Management Implications
Introduced annual grasses have caused severe ecological damage in native rangeland ecosystems across the western United States, and considerable efforts have been made to control introduced species and reestablish native species. However, without an appropriate survey method, it is difficult to evaluate the success of weed management and restoration projects. Visual estimation and line-point intercept (LPI) methods are commonly used to estimate vegetation cover, and researchers have often debated which method is the most reliable. In our study, we compared species richness, diversity (Shannon index), cover, and biomass estimates from one visual (quadrat-based) and two LPI (canopy and basal) methods in four sites in Wyoming, USA. Greater species richness and Shannon index values were measured with quadrats, while both introduced and native graminoid cover was higher in LPI canopy, except for two sites with similar results. Biomass regression models developed using quadrat-based estimates were more reliable than the others. Because quadrats outperformed LPI canopy in number of species and biodiversity and performed at least as well for graminoid cover, quadrats should be the preferred method. However, quadrats may be subject to observer bias in visually estimating cover, while LPI is considered nonbiased. When operators measure species incidence (presence or absence) and cover estimation is secondary, quadrats become bias free. Therefore, visual estimation methods should be adopted when biodiversity is prioritized, while LPI canopy should be used when estimation of cover is the main goal. Generally, we think that cover represents a fast and reliable method to estimate graminoid biomass, and our results indicated quadrat-based estimates produced the most accurate results in terms of coefficient of determination, but these outcomes might be influenced by the fact that the biomass clippings used to develop the models were obtained from the same quadrats.
Introduction
One of the most important components of ecological restoration projects is monitoring of plant communities. Effective monitoring allows land managers to estimate potential changes in community structure and processes over the time, providing a rationale to evaluate the effectiveness of restoration treatments (D’Antonio and Meyerson Reference D’Antonio and Meyerson2002; Davies et al. Reference Davies, Bates and O’Connor2021; Mealor et al. Reference Mealor, Mealor, Kelley, Bergman, Burnett, Decker, Fowers, Herget, Noseworthy and Richards2013; Sutter Reference Sutter, Falk, Millar and Olwell1996). Therefore, choosing an appropriate monitoring methodology becomes a high-priority decision for scientists and natural resource professionals involved in restoration and adaptive management (Mealor et al. Reference Mealor, Mealor, Kelley, Bergman, Burnett, Decker, Fowers, Herget, Noseworthy and Richards2013). In the western United States, introduced annual grasses, such as bromes (Bromus spp.), ventenata [Ventenata dubia (Leers) Coss.], and medusahead [Taeniatherum caput-medusae (L.) Nevski], have invaded large areas, displacing native species and altering ecosystem structure and function (Jones et al. Reference Jones, Norton and Prather2018; Monaco et al. Reference Monaco, Mangold, Mealor, Mealor and Brown2017). Extensive efforts are underway to restore native rangeland ecosystems across the western United States, aiming to reestablish native species; eradicate invasive species; implement best management practices; and ultimately restore ecosystem structure, composition, stability, and functionality (D’Antonio and Meyerson Reference D’Antonio and Meyerson2002; Davies Reference Davies2011; DiTomaso et al. Reference DiTomaso, Masters and Peterson2010; Humphrey and Schupp Reference Humphrey and Schupp2004).
To track restoration goals, monitoring must rely on dependable sampling techniques capable of reliably and consistently detecting spatial and temporal changes in vegetation structure (Chen et al. Reference Chen, Li, Zhang, Zhou, Koehler, Leblanc, Fraser, Olthof, Zhang and Wang2009; Kopecký and Macek Reference Kopecký and Macek2015; Sutter Reference Sutter, Falk, Millar and Olwell1996), with particular attention to vegetation cover, productivity, species richness, and diversity (Elzinga et al. Reference Elzinga, Salzer, Willoughby and Gibbs2009; Herrick et al. Reference Herrick, Van Zee, Havstad, Burkett and Whitford2005; Seefeldt and Booth Reference Seefeldt and Booth2006). These indicators are closely associated with habitat quality, resilience to invasion, grazing, erosion potential, and climate change (Herrick et al. Reference Herrick, Duniway, Pyke, Bestelmeyer, Wills, Brown, Karl and Havstad2012; Pyke et al. Reference Pyke, Herrick, Shaver and Pellant2002). Particularly, plant cover is a key indicator of rangeland condition, and it can be efficiently measured and assessed (Booth and Tueller Reference Booth and Tueller2003). Furthermore, researchers and managers have collaborated to produce standardized protocols to obtain repeatable, statistically defensible, and comparable data (Herrick et al. Reference Herrick, Lessard, Spaeth, Shaver, Dayton, Pyke, Jolley and Goebel2010; Wirth and Pyke Reference Wirth and Pyke2007).
Among the various vegetation sampling techniques described in the scientific literature, visual assessment methods such as Daubenmire and other size quadrats, and line-point intercept (LPI) methods are often used to assess cover at the species or functional group level (Thacker et al. Reference Thacker, Messmer and Burritt2015). Several authors have compared the two methods, trying to identify the most efficient way to sample plant cover both in greenhouse experiments and natural ecosystems (Dethier et al. Reference Dethier, Graham, Cohen and Tear1993; Floyd and Anderson Reference Floyd and Anderson1987; Gregg Reference Gregg2006; Herrick et al. Reference Herrick, Van Zee, Havstad, Burkett and Whitford2005; Hulvey et al. Reference Hulvey, Thomas and Thacker2018; Karl et al. Reference Karl, Karl, McCord and Kachergis2016; Martyn et al. Reference Martyn, Beltz, Palmquist, Pennington, Rottler and Lauenroth2015; Thacker et al. Reference Thacker, Messmer and Burritt2015). In these studies, the two methods were evaluated for rangeland habitat management and conservation, with an emphasis on greater sage-grouse [Centrocercus urophasianus (Bonaparte, 1827)] habitat monitoring (Hulvey et al. Reference Hulvey, Thomas and Thacker2018; Karl et al. Reference Karl, Karl, McCord and Kachergis2016; Martyn et al. Reference Martyn, Beltz, Palmquist, Pennington, Rottler and Lauenroth2015; Thacker et al. Reference Thacker, Messmer and Burritt2015). However, we have found no comparative studies focused on informing decisions around management of invasive annual grasses in rangelands. Quantitative information about pre- and posttreatment conditions is particularly needed, as the demand for more precise data on rangeland condition and restoration effectiveness has increased, while financial resources for monitoring remain, unfortunately, limited (Davies et al. Reference Davies, Boyd, Bates, Hamerlynck and Copeland2020, Reference Davies, Bates and O’Connor2021). Pretreatment assessments help land managers develop effective management strategies, while posttreatment monitoring evaluates whether plant communities are shifting in the desired direction to meet management goals and objectives—optimizing resources invested in the management strategy (Mealor et al. Reference Mealor, Mealor, Kelley, Bergman, Burnett, Decker, Fowers, Herget, Noseworthy and Richards2013). In ongoing discussions with the Wyoming Invasive Grasses Task Force, land managers posed the question of which monitoring methods were best suited to evaluate “recovery potential” (the potential for desirable plants to respond favorably to annual grass control) before annual grass control and to capture plant community response to annual grass treatments. We were concerned that different monitoring techniques may yield inconsistent data regarding species richness, diversity, and cover, affecting the ability to detect changes in the vegetation community response to restoration treatments.
The main goal of this study was to compare three sampling techniques (quadrat based, LPI canopy, and LPI basal) to assess plant community diversity and cover and to estimate biomass production on sites invaded by introduced annual grasses. The specific objectives were to (1) quantify species richness in relation to introduced annual graminoid cover, (2) determine species diversity in terms of the Shannon-Weiner index, (3) measure differences in vegetation cover, and (4) assess the relationship between plant cover (by method) and biomass production of cheatgrass (Bromus tectorum L.) and native perennial graminoids. We were particularly interested in how these sampling approaches, as typically employed in the field, performed in characterizing vegetation in pre- and posttreatment conditions.
Materials and Methods
Locations
We compared vegetation monitoring methods focused on annual grass invasion impacts to plant community composition and pretreatment assessments at four sites in Wyoming: Hyattville Pinedale, Saratoga, and Sheridan (Figure 1). Topographic, climate, and soil characteristics varied among sites and are listed in Table 1. The Hyattville, Pinedale, and Saratoga sites are located within the Cold Desert level II EPA ecoregion, while Sheridan is located within Temperate Prairies. The most prevalent soil series were Neville (Entisols) and Teensleep (Aridisols) in Hyattville, Pinedale (Mollisols) and Noclios (Alfisols) in Pinedale, Dranburn and Kilgore (both Mollisols) in Saratoga, and Workfa (Aridisols) and Samday and Shingle (Entisols) in Sheridan (Soil Survey Staff 2014).
Figure 1. Locations of the four study sites in Wyoming, USA.
Table 1. Descriptive details for four Wyoming sites (each ∼30 ha) where annual grass vegetation sampling methods were evaluated. a
a Mean annual precipitation (MAP) was calculated using PRISM climate group data (www. https://prism.oregonstate.edu/). Soil texture, pH, electrical conductivity (EC), organic matter (OM), NO3-N, P, and K were based on multiple samples aggregated across each field site to a depth of 20 cm.
Native vegetation across the study sites consisted of sagebrush species, such as silver sagebrush (Artemisia cana Pursh), prairie sagewort (Artemisia frigida Willd.), and mountain big sagebrush [Artemisia tridentata Nutt. ssp. vaseyana (Rydb.) Beetle], associated with native perennial grasses, including western wheatgrass [Pascopyrum smithii (Rydb.) Á. Löve], needle-and-thread grass [Hesperostipa comata (Trin. & Rupr.) Barkworth], Idaho fescue (Festuca idahoensis Elmer), bluebunch wheatgrass [Pseudoroegneria spicata (Pursh) Á. Löve], and Sandberg bluegrass (Poa secunda J. Presl). Common introduced annual grasses included B. tectorum, Japanese brome (Bromus arvensis L.), soft brome (Bromus hordeaceus L.), smooth brome (Bromus inermis Leyss.), rattlesnake brome (Bromus briziformis Fisch. & C.A. Mey.), and V. dubia. All sites except Hyattville have had livestock grazing excluded for many years, whereas the Hyattville site was moderately grazed as part of a rotational grazing system through recent history. All sites except the Sheridan site had burned previously, but no burns had occurred within 10 yr before the current research being conducted.
Experimental Design
Because we were particularly interested in evaluating information gathered from three different vegetation monitoring methods across a broad range of annual grass abundance, the overall sampling scheme was developed to be analyzed within an experimental regression framework. Response variables were treated differently based on the question being addressed. Approximate plot density was 6 plots ha−1, which was higher sampling density than many management programs would install, but sampling at this density allowed us to develop sample-based species accumulation curves at each site to understand the relationship between the number of sample locations and detection of species richness (Colwell and Coddington Reference Colwell and Coddington1994; Magurran Reference Magurran2004). Our sampling approach also allowed us to evaluate site-level vegetation characteristics using the three monitoring methods as categorical predictor variables.
Vegetation Surveys and Data Collection
We collected vegetation data at the Saratoga and Pinedale field sites from June 22 to July 1 in 2015, and at the Hyattville and Sheridan field sites from June 6 to 17 in 2016. At each field site, we sampled areas of approximately 33 ha that contained a broad range of B. tectorum abundance. After preliminary site mapping for annual grass cover, we established plots that represented B. tectorum cover across each site ranging from absence (0% cover) to dominance (cover equal to or greater than 50%). A total of 627 circular plots (r = 7.62 m) were established in the four sites, divided as follows: 219 plots at Hyattville (73 per each sampling method), 192 plots at Pinedale (64 per each sampling method), 198 plots at Saratoga (66 per each sampling method), and 180 at Sheridan (60 per each sampling method).
At each plot, we collected plant cover via LPI and quadrat-based methods. Three 0.5 by 0.5 m (0.25-m2) quadrats were randomly placed inside every circular plot, and a 15.24-m transect was established along the plot as well (Figure 2). We collected LPI data (canopy and basal cover) at 50 points (each represented by a 45.72-cm-long by 0.2-cm-diameter pin) placed at 0.3-m intervals along the transect, assessing cover at the species level with the addition of litter, rock, and bare ground as additional cover categories. We visually estimated species-level quadrat cover data in the three randomly placed quadrats by assigning values for each species into one of seven cover classes: (1) 0%, (2) 1% to 5%, (3) 5% to 25%, (4) 25% to 50%, (5) 50% to 75%, (6) 75% to 95%, and (7) 95% to 100%. Vegetation visual cover assessment was done independently by individual species, enabling vertical overlapping and therefore allowing cover values greater than 100%. LPI estimation was conducted recording the numbers of times that a species was “hit” by the pin dropped vertically to the ground. Then, percent cover was calculated by dividing the number of hits for each species by 50 (total number of sampling points along each transect) and multiplying the result by 100. A canopy “hit” is defined when the pin intercepts any part of the plant (leaves or stems) that can intercept raindrops or provide shade from vertical sunlight (gaps not included). A basal “hit” occurs when the pin intercepts the plant basal cover, defined as the area of the ground surface covered by the basal part of plants. Generally, basal cover is considered most stable, because it does not vary as much in relation to climatic variation or grazing.
Figure 2. Plot design.
Vegetation cover assessment was performed by the same observer to ensure consistency; therefore, observer 1 assessed the vegetation cover in the quadrats, observer 2 in the transects (LPI canopy and basal). Within the three quadrats, we clipped and bagged all current-year aboveground herbaceous biomass for B. tectorum and native perennial graminoids. Biomass samples were dried in a forced-air oven at 60 C for 72 h, weighed to the nearest milligram, and pooled by plot. Within each site, 15 soil cores (20 cm) were collected and sent to the laboratory (www.wardlab.com) to measure soil pH, electroconductivity (EC), total organic matter (OM), nitrate-nitrogen (NO3-N), P, and K. Soil pH and EC were quantified using a 1:1 soil–water solution, while soil OM was estimated by loss-on-ignition. NO3-N was determined using a flow-injection analyzer. K (ammonium acetate) and P (Mehlich-3) were assessed using the inductively coupled plasma-optical emission spectrometry method.
Species Richness and Diversity
We computed species richness (S) and Shannon index (H′) (Hayek and Buzas Reference Hayek and Buzas2010; Jost Reference Jost2006, Reference Jost2007; Magurran Reference Magurran2004) from vegetation cover collected using the three methods. To calculate species richness, vegetation cover was converted into species incidence (presence/absence of a species in form of 1 or 0 respectively), and then each species value was summed to obtain the total number of species per plot. The Shannon index was calculated using vegetation cover as species proportion as follows:
$H' = - \mathop \sum \nolimits_{i = 1}^N {p_i}{\rm{ln}}{p_i}$
where p i represents the relative proportion of the ith species.
We used species richness data to generate sample-based species accumulation curves per each survey method within each site, which describe the cumulative number of species discovered in a community as a function of sampling effort, defined as cumulative number of samples (Colwell and Coddington Reference Colwell and Coddington1994; Magurran Reference Magurran2004). We plotted sample-based species accumulation curves indicating the number of sampling units (circular plots) on the x axis, and the cumulative number of species on the y axis (Gotelli and Colwell Reference Gotelli and Colwell2001, Reference Gotelli and Colwell2011). The number of species in a plot was measured by aggregating the species detected in the three quadrats for the quadrat-based method and the number of species hit by pins along the transects for the LPI methods.
Data Analyses
ANOVA was conducted to test for differences in species richness, Shannon index, species cover, and grass biomass at α = 0.05 significance level. Vegetation survey method, site, and their interaction were used as fixed effects. When ANOVA indicated significant effects, means separations were performed using Tukey’s honest significant difference test. We performed simple linear regression analyses between (1) B. tectorum aboveground biomass and its percent cover and (2) native perennial graminoid biomass and their relative percent cover and tested significance with analysis of covariance.
Shannon index and sample-based species accumulation curves with 95% confidence intervals were estimated using the R package vegan (Oksanen et al. Reference Oksanen, Blanchet, Friendly, Kindt, Legendre, McGlinn, Minchin, O’Hara, Simpson, Solymos, Stevens, Szoecs and Wagner2020). Curves with overlapping confidence intervals were considered the same. All statistical analyses were performed using R (R Core Team 2021), and the plots were produced using the ggpubr package (Kassambara Reference Kassambara2020).
Results and Discussion
Species Richness and Diversity
Results suggested that vegetation survey methods differed in the overall number of species detected, as shown by sample-based species accumulation curves (Figure 3). Quadrat-based estimates detected a higher number of species in all four sites, followed by LPI canopy and LPI basal. ANOVA results (Figure 4, Supplementary Table S2) reflected the outcomes of sample-based species accumulation curves, indicating differences among methods within each site for both species richness and Shannon index. These differences were more evident at Sheridan and Saratoga, where richer and more even communities grew. Means and relative standard errors for species richness and Shannon index are indicated in Table 2. Generally, quadrats outperformed both LPI methods in detecting the number of species and measured more diverse and equally distributed communities in terms of Shannon index.
Figure 3. Sample-based species accumulation curves (solid lines) and 95% confidence intervals (dashed lines) for the three sampling methods in the four study sites. LPI, line-point intercept.
Figure 4. Box plots of number of species, Shannon index, introduced annual, and native perennial graminoid percent cover in the four sites. The lower and the upper parts of the box represent the first quartile (Q1) and the third quartile (Q3) respectively. The horizontal line denotes the median, the whiskers the upper and lower extremes, and the dots the outliers. Different letters indicate differences between vegetation survey methods within a site.
Our results were similar to those from other studies based on the contrast of point intercept versus visual cover methods (Etchberger and Krausman Reference Etchberger and Krausman1997; Godínez-Alvarez et al. Reference Godínez-Alvarez, Herrick, Mattocks, Toledo and Van Zee2009; Kinsinger et al. Reference Kinsinger, Eckert and Currie1960; Stohlgren et al. Reference Stohlgren, Bull and Otsuki1998; Symstad et al. Reference Symstad, Wienk and Thorstenson2008). The reason for these differences lies in the tendency of point intercept methods to miss less frequent and rare species: a species must be hit by at least one pin to be recorded, and the likelihood of this occurrence decreases with cover (Friedmann et al. Reference Friedmann, Pauli, Gottfried and Grabherr2011; Mamet et al. Reference Mamet, Young, Chun and Johnstone2016). Moreover, it is well documented that increasing the sampling area and the minimal scale sampled will increase species richness (Palmer and White Reference Palmer and White1994; Rapson et al. Reference Rapson, Thompson and Hodgson1997). Korb et al. (Reference Korb, Covington and Fulé2003) measured a direct correlation between area sampled and number of species detected, with the low number of species captured in the point intercept transect (0.1 m2), followed by the Daubenmire transect (4 m2), the belt transect (500 m2), and a modified-Whittaker plot (1,000 m2), indicating a direct correlation between sampling area and ability to detect new species. In our study, quadrats had the higher sampling area compared with the LPI methods, confirming the species–area correlation highlighted in past studies. The limited number of measured points in LPI methods decreased the likelihood of detecting rare species. Because of this issue, some authors regarded LPI not suitable when monitoring biodiversity (Korb et al. Reference Korb, Covington and Fulé2003; Leis Reference Leis2003; Prosser et al. Reference Prosser, Skinner and Sedivec2003; Stohlgren et al. Reference Stohlgren, Falkner and Schell1995). Therefore, we suggested that visual methods are more appropriate when the assessment of total number of species, species evenness, and biodiversity are the priorities.
Introduced Annual and Native Perennial Graminoid Cover
There were differences in cover of introduced annual graminoid cover between quadrats and LPI basal, quadrats and LPI canopy, and quadrats and LPI basal at all sites, but not between quadrats and LPI canopy at Hyattville and Pinedale (Figure 3; Table 2). At the remaining two sites, all three methods differed from one another, with higher cover levels detected by LPI canopy. For native perennial graminoid cover, there were differences between quadrats and LPI basal, quadrats and LPI canopy, and quadrats and LPI basal at all the sites, but not between quadrats and LPI canopy at the Hyattville and Sheridan sites. The other two sites had differences between LPI canopy and the other methods, with LPI canopy detecting higher cover values than the other two methods. Our outcomes were in part consistent with other studies in which point intercept methods yielded higher grass cover values than quadrats (Korb et al. Reference Korb, Covington and Fulé2003; Thacker et al. Reference Thacker, Messmer and Burritt2015). Generally, point-based methods are considered more objective and precise than visual-based techniques, because they use pins (rather than a visual estimation) to detect points of contact of plant species (Bonham Reference Bonham2013; Dethier et al. Reference Dethier, Graham, Cohen and Tear1993; Elzinga et al. Reference Elzinga, Salzer, Willoughby and Gibbs2009), while visual estimation techniques may be subject to bias, because the observers need to mentally integrate the cover of individual species (Bonham Reference Bonham2013; Elzinga et al. Reference Elzinga, Salzer, Willoughby and Gibbs2009; Floyd and Anderson Reference Floyd and Anderson1987; Godínez-Alvarez et al. Reference Godínez-Alvarez, Herrick, Mattocks, Toledo and Van Zee2009; Hanley Reference Hanley1978; Korb et al. Reference Korb, Covington and Fulé2003; Sykes et al. Reference Sykes, Horrill and Mountford1983). On the other hand, our results indicated no differences between quadrats and LPI canopy, partially supporting the recommendation of other authors (Korb et al. Reference Korb, Covington and Fulé2003; Leis Reference Leis2003; Prosser et al. Reference Prosser, Skinner and Sedivec2003; Stohlgren et al. Reference Stohlgren, Falkner and Schell1995).
Table 2. Mean (±SE) and sample size relative to number of species, Shannon index, and graminoid percent cover in the four study sites.
a LPI, line-point intercept.
Our outcomes suggested that quadrats should be the recommended method when trained observers are employed. In our study, highly trained personnel conducted the vegetation cover assessment, which can explain the consistency between point- and visual-based methods. This homogeneity in results can be evinced by the correlation coefficients for introduced annual and perennial native graminoid cover measured with method versus LPI canopy—0.8 and 0.68, respectively (Figure 5)—suggesting consistency in visual cover assessment. The level of training of the surveying crew can affect the quality of measurements using visual methods. Anderson and Kothmann (Reference Anderson and Kothmann1982) recommended that observers practice reading cover for species before sampling to increase consistency and precision. We think that the LPI method remains the most successful and least biased method to monitor changes in vegetation cover, although similar results can be achieved with visual methods when properly trained personnel are employed.
Figure 5. Scatter plots and Pearson’s correlation coefficient between the three methods relative to introduced annual and native perennial graminoid percent cover. LPI, line-point intercept.
Introduced Annual and Native Perennial Graminoid Cover Biomass
Among invasive annual graminoids, B. tectorum was the most abundant species in terms of cover (Table 3), and we included a list of native perennial graminoids in Supplementary Table S1. For quadrats, there were significant direct linear relationships between B. tectorum cover (%) and its aboveground biomass and between native perennial graminoid cover (%) and aboveground biomass (Figure 6; Table 4). The relationship was similarly positive for LPI methods, but with lower R2 values. Slopes were different: greater for LPI basal with an exception for the native perennial graminoid model. In all models, LPI basal overestimated aboveground biomass. When simple regression models were fit separately according to the sites, the coefficient of determination increased in some cases (Figure 7), and the analysis of covariance (Table 5) revealed a cover by method by site interaction only for the native perennial graminoid cover model. Overall, quadrats delivered a better estimate of aboveground grass biomass in terms of coefficient of determination, probably because biomass clippings were obtained directly from the quadrats. Cover has been regarded as a fairly effective, fast, inexpensive, and nondestructive method to estimate aboveground biomass (Abella Reference Abella2020; Axmanová et al. Reference Axmanová, Tichý, Fajmonová, Hájková, Hettenbergerová, Li, Merunková, Nejezchlebová, Otýpková, Vymazalová and Zelený2012; Casady et al. Reference Casady, Van Leeuwen and Reed2013; Chieppa et al. Reference Chieppa, Power, Tissue and Nielsen2020; Flombaum and Sala Reference Flombaum and Sala2007; Goslee Reference Goslee2020; Humphrey Reference Humphrey1985). Site characteristics, such as precipitation, elevation, soil, and topography, may affect aboveground biomass productivity and were recommended to be taken into account when developing biomass-predicting models (Abella Reference Abella2020; Abella et al. Reference Abella, Embrey, Schmid and Prengaman2012; Beatley Reference Beatley1966). In our study only the native perennial graminoid cover model presented a cover by method by site interaction indicating different regression slopes, while similar slopes were obtained for B. tectorum models, suggesting an inconsistent effect of site characteristics. However, exploring the effect of site characteristics may help to generate and calibrate site-specific models, considering site productivity.
Table 3. Introduced annual graminoid mean percent cover by species indicated by site and survey method.
a LPI, line-point intercept.
Figure 6. Scatter plots of (A) Bromus tectorum biomass against its percentage cover and (B) native perennial graminoid biomass against its percentage cover. Simple regression models were fit for each vegetation survey method. LPI, line-point intercept. LPI, line-point intercept.
Table 4. Simple regression parameters for quadrat and line-point intercept (LPI) canopy and basal survey methods.
Figure 7. Scatter plots of (A) Bromus tectorum biomass against its percentage cover and (B) perennial graminoid biomass against its percent cover. Simple regression models were fit for each vegetation survey method within each site. LPI, line-point intercept.
Table 5. Analysis of covariance table relative to simple regression models fit by sites. Covariates: Bromus tectorum, introduced annual, and native perennial graminoid percent covers. Grouping variables: vegetation survey method and site.
Finally, another potential issue is the use of fixed number cover classes and the minimum number of cover classes necessary for reliable biomass estimation (Abella Reference Abella2020). We agree that the use of finer-resolution cover classes may improve the reliability of biomass estimation. We recommend the use of visual methods and LPI canopy methods as estimators when the assessment of aboveground biomass is involved.
Generally, visual cover estimation and LPI canopy methods yielded consistent and comparable data regarding native and nonnative graminoid cover and biomass prediction. Both methods are suited to evaluate the desirable vegetation recovery potential in response to restoration treatments, and their selection should be based on management’s objectives. We recommend visual cover estimation methods when the priority is centered on biodiversity monitoring, while we suggest LPI canopy to detect changes in vegetation cover because of its objectivity and lack of bias.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/inp.2021.36
Acknowledgments
We thank the research support team from the University of Wyoming Sheridan Research and Extension Center for their work collecting field data, including Clay Wood, Peter Adams, Sara Eller, Jordan Skovgard, Jaycie Arndt, Tyler Jones, Julia Workman, Will Rose, B. J. Bender, Tevyn Baldwin, Nolan Hicks, James Burford, Justin White, and Caleb Gray, and the landowners and managers who cooperated with us for this work. Additional thanks goes to Reese Irvine, Katie Cheesebrough, Hadley Pape, Bryan Christiansen, Paige Hellbaum, and Ann Danielson for field research assistance. This research was supported by the Wyoming Game and Fish Department, Bighorn Basin and Northeast Wyoming Sage Grouse Local Working Groups, Carbon and Sublette County, WY Weed and Pest Districts, USFWS Grant F20AC11493-00, and USDA-NIFA intramural research program Hatch accession 1013280. No conflicts of interest have been declared.
Open access
