2009 SOUTHEASTERN NATURALIST 8(4):695–708 Military Training and Road Effects on Imperata cylindrica (L.) Beauv. (Cogongrass) Lisa Y. Yager1,*, Jeanne Jones2, and Deborah L. Miller3 Abstract - Type, level, and intensity of human activities may facilitate establishment and spread of invasive plant species. A better understanding of how human activities infl uence invasion can assist land managers in developing strategies for control and monitoring of invasive plants. Spread of the invasive species Imperata cylindrica (Cogongrass) has been attributed to human activities. During 2002−2004, on Camp Shelby Training Site, MS, we investigated relationships between military activity and establishment and growth of Cogongrass. In areas of soil disturbance from military equipment, vegetative linear growth rates of 7−10 m yr-1 were recorded on firing points. There was a positive relationship between military troop use and Cogongrass establishment on firing points for one of the 2 years of the study (P = 0.023). Thus, steps to minimize soil disturbance in and near Cogongrass may reduce spread. We examined frequency of Cogongrass infestation and vegetative growth rates for roadside areas along gravel roads subject to at least annual mowing and grading, and dirt tracks receiving infrequent maintenance. Cogongrass spread and establishment on roadsides did not differ for the two road types (P ≥ 0.116). These results may refl ect activities already in place to reduce disturbance of Cogongrass patches. Introduction Less than 3% of the 38 million ha of the highly diverse Pinus palustris Mill. (Longleaf Pine) forests remain in the southeastern United States (Frost 2006). A large portion of these remaining well-maintained Longleaf Pine communities occur on military bases (Frost 2006). However, managers seeking to maintain and protect these plant communities must consider threats from invasive species that may detrimentally impact species diversity or ecosystem processes (Walker and Silletti 2006). Imperata cylindrica (L.) Beauv. (Cogongrass) has negatively affected natural and managed ecosystems in the southeastern United States since introduction from southeast Asia in the early 1900s (MacDonald 2004). In Longleaf Pine forests, Cogongrass displaces native vegetation, alters soil processes, reduces recruitment of planted native woody seeds and seedlings, and produces higher fine fuel loads that stimulate hotter fires than native grasses (Lippincott 1997, 2000; Platt and Gottschalk 2001). Species richness of native plants in Longleaf Pine communities is negatively impacted by Cogongrass (Brewer 2007, 1Mississippi Museum of Natural Science, Mississippi Department of Wildlife, Fisheries, and Parks, 2148 Riverside Drive, Jackson, MS 39202. 2Mississippi State University, Department of Wildlife and Fisheries, Mississippi State, MS 39762. 3University of Florida, Department of Wildlife Ecology and Conservation, West Florida Research and Education Center, Milton, FL 32483. *Corresponding author - Lisa. yager@mmns.state.ms.us. 696 Southeastern Naturalist Vol. 8, No. 4 Yager 2007). Forage quality of Cogongrass for domesticated animals and wildlife is lower than native grasses or agronomic pasture grasses (Lippincott 1997, MacDonald 2004). In addition, on military installations, training operations may be adversely impacted because firing operations may create wildfires in Cogongrass as a result of its high fl ammability (Yager 2007). A better understanding of factors that infl uence Cogongrass establishment and spread would benefit military and other land managers tasked with developing appropriate control strategies that protect important habitats and military training operations from cogongrass infestation and establishment. Rates of establishment and spread of invasive species are dependent on factors which affect the number and frequency of propagules introduced into an area, and once introduced, subsequent growth and establishment of the plants (Radosevich et al. 2003). Human activity is often cited as an important factor in invasive species establishment (Hobbs and Humphries 1995, Rodgers and Parker 2003). Disturbance from human activities may enhance establishment of invasive species by reducing competition from native species, altering surface texture and microclimate, and increasing availability of resources such as light, water, and nutrients (Davis et al. 2000, Greenberg et al. 1997). Additionally, propagule pressure may be increased where humans transport seeds and vegetative fragments on vehicles, in contaminated soil, or on equipment (Greenberg et al. 1997, Hodkinson and Thompson 2002). Cogongrass establishment into new areas has been linked to rhizomes that are moved by land-management activities, such as disking, mowing, and transport of contaminated soil (Shilling et al. 1997). These disturbances also may facilitate seedling establishment. For example, Shilling et al. (1997) reported enhanced establishment of Cogongrass seedlings when Paspalum notatum Fluegge (Bahiagrass) cover was reduced. Areas of more frequent rights-of-way maintenance operations (e.g., intersections and interstate interchanges), were more likely to be infested with Cogongrass along roadsides than areas subject to less maintenance (Patterson et al. 2004). Eussen (1980) reported that one rhizome of Cogongrass can produce 350 shoots in 6 weeks and cover up to 4 m2 in 11 weeks. At these growth rates, rhizomatous growth plays an important role regarding localized spread of Cogongrass. Human activities may increase growth rates and vegetative spread of cogongrass. For example, growth of Cogongrass seedlings was enhanced with soil disturbance (tilling) in a coastal wet pine savanna (King and Grace 2000). However, type and frequency of disturbance may be important. Disking and mowing Cogongrass have been reported to reduce Cogongrass biomass, if only temporarily (Gaffney 1996, Johnson 1999, Willard et al. 1996). Repeated deep tillage has even been suggested as an effective method to control Cogongrass, and Cogongrass is generally not a problem in intensively tilled crops (Willard et al. 1996). Cogongrass is widespread on and near the military installation, Camp Shelby Training Site (CSTS), MS. Control of Cogongrass is considered essential to protect military training operations, along with Longleaf Pine 2009 L.Y. Yager, J. Jones, and D.L. Miller 697 and other habitats important to federally listed species, such as Gopherus polyphemus Daudin (Gopher Tortoise). On CSTS, Cogongrass patches (see Methods for definition) appear to be concentrated in human-created ruderal areas, such as roadsides and firing points (a firing point is here defined as an upland area cleared for military training operations), which receive regular disturbance from human activity. Therefore, it is important to understand how human activities associated with these areas infl uence spread and growth of Cogongrass. The first objective of this study was to determine if frequency and area of Cogongrass patches and rates of patch formation and growth differed on firing points with different training regimes. If propagule pressure and other environmental factors were equal across CSTS, then larger firing points should be more likely to be infested with Cogongrass. As Cogongrass has become more prevalent near Camp Shelby, firing points constructed more recently may have had a greater probability of propagule introduction from equipment or fill dirt. Military tanks which could disturb the soil and transport rhizomes from one area to another may be another potential source of infestation. From these considerations, we developed a working hypothesis that presence, numbers, and area of Cogongrass patches on firing points were related to date of construction and size of firing points as well as use of firing points for tank training. We also hypothesized that changes in numbers and area of Cogongrass patches on firing points was related to days of use for military training activities involving tracked vehicles and other equipment that could potentially result in soil disturbance. The second objective of this study was to determine if frequency of Cogongrass patches and rate of patch formation and growth differed on maintained roadsides (those with annual mowing and frequent grading) of gravel roads compared to roadsides on dirt tracks that were infrequently maintained. We hypothesized that the more highly maintained graveled roads would be more likely to be infested with Cogongrass and that rates of vegetative growth of Cogongrass would be greater along their adjacent roadsides. Field-site Description Camp Shelby Training Site is located in Perry, Forrest, and George counties, MS, and at 54,315 ha, is one of the largest National Guard Training installations in the United States (Yager 2007). Most of the land is owned and managed by the United States Forest Service, but about 6900 ha are owned and managed by the Department of Defense, State of Mississippi, or Department of the Army (Yager 2007). Land is managed for timber production, wildlife, recreation, and water protection, in addition to military training (Yager 2007). Camp Shelby is located within the Gulf Coastal Plain Physiographic Region and within the Longleaf Pine-Bluestem portion of the Longleaf Pine ecosystem. As a result of past fire regimes and silvicultural practices, a variety of upland Longleaf, Pinus taeda L. (Loblolly), and Pinus elliottii Engelm. (Slash) Pine forest associations occur on CSTS (Yager 2007). Embedded within these forests are isolated areas of disturbance, including 698 Southeastern Naturalist Vol. 8, No. 4 roadsides, military training areas (firing points and ranges), and powerlines, which have a predominantly herbaceous groundcover and are maintained by mowing and/or burning. Military training areas have been cleared of trees and are used for a variety of training purposes. Training areas are mostly upland sites with well-drained sandy loam soils. Non-native species, such as Paspalum notatum Fluegge (Bahiagrass), Cynodon dactylon (L.) Pers. (Bermudagrass), and Lolium perenne L. (Ryegrass), have been planted for erosion control during construction and to rehabilitate areas of soil disturbance. Native species, such as Schizachyrium scoparium (Michx.) Nash. (Little Bluestem), Schizachyrium tenerum Nees (Slender Bluestem), Andropogon virginicus L. (Broomsedge), Pityopsis graminifolia (Michx.) Nutt. (Silkgrass), Desmodium spp. (beggar’s tick), and Rhus copallina L. (Winged Sumac), also occur within these areas (Yager 2007). Roads of CSTS, which include paved, graveled, and dirt tracks, are variously maintained and vegetated. Roadsides of paved and graveled roads are maintained with a predominantly herbaceous groundcover by mowing at least annually including periodic prescribed fires. These herbaceous roadsides are about 3 m wide and easily differentiated from adjacent forest or other vegetation associations. Graveled roads receive grading multiple times per year, as needed. Dirt tracks are unimproved roads without gravel that receive mowing and grading less than once a year, if ever. Roadside areas of dirt tracks transition rapidly into the adjacent vegetation associations and may have substantial woody vegetation mixed with early successional herbaceous species. Climate is warm-temperate to subtropical. Mean monthly normal temperatures are least in January (9 °C) and greatest in July and August (27 °C) (National Oceanic and Atmospheric Administration 2006). The 30-year mean annual rainfall at the nearest weather station (Wiggins, MS), is 163 cm/year (National Oceanic and Atmospheric Administration 2006). Annual rainfall recorded on Camp Shelby Training Site from 2002−2004 ranged from 179 to 206 cm (Yager 2007). Soils are predominantly silty loams, sandy loams, and loams, and most are classified as Ultisols or Alfisols (USDA Natural Resource Conservation Service 1999, USDA Soil Conservation Service and Forest Service 1979). Methods Definitions Cogongrass is a clonal plant that generally produces closely aggregated tillers with an extensive rhizome system. For this research, Harrington’s (1977) definition of a tiller as an erect, lateral grass shoot was used. A Cogongrass patch was defined as the area of ground occupied by an aggregation of tillers. Tillers within 2 m of each other were considered part of one patch. Patches included more than one plant (i.e., one or more genets). Although rhizomes may extend beyond the tillers of the patch underground, a patch edge was defined as the outermost surface-occurring tillers along the perimeter of the patch. 2009 L.Y. Yager, J. Jones, and D.L. Miller 699 Military training activity effects on Cogongrass growth The following data were recorded for military firing points: 1) area (m2), 2) number and area of Cogongrass patches, 3) soil disturbance from military use as evidenced by rehabilitation (presence or absence of rows of recently planted Ryegrass and Bahiagrass), 4) designation for tank use—restricted or permitted, 5) age class—construction pre- or post-1968, and 6) military use—number of days a firing point was requested for training involving tracked vehicles and other equipment that could potentially result in soil disturbance. Cogongrass patches on extant artillery firing points (n = 99) on CSTS were mapped with a Trimble Pro−XR GPS unit (Trimble Navigation Inc., Sunnyvale, CA) during winters of 2002, 2003, and 2004. Evidence of recent rehabilitation (revegetation of plants after soil disturbance) on firing points also was recorded during mapping in 2002. A few firing points were not mapped in each year because of entry restrictions or problems with equipment. Patches were considered separate if there was more than 2 m distance between patch edges. Firing-point size and potential to be used for tank training were determined using the Mississippi Army National Guard (MSARNG) GIS database. Aerial photography from the MSARNG GIS database was used to classify firing point construction as pre- or post-1968. Records for military use were obtained from MSARNG Range Control (2005). The relationships of Cogongrass presence, numbers of patches, and patch area on firing-points in 2002 to firing-point age, size, and tank use were examined using regression analyses appropriate to the type of data (binomial or continuous data) (Quinn and Keough 2002). For Cogongrass presence or absence data, logistic multiple regression was used because the data followed a binomial distribution (Quinn and Keough 2002). A general linear model was used to perform multiple regression for the continuous data of total area of Cogongrass (Quinn and Keough 2002). Changes in numbers of Cogongrass patches on firing points from 2002 to 2003 and 2003 to 2004 were analyzed using a SAS PROC GENMOD poisson regression for each year (SAS Institute, Inc. 2004). The model used was: Increase in number of Cogongrass patches = military use + presence of Cogongrass patches in previous year + area of Cogongrass patches in previous year. Data were underdispersed, and therefore quasi-likelihood analysis was determined to be appropriate (Quinn and Keough 2002). The PSCALE option was added to the analysis to calculate parameter estimates (SAS Institute, Inc. 2004). Changes in area of Cogongrass for each of the 2 years were analyzed using PROC GLM to perform multiple regression (SAS Institute, Inc. 2004). The model used was: Increase in area of Cogongrass patches = military use + presence of Cogongrass patches in previous year + area of Cogongrass patches in previous year. Infestation data from previous years were included in all models as a covariate because existing infestations were a potential source of propagules and growth. 700 Southeastern Naturalist Vol. 8, No. 4 Growth and disturbance from tracked vehicles In another study on CSTS which examined the effect of patch size on rates of linear growth of Cogongrass (Yager 2007), baseline size data on 30 Cogongrass patches located on firing points were collected using a S.P. Constructor 50 Total Station GPS unit (Trimble Navigation, Inc., Sunnyvale, CA) in April 2002, 2003, and 2004. Universal Trans Mercator coordinates were collected every 30−50 cm along patch edges, and ArcView 3.2 (ESRI, Inc., Redlands, CA) was used to calculate patch area, distances of 2003 points from the 2002 perimeters of each patch, and distances of 2004 points from the 2003 perimeters of each patch. Mean linear growth was determined for each patch by averaging distances measured. During the study period, 3 of the 30 patches showed evidence of soil disturbance from tracked vehicles. Linear growth rates for the 3 patches with evidence of soil disturbance are reported because they are examples of effects of soil disturbance on linear spread of Cogongrass growth. Linear growth rates for Cogongrass patches without soil disturbance are reported for comparison purposes. Infestation rates and growth on roads and tracks To determine if graveled roads or tracks were more likely to exhibit infestation, thirty 100-m sections of different graveled roads were randomly selected. For each selected gravel road, an additional 100-m section of track was selected randomly from the track nearest the selected road section. Occurrence of Cogongrass infestations from road or track edge to 15 m into adjacent vegetation association was recorded using a Trimble ProXT GPS device during January 2003 and March 2004. Chi-square analysis was performed to evaluate differences in frequency of infestation for graveled roads compared to tracks (Ott 1988). To determine differences in vegetative linear growth between these two road types, 11 patches along roads and 4 patches along tracks were mapped using a GPS device (S.P. Constructor 50 Total Station) to record points in 30- to 50-cm intervals along the patch edge during April 2002, 2003, and 2004. Cogongrass patches on tracks ranged in size from 18 m2 to 286 m2, with a mean of 104 m2 and a median of 56 m2. Patches along graveled roads ranged in size from 24 m2 to 302 m2, with a mean of 113 m2 and a median of 76 m2. Points were classified as being within adjacent forest or roadside (mowed area adjacent to road). Roadside patches were selected randomly from graveled roads which had been surveyed east of Mississippi Highway 29 during 2001. Because tracks had not been as extensively surveyed, track patches were selected based on knowledge of where they occurred. Areas adjacent to tracks were not mowed and thus did not have an area that was clearly defined as roadside. Therefore, for comparison purposes between the two road types, points collected for patches along tracks were classified as roadside if they fell within 2.85 m of the track edge. This width equaled the mean width of roadsides of graveled road patches. ArcView 3.2 was used to calculate patch area, distances of 2003 points from the 2002 perimeters of each patch, and distances of 2004 points from the 2003 perimeters of each patch. Mean linear roadside growth was determined for each patch 2009 L.Y. Yager, J. Jones, and D.L. Miller 701 by averaging distances measured (>8 measurements per patch) within the roadsize zones. Patch growth data was analyzed using a repeated measures analysis of variance with year and road type as explanatory variables (Ott 1988; SAS Institute, Inc. 2004). Markers were lost for two of the graveled road patches during 2002, hence, 2002−03 linear growth analysis only included 9 data from 9 graveled road patches. Results Military activity effects on Cogongrass Infestation rates based on firing-point age and tank use ranged from 36 to 55% (Table 1). Logistic regression did not indicate a relationship between firing-point age, size, or use by tanks and probability of Cogongrass infestation (likelihood ratio: χ2 3= 1.91, P = 0.592). The general linear model also did not indicate a relationship between size of Cogongrass infestation on firing points and firing-point age, size, or use by tanks (P = 0.173). Twenty-two new infestations on 14 firing points were mapped in 2003. In 2004, 21 new infestations were mapped on 16 firing points. Regression analyses indicated positive relationships between new infestations with prior infestations in 2003 and with military use in 2004 on firing points (Table 2). Increase in Cogongrass area for 2003 and 2004 on firing points was related positively to area of infestation in the previous year, but not to number of days of military use (Table 3). Table 2. Summary of regression analysis for numbers of new Cogongrass infestations on military firing points on Camp Shelby Training Site, MS in 2003 and 2004 related to independent variables: Cogongrass surface area in previous year, presence of infestation in previous year, and days of military use in previous year. Variable NDFA DDFB Parameter S.E. χ2 P < χ2 Change in number of infestations 2003 Cogongrass area 2002 1 92 -0.73 1.05 0.48 0.489 Infested 2002 1 92 0.31 0.11 7.80 0.005 Days of use 2001 1 92 -0.30 0.03 1.28 0.259 Change in number of infestations 2004 Cogongrass area 2003 1 92 0.05 1.08 0.00 0.960 Infested 2003 1 92 0.15 0.11 1.94 0.164 Days of use in 2002 1 92 0.03 0.01 5.12 0.023 ANominator degrees of freedom. BDenominator degrees of freedom. Table 1. Percentage of firing points (n) infested with Cogongrass and mean area infested categorized by date of construction and tank use on Camp Shelby Training Site, MS in 2002. Constructed pre-1968 Constructed post-1968 % Mean area % Mean area Firing points n infested infested (m2) n infested infested (m2) Designated for tank use 34 53 332 20 45 507 Restricted for tank use 31 55 202 11 36 524 702 Southeastern Naturalist Vol. 8, No. 4 Growth and disturbance from tracked vehicles Mean linear growth rates for undisturbed patches were 0.54 m (n = 19) for 2002−03 and 0.90 m (n = 25) for 2003−04, whereas mean linear growth rates for the two patches that exhibited soil disturbance in 2003−04 were 1.46 and 3.22 m, respectively (Fig. 1). Maximum growth rates recorded for the undisturbed patches were 0.92 for 2002−03 and 1.51 for 2003−04. Maximum growth rates for the two disturbed patches were 6.35 and 9.85 m, respectively. During the 2002−03 period, one patch received significant disturbance from military equipment, but the site of disturbance was subsequently rehabilitated by disking and planting with ryegrass. Similar activities followed by rehabilitation occurred during the 2003−04 period. For this rehabilitated site, the area occupied by the Cogongrass patch decreased in the first year; and by April 2004, no Cogongrass shoots were observed at the site (Fig. 1). However, a small amount of regrowth was observed in August 2004. Infestation rates and growth on roads and tracks Of the 30 sections surveyed on graveled roads and tracks, frequency of Cogongrass infestation was 33.3% on graveled roads compared to 13.3% on tracks. These frequencies did not change from 2003 to 2004. Infestation rates for the two road types did not differ significantly ( χ2 1,59 =3.17, P < 0.075). Linear vegetative growth along roadsides also did not differ between the two road types (F1,24= 2.66, P < 0.116), or years (F1,24 = 2.41, P < 0.133); nor did road type interact with year (F1,24 = 0.83, P < 0.371) (Table 4). Discussion Firing-point size, construction period (pre- or post-1968), and tank use were not related to the probability of infestation by Cogongrass. Probability of infestation of an area is determined by both propagule pressure and biotic and abiotic conditions occurring within the community (Radosevich et al. 2003, Rejmanek et al. 2005). Propagule pressure is a function of extent of Table 3. Summary of regression analysis for increase in area of Cogongrass infestations in 2003 and 2004 on military firing points on Camp Shelby Training Site, MS related to independent variables: Cogongrass area in previous year, presence of infestation in previous year, and days of military use in previous year. Variable NDFA DDFB Parameter S.E. t−value P < t Change in Cogongrass area 2003 (R2 = 0.48) Cogongrass area 2002 1 89 0.05 0.02 2.08 0.041 Infested 2002 1 89 0.01 <0.01 4.69 <0.001 Days of use 2001 1 89 <0.01 <0.01 1.40 0.165 Change in Cogongrass area 2004 (R2 = 0.41) Cogongrass area 2003 1 84 0.25 0.07 3.62 <0.001 Infested 2003 1 84 0.01 0.01 1.77 0.081 Days of use 2002 1 84 <0.01 <0.01 1.06 0.291 ANominator degrees of freedom. BDenominator degrees of freedom. 2009 L.Y. Yager, J. Jones, and D.L. Miller 703 nearby populations and numbers of propagules dispersed into or near the plant community (Rejmanek et al. 2005). If abiotic and biotic conditions and propagule pressure do not vary within a site (for example, CSTS), then the probability of infestation per unit area should be the same across the site. Under these homogenous conditions, larger areas within the site will have a greater probability of infestation than smaller areas within the site (e.g.,, if probability of infestation = 0.1 per m2 then the probability of a 10-m2 area being infested will be greater than the probability of a 2-m2 area being infested: Figure 1. Patterns of vegetative growth of cogongrass patches on military firing points on Camp Shelby Training Site, MS from April 2002─03 and April 2003─04 which: A) received no obvious soil disturbance during study period, B) received soil disturbance from equipment in area of maximum growth during 2003 and 2004, C) received soil disturbance from equipment in area of maximum growth during 2003 and 2004 (top patch) or showed no evidence of soil disturbance (bottom patch), and D) received soil disturbance followed by disking and planting ryegrass in winter 2002/2003 and 2003/2004. Patch area did decrease for patch shown in D. Table 4. Annual linear growth (m) of Cogongrass patches on roadsides next to graveled roads or tracks on Camp Shelby Training Site, MS from April 2002 to April 2003 and April 2003 to April 2004. 2002−03 2003−04 Gravel roadside Initial patch sizes (m2) 18−286 n 9 11 Mean (S.E.) linear growth (m/yr) 0.80 (0.09) 0.88 (0.10) Maximum linear growth (m/yr) 2.11 2.16 Track Initial patch sizes (m2) 24−302 n 4 4 Mean (S.E.) linear growth (m/yr) 0.48 (0.13) 0.80 (0.17) Maximum linear growth (m/yr) 1.53 3.48 704 Southeastern Naturalist Vol. 8, No. 4 10 m2 x 0.1 m-2 > 2 m2 x 0. 1 m-2). The lack of a positive relationship between firing-point size and probability of infestation suggests that at least some of the factors that affect propagule pressure and Cogongrass establishment vary among firing points. Results of this study suggest that factors other than construction period or designation of tank use may be important for explaining infestation patterns. Land managers may view conditions on military firing points as fairly homogenous based on their similar topographical locations, physiognomies, species composition, and land use. However, proximity to nearby Cogongrass infestations, forestry activities, recreational uses, and other conditions vary among firing points and may result in increased likelihood of infestation. Military training includes many activities other than tank training that may bring in propagules or result in soil disturbance, such as paladin training and live-fire exercises. Additionally, factors that affect likelihood of infestation may change annually, or even weekly, as environmental conditions vary and interact with human use (Davis and Pelsor 2001, Davis et al. 2000). A more complete analysis of past use and infestation patterns was not feasible because a complete history of military or other use on firing points was not available. Future studies could be strengthened if, in addition to accurate records of actual amounts and types of military use, records of amounts and locations of ground disturbance from military training, rehabilitation, or other actions were maintained. Analysis indicated a positive relationship between new infestations of Cogongrass in 2004 with days of military use in the previous year, but a similar relationship was not found for new infestations in 2003. One possible explanation of these results may be that timing of activities is important. Activities occurring when soils are wet are more likely to result in ground disturbance, such as rutting. Disturbance in Cogongrass patches may result in transport of rhizomes on vehicles, whereas disturbance in uninfested areas may provide better sites for establishment. Activities occurring during seed production may be more likely to transport seeds. Thus, under some conditions related to seasonal infl uences, military training would be more likely to result in spread of Cogongrass. Another explanation for results observed may be that certain types of training activities were more likely to result in ground disturbance or transport of propagules. For instance, although wheeled vehicles may result in ruts under wet conditions, they would generally be less likely to result in soil disturbance than tracked vehicles Thus, data on weather conditions, specific duration and types of training activities, and amounts of ground disturbance would be helpful for future studies. Cogongrass growth may be affected by the type of soil disturbance. One-time tilling increased growth of Cogongrass seedlings in a Mississippi fl atwood site (King and Grace 2000). However, other studies indicated that disking reduced Cogongrass biomass (Gaffney 1996, Johnson 1999, Willard et al. 1996). Two possible explanations for this result were proposed: 1) exposure of rhizomes by disking led to rhizome dessication, or 2) rhizome fragmentations reduced available carbohydrates for growth (Gaffney 2009 L.Y. Yager, J. Jones, and D.L. Miller 705 1996, Johnson 1999, Willard et al. 1996). Observations of linear growth from 3 Cogongrass patches, each of which received disturbance from military equipment within and adjacent to the patch at 2 unrehabilitated and 1 rehabilitated site, also suggested that type of soil disturbance may result in different growth patterns. For the 2 Cogongrass patches at the unrehabilitated site where much of the patch remained intact but surface vegetation adjacent to the patch was removed, one-time soil disturbance appeared to enhance Cogongrass growth in the area of disturbance. This effect may have been because of reduced competition from surrounding vegetation, along with the remaining intact shoot and rhizome systems benefitting from increased resources in the disturbed area. It should be noted, however, that if the response observed was a result of the disturbance, then other types of equipment used for logging, rehabilitation, or rights-of-way maintenance could have a similar effect. For the patch where rehabilitation occurred following initial disturbance, Cogongrass growth appeared to be impeded and total area covered by the patch was reduced. Rehabilitation included disking the entire patch and surrounding area, followed by planting ryegrass. In this instance, rehabilitation would have fragmented the rhizome system and killed shoots, while offering increased competition from ryegrass. Revegetation, with species such as ryegrass, may play an important role in integrated management programs for certain invasive species (Miller 2003). Greenberg et al. (1997) reported greater exotic species richness and cover on roads with greater disturbance (vegetation removal and soil fill, versus vegetation removal only) in Ocala National Forest, FL. However, exotic species richness was similar on main roads and backcountry trails in Glacier National Park (Tyser and Worley 1992). These differing results suggest establishment of a given exotic species is affected by its dispersal mechanisms and biological characteristics, and also the amount and type of the particular disturbance (Parendes and Jones 2000, Tyser and Worley 1992). Patterson et al. (2004) reported greater probability of Cogongrass infestation in areas adjacent to highways which were subject to more maintenance and disturbance. Parendes and Jones (2000) observed greater frequencies of several exotic species on roads receiving high road traffic and road maintenance activities compared to roads without severe disturbance, but this relationship was more apparent for species with frequencies >30% for at least one road type. Thus, differences between infestation frequencies on gravel roads and dirt tracks on Camp Shelby may not have been statistically significant because of the low infestation frequencies observed at the time. If so, differences may become more pronounced through time and with advancing infestation. Alternatively, the lack of difference may indicate that the level of roadside disturbance and transport of propagules was not sufficiently different between road types to affect Cogongrass establishment. Level of maintenance may be important for establishment of Cogongrass on paved highways in Alabama with presumably different habitat features and humaninitiated disturbances (Patterson et al. 2004), but less so for the unpaved 706 Southeastern Naturalist Vol. 8, No. 4 roads on Camp Shelby. Annual mowing may not increase spread if it does not occur during the seed-production period or disturb the soil to transport rhizome fragments. Mowing which does not disturb the soil or occur during seed production would not likely transport propagules, including seeds or rhizome fragments. During the past several years, mowing had been restricted on CSTS if fruiting infl orescences were visible. On CSTS, road grading also may be less likely to transport propagules because grading operations occurred primarily within the gravel roadbed proper and gravel roadbeds did not appear to be susceptible to Cogongrass encroachment. However, grading through Cogongrass patch edges may sometimes facilitate transport of rhizomes to previously uninfested roadside areas. Roads on CSTS also were presumably subject to less vehicular traffic than highways, which may reduce wind-dispersal of seeds. Linear growth rates also did not vary between gravel roadsides and tracks. Although infrequent mowing has been shown to reduce shoot and root biomass (Willard 1988, Willard et al. 1996), results of this study indicate that mowing may have little effect on overall area occupied by the Cogongrass plant or its rhizomatous spread. These results further suggest that a factor other than grading may be important in explaining rhizomatous growth along roadsides. One difference should be noted for Cogongrass growth between the two road types. Cogongrass was able to spread across dirt tracks via rhizomatous growth, whereas gravel roadbeds served as a barrier to rhizome growth, possibly because of greater compaction, grading, or other activities. Conclusion Rates and type of human disturbance and activity can affect spread of invasive species; however, maintenance or traffic levels did not appear to affect rates of Cogongrass spread and growth on unpaved roads on CSTS. This finding may have been due, in part, to precautionary measures that have been implemented to slow spread of Cogongrass on CSTS. Military activity rates and resulting soil disturbance did appear to infl uence spread of Cogongrass on CSTS. This effect was not strong for infestation rates and only evident for rhizomatous growth when equipment disturbed soil adjacent to and within patches. These results suggest that measures to prevent soil disturbance in and near patches may assist managers in minimizing the spread of Cogongrass. Precautionary measures now implemented include, but are not limited to: education of staff and contractors regarding identification of Cogongrass, posting warning signs around Cogongrass patches on military training areas, and fl agging Cogongrass patches on roadsides. Acknowledgments We thank the Mississippi Army National Guard, The Nature Conservancy, and the US Forest Service for their funding and support. We also thank the Mississippi Museum of Natural Science for allowing Y. Yager time to write this manuscript. We appreciate the assistance of Bob Lemire, Colleen Heise, Brian Mitchell, Robin Switzer, and C.J. Sabette for their contributions to this work. We also thank Brett Serviss and two anonymous reviewers for their comments and suggestions which improved this paper. 2009 L.Y. Yager, J. Jones, and D.L. Miller 707 Literature Cited Brewer, J.S. 2007. Declines in plant species richness and endemic plant species in Longleaf Pine savannas invaded by Imperata cylindrica. Biological Invasions 10:1257−1264. Davis, M.A., and M. Pelsor. 2001. Experimental support for a resource-based mechanistic model of invasibility. Ecological Letters 4:421−428. Davis, M.A., J.P. Grime, and K. Thompson. 2000. Fluctuating resources in plant communities: A general theory of invasibility. Journal of Ecology 88:528−534. Eussen, J.H.H. 1980. Biological and ecological aspects of Alang-alang (Imperata cylindrica (L.) Beauv). BIOTROP 5:15−22. Frost, C.C. 2006. History and future of the Longleaf Pine ecosystem. Pp. 9−42, In S. Jose, E.J. Jokela, and D.L. Miller (Eds.). The Longleaf Pine Ecosystem: Ecology, Silviculture, and Restoration. Springer Science + Business Science Media, New York, NY. 437 pp. Gaffney, J.F. 1996. Ecophysiological and technological factors infl uencing the management of Cogongrass (Imperata cylindrica). Ph.D. Dissertation. University of Florida, Gainesville, FL. 114 pp. Greenberg, C.H., S.H. Crownover, and D.R. Gordon. 1997. Roadside soils: A corridor for invasion of xeric scrub by nonindigenous plants. Natural Areas Journal 17:99−109. Harrington, H.D. 1977. How to Identify Grasses and Grasslike Plants. Ohio University Press, Athens, OH. 164 pp. Hobbs, R.J., and S.E. Humphries. 1995. An integrated approach to the ecology and management of plant invasions. Conservation Biology 9:761−770. Hodkinson, D.J., and K. Thompson. 2002. Plant dispersal: The role of man. Journal of Applied Ecology 34:1484−1496. Johnson, E.R.R.L. 1999. Management of the non-native, invasive weed Cogongrass (Imperata cylindrica (L.) Beauv.): Utilizing an integrated management program. M.Sc. Thesis. University of Florida, Gainesville, FL. 87 pp. King, S.E., and J.B. Grace. 2000. The effects of gap size and disturbance type on invasion of wet pine savanna by Cogongrass, Imperata cylindrica (Poaceae). Amererican Journal of Botany 87:1279−1286. Lippincott, C.L. 1997. Ecological consequences of Imperata cylindrica (Cogongrass) invasion in Florida sandhill. Ph.D. Dissertation. University of Florida, Gainesville, FL. 165 pp. Lippincott, C.L. 2000. Effects of Imperata cylindrica (L.) Beauv. (Cogongrass) invasion on fire regime in Florida sandhill (USA). Natural Areas Journal 20:140−149. MacDonald, G.E. 2004. Cogongrass (Imperata cylindrica): Biology, ecology, and management. Critical Reviews in Plant Sciences 23:367−380. Miller, J.H. 2003. Nonnative invasive plants of southern forests: A field guide for identification and control. General Technical Report SRS−62. USDA Forest Service Southern Research Station, Auburn, AL. 95 pp. Mississippi Army National Guard (MSARNG) Range Control. 2005. Records of firing-point use on Camp Shelby Training Site. Located at: Camp Shelby Training Site, MSARNG Range Control, Camp Shelby, MS. National Oceanic and Atmospheric Administration. 2006. NOWData online weather data. Available online at http://nowdata.rcc−acis.org/MOB/pubACIS_results. Accessed 5 October 2006. Ott, L. 1988. An Introduction to Statistical Methods and Data Analysis, 3rd Edition. PWS−KENT Publishing Company, Boston, MA. 835 pp. 708 Southeastern Naturalist Vol. 8, No. 4 Parendes, L.A., and J.A. Jones. 2000. Role of light availability and dispersal in exotic plant invasion along roads and streams in the H.J. Andrews Experimental Forest, Oregon. Conservation Biology 14:64−75. Patterson, M., D. Teem, and W. Faircloth. 2004. Mapping, control, and revegetation of Cogongrass infestations on Alabama rights of way. ALDOT Research Project 930−486. Alabama Department of Transportation and Auburn University Alabama Agricultural Experiment Station, Auburn, AL. 59 pp. Platt, W.J., and R.M. Gottschalk. 2001. Effects of exotic grasses on potential fine fuel loads in the groundcover of south Florida slash pine savannas. International Journal of Wildland Fire 10:155−159. Quinn, G.P., and M.J. Keough. 2002. Experimental Design and Data Analysis for Biologists. Cambridge University Press, New York, NY. 537 pp. Radosevich, S.R., M.M. Stubbs, and C.M. Ghersa. 2003. Plant invasions: Process and patterns. Weed Science 51:254−259. Rejmanek, M., D.M. Richardson, S.I. Higgins, M.J. Pitcairn, and E. Grotkopp. 2005. Ecology of invasive plants: State of the art. Pp. 104–161, In H.A. Mooney, R.N. Mack, J.A. McNeely, L.E. Neville, P.J. Schei, and J.K. Waage (Eds.). Invasive Alien Species: A New Synthesis. Island Press, Washington, DC. 368 pp. Rodgers, J.C., and K.C. Parker. 2003. Distribution of alien plant species in relation to human disturbance on the Georgia Sea Islands. Diversity and Distribution 9:385-398. SAS Institute, Inc. 2004. SAS 9.1.3 Help and Documentation: Your Complete Guide to Syntax, How to, Examples, Procedures, Concepts, What’s New, and Tutorials. SAS Institute, Inc., Cary, NC. Shilling, D.G., T.A. Bewick, J.F. Gaffney, S.K. McDonald, C.A. Chase, and E.R.R.L. Johnson. 1997. Ecology, physiology, and management of Cogongrass (Imperata cylindrica). Publication Number: 03−107−140. Florida Institute of Phosphate Research, Bartow, FL Tyser, R.W., and C.A. Worley. 1992. Alien fl ora in grasslands adjacent to road and trail corridors in Glacier National Park, Montana (USA). Conservation Biology 6:253−262. USDA Natural Resource Conservation Service. 1999. Soil Survey of Perry County, MS. USDA Natural Resource Conservation Service, Lincoln, NE. 206 pp. USDA Soil Conservation Service and Forest Service. 1979. Soil Survey of Forrest County, MS. USDA Soil Conservation Service and Forest Service, Lincoln, NE. 103 pp. Walker, J.L., and A.M. Silletti. 2006. Restoring the overstory of Longleaf Pine ecosystems. Pp. 297−325, In S. Jose, E.J. Jokela, and D.L. Miller (Eds.). The Longleaf Pine Ecosystem: Ecology, Silviculture, and Restoration. Springer Science + Business Science Media, New York, NY. 437 pp. Willard, T.R. 1988. Biology, ecology, and management of Cogongrass (Imperata cylindrica (L.) Beauv.). Ph.D. Dissertation. University of Florida, Gainesville, FL. 113 pp. Willard, T.R, D.G. Shilling, J.F. Gaffney, and W.L. Currey. 1996. Mechanical and chemical control of Cogongrass (Imperata cylindrica). Weed Technology 10:722−726. Yager, L.Y. 2007. Watching the grass grow: Effects of habitat type, patch size, and land use on Cogongrass (Imperata cylindrica (L.) Beauv.) spread on Camp Shelby Training Site, Mississippi. Ph.D. Dissertation. Mississippi State University, Mississippi State, MS. 178 pp.