**2. Study sites**

364 Herbicides – Properties, Synthesis and Control of Weeds

but it may not carry in areas with insufficient quantity or quality of fuels, and it may not be appropriate or acceptable on some ownerships. Mechanical weed control methods include portable saws and machine-mounted mowers or masticators. These methods are more expensive than prescribed fire treatments, but can have similar effects: competing vegetation is disturbed above ground but not always killed; much of it re-sprouts. Herbicides can provide effective and economical control of competing vegetation, but their use may not be appropriate or acceptable in some areas amid concerns over effects on non-target organisms, movement and drift, and persistence in the environment. Fire or broadcast herbicide treatments can eliminate live vegetation cover, exposing soil to erosive forces and temporarily reducing biodiversity. Applying herbicide in spots as opposed to broadcast applications has the advantage of reducing chemical usage while maintaining some continuity of vegetation cover and preserving biodiversity between treated spots

Research into longleaf pine forest establishment and weed control has focused on the Coastal Plain region of the southeastern USA. Field research on the Coastal Plain indicated that mechanical weed control treatments were inferior to chemical weed control in terms of enhancing longleaf pine seedling survival and growth (Knapp et al., 2006). Chemical weed control with herbicide has proven effective in several longleaf pine restoration studies on the Coastal Plain (Brockway & Outcalt, 2000; Ramsay et al., 2003; Knapp et al., 2006; Haywood, 2007; Freeman & Shibu, 2009; Shibu et al., 2010). Longleaf pine is native to the Coastal Plain, but also occurs naturally in the mountainous regions further inland, and across the Piedmont Region. The Piedmont is a physiographic region extending from the State of New Jersey down to central Alabama, spanning over 200,000 km2 of rolling foothills between the Appalachian Mountains and the Coastal Plain (Anon, 2000). Little has been reported on longleaf pine restoration in the Piedmont, but restoration experiments have

Data from a replicated field experiment established on degraded Piedmont forest sites are presented here. To our knowledge no other experiment simultaneously addresses questions of repeat herbicide applications versus single treatments each of varying spot sizes, and compares all these weed control treatments to non-herbicide management options. We established non-contiguous single-tree plots in a randomized complete block design with multiple treatment levels nested in a split-plot arrangement within contiguous fixed-area treatment plots. Our objective was to determine the influence of frequency and extent of chemical weed control on planted trees and competing vegetation using commonly-used, widely-available herbicides, and to compare herbicide treatments with mechanical weed control and a no-treatment control. Specifically, we sought to answer the following four

i. How does planted seedling survival and growth differ between various herbicide treatments and two alternative experimental treatments: mechanical weed control, and

ii. Is one herbicide treatment sufficient for control of vegetation competing with tree seedlings planted for restoration? Or, will a second 'repeat application' treatment be

iii. How large of an area needs to be treated with herbicides around each planted seedling (when making a single herbicide treatment, and/or when making a repeat application)?

(Richardson et al., 1996a).

been established (Berrill & Dagley, 2009).

questions:

zero weed control?

required?

The restoration experiment was established at four disturbed sites on the 1,900 ha Hitchiti Experimental Forest (N 330 02' W 830 42') in Jones County, Georgia, USA. Southern pine beetles (*Dendroctonus frontalis* Zimmerman) had killed patches of even-aged conifer plantation throughout the forest in 2007. The kill areas totalled 10% of the forest area. Salvage harvesting in 2007 was followed by broadcast burning that consumed most of the scattered woody debris and residual hardwoods. Fire failed to carry through some areas due to lack of fuels. Containerized 1-0 'mountain variety' longleaf pine seedlings were hand planted in late March 2008 at a spacing of approximately 3.65 x 3.65 m (740 stems/ha).

Vegetation naturally regenerating throughout the study sites consisted primarily of hardwood stump sprouts and root suckers, vines, forbs, and various grasses. Natural regeneration of 22 tree species was recorded, including an abundance of dogwood (*Cornus florida* L.), loblolly pine (*P. taeda* L.), persimmon (*Diospyros virginiana* L.), sweetgum (*Liquidambar styraciflua* L.), and water oak (*Quercus nigra* L.). Five shrub species, 49 forb species, and eight vine species were recorded. The most common forb was American burnweed (*Erechtites hieracifolia* (L.) Raf.). Throughout the four sites selected as experimental replicates for the restoration study, muscadine grapevines (*Vitis rotundifolia* Michx.) were abundant and expanding laterally to occupy the disturbed sites.

Elevation of the four study sites ranged from 120-150 m above sea level. Soils were classified as a mixture of Davidson and Vance soil series with remnants of loamy surface layer over clay subsoil. The rolling hills were incised by a series of narrow, shallow gullies (Brender 1952). Before the beetle attack in 2007, the four study sites were forested with planted stands of loblolly pine 24-100 years old. Site index ranged from 24.4 m to 27.4 m at base age 25 years for loblolly pine (Clutter & Lenhart, 1968).

Climate at the study site is humid and warm in summer months, and cool in winter. Monthly average low temperatures range from -1°C in January to 19°C in July, and monthly highs range from 13°C in January to 32°C in July. Extreme temperatures were the record high of 40°C in July 1986 and the record low of -20°C in January 1985. The average annual rainfall of 1180 mm is distributed throughout the year; March being the wettest month with 140 mm, and October the driest with 70 mm average monthly rainfall (www.weather.com).

### **2.1 Experimental design**

One experimental replicate block was established in each of four beetle-killed areas at different locations across the forest. Within each replicate block (study site), four treatment plots were established. The 25 x 25 m square treatment plots were surrounded by 4 m wide buffers. Treatments applied to each plot were either mechanical weed control, chemical weed control (repeated in two plots), and control (i.e., no weed control). In a split-plot arrangement, each chemical weed control treatment measurement plot (considered the experimental unit for main treatments) was divided into approximately 12 replicates of

Vegetative Response to Weed Control in Forest Restoration 367

Mowing in the chemical weed control treatment plots extended beyond the circular spots to cover the entire plot area to uniformly reduce aboveground competition. Vegetation in the mechanical treatment plot was also mowed close to ground level manually using motorized

Prior to treatment in late June 2008, the following data were collected in all 25 x 25 m measurement plots: longleaf pine seedling status (live/dead), health and physical condition (brown spot infection, sparseness of live foliage, damaged/covered), and total height (if emerging from grass stage). Within 30 cm of each longleaf seedling (0.3 m2 sample area), herbaceous ground cover percent was estimated occularly and maximum height of herbaceous cover was measured. Within approximately 50 cm of each longleaf seedling (1 m2 sample area), vine cover percent and woody vegetation cover percent were recorded, and the maximum height of woody vegetation measured. Survival was also assessed at the end of the first growing season, in October 2008. This did not include assessment of competing vegetation due to seasonal discrepancies in cover caused by loss of leaf area

The vegetation assessments were repeated in early June 2009, 11 months after the first assessment and the first set of weed control treatments were applied. All competing vegetation within 1 m2 quadrats centred on each longleaf pine seedling was assessed. Immediately after the year-two assessment, chemical weed control was re-applied in one of the two chemical treatment blocks at each study site. This repeat herbicide application treatment was named treatment "H2". No treatments were applied in year two to the other chemical weed control plot at each study site. This 'single application' herbicide treatment was named treatment "H1". The mechanical weed control treatment (named "M") was repeated at each study site in year two, reducing aboveground competition from herbaceous vegetation, vines, and woody perennials in the measurement plot and surrounding buffer. Mowing was also applied in the H2 treatment in year two, completing reduction of aboveand belowground competition. No treatments were applied to control plots (named "C"). We returned annually thereafter to monitor the development of planted longleaf pine seedlings and competing vegetation, assessing longleaf pine seedling survival, emergence from the grass stage, height of emerged longleaf pine seedlings, and competing vegetation

Seedling survival and growth data were subjected to monthly growth adjustment assuming an 8-month growing season from April to November. This procedure gave seasonallyadjusted age estimates for seedlings at each assessment event i.e., data for assessments in the first growing season were assigned age 0.5 years (end of June) and 0.875 years (October), with subsequent assessments at age 1.375 years in June of the second growing season, age 2.25 years in May of the third season, and age 3.5 years in July of the fourth growing season. Seedlings were assigned age 0 years at the time of planting in the winter month of March

Survival of longleaf pine seedlings was assessed post-treatment, twice in the first growing season, and annually thereafter. Survival over the year immediately following the first treatment (herbicide and mechanical) was highest following chemical control of competing

among annual plants and deciduous perennials (Fig. 1).

brush saws.

height and cover percent.

**3. Survival of planted seedlings** 

2008.

three single-tree plots where a single longleaf pine seedling became the experimental unit. Within a split-plot replicate of three adjacent longleaf pine seedlings, each of the three seedlings was randomly assigned a different 'spot size' spray area treatment: a small, medium, or large circular herbicide spot sprayed around the planted seedling.

#### **2.1.1 Weed control treatments**

Chemical and mechanical treatments were applied approximately three months after planting, in late June 2008. The objective of the chemical weed control treatment was to reduce above- and belowground competition in the vicinity of longleaf pine seedlings. Glyphosate in the form of isopropylamine salt of N-(phosphonomethyl) glycine was delivered using a backpack sprayer with 2% active ingredient in water at a rate of 6.9 liters active ingredient in 360 liters of solution per ha (D'Anieri et al., 1990). Longleaf pine seedlings were covered with large paper cups prior to spraying. One week after glyphosate application, competing vegetation was mowed close to ground level, and cut stumps of woody species within each randomly-assigned spot treatment area immediately treated with an 8% triclopyr water-based solution of triethylamine salt (5.74% triclopyr acid equivalent). Triclopyr was only used when woody vegetation was present within the treatment spots. Therefore the volume of triclopyr applied differed between small, medium, and large spots, and due to variations in density of woody vegetation within and between study sites. Across all sites, the sum of all spot areas in herbicide plots (0.133 ha) and surrounding buffers (0.060 ha) was 0.193 ha. A total of 0.132 liters of triclopyr active ingredient was applied in these spots, giving an average application rate of 0.69 liters per hectare. These application rates would equate to the volume applied per hectare if the entire area was treated. We applied much less volume to our herbicide treatment plots (total area 0.89 ha at four sites) because it was only applied in spots. The chemical weed control treatment applied in circular spots around each longleaf pine seedling resulted in very different volumes of active ingredient being applied in small, medium, and large spots. We calculated that if, for example, three land managers each prescribed one of the spot size treatments we tested, then the prescription with medium size spots would require approximately four times more active ingredient per hectare than the small spots we tested, and four times less herbicide than if the large spots were prescribed (Table 1). Therefore, even with a second 'repeat' application of herbicide in the same spot size, total chemical usage in small spots sprayed a second time would be half the volume used in a single application in medium size spots, and so forth. Implementing the largest spot size across an area would result in 74% of the ground area being treated if 740 stems/ha were planted (Table 1).


Table 1. Herbicide spot treatment sizes, and comparison of anticipated chemical usage assuming each spot size treatment was applied to 740 seedlings planted on one hectare i.e., treated area is the combined area of 740 spots, and glyphosate usage is the total volume of active ingredient (ai) needed to implement 740 small, medium, or large herbicide spots.

three single-tree plots where a single longleaf pine seedling became the experimental unit. Within a split-plot replicate of three adjacent longleaf pine seedlings, each of the three seedlings was randomly assigned a different 'spot size' spray area treatment: a small,

Chemical and mechanical treatments were applied approximately three months after planting, in late June 2008. The objective of the chemical weed control treatment was to reduce above- and belowground competition in the vicinity of longleaf pine seedlings. Glyphosate in the form of isopropylamine salt of N-(phosphonomethyl) glycine was delivered using a backpack sprayer with 2% active ingredient in water at a rate of 6.9 liters active ingredient in 360 liters of solution per ha (D'Anieri et al., 1990). Longleaf pine seedlings were covered with large paper cups prior to spraying. One week after glyphosate application, competing vegetation was mowed close to ground level, and cut stumps of woody species within each randomly-assigned spot treatment area immediately treated with an 8% triclopyr water-based solution of triethylamine salt (5.74% triclopyr acid equivalent). Triclopyr was only used when woody vegetation was present within the treatment spots. Therefore the volume of triclopyr applied differed between small, medium, and large spots, and due to variations in density of woody vegetation within and between study sites. Across all sites, the sum of all spot areas in herbicide plots (0.133 ha) and surrounding buffers (0.060 ha) was 0.193 ha. A total of 0.132 liters of triclopyr active ingredient was applied in these spots, giving an average application rate of 0.69 liters per hectare. These application rates would equate to the volume applied per hectare if the entire area was treated. We applied much less volume to our herbicide treatment plots (total area 0.89 ha at four sites) because it was only applied in spots. The chemical weed control treatment applied in circular spots around each longleaf pine seedling resulted in very different volumes of active ingredient being applied in small, medium, and large spots. We calculated that if, for example, three land managers each prescribed one of the spot size treatments we tested, then the prescription with medium size spots would require approximately four times more active ingredient per hectare than the small spots we tested, and four times less herbicide than if the large spots were prescribed (Table 1). Therefore, even with a second 'repeat' application of herbicide in the same spot size, total chemical usage in small spots sprayed a second time would be half the volume used in a single application in medium size spots, and so forth. Implementing the largest spot size across an area would result in 74% of the ground area being treated if 740 stems/ha were planted

Spot size Small Medium Large

Glyphosate usage (liters ai/ha) 0.332 1.280 5.110 Table 1. Herbicide spot treatment sizes, and comparison of anticipated chemical usage assuming each spot size treatment was applied to 740 seedlings planted on one hectare i.e., treated area is the combined area of 740 spots, and glyphosate usage is the total volume of active ingredient (ai) needed to implement 740 small, medium, or large herbicide spots.

0.455 0.650 0.048 0.892 2.500 0.185  1.784 10.000 0.740

medium, or large circular herbicide spot sprayed around the planted seedling.

**2.1.1 Weed control treatments** 

(Table 1).

Spot radius (m) Spot area (m2) Treated area (ha) Mowing in the chemical weed control treatment plots extended beyond the circular spots to cover the entire plot area to uniformly reduce aboveground competition. Vegetation in the mechanical treatment plot was also mowed close to ground level manually using motorized brush saws.

Prior to treatment in late June 2008, the following data were collected in all 25 x 25 m measurement plots: longleaf pine seedling status (live/dead), health and physical condition (brown spot infection, sparseness of live foliage, damaged/covered), and total height (if emerging from grass stage). Within 30 cm of each longleaf seedling (0.3 m2 sample area), herbaceous ground cover percent was estimated occularly and maximum height of herbaceous cover was measured. Within approximately 50 cm of each longleaf seedling (1 m2 sample area), vine cover percent and woody vegetation cover percent were recorded, and the maximum height of woody vegetation measured. Survival was also assessed at the end of the first growing season, in October 2008. This did not include assessment of competing vegetation due to seasonal discrepancies in cover caused by loss of leaf area among annual plants and deciduous perennials (Fig. 1).

The vegetation assessments were repeated in early June 2009, 11 months after the first assessment and the first set of weed control treatments were applied. All competing vegetation within 1 m2 quadrats centred on each longleaf pine seedling was assessed. Immediately after the year-two assessment, chemical weed control was re-applied in one of the two chemical treatment blocks at each study site. This repeat herbicide application treatment was named treatment "H2". No treatments were applied in year two to the other chemical weed control plot at each study site. This 'single application' herbicide treatment was named treatment "H1". The mechanical weed control treatment (named "M") was repeated at each study site in year two, reducing aboveground competition from herbaceous vegetation, vines, and woody perennials in the measurement plot and surrounding buffer. Mowing was also applied in the H2 treatment in year two, completing reduction of aboveand belowground competition. No treatments were applied to control plots (named "C"). We returned annually thereafter to monitor the development of planted longleaf pine seedlings and competing vegetation, assessing longleaf pine seedling survival, emergence from the grass stage, height of emerged longleaf pine seedlings, and competing vegetation height and cover percent.

Seedling survival and growth data were subjected to monthly growth adjustment assuming an 8-month growing season from April to November. This procedure gave seasonallyadjusted age estimates for seedlings at each assessment event i.e., data for assessments in the first growing season were assigned age 0.5 years (end of June) and 0.875 years (October), with subsequent assessments at age 1.375 years in June of the second growing season, age 2.25 years in May of the third season, and age 3.5 years in July of the fourth growing season. Seedlings were assigned age 0 years at the time of planting in the winter month of March 2008.
