Initial invasion of cheatgrass (Bromus tectorum) into burned pinon-juniper woodlands in western Colorado.
Invasion by nonnative species is a growing concern throughout the United States, as these species, once introduced, can out-compete native species, causing a loss of biodiversity (Billings, 1990; DiTomaso 2000). In fact, 400 of 958 endangered or threatened species in the U.S. are primarily so because of competition with invasive species (Pimentel et al., 2000). Invasive species cause $137 billion per year in environmental damage (Pimentel et al., 2000).
Cheatgrass, one of the most problematic invasive species in the West, has invaded >40 million ha (DiTomaso, 2000), out-competing native species. Once introduced, cheatgrass quickly invades burned, disturbed or degraded land (Morrow and Stahlman, 1984), and even undisturbed grasslands (Young and Allen, 1997). Cheatgrass increases fine dry fuels, which increase fire, favoring continued cheatgrass dominance (Billings, 1990; Knapp, 1996). Once the ecosystem is modified in this manner, it becomes difficult to restore to its original composition (Hobbs, 2000). Cheatgrass controls, including herbicides, controlled grazing and biological control, must be repeated over several years and are expensive (Mosley et al., 1999). A key step in controlling invasive species may be to determine preferred locations of initial invasion and control plants in these areas before they expand and dominate (Byers et al., 2002). Yet, little is known about the initial invasion process for many invasive plants.
Cheatgrass flourishes in semi-arid environments, such as on the Uncompahgre Plateau in western Colorado, where this study of initial invasion was completed. This Plateau, part of the Colorado Plateau, is on the fringe of the most extensive areas, in the Great Basin, where native vegetation has been lost to cheatgrass and burned areas on the Plateau are particularly vulnerable to cheatgrass. The pinon-juniper zone on the Plateau typically burns in high-intensity, stand-replacing fires with a 400-600 y rotation (Shinneman, 2006). These burns are susceptible to cheatgrass invasion, as usually only a few native annuals are established by the next growing season and perennial grasses often do not recover before cheatgrass and other nonnative species can take hold (Arnold et al., 1964; Shinneman, 2006). Cheatgrass may have low but variable percent cover (i.e., 1-6%) 1-2 y after fire, but these minor differences in cheatgrass cover are followed by large differences 4-8 y after fire when cheatgrass cover can remain low or can surpass that of native plants (Shinneman, 2006).
Cheatgrass does not become dominant on all burns in pinon-juniper woodland on the Plateau (Beavers, 2001; Shinneman, 2006), but where it does, it is unclear whether particular features enhance invasion. Roads, for example, are widely recognized as dispersal pathways for nonnative species (Gelbard and Belnap, 2003) and roads run along or through many burned areas on the Uncompahgre Plateau. Off-road vehicles (ORVs) may also be driven through burns, carrying cheatgrass seed. Also, seed may survive in lightly-burned areas (Young et al., 1976). Seed may be blown in from the edge of adjacent unburned woodlands, or it may be a contaminant in seed mixes applied to decrease erosion after fire. For example, >1 billion cheatgrass seeds were found in seed applied after the 2000 Cerro Grande fire in New Mexico (Barclay et al., 2004). It also may be brought in by cattle or wildlife; generally, burned areas are not grazed by livestock for at least two growing seasons after fire, but in practice this is not always the case. Also, cheatgrass is initially patchy in burns, and the question arises: do certain plants inhibit or promote cheatgrass invasion? In Utah, for example, squirreltail (Elymus elymoides) may compete effectively with cheatgrass after fires (Ott et al., 2003).
We studied the pattern of cheatgrass invasion 2 y after wildfires on the Uncompahgre Plateau to address three questions about the initial invasion process: (1) is initial postfire cheatgrass more abundant in particular features within fires, (2) is cattle or deer and elk use associated with features that have more initial cheatgrass and (3) are particular grasses, that might retard or enhance cheatgrass invasion, associated with cheatgrass after fires?
The Uncompahgre Plateau extends 113 km northwest to southeast in southwestern Colorado, covering about 8000 [km.sup.2], with elevations ranging from about 1500-3200 m. Precipitation is 2550 cm annually, mostly as rain in late summer (Western Regional Climate Center, 2004). The Plateau has largely sedimentary substrata, with moderately shallow well-drained soils (USDA Natural Resource Conservation Service, 1995). The pinon-juniper zone we studied is from 1525-2440 m in elevation and mostly twoneedle pinyon (Pinus edulis) and Utah juniper (Juniperus osteosperma), whereas Rocky Mountain juniper (Juniperus scopulorum) occurs occasionally at higher elevations. Sagebrush shrublands occur within and below the zone, and ponderosa pine and Douglas-fir forests above it.
In the summer of 2004 we sampled four 2002 fires on the Plateau: Dierich, Coates Creek, Bucktail Buck and Burn Canyon (Fig. 1, Table 1), which have similar elevation and geology (Table 1). Two-yearold burned areas were chosen for sampling because a related study (Shinneman, 2006) discovered that by years 2-3 after a fire in this area, cheatgrass invasion is underway, but not complete, and the early invasion process could be studied.
Within each burned area, we sampled transects within four different features: roads, edges of fires adjacent to unburned woodland, aerially seeded fire interiors and unseeded fire interiors. Maps of seeded (areas were seeded after fires with similar seed mixes) and unseeded areas came from the Bureau of Land Management. Environmentally similar areas, adjacent to burned areas, were sampled to estimate pre-fire cheatgrass cover in burned areas.
We analyzed data by feature, not by fire. Fires could only be used as a stratum in stratified sampling spanning the fires because the number of transects per feature per fire could not be balanced, due to varying fire sizes and availability of features. Thirty transects of each feature and adjacent unburned areas were sampled for 150 total transects (Table 1).
Transects were randomly placed in fires by traveling to an access point near a particular feature to be sampled. We located a random starting point by choosing a number between 1 and 99 from a random number table, and then pacing that number of steps from the access point. For roads and edges, we paced steps along the side of the road or edge. Subsequent transects were systematically spaced by hiking in a randomly chosen direction until about 0.4 km from the previous transect. For the interior seeded, interior unseeded and adjacent transects, we randomly chose a direction and then paced the steps in that direction.
Each transect consisted of at least one, and up to six contiguous 10 x 10 m plots. As many as six plots were necessary to accurately estimate the mean cheatgrass cover of an area, based on a running mean, because cheatgrass sometimes varied widely over a small area. If cheatgrass cover varied little, one plot or at least 25 cheatgrass patches were measured. The total area of cheatgrass cover was measured in each plot separately, by breaking the plot into 2-m strips and tallying, by size class, each contiguous patch of cheatgrass in each 2-m strip. We calculated cheatgrass percent cover for each plot by dividing total cheatgrass cover, estimated from the size-class tallies, by the total area of the plot. Percentages for each plot were averaged among plots in each transect. Mean percent cover of cheatgrass for each feature was calculated by averaging estimates for each transect by feature. The standard error for each feature was then calculated to estimate how closely mean percent cover in transects approximates mean percent cover across each feature (Clark and Hosking, 1985).
To test the null hypothesis of no statistically significant difference between mean percent covers of cheatgrass across features, we used one-way analysis of variance (ANOVA; Ott, 1988). Before performing the ANOVA, we arcsine transformed the data, because they were percentages, containing both extreme and moderate values, resulting in high variances (Ott, 1988). Fisher's multiple comparison procedure was subsequently used to determine whether individual means differed (Toothaker, 1993).
[FIGURE 1 OMITTED]
We used finer sampling to analyze cheatgrass spread from roads and edges. For road and edge transects, the first 10 x 10 m plot (that directly adjacent to the road or edge) was divided into five 2 x 10 m strips with length parallel to the feature. We estimated cheatgrass percent cover separately for each strip. The 2-m strips were categorized by distance from the feature (0-2 m, 2--4 m, etc). We averaged percent covers within each distance category, arcsine transformed them and then used oneway ANOVA to test the null hypothesis of no difference in mean percent cover of cheatgrass among the 2-m strips (Ott, 1988).
We counted groups of wildlife (deer and elk) and cattle droppings in each plot to estimate wildlife and cattle use. We totaled cattle and wildlife droppings separately for each transect, then divided by the number of plots per transect to estimate density (groups of droppings per 100 [m.sup.2] plot). We then calculated mean density of droppings in each feature and used one-way ANOVA to test the null hypothesis of no difference in mean densities of droppings among features. Fisher's multiple comparison was used to determine which features had significantly differing mean densities of droppings (Toothaker, 1993).
To explore potential within-plot association of cheatgrass with other grasses, in 30 of the 150 transects we sampled 15, 1-m diameter microplots centered on a cheatgrass plant and 15, 1-m diameter microplots containing no cheatgrass. Cheatgrass microplots were chosen by walking methodically through the plot and centering a microplot on the first 15 cheatgrass. Microplots without cheatgrass were chosen by walking a randomly chosen number (1-9) of steps perpendicular to the central meter tape at each meter marker. If there was a cheatgrass plant in that area, we took additional steps until there was an area without cheatgrass. In each microplot, all grasses were identified and their presence recorded.
We used logistic regression (Hosmer and Lemeshow, 1989) to determine if cheatgrass was positively or negatively associated with a particular grass or set of grasses at a frequency significantly greater than could occur by chance. Stepwise logistic regression was used to determine AIC (predictive) and BIC (descriptive) values for increasing predictors (Shtatland et al., 2001). The lowest AIC and BIC values show the optimal number of predictors for either a predictive or descriptive model (Shtatland et al., 2001). Best subsets regression was used to determine the best group of species for each number of predictors (1-27). Once the best number of predictors was determined, and the species determined from the best subsets, a binary logistic regression was run with those predictors (Hosmer and Lemeshow, 1989).
The Coates Creek fire had exceptionally high mean percent cover of cheatgrass, had unusually high livestock use and was too small to contain more than one or two transects for each feature. These transects were treated as statistical outliers, included in the interpretation, but excluded from formal statistical analysis.
The ANOVA performed on mean cheatgrass percent cover among features showed a significant difference among means (F = 8.24, P = 0.000) and the null hypothesis of equal means was rejected. Fisher's multiple comparison showed that mean percent cover of cheatgrass did not differ significantly among edge, road and interior seeded features. Mean percent cover of cheatgrass for interior unseeded areas and adjacent woodlands also was not significantly different, but mean percent cover of cheatgrass for roads, edges and interior seeded features did differ significantly from those of interior seeded and adjacent features (Fig. 2). ANOVA for 2-m strips in first plots of the road and edge transects lacked significant F-values (roads: F = 0.17, P = 0.951; edges: F = 0.33, P = 0.854); the null hypothesis of no difference in mean percent cover of cheatgrass between strips could not be rejected.
The ANOVAs for both cattle and wildlife showed statistically significant F-values (cattle: F = 2.51, P = 0.044; wildlife: F = 6.36, P < 0.000), indicating that the null hypothesis of no difference in mean density of droppings between features can be rejected. Fisher's multiple comparison procedure showed that, for cattle, there was a higher mean density of droppings in road areas than in other features (Fig. 3a). For wildlife, mean density of droppings was significantly higher in adjacent woodlands than in other features, and mean density of droppings in edges was significantly higher than densities in road, interior seeded and interior unseeded areas (Fig. 3b).
[FIGURE 2 OMITTED]
The optimal number of predictor species indicated by AIC values was 7 and by BIC values was 1. The logistic regression was run with seven predictors, as one is not very informative. The 7-predictor model identified Japanese brome (Bromus japonicus) and prairie junegrass (Koeleria macrantha) as substantially increasing the odds of cheatgrass occurrence, but only prairie junegrass was significant at P < 0.10 (Table 2). Remaining grasses decreased odds of cheatgrass occurrence, but only the native, James' galleta (Pleuraphis jamesii) and the non-native, intermediate wheatgrass (Thinopyrum intermedium), were significant at P < 0.10 and P < 0.05 respectively.
Edges, roads and seeded interior did not significantly differ from each other in mean percent cheatgrass cover, but had significantly higher cheatgrass cover than unseeded interior and adjacent transects (Fig. 2). Small differences in cheatgrass cover during years 1-2 after fire appear to result in much larger differences 4-8 y after fire (Shinneman, 2006). Roads are a significant location of invasion; vehicles can transport seed, disturbance occurs along roads from grading, ditching and mowing and there is increased water availability due to run-off (Gelbard and Belnap, 2003). Also, there was a high presence of cattle along roads relative to other parts of the fire area (Fig. 3a). Cattle may contribute to spread of cheatgrass seed by carrying it in their hide and by increasing disturbance (both trampling and grazing).
It was unexpected to find the edge of fires to be a principal location of cheatgrass invasion, but there are several possible explanations. Initially it was assumed that because of the high intensity of the fires, cheatgrass seed in the seedbank would be largely destroyed; however, the fire is going out along the edge, and is likely less hot, allowing more survival in the seedbank (Young et al., 1976). Cheatgrass seed remains viable in the seedbank for at least 1 y, whereas seed stored in paper sacks for up to 11.5 y had 95-100% germination (Hulbert, 1955; Billings, 1994). Moreover, cheatgrass seed exposed to smoke results in large, vigorous plants with more tillers and seed than plants from seed not exposed to smoke (Young and Evans, 1978; Bradley, 1999). If seed survived in the burn edges, high seed production from these plants could result in increased cover of cheatgrass within 2 y.
[FIGURE 3 OMITTED]
Another potential reason for high percent cover of cheatgrass in edge transects was the high presence of wildlife (deer and elk) in the edge (Fig. 3b). Wildlife may disturb already vulnerable ground and increase the favorability of the site for cheatgrass invasion. It seems unlikely that wildlife are spreading cheatgrass seed into the edge from adjacent unburned woodlands, because there is little cheatgrass in the unburned woodlands (Fig. 2, Fig. 3b). However, wildlife could possibly be spreading seed from other locations not sampled, such as meadows or sagebrush, which are often interspersed with woodlands.
The significant difference in mean percent covers of cheatgrass between the interior seeded and interior unseeded areas (Fig. 2) suggests that aerial seeding, while meant to help prevent invasion by weeds such as cheatgrass, is actually associated with increased cheatgrass invasion in burn interiors. This may happen either through cheatgrass seed contaminating the seed mix (e.g., Barclay et al., 2004), or through seeding of plants which favor cheatgrass, rather than inhibiting it. Other studies have shown mixed results in suppressing cheatgrass by seeding (Beavers, 2001; Beyers, 2004). Of course, some seed mixes may be well suited to suppressing weed invasions, whereas others have lower weed-suppressing capabilities.
There was no significant difference in mean percent cover of cheatgrass between unseeded interiors and adjacent woodlands (Fig. 2). This shows that, in the interior of the burned area which is not seeded, there is not an increase in cheatgrass simply due to the fire and seeding may not be needed any more than it is in unburned woodlands. Seeding also was ineffective in reducing cheatgrass below levels found in unburned woodlands.
There was no significant difference in mean percent cover of cheatgrass among 2-m strips in the first 10 x 10 m plot along either roads or edges. Thus, if there is a change in cover of cheatgrass with distance from the edge or road, it is not found within the first 10 m.
Given the high cover of cheatgrass in edge areas and the low cover of cheatgrass in the unseeded interior of the burned areas, small fires with a high edge-to-interior ratio may be most vulnerable to complete invasion and dominance by cheatgrass. Coates Creek, which was only 2.5 ha in area and was not seeded, had very high cheatgrass cover, possibly partially from the high edge-to-interior ratio. The Coates Creek fire, and other small fires, may also have been less hot than large fires, resulting in both a small area burned and high survival in the seedbank. There was also very high cover of cheatgrass in woodlands surrounding the Coates Creek burn, so there was likely a very large seedbank prior to the fire. Small bums, especially if cheatgrass is abundant nearby, may be most vulnerable to cheatgrass invasion. Larger bums, while having more total edge area, may be less vulnerable to cheatgrass dominance because perennial grasses can recover in the interior before cheatgrass dominates. Perhaps in larger bums, direct control of cheatgrass could be concentrated around edges and along roads where invasion is most abundant. Because present seeding may be enhancing cheatgrass invasion, seeding as a method of cheatgrass control is not justified at this time.
However, additional research is warranted to identify specific plants that might resist cheatgrass invasion. Burned areas are often widely seeded to prevent erosion and weed invasion (Beyers, 2004) and it is therefore important to know which plant species are actually likely to prevent or enhance cheatgrass invasion. In this study only three grass species were significantly associated with cheatgrass, either positively or negatively. Other grasses, however, were close to significant at [alpha] = 0.10, but had low sample size; with an increased sample, these grasses might be significantly associated with cheatgrass. This part of the study is exploratory and small sample sizes underscore the need for further study.
Nonetheless, cheatgrass was about seven times more likely to be found around prairiejunegrass than expected by chance. This association could be one of either mutualism or competition. If it is competition, prairie junegrass may be a wise choice for postfire seed mixes because it appears to successfully establish and grow near cheatgrass. If the association is mutualism, however, prairie junegrass may enhance, rather than prevent invasion. Similarly, James' galleta decreases the odds of finding cheatgrass by a factor of about six, and galleta could be suppressing cheatgrass, but cheatgrass could instead be excluding James' galleta. Further study is necessary to determine the nature of these associations.
This study shows that seeding appears to be promoting cheatgrass invasion rather than suppressing it. Other studies, however, have shown that some non-native grasses, such as crested wheatgrass (Agropyron cristatum) and intermediate wheatgrass, when seeded, will compete with cheatgrass (Ott et al., 2003). In this study intermediate wheatgrass decreased the chance of finding cheatgrass when it was present, but only by a factor of two. Native grasses, in this study, appear to have more promise, given their stronger associations with cheatgrass. Native perennial grasses, such as squirreltail (Elymus elymoides), have been shown to effectively compete with cheatgrass following a fire, if they were abundant before the fire (Ott et al., 2003). As more information about plant associations with cheatgrass and other weeds is collected and analyzed, choosing the appropriate grass seed for seed mixes and including native species which suppress cheatgrass invasion becomes more practical.
There is not one specific location of cheatgrass invasion into these burned areas. Rather, edges of fire areas, roads and seeded interiors in fire areas are important locations of cheatgrass invasion. This is somewhat discouraging, as these three features in burned areas cover a wide area. Edges and seeded interior are especially problematic because they not only cover a large area, but may be difficult to reach, limiting the type and amount of management feasible to control cheatgrass. There was no significant trend in mean cheatgrass cover in the first 10 m of either roads or edges. Unfortunately, cheatgrass abundance is quite variable along these features, making control efforts more difficult.
Unseeded interiors were not significantly different in mean cheatgrass cover compared to adjacent unburned woodlands and do not appear to be enhanced locations of cheatgrass from seed surviving in the seedbank. Seeding of burn interiors could be suspended, as current methods of seeding are associated with enhanced cheatgrass invasion. In small fires surrounded by cheatgrass, cheatgrass is likely to dominate in any case, and in these areas it seems sensible to experiment with seeding promising native grasses and forbs, such as James' galleta and prairie Junegrass, to determine if they will resist cheatgrass invasion. Along roads and edges where cheatgrass is very abundant, direct control of cheatgrass may be the most effective method of control until more successful seed mixes or seeding methods can be developed. Further research is needed to understand the invasion process and to identify native species that may compete effectively with cheatgrass if effective plans are going to be made to counteract cheatgrass invasion into burned areas of pinon-juniper woodlands.
Acknowledgments.--This research was funded in part by the Bureau of Land Management (BLM) under Agreement ESA020016. Thanks to A. Clements and the BLM Uncompahgre Field Office in Montrose, Colorado, for providing housing and transportation during field work, D. Selby for assistance with field work, M. Andersen for assistance with logistic regression, and D. Shinneman and S. Lanning for assistance with GIS.
ARNOLD, J. F., D. A. JAMESON AND E. H. REID. 1964. The pinyon-juniper type of Arizona: effects of grazing, fire, and tree control. USDA Production Research Report No. 84, Rocky Mountain Forest and Range Experiment Station, Fort Collins, Colorado. 28 p.
BAKER, W. L. AND D. J. SHINNEMAN. 2004. Fire and restoration of pinyon-juniper woodlands in the western United States: a review. For. Ecol. and Manage., 189:1-21.
BARCLAY, A. D., J. L. BETANCOURT AND C. D. ALLEN. 2004. Effects of seeding ryegrass (Lolium multiflorum) on vegetation recovery following fire in a ponderosa pine (Pinus ponderosa) forest. Intern. J. Wildl. Fire, 13:183-194.
BEAVERS, A. M. 2001. Vegetation recovery following fire in west-central Colorado. Master's thesis, Colorado State University, Fort Collins. 31 p.
BEYERS J. L. 2004. Postfire seeding for erosion control: effectiveness and impacts on native plant communities. Cons. Biol., 18:947-956.
BILLINGS, W. D. 1990. Bromus tectorum, a biotic cause of ecosystem impoverishment in the Great Basin, p. 310-322. In: G. M. Woodwell (ed.). The earth in transition: patterns and processes of biotic impoverishment. Cambridge University Press, New York, New York.
--. 1994. Ecological impacts of cheatgrass and resultant fire on ecosystems in the Western Great Basin. U.S. Department of Agriculture, Forest Service Intermountain Research Station, Ogden, UT. General Technical Report INT-GTR-313.9 p.
BRADLEY, T. S. 1999. Prescription fire in western ecosystems: microsite recovery of vegetation in a pinyon juniper woodland and bark beetle response to fire injured trees in the Lake Tahoe Basin [thesis]. University of Nevada, Reno, NV. 87 p.
BYERS, J. E., S. REICHARD, J. M. RANDALL, I. M. PARKER, C. S. SMITH, W. M. LONSDALE, I. A. E. ATKINSON, T. R. SEASTEDT, M. WILLIAMSON, E. CHORNESKY AND D. HAYES. 2002. Directing research to reduce the impacts of nonindigenous species. Cons. Biol., 16:630-640.
CLARK, W. A. V. AND P. L. HOSKING. 1986. Statistical methods for geographers. John Wiley and Sons, New York, New York.
DITOMASO, J. M. 2000. Invasive weeds in rangelands: species, impacts, and management. Weed Sci., 48:255-265.
GELBARD, J. L. AND J. BELNAP. 2003. Roads as conduits for exotic plant invasions in a semiarid landscape. Cons. Biol., 17:420-432.
HOBOS, R.J. 2000. Land-use changes and invasions, p. 55-64. In: H. A. Mooney and R.J. Hobbs (eds.). Invasive species in a changing world. Island Press, Washington, D.C. HOSMER, D. W. AND S. LEMESHOW. 1989. Applied logistic regression John Wiley & Sons, New York, New York. HULBERT, L. C. 1955. Ecological studies of Bromus tectorum and other annual bromegrasses. Ecol. Monogr., 25:181-213.
KNAPP, P. A. 1996. Cheatgrass (Bromus tectorum) dominance in the Great Basin Desert: history, persistence, and influences to human activities. Glob. Environ. Change, 6:37-52.
MORROW, L. A. AND P. W. STAHLMAN. 1984. The history and distribution of downy brome (Bromus tectorum) in North America. Weed Science, Supplement, 32:2-6.
MOSLEY, J. C., S. C. BUNTING AND M. E. MANOUKIAN. 1999. Cheatgrass. p. 175-188. In: R. L. Sheley and J. K. Petroff (eds.). Biology and management of noxious rangeland weeds. Oregon State University Press, Corvallis, Oregon.
OGLE, S. M. AND W. A. REINERS. 2002. A phytosociological study of exotic annual brome grasses in a mixed grass prairie/ponderosa pine forest ecotone. Am. Midl. Nat., 147:25-31.
OTT, J. E., E. D. McARTHUR AND B. A. ROUNDY. 2003. Vegetation of chained and non-chained seedings after wildfire in Utah. J. of Range Manage., 56:81-91.
OTT, L. 1988. An introduction to statistical methods and data analysis. (3rd ed). PWS-Kent, Boston, Massachusetts.
PIMENTEL, D., L. LACH, R. ZUNIGA AND D. MORRISON. 2000. Environmental and economic costs of nonindigenous species in the United States. Bioscience, 50:53--64.
SHINNEMAN, D. J. 2006. Determining restoration needs for pinon-juniper woodlands and adjacent semiarid ecosystems on the Uncompahgre Plateau, western Colorado. Ph.D. dissertation, University of Wyoming, Laramie, WY.
SHTATLAND, E. S., E. CAIN AND M. B. BARTON. 2001. The perils of stepwise regression and how to escape them using information criteria and the output delivery system. Paper 222-26, SAS Institute, Inc. Cary, NC.
TOOTHAKER, L. E. 1993. Multiple comparison procedures. Sage Publications, Newbury Park, California.
USDA NRCS. 1995. Soil Survey of Uncompahgre National Forest area, Colorado: parts of Mesa, Montrose, and San Miguel Counties. USDA Forest Service and Soil Conservation Service.
--. 2005. The PLANTS database, Version 3.5. Available at: http://plants.usda.gov. Accessed 12 May 2005.
WESTERN REGIONAL CLIMATE CENTER. 2004. Western U.S. Climate Historical Summaries. Available at: http://www.wrcc.dri.edu/climsum.html. Accessed 10 Dec. 2004.
YOUNG, J. A. AND F. L. ALLEN. 1997. Cheatgrass and range science: 1930-1950. J. of Range Manage., 50:530-535.
-- AND R. A. EVANS. 1978. Population dynamics after wildfires in sagebrush grasslands. J. of Range Manage., 31:283--289.
--, --, R. E. ECKERT, JR. AND B. L. KAY. 1987. Cheatgrass. Rangelands, 9:266-270. --, -- AND R. A. WEAVER. 1976. Estimating potential downy brome competition after wildfires. J. of Range Manage., 29:322-325.
HILARY L. GETZ AND WILLIAM L. BAKER (l), Department of Geography, Dept. 3371, 1000 E. University Avenue, University of Wyoming, Laramie 82071. Submitted 29 January 2007; accepted 8 November 2007.
(1) Corresponding author: e-mail: email@example.com
TABLE 1.--Sampled fires, fire areas and the number of transects sampled in each feature Fire area Interior Interior Unburned Fire name (ha) Roads Edges seeded unseeded adjacent Bucktail-buck 908 6 5 5 3 5 Burn canyon 12,260 21 18 17 23 17 Coates creek 3 1 1 0 1 2 Dierich Creek 1,025 2 6 8 3 6 Totals 14,196 30 30 30 30 30 TABLE 2.--Grasses and the sample size, significance, and odds ratio measuring their association with cheatgrass, based on the [AIC.sub.k] (predictive) logistic regression model Predictor n p Odds ratio Constant 0.802 Bromus japonicas 6 0.998 1.96E+09 Koeleria macrantha 8 0.072 6.88 Pleuraphis jamesii 8 0.067 0.14 Bouteloua gracilis 3 0.999 0.00 Thinopyrum intermedium 44 0.034 0.49 Poa pratensis 30 0.114 0.28 Elymus elymoides 41 0.180 0.81
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Notes and Discussion|
|Author:||Getz, Hilary L.; Baker, William L.|
|Publication:||The American Midland Naturalist|
|Date:||Apr 1, 2008|
|Previous Article:||Effects of species, density, season and prairie-type on post-dispersal seed removal in Oklahoma.|
|Next Article:||The Cumberland Plateau disjunct paradox and the biogeography and conservation of pond-breeding amphibians.|