Made in the United States of America Reprinted from Forest Science Vol. 21, No. 4 December 1975 pp. 365-370 6 �. � \ � � �. � Q I ro .� Out..of.·Roundness in Douglas..fir Stems Vl o Out-of-roundness 806 that in upper stem positions. 0.. 7'" ro Vl trees felled and bucked according to current ) Out-of-roundness (OOR at stump height was greater than Orientation of longest diameter (OLD) on gentle slopes of slope. No such significant, or even apparent, tendencies were found on moderate or Because OLD was random in most circumstances, large-scale cruises o.> ro The bias in volume estimates of individual trees, however, can be appreciable. menziesii. �. ro ro o.. -o :l c I� n d� Pseudotsuga 8 �I -- ' i" ) ro that observers had to go uphill from subject � trees to avoid overly large vertical angles. Obviously, the amount of bias will depend on both OLD and amount of OOR. This bias was studied in data collected for new log grading rules for coastal Douglas-fir (Pseudotsuga menzicsii (Mirb.) Franco var. menziesii). The data and ru1es (Lane and others 1973) are intended for grading logs in standing trees, but data such as length and direction of longest diameters were also collected. With these data, I hoped to answer these questions about cross-section OOR: 1. Is OOR related to factors such as diameter breast high ( dbh) and height of tree, degree of slope, and amount of tree lean? 2. Is OLD related to other directional variables such as aspect, direction of tree lean, shape of crown, or any particular azimuth(s)? 3. Do OOR and OLD differ at various relative heights in trees? It is important to note that the available data do not include OOR measurements The author is Mensurationist, Pacific North­ west Forest USDA and Forest Range Service, Experiment Olympia, Station, ,}�,g§h. The author thanks David Bruce and Ca�;ryanq�, V�n­ coevering for assistance with statistical analyses. Manuscript received September · �· 1:, � o· ro --- ----.,----� OPTICAL CALIPERS and rangefinder den­ drometers permit upper-stem measurements on trees without climbing or felling. Such dendrometer measurements can provide tree volumes, form factors, and other tree characteristics. Dendrometer measure­ ments lead to direct tree volume estimates, bypassing local or standard volume tables and their possible biases. When coupled with a suitable sampling plan for deciding which trees to measure, dendrometer cruises offer increased precision over other types of standing tree cruises (Grosenbaugh 1965, Hazard and Berger 1972). One possible difficu1ty in using dendrom­ eter measurements to get tree volume esti­ mates is their inability to detect, or com­ pensate for, out-of-roundness (OOR) when stems are viewed from only one direc­ tion. In this study, OOR was defined as the quotient of the difference in length be­ tween the longest diameter of a stem cross section and the diameter at right angles to it, divided by this shorter diameter. Neither diameter necessarily passed through the pith. If there happened to be a consistent relationship between orientation of longest diameter (OLD) in azimuth degrees and that of tree to instrument direction, bias in volume estimates would resu1t. This would happen if, for instance, the longest diam­ eters were oriented in the same direction as slope, and if slopes were steep enough ro < Forest Sci. 21:365-370. Additional key words. Stem diameter measurement, stem cross-section area, Vl s· � .-o using optical dendrometry for upper-stem measurements should be biased very little by OOR. 0 c Vl ro -· showed significant, but slight, tendencies toward north and south directions and direction steep slopes. ro ro (stem circumference not circular) of Douglas-fir stems was measured on cross sections of merchantability standards. � 0:. e; ro n 3 ;:jd. "'"' 3Vi' ��� ol'�o-� ,....-< ::T Q �· ·� �;::+' 5 � -< @ � � -;g 3 � &� RICHARD L. WILLIAMSON Abstract. :=; ro ;:j Vl 30, 1974. volume 21, number 4, 1975 I 365 ;:j · 1 j i at breast height. Basal OOR measurements are those at stump height. Data Collection and Preparation Trees were selected from 72 sample areas distributed throughout northern California and western Oregon and Washington to ob­ tain a wide range of tree size and quality in coastal Douglas-fir sawtimber. Tree size, stem quality, and site index were the principal stratifications used in selecting sample trees. The sample areas were also located to sample the environmental factors of forest type, stand density, elevation, aspect, and geographic location. Before trees were felled, a mark was made on the north side of each trunk, and a vertical projection of the crown sketched. This yielded information on azimuth and distance of crown center from center of stump and orientation of the longest crown diameter in a horizontal plane. Diameter breast high was measured to the nearest 0.1 inch (0.25 em) with a diameter tape. Basal area of the surrounding stand was estimated and tree lean was measured to the nearest degree. Direction of lean and aspect azimuth were measured to the nearest 45 o. After each tree was felled, an orientation line was painted the full length of the merchantable part of the bole and on top of the trunk as it lay. The direction this orientation line would face if the tree were standing was determined. Logs were bucked to achieve maximum value of end­ product; thus, they were of variable length. Log-end measurements were made in mill yards. At each end of each log, the longest diameter and diameter at right angles to the longest were measured to the nearest 0. 1 inch (0.25 em). Neither diam­ eter necessarily passed through the pith. The angle formed by the longest diameter and the diameter passing through the orientation line and through the approxi­ mate geometric center of the section was recorded and used to calculate azimuth of OLD to the nearest 5 degrees. Azimuth of the orientation line was unknown to the people making these log-end measure­ ments. The 806 trees available for analysis of 366 I Forest Science OOR averaged 36 inches (91 (ranging from 10 to 78 inches, 198 em) and 153 feet (47 m) (ranging from 68 to 258 feet, 79 m). em) dbh or 25 to in height or 2 1 to Analys�s, Results, and Discussion OOR at Stump Height-Magnitude. The dependent variable (OOR) ranged between 0 and 0.644 with mean and standard devia­ tion of 0. 123 and 0.088. Numerous step­ wise regressions of pooled data from all sampling areas yielded no combination of significant independent variables (includ­ ing slope, aspect, basal area, lean, and crown characteristics) which accounted for more than 4 percent of total variation in OOR. Even after data were sorted into three slope classes (0°-14 o, 15°-29°, 30°+ ), maximum variation accounted for in any slope class was only 10 percent. I sorted data into slope classes because I expected more OOR on steep slopes and thought a possible lack of OOR on gentler slopes might desensitize analysis when all data were pooled. Average basal OOR� however, was significantly {p < 0.05) greater (0. 135) in the 0°-14° slope class than in either the 15 -29° or 30°+ classes ·(0.114 and 0.117, respectively). This re­ sult may be hard to explain because of the commonly observed increase of eccentricity of pith with increasing slope. However, trees with eccentric pith are not necessarily out-of-round (Pawsey 1966). Throughout all multiple regression trials, consistently significant variables were those including dbh. On the average, the bigger the dbh, the greater the OOR. This finding is consistent with other work (Muller 1958). Over all slope classes, I was able to account for only 4 percent of the variation in OOR. Other areas for further study of OOR are its relationships to detailed stand and crown structures surrounding subject trees, and to direction of prevailing winds. Although prevailing winds may control amount of OOR and OLD, these winds may become randomly oriented duejo hilly or mountainous terrain. Almost au.;;fih.�e . data were collected in such terrain. o TABLE 1. Means standard deviations and sample sizes (n) for basal azimuth difference between aspect and longest diameter (DEV), by octants of aspect, for trees in the 2°-14° slope class. Aspect Statistic Mean, degrees oo 270° 3 15° 46 44 44 26 28 31 16 85 17 41 18 45° 90° 135° 38 55 30 47 42 30 23 28 24 59 30 14 11 . 180° 225° (expected, 45 o ) Standard deviation, degrees (expected, 26°) 11 Orientation. Analyses of basal OLD tested the hypothesis that this orientation was random, that is, that OLD could occur in any direction with the same probability that it could occur in any other direction. I used three methods of testing. First, visual inspection of plottings of OLD and of the angular deviation (DEV) of OLD from aspect azimuth revealed no coqsistent relationship of OLD with com­ pass direction or with aspect azimuth. Since one has a choice of two directions in de­ fining OLD, each 180° from the other, I arbitrarily chose the direction that was within ±goo of slope direction. Since OLD was calculated to the nearest 5 degrees, DEV was also rounded to the nearest 5 degrees. The second method involved segregating basal OLD, by slope classes, into equal subdivisions (15°) of the compass semi­ circle. Means and standard deviations for basal OLD were reasonably close to the values expected if counts were distributed uniformly in the oo to 180° range. How­ ever, a chi-square test of goodness of fit for each slope class showed that, for trees on gentle ground, there was a slight but significant (p < 0.05) tendency for OLD to occur in approximately north-south di­ rections. Basal OLD apparently occurred randomly on moderate and steep slopes. The third method involved analysis of DEV. DEV was expressed on an absolute (sign deleted) basis; thus, it had a possible range of oo to goo. Means and standard deviations of DEV were close to the values expected if DEV occurred at random (Table 1). However, DEV was categorized into 10° cells, and chi-square tests, by slope classes, showed a moderate but highly sig­ nificant (p � 0.01) tendency on gentle ter­ rain for OLD to lie in the direction of slope. DEV, in common with OLD, is ap­ parently random in the 15°-2go and 30°+ slope classes. Similar results were obtained when angle between OLD and direction of tree lean was analyzed the same way. For trees on gentle terrain, there is a moderate ann-sig­ nificant tendency for OLD· to lie in the direction of tree lean. No such tendency is evident on moderate and steep slopes. The significant tendency for OLD to agree with slope and lean directions in the 2 o-14 o slope class is not likely important. DEV means and standard deviations on each class of aspect appear reasonably close to values expected for a uniform (listribu­ tion (45° and 26° , respectively) on all octants (0°, 45°, goo, and so on) of aspect, except east (goo) and northwest (315 °) (Table 1). These most severe de­ partures from expected values occur with small sample sizes and are not statistically significant (p > 0.05). OOR Above Stump Level-Magnitude. The amount of OOR in portions of trees from the top of the first log through go percent of total height (measured from ground level) averages about half of average OOR at stump height (Table 2). Average mean and standard deviation for OOR in these upper portion��e 0.0615 �h and 0.0477, respectively. · • volume 21, number 4, 1975 I 367 TABLE 2. Average out-of-round (OOR), azimuth difference between aspect and longest diameter (DEV), and orientation of longest diameter (OLD), by slope classes, for stump height and for positions above stump height. Variable Slope class D EV OOR ��-- Basal U pper OLD Basal Upper Basal Upper 11 Degrees Degrees Degrees Degrees 0°-14° 0.135 0.0604 45.4 43.3 92.4 100.0 295 15°-29° .114 .0578 44.6 45.8 89.5 97.0 247 30°+ .117 .0663 44.5 42.7 90.8 90.7 264 All classes .123 .0615 44.9 43.9 90.9 96.0 806 Section data above stump level were mea­ sured on ends of logs bucked according to current merchantability standards practiced at the mill for which the logs were destined. Section data were sorted into relative height classes, with each class 1 0 percent of total height. Each relative height class contained sections from different sets of trees be­ cause of the chance location of log buck­ ing points. A separate one-way analysis of variance was made for each relative height class, except 0-10 and 91-100, to test for significant differences between average OOR by slope classes. These latter classes were judged to have insufficient numbers of observations to yield reliable averages. These tests were inconclusive-a few with high significance (greatest OOR on steep slopes), but the rest insignificant (p ;? 0.20). Regardless of relative position in trees, OOR varies little between slope classes, and the minor occurrences of statistical signifi­ cance may well be happenstance. Orientation. Overall average and stan­ dard deviation (90.9o and 54o, respec­ tively) for upper stem OLD closely ap­ proach expected values for a uniform dis­ tribution. The same holds true for DEV (mean= 44.9, standard deviation= 28.3). Longest diameter orientation and DEV above stump level were analyzed by chi­ square tests similar to those for stump level data. Separate chi-square analyses were done for each relative height class, except 0-10 and 91-100, and for each slope class. 368 I Forest Science Significance occurred primarily for trees on gentle ground and in the upper portions of trees, where effects on merchantable volume would be minimal. Where signifi­ cant, the basic data consistently show a moderate tendency for OLD to lie in car­ dinal directions, principally north-south, and in the direction of slope. This agrees with basal OLD. And, as with basal OLD, these tendencies are not likely to be im­ portant since means and standard deviL tions for upper stem OLD and DEV are reasonably close to expected 1values, that is, orientation appears to be a nearly ran­ dom occurrence. OOR Influences on Tree Volume Esti­ mates. If OLD were c�nsistent throughout tree lengths, the worst cases of bias in volume estimation with an optical den­ drometer should be when measurement is from a direction parallel or perpendicular to OLD at stump height. To gain insight into the possible extent of this bias, I calculated the volume of each tree in four ways as viewed ( 1) from a direction parallel to basal OLD, (2) from a perpendicular direction, (3) from 45o, and ( 4) from 13 5o. Individual log volumes in each case were calculated with Smalian's formula. For the parallel direction, apparent diameter for determining a sectional area above stump level was determined by: ( 1) multiplying the difference between long and short diameters of that section by the cosine of the angle between'"''LD at -, stump height and OLD of the section: ' (2) adding the product of (1) to the short diameter of the section. Apparent diameter for the perpendicular direction was determined by a similar pro­ cedure. These steps give correct caliper values for longest diameter and diameter at right angles to the longest and allocate the difference smoothly for other view angles. Because tree sections are irregular, no single function will be exact. This procedure should yield the maximum range in estimation of volume per tree if OLD is consistent within trees. Finally, volume ratio-parallel over per­ pendicular lines of sight-was calculated for each entire tree, and for the butt log and upper portion above butt log separately. If OLD occurs randomly, average ratio for upper portions of trees should be nearly 1.0. Average volume ratios and average stump and top of butt log parallel: per­ pendicular basal area ratios are: Section of tree Basal area Volume Mean Standard deviation Stump Top of butt log 0.81 0.11 0.97 0.08 Butt log Upper Entire 0.86 0.98 0.92 0.08 0.06 0.06 The mean basal area ratio for tops of butt logs shows almost no agreement with mean stump ratio and indicates almost random OLD at that position. Likewise, relation­ ship between stump basal area ratios and tree volume ratios, though statistically highly significant (p � 0.01), is very weak (r2 = 0.0102). Volume ratios for individ­ ual entire trees ranged from 0.70 to 1.09, and, for upper portions of trees, from 0.76 to 1.21. Since these ratios are derived from extremes, and true volume probably lies about midway between extremes, maximum deviations of dendrometer estimates from true volume probably lie within 10 to 12 percent of true volume. Much of the butt log bias in volume estimation of individual trees can be elimi­ nated by using diameter tape measurements at breast height and for lower positions if required. One may wonder, however, about the relative advantages of dendrometer and D-tape when many trees are to be mea­ sured. Dendrometry from one position is approximately equivalent to calipering from one direction; If this direction is assumed to be randomly oriented, calipering will re­ sult, in average positive bias in cross-sec­ tional area estimation of 0 to 3.4 percent, depending on the geometric shape as­ sumed for the cross-section (Matern 1956). This bias is approximately equivalent to that from diameter-tape measurements, so the choice between diameter-tape and opti­ cal dendrometer should probably depend on the precision of each. Diameter-tape measurements are much more consistent than measurements made by calipering in two directions ( McArdle 1928) and would logically be even more consistent than calipering in one direction. The greater precision of the diameter tape will lead to greater sampling efficiency. For the two measurement directions of 45° and 135° to basal OLD, volume and basal area. ratios relative to perpendicular direction values are almost identical: Portion of tree Basal area Volume ;·.-· . Mean Standard deviation Stump Top of butt log 0.94 0.04 0.99 0.05 Butt log Upper Entire 0.96 0.99 0.97 0.03 .0.04 0.03 All the above indicate nearly random orientation of upper stem OLD and basal OLD, and little, if any, relation between them. In addition, where cardinal or slope direction tendencies do seem to occur­ on gentle terrain---dendrometer users may place themselves to have the best views possible of subject trees. This placement is likely to be in random directions from subject trees. Thus, stand volume estimates by dendrometry should have small average bias due to OOR. However, the two tabu­ lations show that errors in indi-vdQual tree estimates can be large. volume 21, number 4, 1975 I 369 Conclusions Out-of-roundness generally increases with tree size but is little affected by slope of ground, amount of tree lean, or density of the surrounding stand. OOR, from the top of the butt log on up the tree, averages only about half that at stump level. OLD does not appear to depend on orientation or position of crown. There is a slight tendency for the longest diameter to be oriented north-south, in direction of slope, and in direction of tree lean on ground with gentle slopes. On gentle ter­ rain, dendrometer users may place them­ selves so as to have the best views of subject trees possible. This placement is liable to have a random orientation with respect to cross-section OLD and direction of slope. On steep ground, where dendrometer users must go uphill from subject trees to avoid large vertical angles, OLD appears to be a random occurrence and no appreciable bias in volume estimates should result from OOR. Evaluation of tree volumes derived as though viewed from various directions indicate that average stand volume biases due to OOR will be negligible (about 2 or 3 percent) under any circumstances. Volume bias of individual trees can be large, however. Literature Cited GROSENBAUGH, 'L. R. 1965. Three-pee sampling theory and program "THRP" for computer gtlneration of selection criteria. S�rv Res Pap PSW-21, 53 p. Forest & Range Exp Stn, Berkeley, Calif. H�ZARD, JoHN W., and JoHN M. BERGER. 1972. ·Volume tables vs. dendrometers for forest sur­ veys. J For 70:216-2 19. LANE, PAUL H., RICHARD 0. WooDFIN, JR., JOHN W. HENLEY, and MARLIN E. PLANK. 1973. New timber cruising grades for coast Douglas­ fir. USDA Forest Serv Res Pap PNW-151, 12 p. Pac Northwest Forest & Range Exp Stn, Portland, Oreg. McARDLE, RICHARD E. 1928. Relative accuracy of calipers and diameter tape in measuring Douglas fir trees. J For 26:338-346. MATERN, BERTIL. 1956. On the geometry of the cross-section of a stem. Medd Fran Stat Skogs­ forsk 46(1 1), 28 p. MULLER, G. sectional 1958. shape Investigations on the cross­ of tree trunks. Forstwiss Centralbl 77( 11/12) :374-381. PAWSEY, C. K. 1966. Lean and-�ccentricity in Pinus radiata (D. Don) in the southeast of South Australia. Aust For Res 2(3):22-35. PURCHASED BY THE FOREST SERVICE, U.S. DEPARTMENT OF AGRICULTURE, 370 I Forest Science USDA For Pac Southwest for officia�u.�e.,