8 4 Journal of Tropical Forest Science 5 5 2(1) - 8 4 :

YIELDS OF RUBBERWOOD SAWN TIMBER

H. C. Sim

Forest Research Institute Malaysia, Kepong, 52109 Kuala Lumpur, Malaysia

Received September 1989

SIM.H 1989. .C . Yield rubberwoof so d sawn timber. Yield sawf so n timber from rubber trees were studied in a processing mill. Significant differences were detected when yields were categorized by diameters of logs and sizes of sawn timber produced. A relative frequency weighted approach was proposed as a better estimation techniqu asseso et average sth e yields. Yields categorize diameterg lo y d b sawd san n timber sizes, and weighted with relative frequencies would be a more accurate and useful input for planning and controlling of the sawmilling operation.

Key words: Rubberwood -yields-sawn timber - recovery - sawmilling - weighted means

Introduction

Within the last two decades, rubberwood has evolved from being relatively unknow highle on o nyt sought afteprincipaa s ra materia w furniture lra th y b l e industry, both in Malaysia and abroad. As such, the trees, originally planted for latex, have gained much importance contributin competitivenese th go t locaf so l woo wood dan d based industries. Productio sawf no n timber from rubber tree bees sha n studie numbea y db r of researchers (e.g. Lopez. Cho1985) . & 1980al al t o t oe e H ., 1982 n Ga , However, these studies have reported results that varied very widely. For example, recover f rubberwooo y d sawn timber from sawmills equipped with band saws was reported to vary from as high as 46% (Lopez et al. 1980) to as low (Ga. 1985)% al recovers t 22 ne A .s a f sawo y n timber depends very mucn ho factors suc relativs ha e frequenc sizef yo d qualit san logsf o y , besides sizd ean qualit sawf yo n timber produce sawyed dan r skills, such variations reportee dar o extraordinaryto t no . Nevertheless, these widely varied resultlimitef o e sar d milo t le managementus . Althoug e effecthth recovern so wele yar l recognised, attempo littln r bees eo ha tn mad ecategoricallo t thur fa s y study them. Accurate assessmen f sawto n timber yield s criticai s efficiene th o lt t operation of a mill which determines to a large extent the mill's profitability. Conventional sampling techniques are difficult to apply in mill studies. Most mills woul t entertaidno n properly designed mill o perceivestudiet e du s d disruptions. On the other hand, a 100% sampling is both unpractical and costly. Moreover, influencing factors such as the distribution of log diameter and sawn Journal of Tropical Forest Science5 5 (1)- 2 8 :4 49 output vary not only between mills but also within a mill. Accurate assessment of sawn timber yield onln e achieveca sy b f sucdi h nuisance variablee b n ca s minimized. It is therefore useful to categorize the output of sawn rubberwood, and adop simplta accuratt eye e approac asseso t h yieldse sth . This paper describes studies carried out at a rubberwood sawmill in Peninsular Malaysia. A relative frequency weightage approac proposes hi achievo dt betteea r assessmenf o t overall average output.

Mill description

The mill was equipped with five band saws without carriages. One of these saws was used as a head rig to break down logs into halves. Logs of diameter smaller than 175 mm (7.0 in) were generally not sawn at the headrig. The other four saws served as resaws, and converted the halves as well as small logs into sawn timbe variouf ro s sizes. Figur showe1 simplifiea s d layoumille th f . to Journal of Tropical Forest Science 2(1): 48 - 55 50

Sawn timber resawe th f sof were separated intbatcheso o tw generall e on , y free of defects, and the second containing timber having some defective portions lattee cross-cue b Th . ro t batc s t hwa int o shorter length frof so 3 m0. to 1.5 m (1 to 5 ft) to remove most of the defects to upgrade them to the same fule gradth l lengts ea h pieces sawe Th n. timbers were then tallied according o lengtht sizesd an s . Rubberwood sawn timbe generalls i r y being sol lotn di s of random lengths wels ,a rando s la m widths.

Study procedures

Diameters of both ends of each log were measured as the logs were being rame mille th th t p fea , o leadindt e headrigth o gt . Diameter measurements were rounde nearese th do t t half inch. Thi conforo t s i milmo t l practices where measurement stile Imperiasn ar i l l Units. Since rubberwood logs wer crossl eal - ) lengthft 6 ( standaro t m , measurin t 8 cu 1. d f go lengt s deemewa h d unnecessary. Each log was also marked at both ends to facilitate tracing loge th s through subsequent operations. As it was costly and impossible to cover all the four resaws simultaneously mille th ,t sawa n timber talls recordeywa t onle resawa d on y . e Sincth l eal resawe samth ey s b headrig wer d fe convertined an , g flitche sawo st n timber according to a single cutting bill, the data collected from a single resaw would be a good representative of all the resaws. In addition, since flitches from a t differena singl p u mighg elo d t en t resaws, sawn timber recordes tallwa y d onl r logyfo l sflitche al whic d sha h processe same th et a resawd . This approach would generally filter out variations introduced by different resaws d differenan t sawyers, which otherwis distory ee ma data th t . Sawn timber tally and subsequent computation were based on nominal sizes of the lumber which smallee ar r thae actuath n l sizes (Anonymous 1984). Although this will indicate lower recovery than actually achieved, this approach was chosen to conform to sawmill practices.

Results and discussion

Distribution of log input into the sawmill recorded during the studies and classified by the logs' small end diameters is as shown in Figure 2. To isolate the variations introduced by log sizes, subsequent analysis grouped data into diameter classes according to the logs' small end diameters. The diameter classes comprised three classe equaf so l class intervals representing about three quarter inpue classeth o f tstw o logsd containinsan e extremelgth y smald an l extremely large logs respectively. Likewise differen2 1 e th , t sawn timber sizes Journal of Tropical Forest Science 2(1): 48-55 51

produced by the mills were grouped into three lumber size categories based on their thicknesses othey An r. mean categorizinf so lumbee th g r sizes, although may improve the accuracy of the analysis, will result in unpractically large numbe f categoriesro compositioe Th . thesf no e categorie shows i Tabln i e 1.

5 22 0 20 5 17 ISO 250 275 3000 325 350 375

Small end diameter (mm)

Figure 2. Composition of log input into the mill classified by small end diameters

Table 1. Composition of lumber size categories

Lumber size Size of sawntimber included category (mm)

Size 1 5 4 x 5 4 , 65 100x x 0 5 ,6 10

Size 2 30 x 30, 30 x 50, 30 x 75, 30 x 150

Size 30 15 100 125x x x 5 5 ,5 2 , 2 2 , 65 x 5 2 , 50 2 5x

The yields of sawn timber per log categorized by diameter classes and size grouping showe sar Figurn i . While3 e there were significant differencey (b s Scheffe's metho 0.0e th 5t da level r sawntimbefo ) r output siz betwee1 e l nal Journal of Tropical Forest Science 2(1): 48 - 55 52

diameter classes, no significant difference was detected for the other two size groupings e usua.th Thio ls mainlt practicwa se du y recoverinf o e g larger sizes (Size 1) first before attempting to recover other smaller sizes (Sizes 2 & 3) from sawing the logs. As the number of pieces of large sizes that could be recovered from a log is largely governed by the size of the log, the larger the log biggee th volume r th wil e Sizlf b eo thae1 t coul recoverede db majoA . r portion of the large logs were recovered into the 'prime' sizes (Size 1). Small sizes (Sizes 2 & 3) were recovered from the remaining portion. For the smaller logs, very little size 1 lumber could be recovered, hence the major portion had to be converted into smaller size lumber. This contributed to the lack of significant difference between the diameter classes on yields of smaller size lumber. Figure 3 also shows the total volume of lumber that could be recovered from each log categorized by its small end diameter. Average sawntimber output per log varied from 0.01 m3 (0.36 ft3) for logs having small end diameter smaller than 200 mm 0.0o t (8.(2.73 ) 8m 0 in 4 logr fo fts3) having smal(14. m diameted m 0 len 0 35 f ro d largerinan ) . Statistical comparison using Scheffe's method indicatee th d outputs were significantly different at the 0.05 level for all classes except those from logs of smaller end diameter less than 250 mm (10.0 in). These significant differences were introduce lumbey db Sizn ri categore1 volume th s ya e outputs in the other two categories did not show any significant difference.

<200m m

200 MM - 237 fl

6.17 250 mm - 2«7 « | 300 on-| . 337 > 8 ^ I

•"«„,

AAAAA Size 2

Figure 3. Yields of rubberwood sawn timber categorized by diameter classes and lumber size groupings [bars with the same alphabet(s) within each size grouping are not significantly different by Scheffe's method at 0.05 level]

As logs are typically traded in volumetric units such as m3 or ft3 rather than number of logs, the analysis presented above may not be familiar and Journal of Tropical Forest Science5 2(1)5 - 8 4 : 53

understood by sawmill management. Recovery of sawntimber is usually reported as percentage of input log volume, with the log volume computed using Smalian's formula:

volumg Lo e=

where Dl is the large end diameter, D2 is the small end diameter and L is the length of log. Figure 4 shows the recovery of sawntimber categorised by diameter class and lumber size. Logs of the first two diameter classes (those having small end ) havdiametemm e 0 significantl25 < r y smaller percentage recoverie f sizso e1 and significantly larger percentage recoveries of Size 2 lumbers. There was, however significano n , t difference amon percentage gth e recoverie f Sizso 3 e lumber from logs of all diameter classes. As would be expected, results of comparin overale gth l recoverie e diameteth r fo s r classes followe same dth e pattern as those for Sizes 1 and 2 categories, the no significant difference in Size categoroverale 3 o th effecn n d o l ty recoveriesha trende Th . s indicatey b d these results reinforce the explanation presented for the trends suggested by the volume recovered per log.

32.22 Diameter classes 7 30.00 . fX> yj < 200 mm / 29.08 ————— * * B [\\\1 20 023- m 7m / d y X l~ 2587 250 -287 mm R< / 7~~ *\< / 25.00 - J 300 [ - 337 mm / 23.94 •yC .X / EZ1 > 350 mm ^ / 1.22 § 20.40 / 20.00 - 18.36 /

15.00 _

12.12

/ W//////////A \\\\\\ v

10.00 _ 8.35 / ^rn '•" 6.58 5.34 5.39 5.00 _ / fc-S 1 2.61 I| / 0.00 _ ^/////////Al / li~a_l AABBB AABBA B A A A A A AB AB AB B Size 1 Size 2 Size 3 Overall

Lumber size category Figur . Percentage4 e recover f rubberwooo y d sawntimber categorize diametey b d r classes and lumber size groupings [bars with the same alphabet(s) within each size grouping are not significantly different by Scheffe's method at 0.05 level] Journal of Tropical Forest Science 2(1): 48 - 55 54

Weighted means approach

o computT average eth e recover mille th r a ,simpl yfo e average basen o d total log volume and total sawntimber produced is frequendy used by saw millers. This method, although simpl computeo et fror fa ms i ,accurat s ea sawe th n timber output depends substantiall n size o ylogf o s indicate sa d above. Without knowin frequence gth y distributio sizeg lo s e sucth f hno simple estimate n onl accurate ca b sy mila r le fo thas uniformlha t y distributeg dlo inputs. Stratified random sampling, which gives a better estimate of means than random sampling (Cochran 1963), would be a good solution to this problem. However, due to the perceived disturbances the mill managers would resist conductin ga properl y designed experiment closA . e approximatioo t n statified random sampling which would improve the accuracy of estimation of population mea estimatins ni summin y gclase b th s p meanu g s weighted with clase th s relative frequencies (Dixo Massen& y 1969) computatioe Th . n formula is as follows:

Estimated population mean,

N, y- - xi

N

h where i equals to 1,2,...... C. N, is the total number of logs in the clasit s in the population, N is the total number of logs in the population, C is the total number of classes in the population and x, is the sample mean for logs in the ith class.

Table 2. Work example to estimate expected yield using the weighted mean approach

(1) (2) [(l)x(2)] Diameter Relative Estimated average Weighted average class frequency volume recovery volume recovery (mm) (m3/100 logs) (m3/100 logs)

Batch Batch Batch Batch 1 2 1 2

<200 0.20 0.10 0.42 0.084 0.042 7 2023 0- 0.25 0.15 0.71 0.177 0.106 250-287 0.30 0.30 2.27 0.680 0.680 300-337 0.15 0.25 2.69 0.404 0.673 >350 0.10 0.20 3.12 0.312 0.623

Weighted total 1.657 2.124

(For conversio Imperiao nt l Units conforo t , normamo t l mill practices, calculat value e th 35.32 y eb ) Journal of Tropical Forest Science 2(l):48-55 55

above Th e formula coul appliee db estimato dt meae eth n recovery r yieldo , s of certain groups of lumber produced for a batch of logs with known composition (frequency distribution). This approach coul extendee b d o dt compute expected yields fro batcma logf h o knowf so n composition using yield data categorize g diametelo y dlumbeb d an r r sizes that have been estimated d maintainean e mill worth A y . b kd exampl shows a e s ei TablTh n ni . 2 e example also illustrates the substantial effects introduced by a slight change in the input log composition.

Conclusion

Recover f sawo y n timber from rubberwoo significantle founs b dwa o t d y affecte sizey b df bot o s e log d sawth h an s n timber. Categorizing yieldy b s diameter classe sizd sean groupings could provid betteea r too assistinn i l g mill management in production planning and control. Such a data base would assist e milth l manage computinn i r mucw g sawinwoul y e ho whah b h d t a an tdgge batc f log o hknowf o s n composition (size distribution) thereford an , e ablb e e to plan accordingl a buildesireo yt p u d d inventor r schedulinyo g production to meet delivery dates. Yield f sawo s n timber categorize diameterg lo y d b d lumbean s r sizes could usee b d together wit weightee hth d means approac estimato t h e expecteth e d sawn timber output and assist in better planning and control of sawmilling operations.

References

ANONYMOUS. 1984 Malaysiane Th . Grading Rules r sawnfo hardwood timber (1984 Ed.). Malaysian Timber Industry Board, Ministr Primarf o y y Industries. COCHRAN, W.G. 1977. Pp. 90 - 95 in Sampling Techniques (3rd Ed.) John Wiley, New York. DIXON, W.J. & MASSEY, F.J.Jr. 1969. Pp. 184 -187 in Introduction to statistical analysis. Mc Graw- YorkHillw Ne , . GAN, L.T., HO, C.Y & CHEW, O.K. 1985. Rubberwood: Sawntimber production and recovery studies. Pp. 97 -122 in Proceedings of the Second Rubberwood Seminar. November 19-20, 1985 Forest Research Institute Malaysia, Kepong, Malaysia. , K.SHO & . , K.T. 1982. Processing of rubberwood. Malaysian Forester 45 (3): 290 - 298. LOPEZ, D.T., MOHD. ARSHAD SARU & TAN, A.G. 1980. Rubberwood - a study of recovery in production mills. Malaysian Forester (1)3 4 : 74-80.