Water quality effects of short-rotation pine management for bioenergy feedstocks in the southeastern United States

Forest Ecology and Management 400 (2017) 181–198

Contents lists available at ScienceDirect

Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco

Water quality effects of short-rotation pine management for bioenergy feedstocks in the southeastern United States q Natalie A. Griffiths a,⇑, C. Rhett Jackson b, Menberu M. Bitew b,1, Allison M. Fortner c, Kevin L. Fouts b, Kitty McCracken c, Jana R. Phillips c a b c

Climate Change Science Institute and Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA Warnell School of Forestry and Natural Resources, University of Georgia, Athens, GA 30602, USA Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA

a r t i c l e

i n f o

Article history: Received 9 March 2017 Received in revised form 1 June 2017 Accepted 5 June 2017 Available online 12 June 2017 Keywords: Intensive silviculture Short-rotation woody crops Nitrogen Concentrated flow tracks Interflow Surface water

a b s t r a c t Growing interest in renewable and domestically produced energy motivates the evaluation of woody bioenergy feedstock production. In the southeastern U.S., woody feedstock plantations, primarily of loblolly pine (Pinus taeda), would be intensively managed over short rotations (10–12 years) to achieve high yields. The primary differences in managing woody feedstocks for bioenergy production vs for pulp/sawtimber production include a higher frequency of pesticide and fertilizer applications, whole-tree removal, and greater ground disturbance (i.e., more bare ground during stand establishment and more frequent disturbance). While the effects of pulp/sawtimber production on water quality are wellstudied, the effects of growing short-rotation loblolly pine on water quality and the efficacy of current forestry Best Management Practices (BMPs) have not been evaluated for this emerging management system. We used a watershed-scale experiment in a before-after, control-impact design to evaluate the effects of growing loblolly pine for bioenergy on water quality in the Upper Coastal Plain of the southeastern U.S. Intensive management for bioenergy production and implementation of current forestry BMPs occurred on
50% of two treatment watersheds, with one reference watershed in a minimally managed pine forest. Water quality metrics (nutrient and pesticide concentrations) were measured in stream water, groundwater, and interflow (i.e., shallow subsurface flow) for a two-year pre-treatment period, and for 3.5 years post-treatment. After 3.5 years, there was little change to stream water quality. We observed a few occurrences of saturated overland flow, but sediments and water dissipated within the streamside management zones in over 75% of these instances. Stream nutrient concentrations were low and temporal changes mainly reflected seasonal patterns in nitrogen cycling. Nitrate concentrations increased in groundwater post-treatment to <2 mg N L1, and these concentrations were below the U.S. drinking water standard (10 mg N L1). Applied pesticides were almost always below detection in streams and groundwater. Overall, these findings highlight that current forestry BMPs can protect stream water quality from intensive pine management for bioenergy in the first 3.5 years. However, groundwater quality and transit times need to be considered in these low-gradient watersheds of the southeastern U.S. that are likely to become an important location for woody bioenergy feedstock production. Ó 2017 Elsevier B.V. All rights reserved.

1. Introduction q

This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http:// energy.gov/downloads/doe-public-access-plan). ⇑ Corresponding author. E-mail address: [email protected] (N.A. Griffiths). 1 Present address: USDA-ARS Southwest Watershed Research, Tucson, AZ 85719, USA. http://dx.doi.org/10.1016/j.foreco.2017.06.011 0378-1127/Ó 2017 Elsevier B.V. All rights reserved.

Forestry is a large part of the economy in the southeastern U.S. (Prestemon and Abt, 2002). Over half of the timber harvested in the U.S. comes from southern forests, and loblolly pine (Pinus taeda) is the dominant timber species in this region (Smith et al., 2009; Wear and Greis, 2002). Forest products are mainly used for sawtimber and pulpwood. However, due to the growing interest in renewable and domestically produced energy (e.g., the Energy Independence and Security Act of 2007), there is an increasing potential to utilize woody feedstocks for bioenergy (i.e., biofuels, biopower) (U.S. Department of Energy, 2016). There are two

182

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

silvicultural strategies for producing woody biomass for bioenergy: (1) the utilization of tree tops and branches that are not removed for sawtimber and pulpwood production, and (2) the production of short-rotation plantations where the whole tree is harvested and utilized for bioenergy. Under the latter scenario, trees would be grown on a short rotation (10–12 years; Munsell and Fox, 2010) with more intensive management (mechanical and chemical site preparation, multiple pesticide and fertilizer applications) than for sawtimber or pulpwood in order to achieve high yields (Fox et al., 2007; Hinchee et al., 2009; Scott and Tiarks, 2008; Zhao et al., 2016). Whole-tree harvest can result in lower residual forest floor biomass relative to roundwood harvest, but quantification is difficult, and the few quantitative studies suggest that the reduction in forest floor biomass can range from 18 to 81% (Fritts et al., 2014; Klockow et al., 2013). In the southeastern U.S., the native, fast-growing, and resilient loblolly pine is the primary candidate species for bioenergy feedstock production (Kline and Coleman, 2010) and could be grown on a short rotation followed by whole-tree harvesting. However, the environmental effects of intensive production of short-rotation loblolly pine for bioenergy have not yet been evaluated at the watershed scale. Local water quality effects (i.e., increased concentrations and fluxes of nitrogen, phosphorus, suspended sediment, and pesticides) and associated downstream effects (i.e., eutrophication, habitat degradation, impacts to aquatic organisms, and increased water treatment costs) are primary concerns of forest management. The water quality effects of silvicultural operations primarily depend on the amount, connectivity, and duration of bare soils and on the coincidence of chemical application (fertilizers, pesticides) with the presence of bare soils. Harvest and site preparation equipment can expose and compact bare mineral soils, and the resulting surface runoff (i.e., Horton overland flow) can mobilize soil particles and potentially increase concentrations of suspended sediments, sediment-bound nutrients, dissolved nutrients, and pesticides in stream water (Binkley and Brown, 1993; Yoho, 1980). Forest roads, landings, and skid trails have been repeatedly identified as the dominant sources of sediment from silvicultural operations (e.g., Hoover, 1952; Megahan and Kidd, 1972; Rivenbark and Jackson, 2004). Dissolved stream water nutrient concentrations (primarily nitrate) can also increase following timber harvest (Blackburn and Wood, 1990; Likens et al., 1970; Swank and Webster, 2014; Wynn et al., 2000) due to a lack of vegetative uptake and warmer soil temperatures that accelerate residue decomposition and nitrification (Grace, 2005; Vitousek and Melillo, 1979). Fertilization can also increase stream water nutrient concentrations (Binkley and Brown, 1993; McBroom et al., 2008), especially if the fertilizers are applied on or near streams, in the form of ammonium nitrate, at frequent intervals, or shortly before storm events (Beltran et al., 2010; Binkley et al., 1999). Stream water nutrient concentrations that are elevated due to forestry practices tend to return to baseline conditions within months to a few years (Aust and Blinn, 2004; Boggs et al., 2016). As the newly planted forest grows, the expanding canopy and associated litter fall reduce bare soils. Thus, in the rapidly growing pine plantations of the southeastern U.S., overland flow and associated water quality issues are expected to occur during the first 2–3 years after harvest. These water quality effects can be minimized if forestry Best Management Practices (BMPs) are implemented (Anderson and Lockaby, 2011a, 2011b; Aust and Blinn, 2004; Cristan et al., 2016; Witt et al., 2013). Forestry BMPs include leaving a vegetated streamside management zone (SMZ) between the field and the stream, and minimizing bare soils and soil compaction during silviculture activities (South Carolina Forestry Commission, 1998). Forestry BMPs are state-specific, and in the southeastern states, BMP implementation is voluntary, but adoption is very high (87% overall implementa-

tion) (NCASI, 2009). Many studies have evaluated the effectiveness of BMPs, and most have found beneficial effects on water quality compared to forestry without BMPs (Anderson and Lockaby, 2011a, 2011b; Aust and Blinn, 2004; Cristan et al., 2016; Witt et al., 2013). For instance, forestry BMPs are effective at minimizing herbicide transport to streams (Scarbrough et al., 2015), with <1– 2% of applied herbicides reaching streams during storm events (McBroom et al., 2013). These SMZs can reduce the amount of nutrients transported to streams (Pratt and Fox, 2009; Secoges et al., 2013; Wynn et al., 2000), as nutrients can be taken up by vegetation or nitrate can be denitrified (Hill, 1996; Peterjohn and Correll, 1984; Sweeney and Newbold, 2014). However, SMZs are often not 100% effective at removing nutrients (Marchman et al., 2015; McBroom et al., 2008), especially if rain events occur shortly after fertilization (Beltran et al., 2010). Streamside management zones also reduce sediment inputs to streams (Carroll et al., 2004; Ward and Jackson, 2004) by minimizing the occurrence of overland flow such that the undisturbed soil of the SMZ can disperse and infiltrate flows before reaching the stream (Pinho et al., 2008; White et al., 2007). However, SMZs retain proportionally fewer small-diameter than large-diameter sediments (Sweeney and Newbold, 2014), and breakthroughs can occur, especially in gullies and areas of concentrated flow (Rivenbark and Jackson, 2004). While many studies have evaluated the effectiveness of forestry BMPs at mitigating effects on water quality, it is not known whether current forestry BMPs are adequate to protect water quality during short-rotation pine production or whether bioenergy-specific BMPs are needed (Shepard, 2006). Shortrotation silviculture involves more frequent ground disturbance, greater competition control with herbicides, and potentially higher fertilizer application. For these reasons, it is not clear if current forestry BMPs are sufficient to protect water quality in watersheds supporting short-rotation woody feedstocks. Forestry BMPs minimize the movement of pollutants to streams by overland flow (Edwards and Williard, 2010). However, the effectiveness of BMPs in groundwater-dominated watersheds is not well known. For instance, in low-gradient watersheds of the Upper Atlantic Coastal Plain in the southeastern U.S., the source of streamflow and stream water nitrate is primarily groundwater (Du et al., 2016; Griffiths et al., 2016; Klaus et al., 2015). Shallow subsurface flow can be important, especially during storms (Du et al., 2016), but downslope travel distances are short (
10’s of m), and thus stream flow contributing areas generally originate within the riparian zones (Jackson et al., 2014). Therefore, it is important to understand the effects of intensive forestry for bioenergy not only on stream water quality but also on groundwater quality in low-gradient, groundwater-dominated watersheds. In this study, we used a watershed-scale experiment in a before-after, control-impact design to examine the effects of growing short-rotation loblolly pine for bioenergy on water quality in the southeastern U.S. We selected three watersheds with minimally managed loblolly pine in the Upper Atlantic Coastal Plain of South Carolina, and examined baseline water quality in groundwater, interflow (i.e., shallow subsurface flow), and stream water for two years. Over the next 3.5 years, water quality sampling continued as
50% of two treatment watersheds were harvested, planted with loblolly pine seedlings, and managed for shortrotation pine production (including multiple fertilizer and pesticide applications). The third watershed was not manipulated and served as a reference. For the first 3 years after harvest, we also identified and characterized locations where overland flow was moving from the harvest/plantation units into the SMZs in sufficient quantity to mobilize the litter layer and sediments. All silviculture practices in the two treatment watersheds followed South Carolina Forestry BMPs (South Carolina Forestry Commission, 1998). We predicted that silvicultural practices for

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

bioenergy would have a minimal effect on stream water quality as runoff and interflow are not the predominant flow paths contributing to streamflow in these low-relief watersheds (Du et al., 2016; Jackson et al., 2014; Klaus et al., 2015), and previous studies have demonstrated the effectiveness of BMPs, especially SMZs, in protecting surface water quality from forestry activities (e.g., Anderson and Lockaby, 2011a, 2011b; Aust and Blinn, 2004; Cristan et al., 2016; Witt et al., 2013). However, because groundwater is the dominant flow path for both water and nitrogen in these watersheds (Griffiths et al., 2016; Klaus et al., 2015), we predicted that nutrient concentrations (particularly nitrate) may increase in groundwater in the treatment watersheds following harvest and silvicultural activities. The results from this study will likely be applicable to watersheds of the Upper Atlantic Coastal Plain in the southeastern U.S. that have similar topography, hydrology, and geology and that apply similar BMPs. Evaluating the effects of intensive, short-rotation pine production for bioenergy on both stream and groundwater quality is a necessary first step in assessing the environmental sustainability of woody biomass production for bioenergy in the southeastern U.S. 2. Methods 2.1. Study sites The watershed experiment took place on the Department of Energy’s Savannah River Site (SRS), near New Ellenton, South Carolina, USA (33°160 N, 81°370 W). The 3 study watersheds (watersheds R, B, and C) are at the northern extent of the Upper Atlantic Coastal Plain physiographic region. The climate is humid subtropical, with a mean annual precipitation of 1225 mm, and a mean annual temperature of 18 °C (Kilgo and Blake, 2005). Annual precipitation during the experiment was 918 mm in 2010, 909 mm in 2011, 1075 mm in 2012, 1121 mm in 2013, 1036 mm in 2014, and 1193 mm in 2015. The SRS was established in 1951, and land previously used for row-crop agriculture was reforested by the U.S. Forest Service. Prior to any manipulations associated with this watershed experiment, the vegetation of the uplands was primarily longleaf pine (P. palustris), loblolly pine (P. taeda), and slash pine (P. elliottii), and these planted pines were minimally managed and in the second rotation following the abandonment of row-crop agriculture. The uplands have a flat topography (
2–3% slope) and well-drained, sandy soils with a loamy to clayey subsoil. The riparian areas are

183

composed of a mixed hardwood vegetation (primarily sweetgum; Liquidambar styraciflua) with flat and wide floodplain valleys, and organic-rich, saturated soils. Headwater streams have indistinct channels and flow intermittently, with generally higher stream flow in winter and spring and low or no flow in summer. In April/May 2011, there was a regional drought and all 3 streams dried up for approximately one year. The streams are characterized as blackwater streams due to high concentrations of dissolved organic matter (Meyer, 1990). Blackwater streams can be nitrogen limited (Mallin et al., 2004), with inorganic nitrogen primarily in the form of ammonium (Jager et al., 2011). The characteristics (i.e., topography, geology, vegetation) of the 3 adjacent watersheds are fairly similar, but the watersheds vary in size: watershed R is 45 ha, watershed B is 169 ha, and watershed C is 117 ha. These watersheds are part of the larger Fourmile Branch watershed that drains to the Savannah River. 2.2. Silviculture for short-rotation pine management The watershed experiment used a before-after, control-impact design (Stewart-Oaten et al., 1986). For 2 years beginning in 2010, water samples were collected for chemical analyses from all 3 watersheds to characterize the pre-treatment conditions (sampling details described in Section 2.4). Starting in late February 2012, approximately 50% of the two treatment watersheds (B, C) were harvested for commercial timber (85.4 ha in watershed B, 62.7 ha in watershed C) (Table 1), while the third watershed (R) was left unmanipulated (i.e., reference watershed). There were a total of 11 units harvested (5 in watershed B, 6 in watershed C), some of which were adjacent to one another (Fig. 1). Harvesting in watershed B and C concluded at the end of May 2012, and approximately 2721 metric tons of low-quality and smalldiameter material was chipped on-site for fuelwood for a local bioenergy facility. There was little residue left on the fields after harvest, and the remaining coarse woody debris averaged 1934 ± 171 kg dry mass ha1 (B. Rau, unpublished data). Harvesting and all other silviculture practices complied with the guidance of forestry BMPs for the state of South Carolina (South Carolina Forestry Commission, 1998). In almost all locations, SMZ widths exceeded the minimums recommended in the BMP manual (12.3 m from each bank) because the typical practice is to mark the harvest boundary where the forest stands shift from upland pines to lowland hardwoods (Rivenbark and Jackson, 2004). SMZ widths on streams and wetlands averaged 27 and 25 m in

Table 1 Timeline and details of silviculture activities that occurred on
50% of the two treatment watersheds (B, C). Start Date

End Date

Activity

Details

02/28/2012

05/31/2012

Harvest
50% of extant forest

06/14/2012

09/13/2012

Site preparation

09/14/2012

09/19/2012

Herbicide

01/28/2013

02/03/2013

Planting of loblolly pine

03/18/2013

03/20/2013

Herbicide

04/22/2013

04/24/2013

Fertilizer

03/03/2014 03/10/2014 03/21/2014

03/04/2014 03/12/2014 04/17/2014

Fertilizer Herbicide Pesticide

02/10/2015

02/12/2015

Fertilizer

03/18/2015

04/12/2015

Herbicide

Harvested a total of 148 ha (2721 metric tons) from 11 units in watershed B (85.4 ha harvested) and watershed C (62.7 ha harvested) Sub-soiling/ripped rows with a tractor (45 cm depth, 3.1 m between rows) parallel to the slope Herbicides broadcast applied: imazapyr (1680 mL ha1 as 4 SL) and glyphosate (6720 mL ha1 as Rodeo) Hand-planted bareroot seedlings (ArborGen MCP AGM 37) at 1346 trees ha1, at a spacing of 2.4 m by 3.1 m Herbicides broadcast applied: sulfometuron methyl (140 mL ha1 as OustXP) and imazapyr (280 mL ha1 as Arsenal AC) Fertilizer broadcast applied: diammonium phosphate (281 kg ha1; 50.6 kg ha1 of N and 56.2 kg ha1 of P) Fertilizer broadcast applied: urea (241 kg ha1; 110.9 kg ha1 of N). Herbicide broadcast applied: sulfometuron methyl (175 mL ha1 as Oust XP) Pesticide injected in soil at the base on each seedling: fipronil for tip moth control (equivalent rate of 1365 mL ha1) Fertilizer broadcast applied: blended urea (179.2 kg ha1; 82.4 kg ha1 of N) and diammonium phosphate (134.4 kg ha1; 24.2 kg ha1 of N and 26.9 kg ha1 of P) Herbicide broadcast applied: sulfometuron methyl (175–210 mL ha1 as Oust XP)

184

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

Fig. 1. Map of the 3 study watersheds located in the Upper Coastal Plain of South Carolina, USA. Approximately
50% of the two treatment watersheds (B and C) were harvested and planted with loblolly pine seedlings for bioenergy production, while the third watershed (R) was the unmanipulated reference. In each watershed, stream water was sampled at the outlet (intermittent stream flume) and at a location upstream with ephemeral flow. Riparian groundwater was sampled from two wells located near each stream sampling site and two near the ephemeral site. Groundwater was sampled from 19 wells (FHR; Forest Hydrology Research) located across the 3 study watersheds and within the larger Fourmile Branch watershed. In watersheds B and C, four FHR wells had a deep (D) and shallow pair, and only the deep well was sampled at FHR011D. Interflow was collected from hillslope trenches located in the established pine forest (Rold-pine, Bold-pine, Cold-pine) or downslope of the newly planted pine stands (Bnew-pine, Cnew-pine). Three concentrated flow tracks were sampled for chemistry (two in watershed B and one in watershed C).

watersheds B and C, respectively. No temporary stream crossings were used during harvest, and no new road surfaces were constructed. Roads featured turnouts to disperse runoff, and log decks were located on uplands far from SMZ boundaries. The silviculture plan was designed to achieve high yields in a short (10–12 year) time period, and the same silvicultural activities occurred on both treatment watersheds (B, C). After harvest, the cut areas were prepared for planting. Site preparation included sub-soiling on contour to form planting rows (3.1 m spacing, 45 cm depth), and broadcast application of herbicides as a mixture of imazapyr and glyphosate (Table 1). After site preparation, bareroot loblolly pine seedlings were hand planted along elevational contours in early spring 2013 at a nominal spacing of 2.4 m by 3.1 m (1346 trees ha1). Pesticides and fertilizers were applied annually in 2013, 2014, and 2015 (Table 1). In 2013, sulfometuron methyl and imazapyr herbicides and diammonium phosphate fertilizer (50.6 kg ha1 of N and 56.2 kg ha1 of P) were applied shortly after planting. In spring 2014, urea fertilizer (110.9 kg ha1 of N) and sulfometuron methyl herbicide were applied. Because of an extensive Nantucket pine tip moth (Rhyacionia frustrana [Comstock]) infestation that severely damaged the planted seedlings during the summer of 2013, the pesticide fipronil was injected at the base of each seedling in March-April 2014. The fipronil application was successful and the pine trees recovered from tip moth damage. In early spring 2015, urea and diammonium phosphate fertilizer (82.4 kg ha1 of N from urea and 24.2 kg ha1 of N and 26.9 kg ha1 of P from diammonium phosphate) and the herbicide sulfometuron methyl were applied. The mobility of pesticides depends on many factors, including soil composition (i.e., percentages of clay and organic matter in soil), soil chemistry (i.e., pH), and climate (i.e., timing of pesticide application relative to the next rainfall) (Neary et al., 1993; Scarbrough et al., 2015). In general, all pesticides applied in this study are considered to have little or low mobility in soils (Jackson et al., 2009; Michael, 2004; USDA Forest Service Southern Region, 2007), and are moderately persistent, with half lives in soil on the order of weeks to months (Jackson et al., 2009; Neary et al., 1993; USDA Forest Service Southern Region, 2007).

2.3. Surveys of concentrated flow tracks Concentrated overland flow enhances transport of sediments and chemicals into the SMZs and creates obvious concentrated flow tracks (CFTs) on the forest floor. CFT surveys were conducted per the methods of Rivenbark and Jackson (2004) along the SMZ boundaries of all plantation units in watersheds B and C. Surveys were conducted for 3 years following harvest (2012–2014). Each year, the survey was conducted in late summer, after the season of intense thunderstorms and before leaf-fall covered the tracks with new leaves. CFT surveys can determine whether overland flow processes may contribute to water quality issues (Litschert and MacDonald, 2009), and if so, can identify the dominant sources of problematic overland flows (Lakel et al., 2010; Lang et al., 2015). If a CFT reaches the stream, this is referred to as a ‘‘breakthrough” and represents a failure of the BMPs Based on the material transported and whether it reached the streams, CFTs were categorized as follows: Category I: water, fine sediments, and sand reach the stream; Category II: water and fine sediments reach the stream; Category III: water (without sediments) reaches the stream; and Category IV: water infiltrates and sediment disperses within the SMZ before reaching the stream. At each CFT, the category, size, and characteristics of the contributing area were noted. 2.4. Water quality sampling We collected samples for chemical analysis from CFTs that formed in the uplands, interflow from trenches in the uplands, groundwater from wells installed in the uplands, groundwater from wells installed in the SMZs/riparian zone, and stream water (Fig. 1). Approximately 5.5 years of water chemistry data are presented here, with 2 years of pre-treatment data (January 2010 – February 2012), and 3.5 years of post-treatment data (February 2012 – October 2015). While sampling of stream water began in January 2010, sampling at other sites began later (described below). The sampling methodologies and results are presented in an order that follows the predominant flow path of water and solutes in these watersheds (i.e., from surface water in the planted

185

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

Beginning in May 2010, riparian groundwater was collected approximately monthly from 4 wells per watershed. Two wells were installed in the hydric riparian soils near the outlet of each watershed (one well on each side of the stream) and two wells were installed in the same configuration near the upstream ephemeral sampling site (Fig. 1). The wells collected groundwater at 1.7–2.0 m below the soil surface. Samples were not collected when there was not a sufficient volume available for analyses. Stream water samples were collected by hand approximately weekly at the outlet of each watershed, and at a location upstream where flow was ephemeral (Fig. 1). Stream flow at the outlet was intermittent, and there were periods when samples could not be collected because the stream was dry. Samples for nutrient analyses were collected at the frequencies described above. Samples for pesticide analyses were collected seasonally from CFTs, interflow, groundwater, riparian groundwater, and streams, and from streams during storm events. One to five stream water samples were collected per storm using an automated water sampler, with 3 storms sampled in watershed B, 4 storms sampled in watershed C, and 1 storm sampled in watershed R. These storms occurred after the 2012 and 2013 herbicide applications (Table 1). Since different pesticides were applied in different years (Table 1), water samples were analyzed for a given pesticide for one year after that pesticide was applied. Fipronil was only measured on samples collected from groundwater wells. Pretreatment samples for sulfometuron methyl, glyphosate, and imazapyr analysis were collected from streams and groundwater before the first pesticide application, but no pre-treatment samples were collected prior to the application of fipronil.

areas (CFTs), to shallow subsurface water, to groundwater, to riparian zones, and then to the stream). CFTs that formed in the cut/planted areas in watersheds B and C were sampled beginning in March 2014. When a CFT was visually identified, a grab sample was collected approximately weekly for water chemistry until the CFT dried up. CFTs were not surveyed in watershed R, and no CFT samples were collected from that watershed. Interflow was collected from subsurface flow trenches that were constructed within hillslopes in the 3 watersheds. The design of the trenches is described in detail in Du et al., (2016) and Jackson et al., (2016). Briefly, a trench was dug in the soil, and subsurface drains were installed to collect water flowing into the trench. In each watershed, one trench was installed in the established pine forest (Rold-pine, Bold-pine, Cold-pine; Fig. 1), and sampling from these trenches began in February 2011. One trench was also installed in the clearcut/planted area, with the outlet of the collection drains at the edge of the planted area and SMZ in each treatment watershed (Bnew-pine, Cnew-pine; Fig. 1), and thus collected subsurface water flowing below the newly planted pine stands. These trenches were installed after the pine was planted, and sampling began in February 2014. When the trenches were flowing (primarily during storms), a composite water sample was collected by hand at the trench drain outlet. Occasionally, multiple samples were collected over an event, but because interflow nutrient dynamics during storms have been evaluated previously (Griffiths et al., 2016), we report the mean nutrient concentrations during each event to compare interflow nutrient concentrations across watersheds. Groundwater was sampled approximately monthly from 19 wells that were located across the watersheds and primarily in upland areas. There were 3 wells within watershed R, 4 wells within watershed B, and 7 wells within watershed C (Fig. 1). Five wells were outside the 3 watersheds, and within the larger Fourmile Branch watershed (Fig. 1). In watersheds B and C, there were 2 pairs of adjacent shallow and deep wells installed in the harvested/planted units. The screened sections of the shallow wells were 13.0–19.3 m below the soil surface and the deep wells were 25.6–43.6 m below the soil surface (Table 2). For all other wells, the depths ranged from 2.7 to 13.1 m (Table 2). Sampling of the deep wells began in December 2011, and sampling of the adjacent shallow wells began in September 2012. Sampling of all other groundwater wells began in September 2011. Wells were usually sampled on the same day, but on occasion, sampling occurred over multiple days (all wells sampled within a one week period).

2.5. Water quality analyses Water quality analyses included measurements of water temperature, specific conductivity, nutrient concentrations (nitrateN, ammonium-N, soluble reactive phosphorus, total nitrogen, and total phosphorus), dissolved organic carbon (DOC) concentrations, and pesticide concentrations (fipronil, sulfometuron methyl, imazapyr, and glyphosate). Water temperature and specific conductivity were measured in the field using a hand-held probe (EcoSense EC300, YSI Inc.). Water samples for nutrient chemistry were brought back to the laboratory on ice, filtered (0.7-mm nominal pore size) into polyethylene bottles, and frozen at 20 °C until analysis. An additional unfiltered water sample was collected from

Table 2 Mean (standard deviation) nitrate, ammonium, and soluble reactive phosphorus (SRP) concentrations across all sampling dates in the 19 groundwater wells in watersheds R, B, C, and the larger Fourmile watershed, and depth to the screen in each well (m below soil surface). Wells with ‘D’ in the name indicate deep groundwater wells. Well

Watershed

Nitrate (mg N L1)

Ammonium (mg N L1)

SRP (mg P L1)

Depth of screen (m)

FHR014 FHR015 FHR001 FHR012 FHR013 FHR016 FHR004 FHR005 FHR010 FHR003 FHR006 FHR007 FHR008 FHR009 FHR014D FHR015D FHR011D FHR013D FHR016D

B B C C C C R R R Fourmile Fourmile Fourmile Fourmile Fourmile B B C C C

652 (166) 1137 (468) 168 (205) 108 (121) 887 (369) 1244 (401) 327 (176) 242 (319) 220 (80) 122 (74) 442 (97) 911 (158) 282 (39) 733 (243) 418 (154) 42 (28) 203 (141) 55 (124) 151 (60)

16 23 52 21 21 15 20 35 11 14 15 15 21 20 21 23 22 18 18

3 (2) 3 (3) 17 (23) 2 (1) 2 (2) 5 (12) 3 (4) 2 (2) 2 (1) 103 (92) 2 (2) 3 (3) 2 (2) 4 (16) 4 (3) 3 (2) 8 (19) 2 (2) 4 (4)

14.2 16.6 3.9 10.6 19.3 13.0 13.1 13.0 6.4 2.7 10.5 5.4 6.8 12.4 25.6 37.6 25.6 43.6 28.6

(17) (28) (68) (26) (24) (21) (22) (62) (9) (7) (8) (10) (20) (14) (16) (19) (25) (15) (20)

186

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

the streams for total phosphorus (TP) and total nitrogen (TN) analyses. Water samples for pesticide analyses were filtered into polyethylene bottles, refrigerated at 4 °C for 1–2 days, and then shipped to an analytical laboratory (OMIC USA Inc., Portland OR). Water samples for DOC analysis were filtered into amber glass vials, acidified with 2 drops of 6 N HCl, and refrigerated (4 °C) until analysis. The main focus of this study was to quantify the effects of shortrotation pine production on nutrients (N, P) and pesticides. Temperature, specific conductivity, and DOC concentrations were ancillary measurements used to elucidate spatial and temporal patterns in N and P chemistry. Nitrate-N concentrations were measured using the cadmium reduction method, ammonium-N concentrations were measured using the phenol hypochlorite method, and soluble reactive phosphorus (SRP) concentrations were measured using the molybdateblue method (APHA, 2005) on a SEAL Analytical AA3 autoanalyzer. TP concentrations were measured using the molybdate-blue method on a SEAL Analytical AA3 autoanalyzer following persulfate digestion (APHA, 2005). TN concentrations were measured using the combustion oxidation and chemiluminescence detection method on a Shimadzu TOC-L CSH/CSN analyzer. DOC concentrations were measured using the high-temperature combustion catalytic oxidation method on a Shimadzu TOC-L CSH/CSN analyzer. Pesticide samples were analyzed at the OMIC USA Inc., analytical facility. Samples for fipronil, sulfometuron methyl, and imazapyr were diluted with an equal volume of methanol and then filtered through a 0.2-mm polytetrafluroethylene filter prior to analysis. Samples for glyphosate were passed through an anion exchange column, derivatized using trimethylorthoacetate, and then cleaned using a florisil column. Isotopic glyphosate (13C, 15 N) was added to each sample to correct for sample differences in derivatization efficiency. External calibration standards were used for all analyses. Each analytical batch included a positive control (spike addition to a duplicate sample) and a reagent blank. All samples, standards, and controls were analyzed using Ultra Performance Liquid Chromatography with a C18 chromatographic column and triple quadrupole detectors (Waters [Premier XE] or AB Sciex [6500 QTRAP]) using electrospray ionization. The detection limits were 1 mg L1 for fipronil and sulfometuron methyl, and 2 mg L1 for glyphosate and imazapyr.

Two analyses were used to examine the effect of silviculture treatments on water quality. For riparian groundwater and stream water chemistry, the difference in nutrient concentration between the treatment (B or C) and reference (R) watershed (Dnutrient) on each sampling day was calculated (B-R, C-R), and a one-way ANOVA was used to examine whether Dnutrient differed in the pre-treatment vs post-treatment period. The post-treatment period was designated as the first day of harvest of the extant forest (February 28, 2012; Table 1). Dnutrient was only calculated when data from both the treatment (B or C) and reference (R) watersheds were available (i.e., Dnutrient was not calculated when one stream was dry). For riparian groundwater chemistry, Dnutrient was calculated from the mean concentration from the intermittent wells, and only when a sample was collected from both wells. Fewer samples were collected from the ephemeral streams and riparian groundwater wells because those sites were dry more often, so the ephemeral stream water and riparian groundwater chemistry data were not used in the analysis of Dnutrient. Because most groundwater wells were installed just before or after the start of the post-treatment period, a statistical analysis of Dnutrient between the pre-treatment and post-treatment period could not be conducted. Instead, we examined whether mean nutrient concentration and Dnutrient changed in each watershed during the post-treatment period using regression, with nutrient concentration or Dnutrient as the dependent variable, and days since post-treatment as the independent variable. Dnutrient was calculated using the mean groundwater nutrient concentration for each watershed, and only when samples were collected from all of the wells per watershed. Dnutrient was calculated separately for deep (Bdeep-R, Cdeep-R) and shallow (Bshallow-R, Cshallow-R) groundwater wells, and changes in groundwater nutrient concentrations in the treatment watersheds were also compared to changes in groundwater nutrient concentrations in the wells in Fourmile watershed (Bshallow-Fourmile, Cshallow-Fourmile, Bdeep-Fourmile, Cdeep-Fourmile). Mean groundwater nutrient concentrations in Fourmile were calculated from wells FHR003, FHR007, FHR008, FHR009 (Fig. 1); data from FHR006 were excluded from the analysis due to a limited number of samples collected in 2011–2013.

2.6. Statistical analyses

3. Results

Statistical analyses were used to examine whether nutrient chemistry varied among watersheds, whether nutrient chemistry changed from the pre-treatment to the post-treatment period, and whether nutrient chemistry changed during the posttreatment period (for groundwater). Significance was determined as P  0.05, and all statistical analyses were carried out in SYSTAT (version 13; Systat Software, San Jose, California) A one-way Analysis of Variance (ANOVA) was used to compare interflow nutrient concentrations among trenches (Rold-pine, Cold-pine, and Bnew-pine trenches; Bold-pine and Cnew-pine trenches did not flow). For all ANOVAs, if the main effect(s) were significant, a post hoc (Tukey’s Honest Significance Difference [HSD]) test was conducted to determine which sites were different from each other. A nested ANOVA was used to examine whether nutrient concentrations in groundwater wells varied among watersheds (watersheds B, C, R, and Fourmile, with wells nested within watersheds). Chemistry data from the deep wells were excluded in the ANOVA because deep wells were only present in two watersheds (B, C). Paired t-tests were used to compare nutrient concentrations in adjacent shallow and deep groundwater wells. Regression analysis was used to examine the relationship between mean nitrate concentration in groundwater and depth of the groundwater well screen.

3.1. Concentrated flow tracks In 2012, we did not observe any breakthroughs but did observe 18 category IV CFTs through which overland flow entered the SMZs. Sixteen of the 18 CFTs travelled <6 m into the SMZ before dispersing and infiltrating into the soil. Seven of the 18 CFTs occurred where the ripper tracks travelled up and down the hills at high angles, rather than on contour. Four were caused by the site preparation tractor turning within the SMZ boundary. None of the CFTs originated at haul roads or log landings. The largest amount of mobilized sand observed was from the roadside ditch leading from one of the main paved site roads, and the sediment came from the road embankment, not the harvested area. The other CFTs occurred where sufficient bare soil exposure near the SMZ boundary produced Horton overland flow. In 2013, total annual precipitation was second highest over our 6-year study period, and we observed 24 CFTs (3 category IIs, 8 category IIIs, and 13 category IVs), including 11 breakthroughs. All three of the category IIs and two of category IIIs were formed by flow from variable source areas (Dunne et al., 1975) generated by high water tables surfacing and exfiltrating from within the plantation units. These variable source areas lacked channel features and occurred in convergent parts of the landscape outside of the

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

SMZs. Low slopes produced low water velocities, and only fine sediments were carried with the water. Flow paths in these variable source areas featured obvious algal growth. Two of the category IIIs were created by overland flow from a fire line, a problem that had been noted by Terrell et al. (2011). Three of the category IIIs were attributed to skid trails. As in 2012, the causes of the remaining CFTs were due to Horton overland flow in areas of bare soil. In 2014, we observed 5 category IV CFTs, two of which were from variable source areas that had likely connected to streams earlier in the year when water tables were higher. The remaining CFTs were due to Horton overland flow. In all 5 cases, litter disturbance extended <4 m into the SMZ. 3.2. Chemistry of CFTs draining variable source areas During periods of extended wet weather, water tables rose such that 3 variable source areas within plantations developed for days to weeks, and these were counted among the CFTs (described in Section 3.1). In 2014, CFT0 and CFT3 were sampled for chemistry in watershed B, and in 2015, an additional CFT developed and was sampled in watershed C (CFT4) (Fig. 1). Nitrate concentrations in these CFTs were variable over time and concentrations were high both before and after fertilization in 2015 (Fig. 2). The highest nitrate concentration was measured in CFT4 (mean = 2353 mg N L1), and lower concentrations were measured in CFT0 (mean = 858 mg N L1) and CFT3 1 (mean = 136 mg N L ) (Fig. 2). Ammonium concentrations in the CFTs were often low (<100 mg N L1) (Fig. 2). However, ammonium concentrations increased in all CFTs (except CFT0 in 2014) after fertilization with urea in 2014 and a urea/diammonium phosphate blend in 2015 (Table 1, Fig. 2). Ammonium concentrations were elevated for approximately 1–3 weeks, and then decreased before the CFTs dried up. Soluble reactive phosphorus concentrations were low in 2014 (mean <7 mg P L1), when no phosphorus-based fertilizers were applied (Fig. 2). In 2015, approximately 2–3 weeks after diammonium phosphate fertilizer application, SRP concentrations increased in all CFTs to peak concentrations of 108 mg P L1 (CFT0), 75 mg P L1 (CFT3), and 175 mg P L1 (CFT4) (Fig. 2). CFTs were sampled twice in 2014 and once in 2015 for sulfometuron methyl. Two samples were below the detection limit and one sample (CFT3 in 2015) was at a trace level (a qualitative result but below the detection limit of 1 mg L1). 3.3. Interflow chemistry Interflow was observed in 3 of the 5 trenches over the study period. Most samples were collected from trenches Rold-pine and Cold-pine in the mature pine forests (16 and 7 sampling events, respectively), and 3 samples were collected from the Bnew-pine trench that was installed in a newly planted pine stand in 2014. No flow was measured in Bold-pine and Cnew-pine trenches, possibly because the slopes of the upland contributing areas were shallower for the Bold-pine (6.2%) and Cnew-pine (6.1%) trenches than the other trenches (slopes varied from 8.4 to 10.1%). Nitrate concentrations differed among the 3 flowing trenches (one-way ANOVA, P = 0.0001), with higher concentrations in the Bnew-pine trench than in the Rold-pine and Cold-pine trenches (Tukey’s post hoc test, P < 0.0003) (Fig. 3a). Nitrate concentrations ranged from 2966 to 5138 mg N L1 in Bnew-pine interflow water. Ammonium and SRP concentrations did not differ among the 3 trenches (ammonium: P = 0.09; SRP: P = 0.56) (Fig. 3b,c). Only one interflow sample was collected for analysis of sulfometuron methyl, and that sample from the Bnew-pine trench was below detection.

187

3.4. Groundwater chemistry Groundwater nitrate concentrations varied both spatially and temporally (Fig. 4a). Across all dates (both pre- and posttreatment), nitrate concentrations differed significantly across watersheds (nested ANOVA, P < 0.0001) with nitrate concentrations in watershed B > watershed C > Fourmile > watershed R (all Tukey’s HSD, P  0.003). This spatial pattern was also observed as a negative exponential relationship between nitrate concentration and distance from the edge of a cut area to the well, with wells further from cut areas having lower nitrate concentrations than wells closer to or within cut areas (data not shown). Groundwater nitrate concentrations varied by an order of magnitude among the 14 shallow wells (nested ANOVA, P < 0.0001) (Table 2). Across all 14 shallow wells, there was a positive relationship between groundwater well screen depth and mean nitrate concentration (r2 = 0.31, P = 0.04); however, this relationship was not significant when nitrate concentrations from the 5 deeper wells were included (r2 = 0.06, P = 0.32). Nitrate concentrations were significantly lower in the deep wells than in the adjacent shallow wells (Table 2; paired t-tests, all P < 0.0001). Groundwater nitrate concentrations changed during the posttreatment period (Figs. 4a and 5). The difference in mean groundwater nitrate concentration between the treatment (B or C) and reference (R) watersheds (Dnitrate) increased significantly during the post-treatment period (Fig. 5a; Bshallow-R: r2 = 0.76, P < 0.0001, Cshallow-R: r2 = 0.70, P < 0.0001). Nitrate concentrations also changed over time (Fig. 4a). Mean groundwater nitrate concentrations increased in watersheds B (r2 = 0.63, P < 0.0001) and C (r2 = 0.72, P < 0.0001) and decreased in watershed R (r2 = 0.56, P < 0.0001) over time (Fig. 4a). The increase in nitrate concentration in watersheds B and C occurred during the first few post-treatment sampling dates in 2013. The increase in groundwater Dnitrate in the post-treatment period was also observed when Dnitrate was calculated as the difference in mean groundwater nitrate concentration between the treatment (B or C) and Fourmile watersheds (Fig. 5b; Bshallow-Fourmile: r2 = 0.62, P < 0.0001, Cshallow-Fourmile: r2 = 0.68, P < 0.0001). Similar to watershed R, mean nitrate concentration decreased in Fourmile watershed over time (r2 = 0.20, P = 0.02). Dnitrate also increased significantly over time in the deep groundwater wells (Fig. 5; Bdeep-R: r2 = 0.48, P = 0.0007, Cdeep-R: r2 = 0.85, P < 0.0001, Bdeep-Fourmile: r2 = 0.72, P < 0.0001, Cdeep-Fourmile: r2 = 0.85, P < 0.0001) due to an increase in mean groundwater nitrate concentration in deep wells in watersheds B (r2 = 0.77, P < 0.0001) and C (r2 = 0.61, P < 0.0001). The absolute increase in nitrate concentration in deep groundwater wells was smaller than in the shallow wells. Patterns in specific conductivity in groundwater wells were also analyzed as an additional metric of water quality change over time. Specific conductivity was negatively correlated with nitrate concentration in watersheds B and C (watershed B: R = 0.80, P < 0.0001, watershed C: R = 0.79, P < 0.0001), and specific conductivity decreased over time in watersheds B (r2 = 0.50, P < 0.0001) and C (r2 = 0.59, P < 0.0001). Specific conductivity also decreased over time in watershed R (r2 = 0.58, P < 0.0001), where it was positively correlated with nitrate (R = 0.57, P = 0.004). There was no change in specific conductivity in Fourmile watershed groundwater over time, and there was no significant correlation with nitrate (R = 0.29, P = 0.16). Specific conductivity was negatively correlated with nitrate concentrations in deep groundwater wells in watersheds B and C (watershed Bdeep: R = 0.54, P = 0.003, watershed Cdeep: R = 0.75, P = 0.0001), and specific conductivity in deep wells decreased over time in watersheds B (r2 = 0.50, P < 0.0001) and C (r2 = 0.60, P < 0.0001).

188

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

200 180

B-CFT0 Ammonium B-CFT3 Ammonium C-CFT4 Ammonium B-CFT0 Nitrate B-CFT3 Nitrate C-CFT4 Nitrate B-CFT0 SRP B-CFT3 SRP C-CFT4 SRP

3000

2000

160 140 120 100 80 60

1000

-1

4000

SRP (μg P L )

-1

Ammonium or Nitrate (μg N L )

5000

40 20

0 03/14

04/14

05/14

06/14

07/14

08/14

09/14

10/14

11/14

12/14

01/15

02/15 03/15

04/15

05/15

06/15

0 07/15

Fig. 2. Ammonium, nitrate, and soluble reactive phosphorus (SRP) concentrations in 3 concentrated flow tracks (CFT0, CFT3, CFT4) that developed in the planted areas of watersheds B and C in 2014 and 2015. CFT0 and CFT3 were located in watershed B, and CFT4 was located in watershed C. Vertical grey bars represent the timing of silvicultural activities occurring on
50% of the two treatment watersheds (B, C). In March 2014, urea fertilizer was applied, and in February 2015, blended urea and diammonium phosphate (DAP) was applied.

350

(a)

b

4000

3000

2000

1000

a

0 pin

e

C old

pin

12

250

10

200 150 100 50

a R old

300

e

-pi B new

ne

(c)

-1

-1

Ammonium (μg N L )

-1

Nitrate (μg N L )

5000

14

(b)

SRP (μg P L )

6000

8 6 4 2

0 R old

pin

e

C old

pin

e

-pi B new

ne

0

R old

-pi

ne

C old

-pi

ne

-pi

B new

ne

Fig. 3. Box plots of (a) nitrate, (b) ammonium, and (c) soluble reactive phosphorus (SRP) concentrations in interflow collected from trenches in the mature pine stands of watersheds R (Rold-pine) and C (Cold-pine), and from the trench installed in a newly planted pine stand in watershed B (Bnew-pine). Box plots represent the median (solid line), mean (dashed line), 25th and 75th percentiles (edges of boxes), 5th and 95th percentiles (whiskers), and outliers (dots). Data represented in box plots are the mean nutrient concentrations across each storm event in a given trench (n = 16 events for Rold-pine, n = 7 events for Cold-pine, n = 3 events for Bnew-pine). Different letters above boxes for a given nutrient denote significant differences among trenches based on a Tukey’s post hoc test.

Groundwater ammonium concentrations were low (Table 2), and varied across watersheds (nested ANOVA, P = 0.05) and across wells within watersheds (nested ANOVA, P < 0.0001) (Fig. 6a). The variation in ammonium across watersheds was partially due to the higher ammonium concentrations in well FHR001 in watershed C compared to the other wells in that watershed (Table 2; Tukey’s HSD, all P < 0.02). Ammonium concentrations in deep wells were low and similar (Table 2). There were no significant changes in groundwater Dammonium during the post-treatment period (Bshallow-R: r2 = 0.01, P = 0.66, Cshallow-R: r2 = 0.01, P = 0.62; Bdeep-R: r2 = 0.05, P = 0.36, Cdeep-R: r2 = 0.09, P = 0.31; data not shown). Groundwater SRP concentrations were low (mostly < 5 mg P L1; Table 2), and varied across watersheds (nested ANOVA, P < 0.0001) and across wells within watersheds (nested ANOVA, P < 0.0001) (Fig. 7a). The variation in SRP concentrations was primarily driven by variable and high SRP concentrations in one well (FHR003) that was in the larger Fourmile watershed and near the stream (Table 2,

Fig. 1). The second highest SRP concentration was measured in FHR001 (watershed C) that was also located near the stream. SRP concentrations in deep wells were also low (Table 2). There were no significant changes in groundwater DSRP during the posttreatment period (Bshallow-R: r2 = 0.01, P = 0.59, Cshallow-R: r2 = 0.12, P = 0.11; Bdeep-R: r2 = 0.03, P = 0.47, Cdeep-R: r2 = 0.06, P = 0.40; data not shown). A total of 145 groundwater samples were analyzed for pesticides. All pre-treatment samples were below detection for the 3 herbicides. In the post-treatment period, two groundwater samples collected in October 2012 were above the detection limit for glyphosate (1.8 and 2.3 mg L1 from FHR014 and FHR014D, respectively), but these elevated levels may have been due to contamination of sampling equipment. After October 2012, all samples collected from those two groundwater wells were below detection for glyphosate. Additionally, all post-treatment samples collected from the other groundwater wells were below detection for the three herbicides and fipronil.

189

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

-1

Nitrate (μg N L )

2000

(a) R-FHR004 R-FHR005 R-FHR010 C-FHR001 C-FHR012 C-FHR013 C-FHR016 B-FHR014 B-FHR015

1500

1000

500

0 01/10

3500

-1

01/12

01/13 3

01/14

01/15 5

01/16

01/12

01/13 3

01/14

01/15

01/16

01/12

01/13

01/14

01/15

01/16

(b) Watershed B Watershed C Watershed R

3000

Nitrate (μg N L )

01/11

2500 2000 1500 1000 500 0 01/10

01/11

(c) Watershed B Watershed C Watershed R

-1

Nitrate (μg N L )

8000 2000

1500

1000

500

0 01/10

01/11

Fig. 4. Nitrate concentrations in (a) groundwater, (b) riparian groundwater, and (c) stream water in watersheds B, C, and R. Groundwater was sampled from multiple wells in each watershed, riparian groundwater was sampled from the two wells located near the outlet of each watershed, and stream water was sampled at the watershed outlet. Vertical grey bars represent the timing of silvicultural activities occurring on
50% of the two treatment watersheds (B, C).

3.5. Riparian groundwater chemistry Nitrate concentrations in riparian groundwater were fairly low in all 3 watersheds (usually < 300 mg N L1). Nitrate concentrations tended to be lowest in watershed B, except for 2 sampling dates

during a regional drought in mid-summer 2012, when concentrations were over 2000 mg N L1 (Fig. 4b). Nitrate concentrations were highest in watershed C, and tended to show a seasonal pattern of higher concentrations in winter/spring than summer/ autumn (Fig. 4b). Riparian groundwater Dnitrate did not change

190

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

1500

(a)

1000

Δnitrate-N

500

0

-500

Bshallow - R Cshallow - R

-1000

Bdeep - R Cdeep - R

-1500 01/12

1500

07/12

01/13

07/13

01/14

07/14

01/15

07/15

01/16

(b)

1000

Δnitrate-N

500

0

-500

Bshallow - Fourmile Cshallow - Fourmile

-1000

Bdeep - Fourmile Cdeep - Fourmile

-1500 01/12

07/12

01/13

07/13

01/14

07/14

01/15

07/15

01/16

Fig. 5. The difference (denoted as D) in groundwater nitrate concentrations in the treatment watersheds B and C and groundwater nitrate concentrations in either (a) the reference watershed (R) or (b) Fourmile watershed. Dnitrate for both deep and shallow groundwater are shown. A positive value denotes a higher concentration in the treatment watershed than the reference watershed, and a negative value denotes a lower concentration in the treatment watershed than the reference watershed. Vertical grey bars represent the timing of silvicultural activities occurring on
50% of the two treatment watersheds (B, C).

significantly from the pre-treatment to post-treatment period (one-way ANOVAs, B-R: P = 0.43, C-R: P = 0.48) (Fig. 8a and b). Riparian groundwater ammonium concentrations were of a similar magnitude as nitrate concentrations, and there was a seasonal pattern of higher ammonium concentrations in summer vs winter in 2014 and 2015 (Fig. 6b). However, opposite to the pattern in nitrate, riparian groundwater ammonium tended to be highest in watershed B, and lowest in watershed C (Fig. 6b). Riparian groundwater Dammonium increased in the post-treatment period compared to the pre-treatment period (one-way ANOVAs, B-R: P = 0.005, C-R: P = 0.03) (Fig. 8c and d). Soluble reactive phosphorus (SRP) concentrations in riparian groundwater were low (
2– 4 mg P L1) and more variable in 2010–2012 than 2013–2015 (Fig. 7b). Riparian groundwater DSRP did not change from the pre-treatment to the post-treatment period (one-way ANOVAs, B-R: P = 0.51, C-R: P = 0.67) (Fig. 8e and f). A total of 66 herbicide samples were collected from riparian groundwater wells. All samples were below detection limits except for one sample collected in February 2013 from watershed B that had a detectable (2 mg L1) concentration of imazapyr.

3.6. Stream water chemistry Stream water nitrate concentrations were low (usually < 100 mg N L1) (Fig. 4c). However, there were a few instances when nitrate concentrations were over 1000 mg N L1, with the highest concentration >8000 mg N L1 in watershed R (Fig. 4c). Several of these high concentrations were associated with re-wetting and initiation of stream flow after a regional drought in 2011– 2012 that resulted in dry streams for over a year (Fig. 4c). Stream water nitrate concentrations also varied seasonally, with higher concentrations in winter than in spring through autumn (Fig. 4c). Stream water Dnitrate did not change significantly from the pretreatment to the post-treatment period (one-way ANOVAs, B-R: P = 0.25, C-R: P = 0.31) (Fig. 9a and b). Stream water ammonium concentrations were seasonally dynamic (Fig. 6c). Ammonium concentrations were generally highest in summer and decreased in winter, and this seasonal dynamic was more prominent in 2013–2015 than in 2010 (Fig. 6c). DOC concentrations also exhibited a similar seasonal pattern of higher concentrations in summer than winter, and both ammonium and

191

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

700

(a) R-FHR004 R-FHR005 R-FHR010 C-FHR001 C-FHR012 C-FHR013 C-FHR016 B-FHR014 B-FHR015

-1

Ammonium (μg N L )

600 500 400 300 200 100 0 01/10

700

-1

01/12

01/13 3

01/14

01/15

01/16

01/12

01/13 3

01/14

01/15

01/16

01/12

01/13

01/14

01/15

01/16

(b) Watershed B Watershed C Watershed R

600

Ammonium (μg N L )

01/11

500 400 300 200 100 0 01/10

700

(c)

600 -1

Ammonium (μg N L )

01/11

Watershed B Watershed C Watershed R

500 400 300 200 100 0 01/10

01/11

Fig. 6. Ammonium concentrations in (a) groundwater, (b) riparian groundwater, and (c) stream water in watersheds B, C, and R. Groundwater was sampled from multiple wells in each watershed, riparian groundwater was sampled from the two wells located near the outlet of each watershed, and stream water was sampled at the watershed outlet. Vertical grey bars represent the timing of silvicultural activities occurring on
50% of the two treatment watersheds (B, C).

DOC concentrations were positively correlated with temperature in each watershed (ammonium vs temperature in watershed B: R = 0.69, P < 0.0001, watershed C: R = 0.53, P < 0.0001, watershed

R: R = 0.27, P = 0.0008; DOC vs temperature in watershed B: R = 0.67, P < 0.0001, watershed C: R = 0.68, P < 0.0001, watershed R: R = 0.28, P = 0.0005) (data not shown). Further, there was an

192

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

100

(a) R-FHR004 R-FHR005 R-FHR010 C-FHR001 C-FHR012 C-FHR013 C-FHR016 B-FHR014 B-FHR015

-1

SRP (μg P L )

80

60

40

20

0 01/10

50

01/12

01/13 3

01/14

01/15

01/16

01/12

01/13 3

01/14

01/15

01/16

01/12

01/13

01/14

01/15

01/16

(b) Watershed B Watershed C Watershed R

40 -1

SRP (μg P L )

01/11

30

20

10

0 01/10

35

(c)

30

-1

SRP (μg P L )

01/11

Watershed B Watershed C Watershed R

25 20 15 10 5 0 01/10

01/11

Fig. 7. Soluble reactive phosphorus (SRP) concentrations in (a) groundwater, (b) riparian groundwater, and (c) stream water in watersheds B, C, and R. Groundwater was sampled from multiple wells in each watershed, riparian groundwater was sampled from the two wells located near the outlet of each watershed, and stream water was sampled at the watershed outlet. Vertical grey bars represent the timing of silvicultural activities occurring on
50% of the two treatment watersheds (B, C).

inverse hyperbolic relationship between nitrate and ammonium concentrations in each watershed, with high nitrate concentrations

measured when ammonium concentrations were low (and vice versa). Stream water Dammonium decreased significantly in the

193

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

3000

2500

2500

2000

2000

1500 1000

1500 1000

500

500

0

0

-500

-500

1500

(a) B - R

1000

1000

500

500

Δnitrate-N

3000

1500

(b) C - R

Δnitrate-N

3500

(a) B - R

Δnitrate-N

Δ nitrate-N

3500

0 -500 -1000

PRE 600

POST 600

b

400

0 -500 -1000

-8400

-1500

PRE

(c) B - R

PRE

POST 500

(d) C - R

-8400

-1500 POST

PRE 500

(c) B - R

-200 -400

a

b

0 -200 -400

-600

-600

-800

-800

0

-250

-500

PRE

POST

PRE 40

(e) B - R

30

(f) C - R

12

POST 20

12

a b

4

8 4

0

0

-4

-4

-8

-8 -12 PRE

-10 POST

b a

10

-12

PRE

(f) C - R

16

Δ SRP

20

0

-10

POST

-250

16

Δ SRP

ΔSRP

ΔSRP

0

PRE

0

(e) B - R

8

10

b

POST

30

20

a

-500 PRE

20

40

(d) C - R

250

Δammonium-N

0

200

Δammonium-N

Δammonium-N

Δammonium-N

a

POST

400 250

200

(b) C - R

PRE

POST

Fig. 8. The difference (denoted as D) in nitrate (a,b), ammonium (c,d), and soluble reactive phosphorus (SRP) concentrations (e,f) in riparian groundwater between the reference watershed (R) and the treatment watersheds B (a,c,e) and C (b,d,f) during the pre-treatment (n = 7 for watershed B-R comparison, n = 6 for watershed C-R comparison) and post-treatment (n = 25 for both watershed B-R and C-R comparisons) periods. A positive value denotes a higher concentration in the treatment watershed than the reference watershed, and a negative value denotes a lower concentration in the treatment watershed than the reference watershed. Box plots represent the median (solid line), mean (dashed line), 25th and 75th percentiles (edges of boxes), 5th and 95th percentiles (whiskers), and outliers (dots). Different letters above boxes for a given nutrient and watershed comparison denote significant differences based on a one-way ANOVA.

post-treatment period relative to the pre-treatment period in watershed C and did not change in watershed B (one-way ANOVAs, C-R: P = 0.0002, B-R: P = 0.38) (Fig. 9c and d). Stream water SRP concentrations were low (mean concentration = 2 mg P L1) in all watersheds, with no discernable seasonal

POST

PRE

POST

Fig. 9. The difference (denoted as D) in nitrate (a,b), ammonium (c,d), and soluble reactive phosphorus (SRP) concentrations (e,f) in stream water between the reference watershed (R) and the treatment watersheds B (a,c,e) and C (b,d,f) during the pre-treatment (n = 33 for watershed B-R comparison, n = 67 for watershed C-R comparison) and post-treatment (n = 83 for watershed B-R comparison, n = 84–85 for watershed C-R comparison) periods. A positive value denotes a higher concentration in the treatment watershed than the reference watershed, and a negative value denotes a lower concentration in the treatment watershed than the reference watershed. Box plots represent the median (solid line), mean (dashed line), 25th and 75th percentiles (edges of boxes), 5th and 95th percentiles (whiskers), and outliers (dots). One outlier for Dnitrate is not shown, but is denoted by its value of 8400 (b). Different letters above boxes for a given nutrient and watershed comparison denote significant differences based on a one-way ANOVA.

pattern (Fig. 7c). Stream water DSRP increased in the posttreatment period vs the pre-treatment period (one-way ANOVAs, B-R: P = 0.02, C-R: P = 0.03) (Fig. 9e and f). SRP concentration was on average 0.4 mg P L1 higher in watershed R than watersheds B and C during the pre-treatment period, and then 0.3–0.4 mg P L1

194

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

lower in watershed R than watersheds B and C during the posttreatment period. Stream water TN concentrations varied seasonally, with higher concentrations in summer than in winter (Fig. 10a). There were a few dates when TN concentrations were high (>1000 mg N L1), and these peaks corresponded to high nitrate concentrations. Stream water TP concentrations were generally low, with no clear seasonal pattern (Fig. 10b). DTN and DTP did not change from the pre-treatment to the post-treatment period (one-way ANOVAs, DTN B-R: P = 0.64, DTN C-R: P = 0.27, DTP B-R: P = 0.20, DTP C-R: P = 0.79; data not shown). A total of 52 herbicide samples were collected from the ephemeral and intermittent stream sites during seasonal sampling. Four stream samples collected from all 3 watersheds during the pretreatment period were at the detection level for imazapyr (1 mg L1). All post-treatment herbicide samples were below detection limits. Herbicide concentrations were below detection limits in streams during storms. In autumn 2012, two storms occurred 2 and 3 months after herbicide application, respectively, and the 5 stream water samples collected during each storm in watershed C were below detection for imazapyr and glyphosate (the herbicides applied during site preparation in September 2012). A storm also occurred 4 days after the end of herbicide application in spring 2013, and all stream water samples collected from watersheds B and C during this storm were below detection for all 3 herbicides.

4. Discussion 4.1. Effects of short-rotation pine for bioenergy on water quality After 3.5 years of harvest and management of short-rotation pine, nitrate concentrations (and Dnitrate) increased in groundwater, but concentrations were low and did not change in riparian groundwater or stream water. There were no significant changes in ammonium or SRP concentrations in groundwater. Changes in ammonium and SRP in riparian groundwater and stream water were subtle, and likely not associated with silviculture activities. Almost all stream and groundwater samples analyzed for pesticides were below detection limits. Even though groundwater nitrate concentrations increased post-harvest, the maximum concentrations were less than 2 mg N L1, which is below the limit for drinking water (10 mg N L1) in the U.S., and lower than the median concentration (3.1 mg N L1) in groundwater under agricultural fields across the U.S. (Dubrovsky et al., 2010). Increasing groundwater nitrate concentrations in the posttreatment period are consistent with the finding that groundwater is the predominant flow path for water and nitrate in these lowgradient, Coastal Plain watersheds of the southeastern U.S. (Du et al., 2016; Griffiths et al., 2016; Klaus et al., 2015). Whereas shallow subsurface flow is common (Du et al., 2016), it is not the predominant contributor to stream flow (Klaus et al., 2015). There are several reasons for the disconnect between interflow and stream flow, including long travel times (due to the flat topography), considerable subsurface storage, and the presence of anomalies in the argillic (clay) soil layer allowing water to percolate through to groundwater (Du et al., 2016; Jackson et al., 2014). Nitrate also likely moves via this groundwater flow path, as the high nitrate concentrations measured in CFTs and interflow coming from planted and fertilized units (i.e., Bnew-pine trench) were not observed in riparian groundwater or stream water, but we did observe an increase in groundwater nitrate concentrations in the post-treatment period. It is not surprising that ammonium and SRP concentrations did not increase in groundwater, as ammonium is often nitrified to nitrate (a more mobile ion) and phosphate is usually bound to soil and sediment. The CFT surveys also found

very few occurrences of overland flow, and combined with no change in DTN and DTP in stream water and herbicides below detection limits, these findings suggest that overland flow was not an important flow path for moving water, nutrients, and pesticides directly to streams or that SMZs were effective at protecting surface water from these inputs (described in detail in Section 4.2). The spatial density of breakthroughs in these watersheds was much lower than observed in private timber lands in the Piedmont of Georgia that included SMZs (Rivenbark and Jackson, 2004), and the breakthroughs were driven by high water tables, not by Horton overland flow. Nitrate concentrations began increasing in shallow groundwater (10.6–19.3 m below the soil surface) 6 months after harvest. Nitrate concentrations also increased in deep groundwater, but more slowly. The travel time of surface water and contaminants to groundwater in these watersheds is not known, but other studies show that these flow paths are slow, and water transit times can be years to decades (Böhlke and Denver, 1995). However, transit times often follow skewed distributions (McGuire and McDonnell, 2006), and it is possible that the increased nitrate concentrations reflect downward movement via faster flow paths, as has also been observed in subsurface flow in hillslopes in these watersheds (Jackson et al., 2016). It is also possible that the increase in nitrate concentration was due to processes other than silviculture activities occurring on the surface. While nitrate concentrations increased over time, specific conductivity decreased in groundwater beneath the treatment watersheds. We would expect specific conductivity to increase in response to silviculture activities if mobile ions were leached to groundwater. Thus, this opposite pattern suggests that some other factor was affecting specific conductivity. Overall, the increase in nitrate concentration in groundwater and the determination of groundwater as a predominant flow path in these low-gradient watersheds of the southeastern Upper Coastal Plain (Du et al., 2016; Griffiths et al., 2016; Klaus et al., 2015) suggest that the most likely fate of leached nutrients from harvest and excess fertilizers is to groundwater. 4.2. Importance of forestry BMPs for surface water quality in lowgradient watersheds Nitrate (as Dnitrate) did not change in riparian groundwater or stream water during the post-treatment period, suggesting that the elevated nitrate measured in groundwater, subsurface flow, and CFTs had not reached the streams or was taken up or transformed. Even though groundwater is the predominant flow path in these watersheds, interflow likely transports water and nutrients to the stream from a smaller contributing area and along short flow paths (i.e., 10’s of m from the stream). However, because nutrient and pesticide concentrations in stream water were not elevated post-harvest in the treatment watersheds, it is possible that sorption, degradation (of pesticides), uptake, or transformations (e.g., denitrification) within the riparian zone reduced nutrient and pesticide inputs from the plantations. Streamside management zones have been shown to be effective at retaining and removing nutrient inputs from adjacent forestry activities (Pratt and Fox, 2009; Secoges et al., 2013; Wynn et al., 2000), and the lack of change in riparian groundwater and stream water nitrate in this study also suggests that the SMZs are initially effective for this loblolly pine bioenergy production system in the Upper Coastal Plain of the U.S. Our study was not designed to directly compare the effects of BMPs vs no BMPs on stream water quality, so we cannot definitively determine whether the BMPs resulted in minimal stream water quality issues, or if other factors such as the predominately groundwater-driven system, wide SMZs, a lack of complete watershed harvest and pine management, or a combination of these factors resulted in minimal changes to

195

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

(a) 10000 Watershed B Watershed C Watershed R

-1

Total nitrogen (μg N L )

8000 2000

1500

1000

500

0 01/10

-1

Total phosphorus (μg P L )

250

01/11

01/12

01/13 3

01/14

01/15

01/16

01/12

01/13

01/14

01/15

01/16

(b)

200

Watershed B Watershed C Watershed R

150

100

50

0 01/10

01/11

Fig. 10. Total nitrogen (a) and total phosphorus (b) concentrations in stream water in watersheds B, C, and R. Vertical grey bars represent the timing of silvicultural activities occurring on
50% of the two treatment watersheds (B, C).

stream water quality. However, there are several lines of evidence suggesting that BMPs were important at minimizing stream water quality issues, and these are described below. In the SMZ, nutrients can be taken up by vegetation or denitrified by microorganisms in anoxic microsites (Hill, 1996; Sweeney and Newbold, 2014). Denitrification may be particularly important in subtropical Coastal Plain systems as conditions are favorable for this microbial process (Lowrance et al., 2000; Peterjohn and Correll, 1984; Schaefer and Alber, 2007). Specifically, temperatures are warm, and riparian soils are often saturated, likely with anoxic microsites, and are high in organic matter content. Although denitrification rates were not measured in this study, the seasonal patterns in stream and riparian groundwater nitrate and ammonium concentrations suggest that nitrification and denitrification processes are coupled (Griffiths et al., 2016). In streams, nitrate concentrations were lowest in summer, while ammonium and DOC concentrations were high, possibly because the warm temperatures and high DOC concentrations accelerated denitrification and inhibited nitrification, resulting in a decrease in nitrate and accumulation of ammonium, respectively. These patterns suggest that the stream and surrounding riparian zone may be hotspots for denitrification. Thus, if nitrate is not denitrified within the groundwater system (Tesoriero et al., 2005), and nitrate-rich groundwater upwells near the stream channel or if interflow deliv-

ers nitrate from the edge of the planted areas, then denitrification in SMZs may reduce nitrate inputs to streams. Measurement of denitrification potential in the SMZ and continued analysis of stream and riparian groundwater chemistry will help elucidate whether denitrification is an important fate of nitrate in the SMZs of these Coastal Plain watersheds. The lack of effect of silviculture activities on stream water quality was also likely due to the small contribution of overland flow. A water quality concern associated with surface runoff processes in these watersheds was the potential transport of nutrients and pesticides from variable source areas that formed within plantations during periods of high water tables. The high nutrient concentrations measured in CFTs suggested that there was a potential for nutrients to enter streams via overland flow. However, because the sandy loam topsoils in the study area feature high conductivities (Du et al., 2016; Jackson et al., 2016), Hortonian overland flow was not much of a water quality problem. In 2014, after two years of pine growth, concentrated Horton overland flow was observed entering SMZs in only 5 places and dispersed within 4 m of the SMZ boundary. Therefore, even when breakthroughs occurred, the SMZs were effective at dispersing inputs before these inputs reached the streams. These observations of small contributions of overland flow were corroborated by a lack of change in DTN and DTP in stream water

196

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

post-harvest and herbicides that were below detection when measured both during low-flow and storm conditions. Previous studies on herbicide transport associated with forestry BMPs have also shown fairly small contributions to stream water, especially when compared to forestry that did not apply BMPs (Michael and Neary, 1993; Neary et al., 1993). The herbicides imazapyr, sulfometuron methyl, and hexazinone applied to pine plantations in two watersheds in the Upper Coastal Plain of Georgia were found in measurable quantities in stream water (peak concentration of 7.7 lg L1), but only during the first two stormflow events after herbicide application (Scarbrough et al., 2015). Herbicides were also measured during stormflows in East Texas, USA, with <1–2% of applied herbicides reaching the streams during these storm events (McBroom et al., 2013). Our study found minimal transport of herbicides to streams with storm events. Despite a storm that occurred 4 days after the end of sulfometuron methyl and imazapyr application in spring 2013, all stream water samples collected from watersheds B and C during this storm were below detection for all 3 herbicides. Further, 259 of the 263 stream, groundwater, and CFT samples collected after the first pesticide application were below detection limits, and two of the four samples above detection were likely due to sampling equipment contamination. Overall, our study found little evidence of pesticide transport to streams in these watersheds despite multiple chemical control applications. 4.3. Temporal and spatial variation in nutrient concentrations There were striking seasonal and spatial patterns in nutrient concentrations that, if not well characterized, could constrain the interpretation and understanding of silviculture activity impacts on water quality. For instance, there were small changes in Dammonium and DSRP in riparian groundwater and stream water post-treatment; however, these changes were likely not due to silviculture activities, and rather reflected biological N cycling (i.e., nitrification-denitrification coupling). Dammonium increased in riparian groundwater and decreased in stream water (watershed C-R comparison only) in the post-treatment period, and these changes were likely due to variation in the natural seasonal pattern in ammonium concentrations over time. The magnitude of seasonal change in stream water ammonium concentrations increased in the post-treatment period, especially in watersheds R and B (Fig. 6c), and thus resulted in a decrease in Dammonium during the post-treatment period for the C vs R watershed comparison. It is possible that local factors (e.g., soil moisture, frequency of intermittent flow) affected inter-annual variability in these seasonal patterns, leading to differences among watersheds. DSRP also increased in stream water in the post-treatment period, but the absolute increase in SRP concentration was small (<1 lg P L1), and was likely not ecologically significant. Thus, identifying additional drivers of nutrient dynamics is important when interpreting the effects of silviculture activities on water quality. Stream water nitrate concentrations were generally low, and Dnitrate did not change post-harvest. However, there were short periods when nitrate concentrations spiked in all three streams (Fig. 4c). These high nitrate concentrations were associated with high flow events or re-wetting periods after a year-long period of no flow. For instance, stream water nitrate concentrations were high immediately after streamflow resumed in 2012 following a regional drought; however, these high concentrations only lasted for 2–3 weeks, and the highest concentrations were measured in the reference watershed. The high nitrate concentrations measured immediately after resumption of flow likely reflects the flushing of nutrients that accumulated in and near the stream during the dry period (Bechtold et al., 2003; Creed et al., 1996). Thus, high frequency sampling over multiple years is needed to capture inter-

annual variability, especially in intermittent streams with variable flow patterns. It is important to capture this variability during both pre- and post-treatment periods when evaluating the water quality effects of silviculture in a before-after control-impact design. There was also considerable variation in nutrient chemistry both within and among watersheds. For instance, spatial variation in groundwater nitrate concentrations was partially explained by depth to sampled groundwater. Riparian groundwater nitrogen concentrations also varied among watersheds, with watershed C having higher nitrate concentrations than watershed B. Shallow subsurface flow paths are likely more dominant in watershed B (Griffiths et al., 2016), and these quick-delivery flow paths may contribute to the flushing of nitrate from the riparian zone in watershed B. Spatial variability in nutrient chemistry can make cross-watershed comparisons difficult, and highlights the importance of collecting pre-treatment data. If pre-treatment data are lacking, then comparison of post-treatment responses in treatment watersheds to multiple reference watersheds can be helpful. For instance, despite the lack of pre-treatment groundwater quality data, we found that the pattern of increasing Dnitrate posttreatment was consistent when calculated relative to watershed R and relative to the larger Fourmile watershed. Nevertheless, this highlights the importance of examining both absolute and relative changes, and accounting for spatial and temporal variation. 5. Conclusions After 3.5 years of harvest and management of short-rotation loblolly pine for bioenergy, there has been little change to stream water quality in these southeastern U.S. watersheds. Only a few instances of overland flow were observed, with 11 total breakthroughs occurring during a wetter year, and most CFTs dissipated sediments and water within the SMZs. Nutrient concentrations in streams were low and temporal changes mainly reflected seasonal patterns in N cycling. Peak nitrate concentrations were associated with re-wetting or high flow events and were observed in all 3 watersheds post-treatment. Further, almost all samples analyzed for pesticides were below detection limits, including in stream water collected during storms that occurred days to months after herbicide application. These results suggest that the current forestry BMPs for South Carolina can to protect stream water quality from intensive pine management for bioenergy in the Upper Coastal Plain of the southeastern U.S., at least in the short term. However, nitrate concentrations increased in groundwater posttreatment, and additional research is needed to determine the travel time of water and dissolved nutrients from groundwater to streams. Further, seasonal patterns in nitrate, ammonium, and DOC concentrations suggest that nitrification and denitrification may be important processes in streams and riparian zones. Measurements of denitrification potential along flow paths in the riparian zone and stream are needed determine the fate of nitrate-rich groundwater that upwells near the stream. Overall, these findings highlight that short-rotation pine production for bioenergy in the Upper Coastal Plain can have little effect on stream water quality during the first 3.5 years. However, groundwater quality, transit times, and the denitrification potential of the SMZs need to be considered in these low-gradient watersheds of the southeastern U.S. that are likely to become a predominant location for woody feedstock production for bioenergy. Acknowledgements Data in this manuscript will be available via the Bioenergy Knowledge Discovery Framework (https://www.bioenergykdf. net). This material is based on work supported by the U.S. Depart-

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

ment of Energy, Office of Energy Efficiency and Renewable Energy, Bioenergy Technologies Office. Oak Ridge National Laboratory is managed by UT-Battelle, LLC, for the U.S. Department of Energy under contract DE-AC05-00OR22725. The SRS is a ‘‘National Environmental Research Park”. Logistical and in-kind support was provided by the Department of Energy-Savannah River Operations Office through the U.S. Forest Service Savannah River under Interagency Agreement DE-AI09-00SR22188. We thank J.I. Blake and E. Olson of the USDA Forest Service for organizing the silviculture activities and for technical expertise and assistance, B. Morris and S. Younger for field sampling, and H. Jager, S. Burns, and an anonymous reviewer for manuscript comments.

References Anderson, C.J., Lockaby, B.G., 2011a. Research gaps related to forest management and stream sediment in the United States. Environ. Manage. 47, 303–313. Anderson, C.J., Lockaby, B.G., 2011b. The effectiveness of forestry best management practices for sediment control in the southeastern United States: a literature review. Southern J. Appl. Forest. 35, 170–177. APHA (American Public Health Association), 2005. Standard methods for the examination of water and wastewater, 21st ed. American Public Health Association, Washington D.C, USA. Aust, W.M., Blinn, C.R., 2004. Forestry best management practices for timber harvesting and site preparation in the eastern United States: an overview of water quality and productivity research during the past 20 years (1982–2002). Water Air Soil Pollut. 4, 5–36. Bechtold, J.S., Edwards, R.T., Naiman, R.J., 2003. Biotic versus hydrologic control over seasonal nitrate leaching in a floodplain forest. Biogeochemistry 63, 53–72. Beltran, B.J., Amatya, D.M., Youssef, M., Jones, M., Callahan, T.J., Skaggs, R.W., Nettles, J.E., 2010. Impacts of fertilization on water quality of a drained pine plantation: a worst case scenario. J. Environ. Qual. 39, 293–303. Binkley, D., Brown, T.C., 1993. Forest practices as nonpoint sources of pollution in North America. Water Resour. Bull. 29, 729–740. Binkley, D., Burnham, H., Allen, H.L., 1999. Water quality impacts of forest fertilization with nitrogen and phosphorus. For. Ecol. Manage. 121, 191–213. Blackburn, W.H., Wood, J.C., 1990. Nutrient export in stormflow following forest harvesting and site-preparation in East Texas. J. Environ. Qual. 19, 402–408. Boggs, J., Sun, G., McNulty, S., 2016. Effects of timber harvest on water quantity and quality in small watersheds in the Piedmont of North Carolina. J. Forest. 114, 27–40. Böhlke, 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 Resour. Res. 31, 2319–2339. Carroll, G.D., Schoenholtz, S.H., Young, B.W., Dibble, E.D., 2004. Effectiveness of forestry streamside management zones in the sand-clay hills of Mississippi: early indications. Water Air Soil Pollut. 4, 275–296. Creed, I.F., Band, L.E., Foster, N.W., Morrison, I.K., Nicolson, J.A., Semkin, R.S., Jeffries, D.S., 1996. Regulation of nitrate-N release from temperate forests: a test of the N flushing hypothesis. Water Resour. Res. 32, 3337–3354. Cristan, R., Aust, W.M., Bolding, M.C., Barrett, S.M., Munsell, J.F., Schilling, E., 2016. Effectiveness of forestry best management practices in the United States: literature review. For. Ecol. Manage. 360, 133–151. Du, E., Jackson, C.R., Klaus, J., McDonnell, J.J., Griffiths, N.A., Williamson, M.F., Greco, J.L., Bitew, M., 2016. Interflow dynamics on a low relief forested hillslope: lots of fill, little spill. J. Hydrol. 534, 648–658. Dubrovsky, N.M., Burow, K.R., Clark, G.M., Gronberg, J.M., Hamilton, P.A., Hitt, K.J., Mueller, D.K., Munn, M.D., Nolan, B.T., Puckett, L.J., Rupert, M.G., Short, T.M., Spahr, N.E., Sprague, L.A., Wilber, W.G, 2010. The quality of our Nation’s waters: nutrients in the Nation’s streams and groundwater, 1992–2004. US Geological Survey Circular 1350, USGS, Reston, VA, USA. . (last accessed: February 2017). Dunne, T., Moore, T.R., Taylor, C.H., 1975. Recognition and prediction of runoffproducing zones in humid regions. Hydrol. Sci.-Bull. 20, 305–327. Edwards, P.J., Williard, K.W.J., 2010. Efficiencies of forestry best management practices for reducing sediment and nutrient losses in the eastern United States. J. Forest. 108, 245–249. Fox, T.R., Jokela, E.J., Allen, H.L., 2007. The development of pine plantation silviculture in the southern United States. J. Forest. 105, 337–347. Fritts, S.R., Moorman, C.E., Hazel, D.W., Jackson, B.D., 2014. Biomass harvesting guidelines affect downed woody debris retention. Biomass Bioenergy 70, 382– 391. Grace, J.M., 2005. Forest operations and water quality in the South. Trans. Am. Soc. Agric. Eng. 48, 871–880. Griffiths, N.A., Jackson, C.R., McDonnell, J.J., Klaus, J., Du, E., Bitew, M.M., 2016. Dual nitrate isotopes clarify the role of biological processing and hydrologic flowpaths on nitrogen cycling in subtropical low-gradient watersheds. JGRBiogeosciences 131, 422–437. Hill, A.R., 1996. Nitrate removal in stream riparian zones. J. Environ. Qual. 25, 743– 755.

197

Hinchee, M., Rottmann, W., Mullinax, L., Zhang, C., Chang, S., Cunningham, M., Pearson, L., Nehra, N., 2009. Short-rotation woody crops for bioenergy and biofuel applications. In Vitro Cell. Develop. Biol. – Plant 45, 619–629. Hoover, M.D., 1952. Water and timber management. J. Soil Water Conserv. 7, 75–78. Jackson, C.R., Bitew, M., Du, E., 2014. When interflow also percolates: downslope travel distances and hillslope process zones. Hydrol. Process. 28, 3195–3200. Jackson, C.R., Du, E., Klaus, J., Griffiths, N.A., Bitew, M., McDonnell, J.J., 2016. Interactions among hydraulic conductivity distributions, subsurface topography, and transport thresholds revealed by a multi-tracer hillslope irrigation experiment. Water Resour. Res. 52, 6186–6206. Jackson, D., Cornell, C.B., Luukinen, B., Buhl, K., Stone, D., 2009. Fipronil Technical Fact Sheet; National Pesticide Information Center, Oregon State University Extension Services. . Jager, H.I., Bevelhimer, M.S., King, R.L., Smith, K.A., 2011. Landscape influences on headwater streams on Fort Stewart, Georgia, USA. Environ. Manage. 48, 795– 807. Kilgo, J.C., Blake, J.I., 2005. Ecology and Management of a Forested Landscape: Fifty Years on the Savannah River Site, Island Press. Washington D.C, USA. Klaus, J., McDonnell, J.J., Jackson, C.R., Du, E., Griffiths, N.A., 2015. Where does streamwater come from in low-relief forested watersheds? A dual-isotope approach. Hydrol. Earth Syst. Sci. 19, 125–135. Kline, K.L., Coleman, M.D., 2010. Woody energy crops in the southeastern United States: two centuries of practitioner experience. Biomass Bioenergy 34, 1655– 1666. Klockow, P.A., D’Amato, A.W., Bradford, J.B., 2013. Impacts of post-harvest slash and live-tree retention on biomass and nutrient stocks in Populus tremuloides Michx.-dominated forests, northern Minnesota, USA. For. Ecol. Manage. 291, 278–288. Lakel III, W.A., Aust, W.M., Bolding, M.C., Doloff, C.A., Keyser, P., Feldt, R., 2010. Sediment trapping by streamside management zones of various widths after forest harvest and site preparation. For. Sci. 56, 541–551. Lang, A.J., Aust, W.M., Bolding, M.C., Barrett, S.M., McGuire, K.J., Lakel III, W.A., 2015. Streamside management zones compromised by stream crossings, legacy gullies, and over-harvest in the Piedmont. J. Am. Water Resour. Assoc. 51, 1153–1164. Likens, G.E., Bormann, F.H., Johnson, N.M., Fisher, D.W., Pierce, R.S., 1970. Effects of forest cutting and herbicide treatment on nutrient budgets in the Hubbard Brook Watershed-Ecosystem. Ecol. Monogr. 40, 23–47. Litschert, S.E., MacDonald, L.H., 2009. Frequency and characteristics of sediment delivery pathways from forest harvest units to streams. For. Ecol. Manage. 259, 143–150. Lowrance, R., Hubbard, R.K., Williams, R.G., 2000. Effects of a managed three zone riparian buffer system on shallow groundwater quality in the southeastern Coastal Plain. J. Soil Water Conserv. 55, 212–220. Mallin, M.A., McIver, M.R., Ensign, S.H., Cahoon, L.B., 2004. Photosynthetic and heterotrophic impacts of nutrient loading to blackwater streams. Ecol. Appl. 14, 823–838. Marchman, S.C., Miwa, M., Summer, W.B., Terrell, S., Jones, D.G., Scarbrough, S.L., Jackson, C.R., 2015. Clearcutting and pine planting effects on nutrient concentrations and export in two mixed use headwater streams: Upper Coastal Plain, Southeastern USA. Hydrol. Process. 29, 13–28. McBroom, M.W., Beasley, R.S., Chang, M., Ice, G.G., 2008. Water quality effects of clearcut harvesting and forest fertilization with best management practices. J. Environ. Qual. 37, 114–124. McBroom, M.W., Louch, J., Beasley, R.S., Chang, M., Ice, G.G., 2013. Runoff of silvicultural herbicides applied using Best Management Practices. For. Sci. 59, 197–210. McGuire, K.J., McDonnell, J.J., 2006. A review and evaluation of catchment transit time modeling. J. Hydrol. 330, 543–564. Megahan, W.F., Kidd, W.J., 1972. Effect of logging roads on sediment production rates in the Idaho Batholith. Research Paper INT-123, USDA Forest Service Intermountain Forest and Range Experiment Station, Ogden, UT, USA. Meyer, J.L., 1990. A blackwater perspective on riverine ecosystems. Bioscience 40, 643–651. Michael, J.L., 2004. Best management practices for silvicultural chemicals and the science behind them. Water Air Soil Pollut. Focus 4, 95–117. Michael, J.L., Neary, D.G., 1993. Herbicide dissipation studies in southern forest ecosystems. Environ. Toxicol. Chem. 12, 405–410. Munsell, J.F., Fox, T.R., 2010. An analysis of the feasibility for increasing woody biomass production from pine plantations in the southern United States. Biomass Bioenergy 34, 1631–1642. National Council for Air and Stream Improvement, Inc. (NCASI), 2009. Compendium of forestry best management practices for controlling nonpoint source pollution in North America. Technical Bulletin No. 966. Research Triangle Park, NC, USA. Neary, D.G., Bush, P.B., Michael, J.L., 1993. Fate, dissipation and environmental effects of pesticides in southern forests: a review of a decade of research progress. Environ. Toxicol. Chem. 12, 411–428. Peterjohn, W.T., Correll, D.L., 1984. Nutrient dynamics in an agricultural watershed: observations on the role of a riparian forest. Ecology 65, 1466–1475. Pinho, A.P., Morris, L.A., Jackson, C.R., White, W.J., Bush, P.B., Matos, A.T., 2008. Contaminant retention potential of forested filter strips established as SMZs in the Piedmont of Georgia. J. Am. Water Resour. Assoc. 44, 1564–1577. Pratt, W.A., Fox, T.R., 2009. Streamside management zones effectiveness for protecting water quality after forestland application of biosolids. J. Environ. Qual. 38, 2106–2120.

198

N.A. Griffiths et al. / Forest Ecology and Management 400 (2017) 181–198

Prestemon, J.P., Abt, R.C., 2002. The southern timber market to 2040. J. Forest. 100, 16–22. Rivenbark, B.L., Jackson, C.R., 2004. Concentrated flow breakthroughs moving through silvicultural streamside management zones: southeastern Piedmont, USA. J. Am. Water Resour. Assoc. 40, 1043–1052. Scarbrough, S.L., Jackson, C.R., Marchman, S., Allen, G., Louch, J., Miwa, M., 2015. Herbicide concentrations in first-order streams after routine application for competition control in establishing pine plantations. For. Sci. 61, 604–612. Schaefer, S.C., Alber, M., 2007. Temperature controls a latitudinal gradient in the proportion of watershed nitrogen exported to coastal ecosystems. Biogeochemistry 85, 333–346. Scott, D.A., Tiarks, A., 2008. Dual-cropping loblolly pine for biomass energy and conventional wood products. Southern J. Appl. Forest. 32, 33–37. Secoges, J.M., Aust, W.M., Seiler, J.R., Dolloff, C.A., Lakel, W.A., 2013. Streamside management zones affect movement of silvicultural nitrogen and phosphorus fertilizers to Piedmont streams. Southern J. Appl. Forest. 37, 26–35. Shepard, J.P., 2006. Water quality protection in bioenergy production: the US system of forestry Best Management Practices. Biomass Bioenergy 30, 378–384. Smith, W.B., Miles, P.D., Perry, C.H., Pugh, S.A., 2009. Forest Resources of the United States, 2007. General Technical Report WO-78. U.S. Department of Agriculture, Forest Service, Washington, D.C., USA. South Carolina Forestry Commission, 1998. South Carolina’s Best Management Practices for Forestry. South Carolina Forestry Commission, Columbia, SC, USA, p. 65. Stewart-Oaten, A., Murdoch, W.M., Parker, K.R., 1986. Environmental impact assessment: ‘‘psuedoreplication” in time? Ecology 67, 929–940. Swank, W.T., Webster, J.R., 2014. Long-Term Response of a Forest Watershed Ecosystem: Clearcutting in the Southern Appalachians. Oxford University Press, Oxford, UK. Sweeney, B.W., Newbold, J.D., 2014. Streamside forest buffer width needed to protect stream water quality, habitat, and organisms: a literature review. J. Am. Water Resour. Assoc. 50, 560–584. Terrell, S.B., Summer, W.B., Jackson, C.R., Miwa, M., Jones, D.G., 2011. Harvest, site preparation, and firebreak effects on hydrology and sediment transport in Coastal Plain headwater streams. Trans. Am. Soc. Agric. Biol. Eng. 54, 2117– 2127.

Tesoriero, A.J., Spruill, T.B., Mew Jr., H.E., Farrell, K.M., Harden, S.L., 2005. Nitrogen transport and transformations in a coastal plain watershed: influence of geomorphology on flow paths and residence times. Water Resour. Res. 41, W02008. http://dx.doi.org/10.1029/2003WR002953. USDA Forest Service Southern Region, 2007. Forestry-Use Pesticides: Labels, Material Safety Data Sheets, and Other Supporting Information. USDA Forest Service Southern Region, Atlanta, GA, USA. U.S. Department of Energy, 2016. 2016 Billion-Ton Report: Advancing Domestic Resources for a Thriving Bioeconomy, Volume 1: Economic Availability of Feedstocks. Langholtz, M.H., B.J. Stokes, and L.M. Eaton (Leads), ORNL/TM-2016/ 160. Oak Ridge National Laboratory, Oak Ridge, TN, USA. 448p. doi: http:// dx.doi.org/10.2172/1271651. (last accessed: February 2017). Vitousek, P.M., Melillo, J.M., 1979. Nitrate losses from disturbed forests: patterns and mechanisms. For. Sci. 25, 605–619. Ward, J.M., Jackson, C.R., 2004. Sediment trapping within forestry streamside management zones: Georgia Piedmont, USA. J. Am. Water Resour. Assoc. 40, 1421–1431. Wear, D.N., Greis, J.G., 2002. Southern Forest Resource Assessment. General Technical Report SRS-53. U.S. Department of Agriculture, Forest Service, Southern Research Station, Asheville, NC, USA. White, W.J., Morris, L.A., Pinho, A.P., Jackson, C.R., 2007. Sediment retention by forested filter strips in the Piedmont of Georgia. J. Soil Water Conserv. 62, 453– 463. Witt, E.L., Barton, C.D., Stringer, J.W., Bowker, D.W., Kolka, R.K., 2013. Evaluating Best Management Practices for ephemeral stream protection following forest harvest in the Cumberland Plateau. Southern J. Appl. Forest. 37, 36–44. Wynn, T.M., Mostaghimi, S., Frazee, J.W., McClellan, P.W., Shaffer, R.M., Aust, W.M., 2000. Effects of forest harvesting Best Management Practices on surface water quality in the Virginia Coastal Plain. Trans. Am. Soc. Agric. Eng. 43, 927–936. Yoho, N.S., 1980. Forest management and sediment production in the South: a review. Southern J. Appl. Forest. 4, 27–35. Zhao, D., Kane, M., Teskey, R., Fox, T.R., Albaugh, T.J., Allen, H.L., Rubilar, R., 2016. Maximum response of loblolly pine plantations to silvicultural management in the southern United States. For. Ecol. Manage. 375, 105–111.