Revisiting Robel’s Visual Obstruction Method for Estimation of Standing Crop in Grasslands

Randall D. Jackson, 1575 Linden Drive, Department of Agronomy, University of Wisconsin-Madison 53706; and Laura K. Paine, PO Box 8911, Wisconsin Department of Agriculture, Trade, and Consumer Protection, Madison 53708

Abstract

We compared readings from a Robel pole, a visual obstruction (VO) device for measuring vegetation structure, to standing crop estimates made with the harvested biomass (HB) technique in warm-season grass stands of various composition. The grass stands were single-species seedings rotationally grazed by bison. Re-colonization of cool-season forages that had previously dominated the site significantly reduced warm-season grass abundance at the time of the study. Visual obstruction readings provided good predictions of standing crop as estimated by HB from 0.1-m² quadrats, but predictions were better (r² = 0.76) when measurements were aggregated to the transect level, a spatial scale representative of a typical grazing paddock. The relationship between VO and HB compared at the individual sample level or when aggregated within experimental species plots was very weak (r² = 0.14 and 0.17, respectively). Our results, and a review of the literature, indicate that the Robel pole should perform well when used in relatively uniform vegetation, but the scale where vegetation is uniform is likely to be larger than the individual sampling units.

Introduction

Structural diversity of grassland vegetation makes the accurate estimation of standing crop, the aboveground biomass at a given time, an ongoing challenge for range managers, graziers, and researchers (5,6,14). Typically, aboveground biomass is harvested from small (< 1 m²) quadrats, which is then returned to a laboratory to be dried and weighed, but this approach is cumbersome, destructive, and time-consuming. These shortcomings may require large numbers of samples to provide reasonable uncertainty estimates (3,4).

Development of the visual obstruction (VO) method by Robel et al. (11) provided a non-destructive means of estimating standing crop on tallgrass prairie in Kansas. Moreover, this technique should integrate estimation over a greater area than the harvested biomass (HB) method at the sampling unit level, because the VO method is occurring across several meters whereas the HB method typically covers a much smaller area. Most importantly, compared to the harvest technique, more VO measurements can be made per unit time allowing for larger numbers of samples to be taken and improved estimation.

The VO method is a standard tool of wildlife biologists characterizing the structure of grassland habitat (8,10,12), but it has not been widely adopted for estimation of standing crop by grassland managers and scientists interested in plant productivity. In our research investigating ground-nesting bird use of cool-season and warm-season managed grasslands in Wisconsin, informal observations suggested a poor correlation between VO measurements and HB. Recent studies duplicating the Robel research report a wide range of r² values (0.41 to 0.93) for the relationship between VO and HB (Table 1). These disparate results beg the question: Under what circumstances does the Robel pole accurately estimate grassland standing crop?

For this study, we analyzed data by individual sampling location, species treatments, and transects (i.e., the grazing paddock scale) to explore the relationship between VO and HB, and evaluated our results in relation to species composition and vegetation management. Our results are interpreted in light of other studies to (i) assess the general utility of the Robel method for estimating standing crop in grasslands, (ii) determine what factors contribute to differing results obtained among the studies, and (iii) provide guidance as to under what circumstances this method is likely to be most effective.

Warm-season Grasses Sown into a Tilled, Loamy Sand Site in Southern Wisconsin

Our study was conducted in 2003 on single-species warm-season grass plots seeded in 1998 at the BisonRidge Ranch in central Wisconsin (43°45’11"N, 89°27’6"W). Soils were classified as Metea loamy fine sand (loamy, mixed, active, mesic Arenic Hapludalfs) on 2 to 6% slopes (9). Typical climate conditions for the area are 81 cm rainfall per year and 102 cm/year as snow, growing season (April-October) mean highs and lows of 21.3 and 8.8°C, respectively, and winter (November-March) highs and lows of 0 and -10°C, respectively.

The experimental design was a randomized complete block with three blocks. Treatments were 2.2-ha single-species stands of big bluestem [Andropogon gerardii (Vitman)], little bluestem (Schizachyrium scoparium [Michx.] Nash), sideoats grama (Bouteloua curtipendula [Michx.]Torr.), switchgrass (Panicum virgatum L.), and indiangrass (Sorghastrum nutans [L.] Nash), plus a mixture of all but switchgrass. These seeding treatments were all sown into tilled bare soil in 1998 (Fig. 1). By 2000, all warm-season grass stands maintained ~95% absolute cover of the sown species. Paddocks with an area of ~0.5 ha were grazed by bison consisting of 60 cows, 44 calves, 5 bulls, and 10 yearlings. Plots were grazed by bison for ~2 days (residual stubble height of 5 to 10 cm) two to three times each growing season starting in 2001 as part of a rotational grazing system that included other paddocks outside the study system. Grazing paddocks were laid out perpendicular to treatment strips so that at any one time, the herd grazed across all of the species-treatment strips. Aboveground biomass, estimated by the harvest method, averaged 420 g dry mass per m² in late July 2003, prior to the initial grazing period for that year.

Fig. 1. Randomized complete block design implemented at BisonRidge Ranch, Marquette Co., WI. Six warm-season grass seeding treatments were randomly applied within each of three blocks. Transects were located perpendicular to treatments strips.

Four transects were located perpendicular to the treatment strips across each of the three replications for a total of 12 transects (Fig. 1). Along each transect, observations were made at one location in each treatment strip for a total of six locations per transect, 24 locations per replication, and 72 locations for the entire study. At each location, four VO measurements were made and one clipped biomass sample was collected, for a total of 288 individual VO observations and 72 HB samples (see Fig. 2 for description of Robel pole). In subsequent analyses we used the mean of the four VO readings (n = 72).

Fig. 2. Photograph of Robel pole in use. Visual obstruction readings were made using a 3-m length of 3.5-cm diameter graduated PVC pipe marked at 0.5-dm increments. An observer stood 4 m from the graduated pole and observed the pole at eye-level at 1 m. The lowest visible numbered marking was recorded. At each location, 4 measurements were recorded with the observer remaining in position and the graduated pole being moved to each of the 4 cardinal directions. The clipped biomass sample was taken from in front of 1 of the graduated pole locations using a 20- × 50-cm quadrat. Biomass samples were dried and weighed to estimate standing crop.

Ordinary least squares regression was used to assess the ability of the mean of the four VO readings to predict HB. First, we paired data taken at the subsample level, i.e., one average VO reading vs one clipped plot taken from within that area. We called these the subsample data. Second, we aggregated these data to the transect level by averaging across the six species plots to generate four pairings of VO and HB estimates per experimental block. These comprised our transect data. Finally, we aggregated samples within each combination of species and block creating the species data. At fifty 1-dm intervals along each transect, we recorded the species intercepting a sharpened point lowered from above the vegetation (7) in late July 2003. Absolute cover by species was calculated for each transect from these data.

VO-HB Relationship Strongest when Data were Aggregated to the Transect Level

Between 2001 (when the seeded species comprised > 90% of the stands) and 2003 (when the study was conducted), cover of the sown grasses had declined to ~30% while species diversity increased to an average of 9 species per 50-m transect. In these relatively diverse stands, regressing VO readings on HB at the subsample and species levels revealed very weak relationships between VO and HB (r² = 0.14 and 0.17, respectively) (Figs. 3A and B). Alternatively, pooling data across species treatments and within transects significantly improved the predictive power of the regression equation, resulting in an r² of 0.76 (Fig. 3C).

A review of other Robel pole studies reinforced our observation that the scale at which VO and HB measurements are paired for analysis may have the greatest effect on the relationship between these two variables. Three of the studies we reviewed (Table 1) used the transect as the experimental unit to generate a regression equation (1,2,13). For those studies, the average r² was 0.82. Two studies used the individual sample location as the experimental unit to develop the regression equation (6,15). For those studies, the average r² was 0.52. These differences suggest that there is not always a strong correlation between VO and clipped sample at the same location, but that the average of multiple VO observations over a larger area does correlate with an average of clipped biomass measurements in the same area.

Fig. 3. Relationship between harvested biomass and Robel pole readings for (A) subsample data, (B) species data, and (C) transect data.

Fig. 4. Relationship between harvested biomass and Robel pole readings using subsample data for each species treatment level.

Uniformity of Vegetation Structure Improved Relationship Between VO and HB

We expected that pooling within species treatments would result in a higher r², as a result of greater vegetation homogeneity within the same species strip.However, aggregation to the species level (i.e., averaging transects within sown species treatments) did little to improve the relationship (r² = 0.17) (Fig. 3B). Only one of the species-treatments, sideoats grama, showed a strong relationship between VO and HB (Fig. 4). We searched for patterns in species composition and diversity metrics within each species treatment level. The main differences between sideoats grama plots and the other species were that they had the lowest median VO readings (2.89), while all other species treatments had median VO readings greater than 3.0; and this treatment supported the greatest cover of annual grasses, 13.2%, while all other treatments maintained < 10% annual grass cover.

These results point to the importance of the re-invasion of these warm-season grass stands by cool-season forage species, which appeared to be greater on some transects than others. We can infer that this result stems from greater homogeneity associated with lower overall biomass and higher annual grass cover. Specifically, the annual grass Bromus tectorum L. had filled in much of the interstitial space around the sideoats grama clumps. Annual grasses such as B. tectorum create relatively homogeneous, single-culm stands where the biomass does not carryover from year to year.

Other published work reinforces this observation that overall vegetation height may influence one’s ability to generate a strong relationship between VO and HB. Vermeire and Gillen (14), working on grazed tallgrass prairie range in Oklahoma generated a regression r² of 0.64 at the subsample level, whereas Ganguli et al. (5), working on ungrazed short and mid-grass range in Texas obtained transect-wise and subsample level regression models with r² = 0.87 and 0.85, respectively. It is possible that the greater potential height of tallgrass prairie in combination with grazing results in a more heterogeneous vegetation structure, which is likely to introduce more variability among sampling locations for both VO and SC, making the establishment of a strong regression relationship more difficult.

Data supporting the notion that vegetation species composition affects the relationship between VO and HB is lacking, however. None of the published studies of the Robel method that we reviewed provided detailed information on species composition and its relationship to vegetation structural heterogeneity. Robel et al. (11) described vegetation in their study simply as ‘homogenous.’ Ackerman et al. (1) listed only one species, Old World Bluestem (Botriochloa ischaemum L. Keng), in their site description and did not report the vegetation composition of these apparently seeded pastures. Most of the other studies described vegetation communities as being dominated by two to four grass species and listed five to ten other grasses and forbs present, but neither proportions of those species nor any other measure of species richness was reported for most of the studies. Vermeire et al. (13), who reported high r² values of 0.90 and 0.93, described the vegetation community at their study site as 80% blue grama [Bouteloua gracilis (K.B.K.) Lag. Ex Steud] and buffalograss [Buchloe dactyloides (Nutt.) Engelm.], reinforcing our observation that there is probably a relationship between high r² and relatively high structural homogeneity.

In past work we found that vegetation heterogeneity, driven by management of disturbance regimes, influenced the strength of the relationship between VO and HB (unpublished data, L. Paine). On cool-season pastures in Wisconsin, we regressed VO versus clipped samples by individual sampling location in bi-weekly measurements on 48 transects over two years. Regression analysis was conducted using individual sampling location (analogous to our subsample data in this study) as the experimental unit. We found that mowed paddocks had higher r² values than rotationally grazed paddocks (0.61 vs 0.57) and continuously grazed pastures had the lowest r² (0.46; L. Paine, unpublished data). In Wisconsin, rotational grazing of dairy cattle involves high stocking densities of (24 animals per acre) for periods of 12 to 24 h on a single paddock. This type of management results in a relatively uniform stand compared to continuously grazed pasture where cattle tend to selectively graze certain areas over the course of a growing season.

In the current study, grazing was applied sequentially in blocks across all treatment strips and may have influenced spatial variability in species composition, diversity, and biomass. Transects were located perpendicular to treatment plots so that each transect crossed six different species strips, coinciding with grazing paddocks. When data were grouped by species, i.e. four transects per species treatment, we expected, but did not find that precision was improved as a result of greater uniformity within a single-species stand. Species composition, richness, and biomass were more similar within transects than within species stands, suggesting that management across treatment strips may have had a greater influence on the results.

Summary and Conclusions

This review of Robel methodology suggests that it may be an effective method of estimating standing biomass of pastures and range under certain conditions, but we offer the following cautionary notes and generalizations:

1. Placing a 20- × 50-cm quadrat (0.1 m²) directly in front of the graduated Robel pole, rather than placing the pole in the center of a 0.25-m² quadrat, should yield higher coefficients of determination (r²).

2. Aggregating many samples to scales greater than the quadrat or the area involved in the Robel measurements offers the greatest chance for high r² values.

3. More subsamples improve r² values, but adequate sample size per unit area depends on vegetation variability.

4. Management practices and species composition that tend to homogenize vegetation structure such as burning or mowing, improve the likelihood of generating a strong relationship between VO and HB.

Acknowledgments

Thanks to Georgia Derrick and Jim Atten for providing land, livestock, and labor for this study. Thanks to Jon Bleier, Jesse Rucker, Liz Froelich, Julie Doll, and John Albright for sampling assistance and data processing. This work was funded by a UW-Madison College of Agricultural and Life Sciences Hatch Grant to R. D. Jackson and a North Central Region Sustainable Agriculture Research & Education grant to L. K. Paine.

Literature Cited

1. Ackerman, C. J., Purvis, H. T., Horn, G. W., Payton, M. E., Vermeire, L. T., Paisley, S. I., and Bodine, T. N. 1999. Estimation of forage mass of Old World bluestem using a visual obstruction measurement technique. Anim. Sci. Res. Rep. P-973:194-199.

2. Benkobi, L., Uresk, D. W., Schenbeck, G., and King, R. M. 2000. Protocol for monitoring standing crop in grasslands using visual obstruction. J. Range. Manage. 53:627-633.

3. Briggs, J. M., and Knapp, A. K. 1991. Estimating aboveground biomass in tallgrass prairie with the harvest method: Determining proper sample size using jackknifing and Monte Carlo simulations. Southwest. Nat. 36:1-6.

4. Brummer, J. E., Nichols, J. T., Engel, R. K., and Eskridge, K. M. 1994. Efficiency of different quadrat sizes and shapes for sampling standing crop. J. Range. Manage. 47:84-89.

5. Ganguli, A. C., Vermeire, L. T., Mitchell, R. B., and Wallace, M. C. 2000. Comparison of four nondestructive techniques for estimating standing crop in shortgrass plains. Agron. J. 92:1211-1215.

6. Harmoney, K. R., Moore, K. J., George, J. R., Brummer, E. C., and Russell, J. R. 1997. Determination of pasture biomass using four indirect methods. Agron. J. 89:665-672.

7. Heady, H. F., Gibbens, R. P., and Powell, R. W. 1959. A comparison of the charting, line intercept, and line point methods of sampling shrub types of vegetation. J. Range. Manage. 12:180-188.

8. Kirsch, L. M., Duebbert, H. F., Kruse, A. D. 1978. Grazing and haying effects on habitats of upland nest birds. Trans. N. Amer. Wild. Nat. Res. Conf. 43:486-497.

9. NRCS. 2006. Official soil series descriptions. Natural Resources Conserv. Serv. (NRCS), USDA. Washington, DC.

10. Patterson, M. P., and Best, L. B. 1996. Bird abundance and nesting success in Iowa CRP fields: The importance of vegetation structure and composition. Amer. Midland. Natur. 135:153-167.

11. Robel, R. I., Briggs, J. N., Dayton, A. D., and Hulbert, L. C. 1970. Relationships between visual obstruction measurements and weight of grassland vegetation. J. Range. Manage. 23:295-&.

12. Roth, A. M., Sample, D. W., Ribic, C. A., Paine, L., Undersander, D. J., and Bartelt, G. A. 2005. Grassland bird response to harvesting switchgrass as a biomass energy crop. Biomass Bioenerg. 28:490-498.

13. Vermeire, L. T., Ganguli, A. C., and Gillen, R. L. 2002. A robust model for estimating standing crop across vegetation types. J. Range. Manage. 55:494-497.

14. Vermeire, L. T., and Gillen, R. L. 2001. Estimating herbage standing crop with visual obstruction in tallgrass prairie. J. Range. Manage. 54:57-60.

15. Volesky, J. D., Schacht, W. H., and Reece, P. E. 1999. Leaf area, visual obstruction, and standing crop relationships on Sandhills rangeland. J. Range. Manage. 52:494-499.