forage biomass and quality data were also analyzed using the Wilcoxon rank-sum test because of an inadequate number of observations to verify the assumptions of parametric tests. Exact P value was used for the hypothesis test; probability level of alpha for rejection of the Ho (null hypothesis) was set at 0.05. Average values of weather parameters were tabulated for all observation dates and diurnal periods for each pasture type.