College of Agricultural Sciences and Natural Resources, and Fire. Ecology Center Tech. Pap. 12, Texas .... because of the technical knowledge required for pro-.
Comparison of Four Nondestructive Techniques for Estimating Standing Crop in ShortgrassPlains Amy C. Ganguli, Lance T. Vermeire, Rob B. Mitchell,* ABSTRACT
and Mark C. Wallace
placed on clipping, using it only for calibration and validation within trials. The cano I (CA) ( LI COR L. oln NE)
Nondestructive standingcropestimators are Important d for efficient I t d
py monitonng
of
native
and
agronoDllc
systems.
This
stu
y
eva
ua
anayzer
-,
mc
,
e.
I
.
plot and pastureestimatesof standingcrop using LAI-2000,visual obstruction,canopyheight,and weightedplate measurements. Researchwasconductedin LubbockCounty,Texas,in 1999on areas doDlinatedby Amarillo fine sandyloam (fine-loamy,Dlixed,thermic Aridic Paleustalfs).Five hundredplot estimationsampleswerecollectedfor eachmethodalong25 transects, andeachtransectmeanwas usedfor the pastureestimationtrials. Coefficientsof deterDlination improvedaswemovedfrom plot (0.34,0.85,0.37, and0.70)to pasture (0.67,0.87,0.59, and0.8~)estimationfor LAI-2000,visualo~struction, canopyheight,and weightedplate ~easurements, respectively.The LAI-2000wasthe only purchasedmstrument($4800),whereasthe visualobstruction($6),canopyheight($14),andweightedplate($14) instrumentswereconstructedfrom readilyavailablematerials.Each instrumentprovidedfast measurements, especiallywhenconsidering the time requiredto hand clip the respectivemeasurement areas. Pastureestimationroot meansquareerrors (RMSE) indicatedthe weightedplate andvisualobstructionwerethe mostaccuratemodels (445and446kgha-1)followedby LAI-2000andcanopyheightmodels (613and691kg ha-1).Visualobstructionandweightedplate instrumentsboth providedfast,inexpensive measurements with acceptable accuracy.We recommend.visualobstr.u~ion~or.estimati~gstanding crop (SC)in shortgrass pl3lnsbecause It ISrapid,mexpensIve, andaccurate.
IS a fast, nonde~tructIve Instrument that IndIrect y estImates leaf area Index (LAI; Well.es and Norman, 1991). Direct measurements of LAI USIngthe LAI-3000 ~~ea meter (LI-COR, Lincoln, NE) have shown a poSItIve linear relationship between leaf area and SC (Engel et al., 1987). Since the CA measures foliage area per unit of ground area, we hypothesized a positive relationship would exist between CA measurements and SC. The CA has been used to estimate SC of grasslands in Iowa (Harmoney et al., 1997), Nebraska (MillerG od t 'I 1999' V I k et al 1999) and Texas 0 m~n ea., , 0 es y ., , d. t (Gang~lI et al., 1999). The CA was not a good pre IC or of SC m Iowa (r2 = 0.32; Harmoney et al., 1997) or Nebraska (r2 = 0.42; Miller-Good.man et ~I., 1999) gra~slands. Volesky et al. (1999) explaIned 33 Yoof the vanation in SC with the CA in the Nebraska Sandhills, and 59% of the variation when sampling modifications were made. Ganguli et al. (1999) isolated the specific area that the CA was measuring and produced a coefficient of determination of 0.69. They found the CA was best correlated with a 0.05-m2 plot when using a 900 viewcap in a semiarid grassland.
...tallgrass M EASUR~MENTS o~ SC a~e e~sentIal for determI~mg stock.mg rates, InvestIgatIng ~erbage pr.oductIo~, and eval~at~ng management strategIes. A van~ty of dIrect and IndIrect methods have been ~s~d to estIm~te SC (Catchpole and Whee~er, 199~). TradItIonally, estImates from hand or mechanIcally clIpped quadrats have be.en used to esti~at~ SC for. pastures or management UnIts. Although clIppIng proVIdes accurate measurements!o;r the area measured (Catchpole and Wheeler, 1992), I~ IS a time intensive, laborious. tech~ique that may r~qwre numerous samples to obtaIn relIable pasture estImates (Brummer et al.., 1994). .surements Several technIques use double samplIng proc~dures as an alternative to c~ipping. ~est:.methods functI?n.by developing a regressIon relatIonshIp of SC to p~edICtIve variables such as pla~t height, leaf ~rea, vegetatIon density, age, cover, or vIsual obstructIon (Cochran, 1977). Double sampling .re9uires s?me ~estructive sampling to develop a predICtIve relatIonshIp. Howeve~, after a relationship has been developed, less emphasIs can be
Visual obstruction has been used to estimate SC in prairie (Robel et al., 1970; Vermeire and Gillen, 2001) and improved pastures (Harmoney et al., 1997). Robel et al. (1970) accounted for 95% of the variation in tallgrass prairie SC, whereas Harmoney et al. (1997) accounted for 63% of the variation in improved pasture SC. However, investigations of the performance of this method in different vegetation types are limited. Rapid measurements of canopy heig~t (C~) for SC prediction have been made with measunng stIcks (Harmoney et al., 1997), plastic disks (Sharrow, 1984), and plates (Whitney, 1974). Canopy height can be difficult to measure due to the subjectivity associated with meaand disagreement over which plants or .plant parts should be considered to form a mean CH estImate (Heady, 1957). Researchers have successfully used disks or plates to incorporate an area dimension to the measurement (Whitney, 1974; Sharrow, 1984), whi~h reduces bias associated with structural heterogeneIty. Forage density is the volume. of abovegro~nd fo~age when compressed and is a functIon of vegetatIon heIght, density, and compressibility (Bransby et al., 1977). Sev-
., " Departmentof Range,Wildlife, and Fishene,s ~anagement,Texas Tech
Univ.,
Lubbock,
TX
79409-2125.
Contnbution
T-9-907
of
the
eral instruments have been used to measure forage density for the prediction of SC, but the rising disk and P
late
meters
have
the
CA,
canopy
advantage
of
recording
the
rest-
Collegeof Agricultural Sciencesand Natural Resources, and Fire Ecology Center Tech. Pap. 12, Texas Tech Univ. Received 3 Feb. 2 = 2,000 .Ci '.. :a
~;
a
a
1,000 0 0.6
.U
115,000 3 000
.'
..g2,000
N=23 0.8
1.0
1.2
1.4
1.6
~;
1,000 0
1.8
N=25 7
10
Plant Canopy Analyzer (LAI) ~- 7,000 ~6000 CI> '
~ ~
~- 7000 '~6' -= ,000
C
:::::~:ยง:~~~::~. ~;::~~~~ :..
..~
;
16
19
22
y=328.39x-515.44 r2=0.83
25
.U
(d)
~~~:::~~~~~
5,000 4,000 3,000
.go
CI> =2000.~ .~ ,
"=
13
Visual Obstruction (cm)
y=144.76x-1179.8 ~=0.59
:.~ 4,000 5,000 U~ 3,000
.
.
.
~
1,000
0
N=25
rI1 20
24
28
32
36
40
44
Canopy Height (cm)
= 2000 , 'C. = 1 000 :" rI1 0
N=25 5
7
9
11
13
15
17
Weighted Plate (cm)
Fig.3. Relationships between instrument readings and standing crop for pasture estimation using the (a) plant canopy analyzer, (b) visual obstruction, (c) canopy height, and (d) weighted plate. Outer lines represent95% confidence bounds for individual points and were connected for presentation purposes. The open symbols in (a) tested as outliers based on Tietjen et al. (1973) and were excluded from the analysis.
requires a computer for data handling, and sampling may be limited by direct illumination. Manual shading can reduce illumination problems (Hicks and Lascano, 1995), but that would reduce the simplicity of the technique. Although each model generated for plot estimation was significant, it appears that only the VO and WP models provided accurate estimates of SC on small plots, due to the variability associated with CA and CH measurements. Most producers and researchers are concerned with the SC on larger areas. Each method we investigated was able to account for more of the variation in SC as we moved from plots to pastures. The most accurate pasture estimation methods (lowest RMSE values) were the WP and VO, which were only separated by 1 kg ha-l, followed by the CA and CH methods. Previous studies investigating the relationship between CA readings and SC have produced varying resuits. Recently, modifications in sampling procedures have been made to improve the relationship (Ganguli et al., 1999; Volesky et al., 1999). Despite improvements, the variability of instrument readings may limit its use. Further understanding of the sensor's measurements as well as how the environment and different observers influence measurements is essential to assess how this instrument should be properly used to estimate SC. The current study followed the VO measurement protocol used by Vermeire and Gillen (2001), and obtained
similar coefficients of determination (0.87 and 0.90, respectively). Robel et al. (1970) used a similar protocol and obtained a coefficient of determination of 0.95. In studies using VO but with different methodology, Harmoneyet al. (1997) accounted for 63% of the variation in SC on improved pastures, and Volesky et al. (1999) accounted for 36% of the variation in SC on a site dominated by mid and tallgrasses. Lower coefficients of determination reported by Harmoney et al. (1997) and Volesky et al. (1999) may be a function of the sampling protocol or communities sampled. Additionally, results from Robel et al. (1970), Vermeire and Gillen (2001), and our study indicate that VO is a good predictor of SC. Most investigations using CH as a predictor of SC have been done on introduced pastures. Researchers using instruments similar to ours (Whitney, 1974; Sharrow, 1984) reported good success (r2 = 0.72 to 0.94). In the heterogeneous community of the current study, CH only explained 59% of the variation in pasture SC. Canopy height may have limited use as a tool to predict SC in native rangeland situations because of the wide variation of plant heights typical in native pastures. Most trials with WP instruments have been conducted in agronomic situations. We used the instrument described by Rayburn and Rayburn (1998) and accounted for 70% of the variation in SC compared with the 52% they reported. We observed a linear relationship as opposed to the curvilinear relationships reported by Ray-
GANGUU ET AL.: ESllMAllNG STANDINGCROPIN SHORTGRASSPLAINS bum and Rayburn (1998). Despite differences in plant growth form and community structure, our model explained 83 % of the variation in pasture sc.
1215
Plant canopy analyzer and CH techniques produced relatively poor estimates of SC for plots and pastures on native shortgrass plains vegetation. The WP and VO I h h t h . h d . d bl ec mques a consl era y stro~ge~ re atlons IpS WIt SC on plots and pastures. Consldenng VO accounted for 87% of the variation in SC, the VO technique is recommended as the best method for nondestructively
Gourley, C.J.P.,a~d A.A. McGowan..l?91. Assessing differenCl?sin pasture.mass Wl~han automated nsm~ plat~ meter and a direct harvestmgtechnique. Aust. J. Exp. Agnc. Anlm. Husb. 31:337-339. Hannoney, K.R., K.J. Moore, J.R. George, E.C. Brummer, and J.R. Russell.1997.Detennination of pasture biomassusing four indirect methods. Agron. J. 89:665-672. Heady, H.F. 1957.The measurementand value of plant height in the study of herbaceousvegetation. Ecol. 38:313-320. Hicks, S.K., and.R.J. .Lascano.1995.Estimation of leaf area index for cottoncanoplesusmgtheLI-CORLAI-2000plantcanopyanalyzer. Agron. J. 87:458-464. Miller-Goodman, M.S., L.E. Moser, S.S. Waller, J.E. Brummer, and P.E. Reece. 1999. Canopy analysis as a technique to characterize defoliation intensity on Sandhills range. J. Range Manage. 52: 357-362. ..
estimating
Mo~tgomery,
SUMMARY
.
SC
in
the
shortgrass
plains.
..
This
study
indi-
...slon
~.C.,
analysIs.
and
John
~.A. Wiley
Peck. &
1982.
Sons,
Introduction
New
to
lmear
regres-
York.
cates .that VO a~equa~ely .estImates SC for dete.rmtmng stoCkIng rates, InvestIgatIng herbage productIon, determining fine fuel load before burning, and evaluating
National Oceanicand Atmospheric Administration. 1997.Climatological data annual summary,Texas. Vol. 102(13). National Climatic Data Center, Asheville, NC.
management
Neter,
strategies.
J..
Applied
REFERENCES
M.H. linear
Kutner,
C.J.
regression
Nachtscheim, analysis.
and Richard
W. D.
Wassennan.19%. Irwin,
Chicago,
IL.
Aiken, G.E., and D.I. Bransby. 1992. Observer variability for disk
Rayburn,.E.B:, and S.B. Raybw:n. 1998. A standardiz~d plate meter for estlmatmg pasture mass m on-farm research trials. Agron. J. 90:238-241.
meter measurements of forage mass. Agron. Baker, B.S., T.V. Eynden, and N. Bogress.1981.
Ro~l, R;J., J.N. Brigg~, A.D. Dayto.n, tlonshlps betwee~ VIsual obstruction
J. 84:603-605. Hay yield detennina-
and L.C. Hulbert. 1970: Relameasurements and weight of
tions of mixed swards using a disk meter. Agron. J. 73:67-69. Blackstock,D.A.1979.SoilsurveyofLubbockCounty, Texas.USDASCS,Washington, DC. Bransby, D.I., A.G. Matches, and G.F. Krause. 1977. Disk meter for rapid estimation of herbage yield in grazing trials. Agron. J. 69:393-3%. Brummer, J.E., J.T. Nichols, R.K. Engel, and K.M. Eskridge. 1994. Efficiency of different quadrat sizesand shapesfor sampling stand-
grassl~d vegetation. J. Range.Manage. 23:295-297. . SAS Institute. 1985.SAS/SAT guide for personalcomputers. Version 6 ed. SAS Inst., Cary, NC. Sharrow, S.H.1984. A simple disc meter for measurementof pasture height and forage bulk. J. Range Manage. 37:94-95. Tietjen, G.L., R.H. Moore, and R.J. Beckman. 1973. Testing for a single outlier in simple linear regression. Technometrics 15: 717-721.
ing crop. J. Range Manage. 47:84-89. Catchpole, W.R., and C.J. Wheeler. 1992. Estimating
Venneire, L.T., and R.L. Gillen. 2001. Measuring herbage standing crop in tallgrass prairie with the visual obstruction method. J. Range
plant biomass:
A review of techniques. Aust. J. Ecol. 17:121-131. Cochran, W.G. New York.
1977. Sampling
techniques.
John Wiley
Manage. (in press). & Sons,
Engel,R.K.,L.E. Moser,J. Stubbendieck,and S.R. Lowry. 1987.Yield accumulation, leaf area index and light interception of smooth bromegrass. Crop Sci. 27:316-321. Ganguli, A.C., R.B. Mitchell, M.C. Wallace,and L.T. Venneire.1999. C~ grasslandbiomassbe indirectly predicted through light attenuatlon? .Proc. 5th Int. Symp. on the Nutrition of Herbivores, San Antomo, TX. 11-16 Apr. 1999.
Volesky, J.D., W.H. Schacht, and P.E. Reece. 1999. Leaf area, visual obstruction, and standing crop relationships on Sandhills range-
land. J. Range Manage. 52:492-499. Welles, J.M., and J.M. Nonnan. 1991.Instrument for indirect measurement of canopy architecture. Agron. J. 83:818-825. Whitney, A.S. 1974. Measurement of foliage height and its relationships to yields of two tropical forage grasses.Agron. J. 66:334-336. Zar, J.H. 1974. Biostatistical analysis. Prentice-Hall, Englewood Cliffs, NJ.
oJ