Table Of Content

Historic, Archive Document Do not assume content reflects current scientific knowledge, policies, or practices. A 3z 0 7G ~~\ United States A l) Department of Microplot Method for Updating Agriculture Loop Frequency Range Trend Data: Forest Service Theoretical Considerations Rocky Mountain Forest and Range and a Computer Simulation Experiment Station Fort Collins, Ward W. Brady, John W. Cook, and Earl F. Aldon Colorado 80526 CO go r Research Paper RM-295 J-* USDA Forest Service April 1991 Research Paper RM-295 A Microplot Method for Updating Loop Frequency Range Trend Data: Theoretical Considerations and a Computer Simulation Ward W. Brady, Professor Arizona State University John W. Cook, Research Associate Arizona State University Earl F. Aldon, Research Forester Rocky Mountain Forest and Range Experiment Station1 Abstract Quantitative forms ofthe relationships between 3/4-inch loop frequency data (Parker 3-step) and data from the Community Structure Analysis (CSA) sampling method were determined from computer simulations of grassland vegetation populations. Percent basal area (measuredusingtheCSAmicroplot) hadastrongerandsimplerrela- tionship with Parker frequency than did CSA canopy cover, CSA density, orCSAfrequency. Datafromthree ormore 100-foottransects fromhomogeneous locations hadsignificantlyhigherprecisionthan datacollectedfromsingletransects. Atableforconverting Parkerdata to percent basal area was developed for plants with mean basal diameters between 0.3 and 33.6 cm. Conversion of old Parker data to basal area should provide a basis for ecological interpretation of long-term changes in range vegetation. USDA National Agricultural Library NAL Building 10301 Baltimore Blvd. Belt&wilie. MO 20705-2351 Headquarters is in Fort Collins, in cooperation with Colorado State University. A Microplot Method for Updating Loop Frequency Range Trend Data: Theoretical Considerations and a Computer Simulation Ward W. Brady, John W. Cook, and Earl F. Aldon Management Implications vegetation changes that have occurred during the last several decades. Although Parker 3-step data have been collected on westernrangelands sincethe early 1950's, muchofthese data remained underutilized because of difficulties in Background interpretation. The existence ofa relationship between 3/4-inch loop frequency and basal area allows histori- Basal frequency data (Daubenmire 1968), the probabil- cal Parker 3-step data to be interpreted in terms ofthis ity ofencountering the basal portion ofa plant species more ecologically meaningful parameter. Interpreting in a plot frame, are easy to collect and are relatively old Parker frequency records in terms ofbasal area as- free of observer error. The Parker 3-step sampling sumes that mean plantbasal diameter canbe estimated method is characterizedbyanumberofprocedurallyad- and that plant populations are neither completely ran- vantageous attributes. A combination ofsmall plot size domly norvery highly contagiously distributed. Evalu- (a 3/4-inch loop), clustered 100-foot transects on ation ofsubjective information from old Parker records homogeneous sites, sampling times correlated withthe is necessary to determine if these assumptions can be phenological stage ofvegetation, and simple dataforms met. minimize observer and transcription error (Parker and Tables forconverting Parkerdatatopercentbasalarea Harris 1959, Sharp 1954). Johnston (1957) reported the were developed forplants withmeanbasal diametersbe- Parker method to be more rapid than either line inter- tween 0.3 and 33.6 cm. While future sampling withthe cept or vertical point frame methods. Parker 3-step method is not recommended, conversion While frequency data are easyto collect, frequency is ofold Parker data to basal area provides abasis for eco- also dependent on plot size. Density of a species is logical interpretation of long-term changes in range directlyrelatedtofrequency measurementonlywhenthe vegetation. populationexhibits arandom distribution (Greig-Smith 1983). Few species, however, are randomly distributed, and because of dependence on plot size, frequency of Introduction a species is as much a function of sampling design as vegetationcharacteristics. This "artificial" nature offre- Parker 3-step clusters havebeenestablished onnation- quency data often makes it difficult to interpret results al forests throughout the western United States (Parker in relationto rangeland condition and trend (Aberdeen 1951, ParkerandHarris 1959, ReppertandFrancis 1973). 1958, Greig-Smith 1983, West 1985). For many locations these clusters provide the longest Johnston (1957), Parker and Harris (1959), Kinsinger historical record ofapparent vegetation change. Unfor- etal. (1960), and Francis etal. (1972) exploredwhether tunately, data collected fromthese clusters, often dating relationships existed between 3/4-inch loop frequency to the 1950's, havebeen underutilizedbecause ofdiffi- and more ecologically dependent sampling methods. culty inobtaining anecological interpretationofParker Unfortunately, small sample sizes and observationerrors "hits." Other sampling designs, such as the Commu- hindered empirical analysis (Smith 1962). Additional- nity StructureAnalysis (GSA) method (Morris 1973, Pase ly, a method was often defined as "best" because it 1981), provide repeatable, interpretable, and statistically provided the least variability among observations (pre- reliable data on better defined vegetation parameters cision) without knowing which method best approxi- (particularly coverand density) which are more useful mated the true population mean (accuracy). Other , for ecological interpretation and in management attempts at determining accuracy ofsampling and rela- applications. tionships between 3/4-inch loop data and othervegeta- Ifarelationship couldbe establishedbetween datacol- tion parameters were limited to theoretical analyses, lected using the Parker procedure and that collected which are themselves in need of empirical verification usingtheCSAora similarprocedure, thenapotentially (Hutchings and Holmgren 1959). powerful tool would exist that would aid in interpreta- Parker(1950) compared 3/4-inchloopfrequencydata, tionoftheabundance ofParkerdata. Byinputting Parker pointframe sampling, pacedtransect data, and percent "hits" andpredictingvaluesforbetterunderstoodvege- basalareaasmeasuredbythe line interceptmethod. He tationparameters (e.g., cover,basalarea, ordensity), the concludedthat3/4-inchloopfrequencydataoftenshared land manager would have a better tool for interpreting a close relationship with the other data types. 1 Hutchings and Holmgren (1959), using a theoretical flected the inability ofParker plots to detect very small model, found that mean plantbasal diameterand plant plants (0.34 cm). He consideredthe numberoftransects numbers both influenced the relationship between requiredfora given level ofprecisiontobe highand de- Parkerdataandpercentbasal area. They suggested that pendent upon plant size. plant shape and plant distribution may also have an Brady (1988) analyzed ParkerandCSA data collected effect. Specieswithbasal diameters greaterthanaParker onthe Bernalillo WatershedResearchNaturalAreanear plot showed a linear relationship between Parker fre- Albuquerque, New Mexico. The Bernalillo Watershed quencyandpercentbasal area. As meanplantdiameters hasbeenprotected from grazing since 1955. Parkerdata increased above 2.5 inches (6.35 cm), Parkerfrequency were available for the years 1955, 1961, 1965, 1970, readings more closely approximated the calculated per- 1975, 1980, and 1982 through 1987. CSA data were cent basal areas. Species with mean basal diameters available onlyforthe last sevenofthose years: 1980 and smaller than that of a Parker plot showed increasingly 1982-1987. On each Parkertransect one hundred 5- by curvilinearrelationships between Parkerfrequency and 10-cm CSA microplots were exactly centered over the calculated percent basal areas. The authors found no one hundred 3/4-inch diameterParkerloops. Sampling consistent estimates ofpercentbasal areaby Parker fre- was conducted on permanently located transects, and quencydata inanempirical studythatcomparedParker Parker loops were centered 1 foot apart. CSA canopy sampling with percentbasal area measured from popu- cover data were collected using the microplots, and lations of paper disks. Changes in plant size, numbers Parkerfrequencywas collectedusingthe 3/4-inchloops. of plants per unit area, and percent basal area were Ten 0.5-m2 circularplots were used forCSA density and derived from a series of pantographic maps that were CSA frequency determinations; thesewere centered over sampled with Parker loop transects scaled to size. every tenth Parkerplot. Errors in species identification Results revealed misleading information from Parker occurred over the 7 years of data collection and intro- data where interactions occurredbetweenplantsize and duced substantialvariationintothedataset, particularly plant numbers. for species with similar physiognomy, such as the Francis etal. (1972) compared 3/4-inch loop frequency Bouteloua genus. In addition, small sample sizes against percent basal area as measured from line inter- (10<n<48) made interpretationofdatadifficult. Several cept and 3-point vertical frame sampling methods, species (including most nongrass species) did not oc- percentcanopycoverasmeasuredfrom4-by 8-inchplot cur frequently enough to allow relationships between estimates, production estimates, and density from a 1- data types to be confidently estimated. by 20-footplot. They reported no consistentcorrelations Parker "hits" from a transect (the independent vari- between frequency and any of the other sampling able) were plotted against percent CSA canopy cover, methods and recommended against using Parker CSA density, and CSA frequency (the dependent vari- frequency data as an index to plant community ables) on the same transect for each species. Simple characteristics. linearregressionwas usedto determine relationshipsbe- Kinsinger et al. (1960) compared Parker frequency, tween data types. Results indicated that for several line intercept data, and datafromavariable plotmethod species Parkerdatawere significantlyrelatedwithCSA against actual measurements of canopy cover from 20- canopy cover and CSA density. Strength of these rela- by 25- or 25-by 50-foot plots. They reported no signifi- tionships varied over a wide range, at least in part, be- cant differences among methods when calculated can- cause of small sample sizes and species identification opy covers were less than 5%, but differences increased errors. However, available datadidsuggestthatrelation- significantlyascanopycoverincreased, withParkerfre- ships couldbe described that would allow managers to quency increasingly overestimating "true" cover more equate data from Parker 3-step clusters to more eco- than either of the other two methods. logically sound and statistically reliable sampling Johnston (1957) compared Parkerfrequency and per- techniques. cent basal area estimates from line intercept and vertical point quadrat methods. Transects were subdivided such that each data point represented the mean ofonly Methods a 10-foot segment. He reported some of the highest coefficients of variation for Parker data when measur- Computer simulations were used to determine rela- ing "dominant" species (percentbasal area greaterthan tionships between data collected using 3/4-inch 1%). WhiletheParkerprotocolwasoneofthe mostrapid (1.905-cm) diameter loops and both 5- by 10-cm sampling methods once the transect tape was in place, microplots and 0.5-m2 plots. Use of simulations elimi- it detected the least number of species, overestimated nated observationerrorand permitted samplesizes large percent basal area the most, and required the greatest enoughto identifyquantitativerelationships. Computer number of 10-foot transect clusters to meet the desired simulations provided artificial vegetation maps with precision. prescribed characteristics on the computer screen (fig. Smith (1962) compared Parker frequency datawith 1). Only one species was drawn on the screen for any calculatedbasal areas ofmeasuredplants (using the for- given simulation. Vegetation parameters that could be mulaforthe areaofacircle). He reportedbothplantsize modified included: mean plant diameter, variation in and plant density affected the relationship between plant diameter, density orpercentcoverofplants, plant Parkerfrequency and percentbasal area. His results re- distribution (randomto highlycontagious), plant shape, 2 and plant growth form (i.e., the ratio between canopy lected from 0.5-m2 circular plots centered over every size and basal area). Shrubby species were not included tenth Parker loop. Each transect, therefore, had one in the simulation. hundred 3/4-inch loops for Parker frequency, one Samplingunitsofappropriatesizeandshape (asspeci- hundred 5-by 10-cm microplots forbasal area and CSA fied by the Parker and CSA protocols) were used to canopycover, andten 0.5-m2 circularplots forCSA den- samplethese artificialvegetationpopulations. Thesam- sity and CSA frequency. plingunits couldbe visually superimposed onthesimu- For each simulation, basal area was systematically lated population for illustration purposes (fig. 1). varied between a specified upper and lower bound by All maps were drawn using graphics with a screen a specified interval. Each incrementbetweenthe upper resolution of 640 by 480 pixels. Under this graphics and lower bounds was considered a cycle of the simu- mode, each pixel had equal height and width. For any lation. Forexample, ifthe maximumpercentbasal area given simulation, each pixel onthe screen could be as- fora simulationwas setto 5.0, the minimum at 0.1 and signed an equivalent area, and both individual plants theintervalat0.1, thefirstvegetationmaps drawnwould and sampling plotscouldbe "drawn" onthescreenwith have approximately 5% basal area and subsequentmaps dimensionsthatwere scaledto those found inthefield. would have the basal area reduced by 0.1% each cycle Several different scales were used in simulation. For ofthe simulationuntilbasal areawas reduced to 0.1%. example, in one simulation model each pixel on the Whenthe minimumbasal area was reached, thenbasal computerscreenwas defined tobe equivalentto anarea area was reset to the maximum and the procedure re- of0.077 cm2 A circle drawn on the screen with aradius peated until the required number of replications had . of 3 pixels contained 37 pixels and had an equivalent been collected. Recorded vegetation parameters in- area of 2.850-cm2 or the area of a 3/4-inch-diameter cluded Parker frequency, CSA canopy cover, CSA den- Parkerplot. ACSAmicroplot (5 by 10 cm) was assigned sity, CSAfrequency, andbasal covermeasuredwiththe dimensions of18by 36pixels plus 1 pixel oranequiva- CSA microplot. CSA cover, basal cover, and CSA den- lent area of 49.995 cm2 The 0.5-m2 plot used in CSA sity measurements were evaluated foraccuracybycom- . densityandfrequency samples containedthe numberof paringsampleresults againstknownscreenparameters. pixels required for the area according to scale. Each simulationwas continued until approximately 300 Sampling plots were located on the screen with in- data points (each having either 100, 300, or 600 plots) terplot distances equivalentto that ofafield sample. For hadbeencollected. Forasimulated sample ofthis size, example, a Parkerloop had a diameterof 7 pixels along a 16-Mhz 80386 based microcomputer required 6 to 10 the horizontal plane ofthe screen. Placing the centerof hours of run-time. Parkerloops 112 pixelsapartgavespacingsofanequiva- Theeffectofvariousplantcharacteristics onrelation- lent distance of 1 foot. Because ofthe spacing between shipsbetweenParkerdataandotherCSAparameterswas sampling units on the screen, a sample of the equiva- determined byholding all vegetation characteristics ex- lent of a 100-foot transect required that the samples be cept one constant and then systematically varying the spread overseveral independently generatedvegetation characteristic of interest. The effect of plant size, for maps (screens) withidentical parameters. Foreachsimu- example, was determinedbyholdingplant distribution lation a sample size of either 100 plots (1 Parker tran- and other parameters constant while systematically sect), 300 plots (3 Parkertransects), or600plots (6 Parker changing plant size. transects) was chosen. Microplots centered over each Prediction equations relating Parker and other data Parkerloop were usedto collectcanopy coverandbasal were determined using least squares regression pro- area data. Density and CSA frequency data were col- cedures (Neteretal. 1985, Statgraphics 1989). Aprelimi- nary evaluation of these prediction equations was conductedusing standard sampling equipmentand artificial populations composed of paper disks. To evalu- ate adequacy of predicted basal area for management purposes, the procedure of Freese (1960) was used. Results and Discussion CSA Density and Parker Frequency Relationships The relationshipbetweenCSAdensityandParkerfre- quency was dependent on mean plant basal diameter (fig. 2). As mean basal diameter increased, the number of Parker hits observed at any given density also in- Figure1.—Photographofcomputersimulationofavegetationcom- creased. Forexample, atadensityof20 plantspersquare munityandtherelativesize,shape,andspacingofsamplingplots. meter approximately 4 Parker hits would be observed Rectangles located across the center of the photograph are 5- forplantshaving abasal diameterof3-cm, approximate- bceynt1e0r-cofmemaicchrorpelcottasn,gl3e/,4-ainndchthPearlkaergrelcoiorpclseanreearlotchaetecdenitnerthoef ly 8 Parker hits for 6-cm plants and approximately 24 the photograph is a 0.5-m2 density plot. hits for 12-cm plants. 3 Figure2.—EffectofbasaldiameterontherelationshipbetweenCSA Figure3.—Effectofcanopy/basaldiameterratioontherelationship densityand Parkerfrequencyforplantswith basal diametersof betweenCSAcanopycoverandParkerfrequencyforplantswith 3, 6, and 12 cm. canopy/basal diameter ratios of 2, 4, and 6. Larger diameter plants resulted in prediction equa- CSA Frequency and Parker tions with flatter slopes when CSA density (the depen- Frequency Relationships dent variable) was regressed against Parker frequency (the independent variable). Equations designed for The large difference in plot sizes used for measuring prediction ofCSA density must includeboth Parkerfre- CSA frequency and Parker frequency (0.5 m2 and , quency and a measure of mean plant basal size as in- 3/4-inch diameter) and the differences in sample sizes dependent variables. Regression equations fit to CSA ofthe two methods (10 plots and 100 plots pertransect, density and Parkerfrequency datawere statistically sig- respectively) resulted in CSA frequencies near 100% nificant, with R2 ranging from 75.1% to 88.5%. Den- with Parker frequencies near 0%. As a result, increas- sity observations, however, were from 0.5-m2 circular ing basal area caused small changes in Parker frequen- plots centered overeverytenth Parkerloop, while other cy and large changes in CSA frequency (fig. 4). Parker observations were from microplots centered over each frequencycontinuedto increase longafterCSAfrequen- Parker loop. The close association ofParker loops with cy had reached 100%. CSA microplots generally resulted in less variable data than was observed with the fewer but larger CSA den- Basal Area and Parker Frequency Relationships sity plots. Computersimulations showedbothCSAcanopycover andCSA densitytobe relatedto Parkerfrequency. Rela- CSA Cover and Parker Frequency Relationships The relationship between CSA canopy cover and Parker frequency was dependent on the ratio between canopy and basal diameters ofthe plant species drawn on the screen (fig. 3). For any given canopy cover, a sample ofplants having a larger canopy/basal diameter ratio (i.e., equal percentcanopy areabut smallerpercent basal area in the community) resulted in fewer Parker hits than a sample of plants having a smaller canopy/basal diameter ratio. Plants with larger bases relative to size of the canopy had flatter slopes when CSA canopy cover (dependentvariable) was regressed against Parkerfrequency (independentvariable). Description of the relationship betweenCSA canopy cover and Parker frequency required thatboth Parkerfrequency and canopy/basal diameter ratio be included as independent variables. Similarto CSA density, regression equations fittoCSAcanopy coverandParkerfrequency datawere Figure4.—EffectofbasaldiameterontherelationshipbetweenCSA statistically significant, with R2 ranging from 56.3% to frequencyand Parkerfrequencyfor plantswith basal diameters 96.5%. of 3, 6, and 12 cm. 4 tionships, however, were not simple and in both cases 40 were dependent on basal area related parameters. Plot- ting CSA canopy cover, CSA density, and basal area 30.8 cm 1.9 cm measured using the 5- by 10-cm microplot on a relative - scale against Parkerfrequency revealed awidevariation 30 / in relationships for the first two vegetation parameters cCoD i.4. with much less variation for basal area (fig. 5). tC—O Analysis of these relationships indicated that the "co co numberofParkerhits obtained onatransectwas a func- S 20 tionofthebasal areaofthe sample species. As thebasal O area of a species increased, the probability of its being Ql o "hit" also increased. Basal area was the vegetation parameter measurement most strongly and simply re- 0.3 cm lated to Parkerfrequency. However, numerous plant and vegetationcommunitycharacteristics could potentially biansfallueanrceeathaendfunPcatrikoenralfrfeoqrumenocfyt.heTrheelsaetiocnhsahriapctbeertiswteiecns 0 - . " included mean plant size, variation in plant size, and 20 40 60 80 100 grouping of plants within the community (random to Parker frequency (%) highly contagious distributions). Figure 6.—Effect of basal diameter on relationships between microplot basal area and Parker frequency. Relationships are Plant Size Effects shown for plantswith basal diametersof 0.3, 1.9, and 30.8 cm. Meanplantbasal diameterhad a significant influence 40 onthe form ofbasal area and Parkerfrequency relation- Basal diameter: 15 cm ^ 30 a5' s OC>D <clDo_ o CO > "co Q. CO O co 20 c -O 3 CoO o < a. o cOo .2 io J* c CL _1 I I L. 10 20 30 40 Parker frequency (%) cco a> Figure 7.—Relationship between microplot basal area and Parker T<3 frequency for plants with a mean basal diameter of 15 cm and CoO standard deviation of either 0 cm () or 3 cm (3). ships. Large plants (30.8-cm mean basal diameter, fig. 6) had a linear relationship between basal area and Parker frequency with a slope near 1 and a Y-intercept near zero. The smallest plant size tested (0.3-cm diameter, fig. 6) had a curvilinearrelationship. Intermediate CcOo plant sizes had functional relationships between these CO -Q two extremes. o o Q. O Plant-Size Variation Effects o Results ofsimulations in which meanbasal diameter Parker frequency (%) washeldconstantandstandard deviationofbasal diam- Figure5.—RelationshipsbetweenParkerfrequencyandpercentages eter was changed indicated no significant influence on o2famnadx6i);muCmS:ACdSenAsictayno(pmyeacnovbearsa(lcadnioapmye/tbearssalofd3i,am6,etaenrdr1at2iocsm)o;f the functional form of the relationship between basal and microplot basal area (mean basal diameters of 3, 6, and areaand Parkerfrequency. This was consistentwiththe 12 cm). findings ofHutchings and Holmgren (1959). For exam- 5 pie, no differences inCSAbasalareaand Parkerfrequen- Plant Distribution Effects cy relationships were evident for simulations of 15-cm meanbasal diameterplants with standard deviations of Plant species are seldomrandomly distributedwithin 0 and 3 cm. Regressions fitto these data sets were simi- plantcommunities. Rather, theyshowatleastsomecon- lar, and plotting these data on the same graph (fig. 7) tagion ("clumping") in their pattern of distribution. illustrated the overlap of data points that typified this Simulations were run inwhichthe degree ofcontagion and othersimulationtests concerning variation inplant ofindividuals ofa plant species varied from randomto size. highly contagious. Randompopulations showed no contagion (fig. 8a). Populations with intermediate contagion (fig. 8b) showed individuals beginning to form into groups but with less intensity than for heavily contagious populations. Populations with heavy contagion had mostplants gathered into afew groups (3 to 5) with large empty spaces between groups (fig. 8c). The degree of contagion of a population was some- times a significant variable, especially when an ex- tremely contagious population was compared with a completely random population. Therefore, the relation- shipbetween CSAbasal areaand Parkerfrequency may depend on the degree ofcontagion within the community. Simulations of3-cmand9-cmbasal diameterplant communitieswith differentcontagions illustratedthe effects. Simple linear regressions for populations with a mean basal diameter of 3 cm and distributions ranging from random to heavily contagious (fig. 9) resulted in Parker frequency (%) Figure 9.—Relationship between microplot basal area and Parker Figure 8.—Photographs of artificial vegetation mapswith various frequencyforpopulationswithvariousdistributionsof3-cmbasal distributions of plant individuals: (a) random population (no diameterplants: (a)random population(nocontagion); (b) inter- contagion);(b)intermediatelycontagiouspopulation;and(c)heav- mediately contagious population; and (c) heavily contagious ily contagious population. population. 6