See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/233595632

Five Instruments for Measuring Tree Height: An Evaluation

Article in Southern Journal of Applied Forestry · May 1994

DOI: 10.1093/sjaf/18.2.76

CITATIONS READS

67 7,526

3 authors, including:

Michael S Williams
United States Department of Agriculture
113 PUBLICATIONS 1,612 CITATIONS

SEE PROFILE

All content following this page was uploaded by Michael S Williams on 29 September 2014.

The user has requested enhancement of the downloaded file.

 Five Instruments for Measuring Tree Height:
An Evaluation
Michael S. Williams, Multiresource Inventory Techniques,
Rocky Mountain Forest and Range Experiment Station,
USDA Forest Service, 240 W. Prospect St., Fort Collins, CO
80526-2098; William A. Bechtold, SoutheasternForest
Experiment Station, Forest Inventory and Analysis Project,
USDA Forest Service, Asheville, NC 28802-2680; and
V.J. LaBau, Forest Health Monitoring, Forest Sciences
Laboratories, Anchorage, AK 99501.

ABSTRACT. Five instruments weretestedfor reliabilityin measuringtreeheightsunderrealisticconditions.

Four linear modelswere usedto determineif tree heightcan be measuredunbiasedlyoverall treesizesand if
any of the instrumentswere moreefficientin estimatingtree height.The laserheightfinder was the only
instrumenttoproduceunbiasedestimatesof thetrueheightfor allfour linear models.An accuracytestshowed
the clinometerand enbeecoinstruments producedbiasedresultsin the 0-33 and > 66 fi heightclasses
respectively.Theresultsfor thelasermaybe misleadingdueto thelimitedamountof data collectedwith this
instrument.The results(subjectively)confirmthat treesup to 40 fi are measuredvery accuratelywith no
exceptionsfor theclinometer,relaskop,andenbeeco.Thetele-relaskop generallyproducedresultsthatwere
poorer thantheotherinstruments. South.J. Appl. For. 18(2):76-82.

Whenmeasuring
the
height
ofatree,
where
the
base
andtop Palustris Mill.). Seventeenadditionalspeciesincluding a
are well definedandclearlyvisible,the existinginstruments varietyof oaksalsoappearin the dataset.
for measuringthe heightof standingtreesare adequatefor All datacollectedarefromForestHealthMonitoringplots
mostapplications (Hunt 1958,Warren1959,Rennie1979). in Georgia.True heightsfor all treesweretakenwhile crews
However,thevarianceandbiasof theheightestimates could collectedfoliage samplesfrom the tree crowns.One crew
be largefor tall treesin densestandsor for treeswhichdonot memberclimbedashigh aspossibleup eachtree.From that
have well-definedtops.An opportunityto evaluatesomeof point poleswere usedto measurethe remainingdistanceto
thecurrentlyavailableheightmeasurement instruments
(in- thetop of the tree.To determinewhenthe polewasat the top
cluding a laser-driveninstrument)presenteditself in the of the tree, sightingswere taken from the groundby two
summerof 1991. Tree climbers,who were employedto observersandby the crewmemberin thetree.When all crew
collectfoliagesamplesfor chemicalanalysis,measured the memberswerein agreement,thetotalheightfrom theground
true heightof standingtrees.Readingsfrom a numberof to thetop of thepole wascalculated.While somemeasure-
differentheightmeasuring instruments werecomparedto the ment error existsin this method, no alternative method could
valuesobtainedby thetreeclimbers.Thesedatawereusedto be implementedwhichmettime andcostconstraints andstall
determineif any one instrumentwas superiorto the other representedrealisticmeasurementsituations,
suchasvarying
instruments under real world conditions. terrain,canopy,treeheightdistributions,
andspeciesmix
Readingsfrom a laserheightfinder (Jasumback1991),
Data Descriptionand CollectionMethods Suuntoclinometer(Husch et al. 1982), Speigelrelaskop,
Enbeeco
clinometer
1,Speigel
tele-relaskop
(Bitterlich
1978,
The data set consists of 100 hardwood and softwood trees
Husch et al. 1982) were recordedfrom the samelocation. All
with21differentspecies.
Thepredominant speciesareloblolly measurementswere taken by an experiencedfield crew
(PinustaedaL.), slash(PinuselliottiiEngelm.var.elliottii), member.A tripod was usedto steadyeachinstrument.Ttus
shortleaf(Pinus echinata Mill.) and longleaf pine (Pinus requiredadaptinga tripodmountfor the Suuntoclinometer,

No•: Mentionof tradenamesor commercialproductsdoesnot constitute 1H. Steward

Ltd,Enbeeco
House,
Carlton
Park,Saxmundham,
SuffolkIP17
endorsement or recommendation of use. 2NL, U.K.

76 SJAF18(2)1994
which1sdesigned to behandheld.All otherinstrumentshad R2 valueswereusedforan•ndlcatorof goodness
of fit.
thumbscrewsystemsto accommodate tripods.The laser LargerR2 values
indicated
a betteragreement
between
the
heightfinderwasa preproduction modelandwasnot avail- estimated
andactual
heights.
Forthisstudy,R2wasdefined
able until late in the field season.Samplesizeswith this as

instrument were smaller than the other four instruments due

to thelimitedamountof timeit wasavailablefor testing.
The instruments
weresetup at a distancefromthetreeso R2= 1 SSE
SST
that a 45ø measurementangle was never exceeded.The
measurements were takenfrom the uphill sideof eachtree
providedan openline of sightto thetopof thetreecouldbe where
found. For two trees in the data set, measurementscould not
be takenfrom an uphill position.
A combination of scales was used so that no two instru- SSE
=E (ht,-[•mi
)2
ments successivelyutilized the same scale. The Suunto i=1

clinometer
andtheenbeeco usedthepercentscale,therelaskop
usedthetopographic
scale,andthetele-relaskop andthelaser
height
finderused
thedegree
scales.
2Thisprovided
aquasi SST
=E(hti-•)2
"blind"aspecttothestudy,intended tointerruptthetendency i=1

to "drive" the currentreadingsto be the sameas those

obtainedwith the previousinstrument.Distancesfrom the /• istheestimated
height
generated
fromthefourmodels
observerto the tree in the data set were given as slope mi
distances.
Leveldistances werecomputedfor the analysis. hr.isthetrueheight
andh isthemean
ofthetrueheights.
1

Confidenceintervalsfor models(1) and (2) allowedthe

Evaluation Techniques
hypothesis
Four linear models were fit to the data to test the bias and
efficiencyof theheightmeasurements andto determineif
thereis an upperlimit to the reliabilityof the instruments.
Ideally, the correlationbetweentrue height and measured
to betested.Analysisof [3for models(1) and(2) wasusedto
heightwouldbe 1, thusthe first modelwas specifiedas
testif a biasin estimationexistsfor anyof theinstruments.
For models(3) and(4) thehypothesis
ht = [3h
m+ œ (1)

wherehtisthetrueheight
andhmisthemeasured
height.
The
errorterms,e, are assumedto be normallydistributedwith was tested.
mean0 andvariance(52. The four models were fit for two sets of data. First for all
The assumption of equalvariancein the error term is speciesof treestogether,thenfor softwoodtreesonly.Hard-
highlysuspect. A reasonableassumptionis thatthe errorin woodswere removedto examinethe effect of extracting
measurement increasesproportionally
withtreeheight.There- additionalvariabilitycausedby poorlydefinedcentralstems
fore the model
typicalof manyhardwoodspecies.
In additionto the linearmodelcomparison, an accuracy
eat=e•hm
+el (2) testbasedon bothbiasand precisionwas used(Reynolds
1984). Mean error was used as a measure of bias. Then
was fit to removethe effect of the heteroscedasticity.
An confidenceintervalswere generatedaboutthe meanerror.
•ntercepttermwasaddedto (1), yielding The biasis consideredsignificantif the confidenceinterval
doesnot containzero.The confidenceintervalweregener-
atedusing
ht = ct+ [3hrn
+œ (3)

Theerrorterms,e, areassumed
normallydistributed
with
mean
0 andvariance
(52.Anintercept
wasalsofit tothelog Stl__a,n_l
•+ 2
transformed
data,yielding

eh'=ea+lShm
+e1 (4) where

2Husch
etal.(1982)givea fulldescription
ofthedifferent
scales
used.

SJAFI8(2) 1994 77
 110.

IO0

g=E mi
n'
i=1

S=
I• (e,-a)2
i=• n-1
70.

t a = Student'
s t distribution
at 1- a/2 with
1-•-,n-1
n - 1 degreesof freedom

n = number of observations

30.
Thistestrequiresa homogeneous errorstructure,
whichis
notpresentin thedata.Thereforethedatasetwasdividedinto
20
three heightclassesto obtainroughlyhomogeneous vari-
anceswithin each height class.The height classeschosen
were 0-33, 33-66 and )66 ft. The accuracytest was per- 10.

formedon the dataset containingall treesandthe dataset

with the hardwood trees removed.
To determineif there is a specifictree height at which 0 10 20 30 40 50 60 70 80 90 100 11o

instrumentmeasurements are no longer accurate,a visual

estimatedheight
comparisonof modelerrorswasperformed.The visualcom-
parisonentailedgraphingthe measuredheightversustrue
heightandlookingfor a point in the scatterplot wherethe Figure 1.
errorsin measurementincreaseddrastically.As anexample, For the clinometerthe modelswithout interceptterms
Figure 1 showsthe relationshipbetweentrue heightand [models(1) and(2)] hadconfidence intervalscontaining• =
measuredheightusingthe clinometer.Both hardwoodand 1, regardlessof whetherall speciesor onlysoftwoodspecies
softwoods areplottedandcanbe differentiated by the sym- wereconsidered.In contrast,for modelswithinterceptterms
bols used.
([models(3) and(4)], no confidenceintervalcontainedeither
A comparison of modelerrorswasperformedby comput- tx= 0 or•= 1.R2values
ranged
from0.9462to0.9739.
When
ing the averageabsoluteerrorbetweenthe model(1) esti- allspecieswereconsidered
theR2 values
wereconsistently
matesand the true heightsby 10 ft heightclasses.These second
largestcompared
to theotherfourinstruments.
When
valueswere computedusingthe formula onlysoftwood
trees
wereconsidered
theR2values
werethird
largest,regardlessof whetherthe data was transformedor
which model was fit. For the accuracytestthe clinometer
showeda significantbiasin the0-33 ft heightclass.In the

• i•h
i=1
m-htr/h
othertwo heightclasses,the biaswas nonsignificant.
The resultsfor the relaskopwere similarto thosefor the
clinometer.For the modelswithoutinterceptterms[models
(1) and(2)] theconfidence intervalsfor • contained1.When
where nh is the numberof trees in a given height class. the intercept term was added ([models (3) and (4)], the
Averageerrorsin eachheightclassprovideda goodindicator confidence intervalscontainedtx= 0 and• = 1 for bothof
of how accuratelytree growthcanbe measuredfor a given thelog transformeddatasets.This mayimply thattheerrors
heightwith eachof the instruments. in fit occurred
because
of heteroscedasticity
in.thedata.R2
values ranged
from0.9363to0.9704.Fortherelaskop
theR2
Results and Discussion valueswerealwaysthirdlargestfor the all speciesdatasets
andR2 values
werefourthlargest
whenonlysoftwood
trees
Tables 1-4 give resultsfor fitting the four modelsto the were included in the data set.
datasetwith all speciesandthe datasetwith only softwood For thetele-relaskop
theconfidence
intervalscontain• =
trees.The accuracytestresultsby heightclassare givenin 1 for the modelswithoutinterceptterms.When intercept
Tables5 and 6. The averageabsoluteerrorsby heightclass terms were added, the confidence intervals contained (x = 0
are given in Tables7 and 8. and • = 1 for the untransformed
datausingboth the all-

78 SJAF18(2)1994
Table
1.Regression
coeffic,ents,
standard
errors,
95%conf,dence
intervals,
andR2 values
w,ththeuntrsnsformed
detaendalltrees.
Instrument Model# Parameter Estimate $td. Err. Cl R2 n

Chnometer [1] • 0.9888 0.0084 (0.9720, 1.0056)* 0.9462 90

[3] (z 3.6659 1.3941 (0.8954, 6.4364) 0.9501
[3] • 0.9329 0.0228 (0.8876, 0.9781)

Relaskop [1] • 0.9835 0.0090 (0.9655, 1.0016)* 0.9372 90

[3] c( 3.6330 1.5225 (0.6074, 6.6586) 0.9410
[3] • 0.9283 0.0248 (0.8791, 0.9776)

Tele-relaskop [1] • 0.9950 0.0160 (0.9628, 1.0272)* 0.8990 51

[3] c( 4.4237 2.5880 (-0.7771, 9.6245)* 0.9047
[3] • 0.9266 0.0430 (0.8403, 1.0130)
Enbeeco
[1] • 0.9746 0.0088 (0.9570, O.9921)* 0.9505 76

[3] (Z 3.3680 1.5062 (0.3666, 6.3693) 0.9537

[3] • 0.8835 0.0372 (0.8093, 0.9576)
Laser
[1] • 1.0008 0.01858 (0.9609, 1.0406)* 0.9250 15

[3] (z 4.4341 4.9531 (-6.2665, 15.1346)* 0.9293

[3] • 0.9387 0.0718 (0.7836, 1.0938)
* Indicates unbiased fit.

species
datasetandthesoftwoods-only
dataset.UsingR2 terms were added,model (4) had confidenceintervalscon-
valuesas an indicationof fit, the tele-relaskopproducedthe taining
a = 0 and[•= 1forthesoftwood
data.R2values
for
worst fit for all models and data sets considered. the enbeecowere consistentlythe largest when all tree
The enbeecowas the only instrumentwhich did not have specieswere considered.When only softwoodtreeswere
confidence intervalscontaining[• = 1 for all modelswithout considered,
theR2valueformodels
(1), (3) and(4)werethe
interceptterms.Whenmodel(1) wasfit to thesoftwooddata second
largestamongallinstruments
andlargestwhenmodel
the assumption that ht = hrnwasrejected.When intercept (2) was fit.

Table2. Regression
coefficients,
standard
errors,
95%confidence
intervals,
andR2 vslues
forlogtransformed
datasndalltrees.
Instrument Model# Parameter Estimate $td. Err. Cl R2 n

Chnometer [2] • 1.0003 0.0021 (0.9961,1.0045)* 0.9702 90

[4] (z O.2074 O.0669 (0.0745, O.3403) O.9731
[4] • 0.9485 0.0168 (0.9151, 0.9819)

Relaskop [2] •, O.9984 O.0021 0.9942, 1.0027)* O. 96 87 90

[4] (z O.1228 O.0727 (-0.02 16,0.2673)* O. 96 97

[4] • 0.9678 0.0182 (0.9316, 1.0041)

Tele-relaskop [2] • 1.0024 O.0048 (0.9926, 1.01 20)* 0.9215 51

[4] (z 0.3822 O.1312 (0.1046, 0.6599) 0.93 21

[4] • O.9065 O.0349 (0.8362, 0.9767)

Enbeeco
[2] • 0.9963 0. 0023 (0.991 8, 1. 0008)* 0.9726 76

[4] a 0.1375 0.07 35 (-0. 0090, 0.2839) 0.97 35

[4] • O.9623 O.0183 (0.9257, 0.9988)

Laser [2] • 1.0017 0. 0041 0. 9597 15

(0.993,1.010)*
[4] c( 0. 2665 0. 2086 (-0.1842,0.7172)* 0.96 42

[4] •, O.9380 O.0501 (0.8297, 1.0462)

* Indicates unbiased fit.

SJAF18(2)1994 79
Table3. Regress,on
coefficients,
standard
errors,95%conf,dence
intervals,
andR2 values
foruntransformed
dataw,thallhardwood
trees removed.

Instrument Model# Parameter Estimate Std. Err. CI R2 n

Clinometer [1] j• 0.9846 0.0096 (0.9655,1.0037)* 0.9480 68

[3] (z 3 9622 1.6137 (0.7403,7.1841) 0.9523
[3] j• 0.9262 0.0255 (0.8753,0.9771)
Relaskop
[1] j• 0.9837 0.0106 (0.9625,1.0048)* 0.9363 68

[3] (z 4.1243 1.8004 (0.5297,7.7189) 0.9410

[3] j• 0.9229 0.0284 (0.8661,0.9797)
Tele-relaskop
[1] • 0.9983 0.0195 (0.9589,1.0377)* 0.8961 41

[3] (z 4.3524 3.1092 (-1.9418,10.6467)* 0.9012

[3] • 0.9335 0.0501 (0.8320,1.0350)
Enbeeco
[1] J• 0.9707 0.0098 (0.9511, 0.9903) 0.9535 59

[3] c• 3.6061 1.7017 (0.1984,7.0138) 0.9569

[3] • 0.9197 0.0258 (0.8679,0.9715)
Laser
[1] • 0.9899 0.01 71 (0.951 7, 1. 0284)* 0.9542 11
[1] (Z 5. 1553 4.31 62 (-4.6086,14.9192)* 0.9605
[3] • 0.91 85 0. 0621 (0.7780, 1.0589)

* Indicates unbiased fit.

Table
4.Regression
coefficients,
standard
errors,
95%confidence
intervals,
andR2 values
forlogtransformed
datawithallhardwood
trees removed.

Instrument Model# Parameter Estimate Std. Err. CI R2 n

Clinometer [2] • 0.9996 0.0025 (0.9946,

1.0047)* 0.9702 68
[4] (z 0.2321 0.0761 (0.0802,0.3842) 0.9739
[4] • 0.9421 0.0190 (0.9042,0.9801)
Relaskop
[2] • O.9989 O.O025 (0.9938,1.O040)* O.9694 68

[4] (z O.1242 O.0834 (-O.0422,0.2907)* O.9704

[4] • 0.9681 0.0208 (0.9266,1.0097)
Tele-relaskop [2] [• 1.0031 0.0060 (0.9911,1.0152)* 0.9220 40

[4] c• 0.3595 0.1596 (0.0364,0.6827) 0.9312

[4] • 0.9133 0.0403 (0.8318, 0.9949)
Enbeeco
[2] • O.9956 O.O026 (0.9904,1.O009)* O.9743 59

[4] (z O.1370 O.0815 (-0.0262,0.3002)* O.9755

[4] I• 0.9620 0.0202 (0.9220,1.0024)
Laser
[2] • 1.0000 0.0042 (0.9907,1.0092)( 0.9730 n:l 1

[4] (z O.3349 O.18O0 (-O.0723,0.7421)* O.9805

[4] • 0.9199 0.0432 (0.8221,1.0177)
* Indicates unbiased fit.

Table5. Confidenceintervalsfor accuracytestand numberof treesin eachof the threeheightclassesfor all trees.
Height
class Clinometer Relaskop Tele-relaskop Enbeeco Laser

0-33 (-1.28,-0.12), 13 (-0.62, 0.64), 13 (-2.43, 0.51), 7 (-0.95, 0.67), 10

33-66 (-1.45, 1.66), 45 (-1.08, 1.74), 45 (-2.71, 3.71), 26 (-1.03, 2.71), 35 (-2.08, 0.68), 5
>66 (-1.42, 2.63), 32 (-1.68, 3.46), 32 (-4.51, 2.05), 18 (-0.25, 3.66), 31 (-4.88, 5.02), 9

* Indicatesnot enoughdata to calculateconfidenceinterval.

80 SJAF18(2)1994
Table 6 Confidence intervals for accuracy test and number of trees in each of the three height claases w•th all hardwoodtrees removed.

Height
class Clinometer Relaskop Tele-relaskop Enbeeco Laser

0-33 (-1.58-0.31), 10 (-0.85 0.79), 10 (-2.43 0.51), 7 (-1.09 0.76), 9 *

33-66 (-1.89 2.64), 30 (-1.81 2.15), 30 (-4.14 5.80), 17 (-1.97 3.75), 22 (-3.45 0.72), 3
>66 (-0.83 2.90), 28 (-1.62 3.69), 28 (-4.95 1.95), 17 (0.51 4.15), 28 (-2.98 5.79), 7

ß Indicatesnot enoughdata to calculateconfidenceinterval.

Table 7. Average absolute model deviation end number of trees in each 10 ft height class for ell treea. The model uaed is ht =/•h m (1).
Heightclass Clinometer Relaskop Tele-relaskop Enbeeco Laser

10-20 0.80, 5 0.71, 5 2.05 0.61, 5

20-30 1.20, 4 0.42, 4 0.50 1.35, 3
30-40 1.33, 14 1.15, 14 3.94 1.51 11 1.37, 2
40-50 2.22, 12 1.76, 12 4.56 2.51 10 0.03, 1
50-60 5.09, 15 4.18, 15 6.03 5.34 11 0.64, 1
60-70 2.89, 11 2.12, 11 2.66 2.06. 9 3.38, 3
70-80 2.69, 13 3.23, 13 3.37 1.97 13 3.32, 4
80-90 5.86, 13 6.96, 13 3.01 5.16 12 6.45, 4
90-100 3.84, 3 7.98, 3 11.31 5.44 3

Table 8. Average absolute model deviation and number of trees in each 10 ft height class with all hardwood trees removed. The model
used is ht=/•hm (1).
Heightclass Clinometer Relaskop Tele-relaskop Enbeeco Laser

10-20 O.97 4 0.70 4 2.02, 3 0.65 4

20-30 1.46 3 0.39 3 0.48, 3 1.45 3

30-40 1.24 11 0.93 11 4.78, 6 1.33 9 1.71 2

40-50 2.74 6 1.25 6 6.78, 4 3.21 4

50-60 6.24 10 4.98 10 8.77, 5 7.48 6

60-70 3.O7 9 2.25 9 2.40, 5 2.29 8 3.56

7O-8O 2.60 12 3.04, 12 3.39, 7 1.65 12 3.30

80-90 5.02 10 6.46, 10 3.08, 5 4.49, 10 4.44

90-100 3.94 3 8.00, 3 11.26, 3 5.29, 3

Thelaserheightfinderhasthedistinctionof beingtheonly valuesin the 40-ft classof 3.94 and4.78 for the all-species
•nstrument
whoseconfidence intervalscontained[•= 1 andct andsoftwood-onlydatasets,respectively.For treesgreater
= 0underevery
modelanddataset.TheR2values
werefourth than 40 ft tall, the error values increase. The error values
largest
whenallspecies
wereconsidered.
R2values
werethe rangefrom 0.03 to 11.31 ft with mostvaluesfalling in the 2
largestfor the softwooddataexceptwhenmodel(2) wasfit. to 5 ft range.In mostcasesremovingthehardwoodtreesfrom
In thatcasetheR2 valuewassecond
largest. thedatareducedthe averageabsoluteerror.The clinometer,
For the accuracytestthe clinometerwasthe only instru- relaskop,Enbeeco,andlaserheightfinderproducedsimilar
ment to show a significantbias using the data set which resultsin all heightclasses.
Thetele-relaskop produced larger
•ncludedall trees.The significantbias was in the 0-33 ft errorsin the 30-60 and90-100 foot heightclasses.
heightclass.In the 33-66 and>66 ft heightclassesthebias The graphicalanalysisalso subjectivelyconfirmsthat
wasnonsignificant foreveryinstrument.Whenthehardwood treeslessthan40 ft aremeasuredaccurately.No substantial
trees were removed from the data set the clinometer still disagreements betweenmeasuredandtrueheightareseenin
showeda significantbiasin the0-33 footclass.TheEnbeeco the datauntil after the 40 ft level. As heightincreases,the
alsoshowedasignificant biasin the>66 ft heightclass.There graphsindicateanincreasingdisagreement betweentrueand
wasonly onetreelessthan33 ft tall measuredwith thelaser measuredheight.
sono accuracytestcouldbe performedin thisheightclass.
The analysisof the absoluteaverageerror by diameter Conclusions
classis givenin Tables7 and8. In all heightclasseslessthan
40 ft theaverageabsolute errorsweregenerallylessthan1.5 Resultsusingthe four linear modelsand averageerror
ft with the exceptionof the tele-relaskop,which had error testsweresimilarfor theclinometer,relaskop,andEnbeeco

SJAF18(2)1994 81
instruments.Both the chnometerandEnbeecoshoweds•g- Literature Cited
nificantbiasesin the0-33 and>66 ft heightclasses. Thetele-
relaskopappearedto be slightlylessaccurate.Even though
theresultsfor thelaserwouldindicatethatit is theleastlikely
BrrrEvmicH,W. 1978. Singletree measurements by the tele-relaskop--a
to producebiasedestimates, theeffectiveness of thisinstru- highlyefficienttoolfor forestinventories.8thWorld ForestryCongress,
ment is still difficult to determine because of the limited Jakarta, Indonesia.
amountof dataavailable.In addition,thelaserheightfinder Hta,rr, E.V. 1959. A time and accuracytest of somehypsometers.
J. For
57:641-643.
usedin thisstudywasanearlypreproduction model.Numer- HuscH,B., MILLER,C.I., and BEERS,
T.W. 1982. ForestmensurationEd. 3
ousdesignimprovements havebeenmadeoncurrentmodels, Wiley, New York. 402 p.
which may improveaccuracy.Additionaltestingwith the JASUMBACK,T. 1991.Laserheightmeasurement. Timbertechtips.USDA For
Serv.Tech.andDev. Prog.Gov.Publ.91242336-MTDC.
laserheightfinder wouldbe prudent.The graphicaland RE•Nm,J.C.1979.Comparison of height-measurement techniques
in a dense
averageerroranalysisindicatesthattreeslessthan40 ft tall loblollypineplantation.South.J. Appl. For. 3(4):146-148.
canbemeasured quiteaccurately.
If errorsof 2 to 5 ft canbe REYNOLDS,M.R., JR.1984.Estimatingtheerrorin modelpredictions.
For.ScI
30:454-469.
tolerated,all of theinstruments,
exceptthetele-relaskop, are WARREN,
W.G. 1958.Testsof someinstruments
for measuringtreeheights
suitablefor measuringtreeslessthan40 ft tall. N. Z. For. Res. Note 12.

Whaddya
Say
ToAGuy
Who's
HadTheSame
JobFor50Years,
HasNeverCalled
In SickOrShowed
UpLate,
NeverTaken
AVacation
OrAHoliday,
Never
Asked
For
ARaise
OrGriped
Abou[
HisBonus
And,Believe
It OrNot,HasNo
PlansForRetirement?

Thanks.
Show
Smokey
howmuchyouappreciate
hismany years
ofvigilance
bybeingcareful
withmatches
andcampfires.
Remember- only
youcanprevent
forest
fires.

Public
S•I'•I•
of
the
USDA
Forest
Service
and
Your
Slate
Forester.

82 SJAF18(2)1994
View publication stats