Calculation of timber yields from North Queensland rainforests

TECHNICAL PAPER
N
NO. 47
CALCULATION OF TIMBER YIELDS
FROM
NORTH QUEENSLAND RAINFORESTS
BY
R. A. PRESTON AND J. K. VANCLAY

ISSN 0155-9664
CALCULATION OF TIMBER YIELDS
FROM
NORTH QUEENSLAND RAINFORESTS
BY
R. A. PRESTON AND J. K. VANCLAY

PREFACE
This paper has been prepared in the interests of community discussion of recent proposals
for World Heritage Listing of the wet tropics of north Queensland. It presents the results of
studies carried out in 1985, which provided the basis for sawmill allocations for the period 1
October 1986 to 30 September 1991. Consequently, the paper does not discuss in detail the
subsequent reduction of the allowable cut for Zone 2 (Innisfail-Tully) from 14000 toL2320
cubic metres per year, in response to damage from Cyclone Winifred. Details of this
reduction are reported in Preston (1987).

ABSTRACT
Calculation of timber yields from north Queensland rainforests indicate that the long term
average yield is in the vicinity of 63 000 cubic metres per year, and that an allowable cut of
60000 cubic metres per year should apply for the period 1986-1991. These calculations
aPply to the 158 000 hectares of Crown land managed for timber production between
Townsville and the Daintrce River.
Estimates were prcpared using cutting cycle analysis, and incorporated a number of
innovations made possible by advances in computing technology. These included simulating
the growth of individual plots rather than of stratum averages, and the use of a dynamic
growth model which accomodated stand dcnsity, composition and site quality.
INTRODUCTION
The tropical rainforests of north Queensland are one of Australia's most valuable natural
resources- They are highly valued for their conservation significance and ecological
diversity, and sustain a harvest of valuable cabinet, vcneer and structural timbers
The study area includes all coastal and hinterland tropical rainforest between Townsville
and the Daintree River. This region is divided into five allocation zones (Figure 1) which
form the basis for regulation of timber supplies. This study appraises the allowable cut of
rainforest timber from the state forests, timber reserves, and other Crown lands in each
allocation zon'e for the pcriod 1 Octobcr 1986 to 30 Scptember 1991. Yields will be
reviewed prior to 1991 in accordance with Departmental policy to review the allowable cut
in all native forests every fivc years.
OVERVIEW OF YIELD REGULATION
The process of yield regulation comprises three major stages. The long term avcrage yieldn
is first calculated to give a sound pcrspectivc of the future resource position. This prlvides
the basis for determination of thc allowable cut, which is set for a five year period ior each
allocation zone. Within each zone, allocations of timber are ofrered to sawmilling firms
entitled to crown supplies, so that they sum to the allowable cut.
Calculation of the long term average yield entails several basic operations:
o the area of forest capable of producing timber is determined. Forest subject to spccial
management (e.g. scientific areas16, bufrer stripsa along creeks) and inaccessible or
unproductivc forest is excluded.
o a detailed inventory of the cxisting forest is prepared by measuring temporary plots and
recording the species, size and merchantability of each tree within ihe plot.
22,16,4 Sce glossary for description of tcrms idcntificd by supencript ngmbers.

- 2 -
" l
!(tl
! r i
bl t l
t ' l q l
- l - l
-l- | '-*
I STUDY AREA I
lsca le
abt .1:3 ooo ooo
II
b
"r,0" I
CAIRNS
-.\
Mareeba
@
Athert
@
a
Ravenshoe
a @
__l_,*i- o
,ri. N
Rainf orest
Sta te Fores ts
Study Area
- - -@ Al locat ion Zones
It II
. n
1 9 0
\
\.
i
t.ll::Tlfl'
Figure 1. Lccation of the study area

- 3 -
o the future condition (number, size and merchantability of trees) of each inventory plot is
predicted by simulating the growth (diameter increment, mortality and recruitmentls) of
the forest over time.
o at selected intervals, a timber harvest is simulated to indicate which stems would be
removed in logging, and to predict mortality to the residual standls arising from fell ing
and snigging damage.
o the anticipated harvest volume is then calculated using volume equations2l.
o the growth and harvesting of each inventory plot is simulated through several cutting
cyclesT to ensure the continuity of future timber harvests.
This procedure provides an estimate of the timber yield which can be sustained under the
specified management regime and assumed economic conditions. Timber harvesting can be
sustained at any level not exceeding this yield.
CALCULATION OF TIMBER YIELDS
Area Estimates
Area estimates are an essential ingredient of the resource forecast, and due account must be
taken of unproductive land such as rock outcrops, stream buffers, and other areas which
cannot be logged. To facilitate the preparation of area estimates, a computerized area
information system was commenced in 1978, and was used in the present calculation. It was
based on the New South Wales FORINS System (Hoschke and Squire L978), and records
management information at each 1000 metre Australian Map Grid (AMG) intersection
within the study region. Although this provides relatively "coarse" information (in that the
best estimate attainable is to the nearest 100 hectares), it is an efEcient mechanism for
dealing with large amounts of data over extensive areas of forest. This system was used to
calculate the area utilized for timber production within the study area (Table t), and within
each allocation zone (Table 2).
Data were drawn primarily from timber management maps and included tenure, allocation
zorre, management intention and logging history. These maps were prepared by field staff
during the period 1978 to 1980 using historical timber sales records dating back to the mid
1950's, and have been regularly updated. Where no records were available, estimates of
accessibility and productivity were prepared from interpretation of 1 : 25 000 scale aerial
photographs.
The gross productive areal (or mapped area) of rainforest was determined by multiplying
the number of sample points in each allocation zone believed to be availablc, accessible and
productive, by 100 hectares per point.
The productive areal was calculated from the gross productive area by applying a correction
factor to adjust for inaccessible or unproductive rainforest misclassified as accessible and
productive on management maps. This factor was determined from inventory by calculating
the proportion of plots located on contiguous areas of inaccessible rainforest, and was found
to be 0.839 for areas logged before L970, and 0.960 for areas logged since Lg7O. The
difference reflects the improved accuracy of information arising from more intensive
management in recent years. A check was also made to determine if any land outside the
gross productive area had been misclassified (i.e. actually productive). About 98 percent
was classified correctly, and about two percent was marginal. Misinterpretation of
inaccessible and unproductive land was therefore assumed to be negligible, and no attempt
was made to establish inventory plots in these areas.

I
I
1
1
:
I
- 4 -
Trblc 1. Rdnlorcrt tcnurc - Towusvlllc lo lhlntrec Rlvcr.
Arcr (hr)
Tenurc rnd mrnegenent Subtotel Subtotel Totd Pcrccnt
Areas where logging is excluded
1. Special management areas 54000
2. Inaccessible and rurproductive
forest and bufrer strips 227000
Produstivc area
281 000
158 000
Statc forcsts, tirnbcr reserveg
and other crown lands 439000 75
National park 105000 18
Frechold 42WO 7
Total 586000 100
TeHc 2. Arcr ol relnlorcrt by rllocrtlon zonc.
Allocrtlon zonc Productlve erer (hr)
1. Ingham- Ravenshoc 52500
2 Innisfail - Tully 31900
3. Tablelards 27 600
4. C-airns - Kuranda 24300
5. Windsor - Lcwis 2l7W
Totat all zoncs 158 000
9
39
48
27

- 5 -
The nett productive areal (or efrective area) was used as the basis for yield calculations, and
was determined by adjusting the productiye area for small areas of rock, road and stream
bufrers. A netting factor of 0.956 was estimated from inventory data by calculating the
proportion of stems on each plot which were assessed as inaccessible or unavailable.
At any given time the loggable areal of rainforest is less than thc nett productive area. This
occurs because areas currently considered uneconomic to log may be logged at a later date.
To take account of this in yield calculations, any inventory plot (and its associated nett
productive area) that would not produce a minimum yield of five cubic metres per hectare at
the midpoint of a cutting cycle was excluded from logging in that cycle, but would remain
eligible to be logged in subsequent cycles. This procedure reflects current resource and
operating conditions, and generally ensures a viable yield of. 12 cubic metres per hectare of
gross productive area. An estimate of the loggable area may therefore be obtained by
multiplying the nett productive area by the proportion of inventory plots yielding more than
five cubic metres per hectare at the timc of (simulated) logging. During simulation, about
10 percent of the nett productive area was only logged every alternate cutting cycle.
In the Department's 1981 allowablc cut calculation, thc loggable area was used as the basis
for calculations, and was defined as the area able to be logged in each of thc next three
cutting cycles. This was calculated as 0.91 of thc nett productive area, and derived from the
observation that tr 7 out of 181 inventory plots then available were considered unproductive
or loggable only within 100 years.
Inventory
Inventory data collected during the period 1978 to 1985 were used in the current
calculation. Temporary plots were establishcd at prcdetermined 1000 mctre AMG
intersections, selected from the area information systcm using stratified random sampling
with logging history as the primary stratum. As most logging of accessible and prodortir"
virginle rainforest will be completed in thc near future, "tl inventory plots in virgin stands
were excluded from these calculations. Thrce difrerent types of plot were used over this
period (Table 3).
Trblc 3. Plot tyFc end nmpllng lntcnrlty.
Plotr pcr rllocrtlon zonct Totet
Yeer Plot typc 1 2 3 4 s p t o t r
1978-81 Point sarnplc
1983 Point sample
1984 Fircd erca
1985 Fircd/point
Totals 109 34 10E 49 t9 319
34 15 98 t9 5 t7r
4 4 0 0 0 0 4 / i
2 1 1 5 8 2 6 t 4 E 4
1 0 4 2 4 0 2 0
t plots in virgin stard$ excluded

6 -
During the period 1-9lg to 1983, clusters of ten point samplesl4 were employed, in which an
optical wedgel3 with a basal area factor3 (BAF) of 10 square metres per hectare was used,
and all stems exceeding three centimetres diameter (at breast height or above buttress, over
bark) w€re measured. During 1984, fixed area plots were favoured. All stems exceeding 30
cm diameter were sampled on a half hectare plot, and stems exceeding 20 cm diameter were
sampled on a 0.725 ha sub-sample. During 1985, a small number of plots were established
using a new approach. These plots sampled all stems exceeding 40 cm diameter over one
hectare, and used four point samples (BAF 2.3 sq m/ha) to sample stems 3 to 40 cm
diameter.
The species, diameter, merchantability and visual thinning2o was recorded for,each tree on
all plots.
There is no compelling statistical advantage in the use of any of these plot types in
preference to the others for description of the current stand or to provide forecasts. For
quantifying the existing stand, there is some advantage in having a large heterogeneous plot
to minimize between-plot variation. Conversely, for simulation studies, a small
homogeneous plot is more appropriate. In practice, cost factors and the preference of field
staff are of greater consequence.
Growth Model
An integral part of yield forecasting is growth prediction. Growth models for plantations
and for monospecific forests have become sophisticated and highly accurate. Rainforests
comprise hundreds of species, posing a much more difficult challenge. Notwithstanding this,
a dynamic growth model for rainforests has been developed and was used in the present
study. The model is described by Vanclay (1987) and the data upon which it is based is
summarized in Queensland Department of Forestry (1983).
The growth model has functions for diameter increment, tree mortality, and recruitment of
new trees into the stand. Each of these functions takes into account the site quality, the soil
parent material, the stand composition and density, and the size of the individual trees. Site
quality, expressed as good or poor, is determined objectively by assessing soil parent
material, presence of indicator specieslr , cstimated residual volume after visual thinning,
and average log length (Vanclay, in press). The model recognises several soil parent
materials: acid volcanic, basic volcanic, coarse granite, Tully granite, sedimentary-
metamorphic and alluvial-colluvial. In general, coarse-grained granite-derived soils support
the most productive forests.
It is impractical to develop individual functions for each of the several hundred tree species
represented in north Queensland rainforests. Accordingly, commercials species were
grouped into four growth groupslT according to their growth habit:
o large fast growing;
o large slow growing;
o (comparatively) small fast growing; and
o small slow growing.

7 -
Practical necessity required the use of a single group for all non-commercial specics. This
grouping has ecological significance, with gap opportunists contained in thc large fast group,
pioneer species in the small fast group, and shadc tolerant spccies in the slow groups.
The model does not retain the species idcntity of each trce, but cmploys species groups
formed from trees having membership of the same growth and harvesting groups17, and
using thc same volume equation. The identity of the inventory data (as distinct from
predicted recruitment) is also retained. This cnables the flagging of yicld forccasts which
include predicted recruitment.
The growth functions cmployed in the model were fitted to thc data using linear rcgrcssion
to ensure that estimatcs are unbiased. Because diamcter increment data wers sparse for
large trees, the functions were constrained to be asymptotic to subjectively determined
maximum diameters (Figure 2).
Because the 1984 data sampled only stems exceeding 20 cm diameter, rccruitment was
predicted at 2O cm, diametcr. However, whcre stems smaller than 20 cm werc measured,
these data were utilized, and the rccruitment function was activated only when obscrved
small stems exceeded 20 cm diameter. Stems smallcr than 10 cm diameter were ignored, as
with point samples (cspecially with BAF 10 sq m/ha), thesc reprcsent very largc numbers of
stems per hectare, and may convcy an unrcalistic impression of stand composition.
Harvestlng Model
Prior to logging, trees thought capable of producing a merchantablel2 log arc marked for
removal in accordance with Departmcntal guidelines (Appendix 1). When felled, some
stems reveal defects not evident when the trec was standing. Depcnding on the amount of
this defect, the log may be classified as compulsory6 or optional. Only compulsory timber is
debited to the sawmill allocation.
Essential components of the harvesting model are the logging rule, which indicates stems to
be removed; an allowance to predict the compulsory proportion of the logged stems, and a
damage function, which predicts mortality arising from felling and snigging damagc to the
residual stand.
To simulatc harvesting, species wcre placed into nine harvcsting groups 17, based on the
treemarking groups defined in Appendix 1 and on merchantability (Table 4). Two diameters
may determine whether a tree is selected for harvesting. Trees smaller than thc cutting
diametere -"y bc removed only if they exceed 40 cm diameter and can be expected to die
prior to the next logging. Stems abovc thc cutting diametcr and up to the retention
diameter9 will generally bc removcd unless thcy have exceptional form or vigour, or are
required as a seed tree. Stcms exceeding the retention diameter must bc markcd for logging.
Tablc 4 also shows that the current (1986) treemarking rules are intermcdiatc to logging
rules A and B, but most closely resemble rule A. Thc logging rule describes thsse
treemarking guidelincs as a series of simple lincar relationships which predict the
perccntagc of stems harvested (Figure 3). Two difierent logging rules werc evaluated.
Logging Rulc A reflccts removals under thc 1985 trcemarking guideline$ Logging Rule B
incorporates a reduced retention intervale.
The logging rules were prepared from visual thinning assessments on the 1985 inventory
plots. Thesc data werc used in prcfcrence to other alternatiyes, as the 1985 inventory plots
were large (1 ha) and werc expressly established by expcrienced field stafr in stands logged
before 1970.

I
I$
----l
iI
7t
I
I
1
l
l
1
o
-f
F{
o
c\r
t{
F{ C{ 6a <r lJ)
aaaae( 5 ( r ( , ( 5 ( ,
o
e
F{
E
c)
O 9 .
@ o
+,
o
E
d
o X
CO
6
og.
qt
o
cd
m
et
6 N
o t r
t a -( o
Fr
(g
f," i
f i8
t ! o
.9p.:t r E
-_. ti
8 ( ,
f i > ,
a =
+t .tr
O F
H -
.l1 >)15 +,
et
=
@
o
a
t-.
oo
A
o(O
o
-f
F{
o
c\r
F{
o
o
F { n
O
t{
e o
O + ,
o
c
d
o A(g
qt
o&r
r-
d
a
d
FA
d
F
-\
sl
d E
9 o
.z (o
E €dt cn
n o
E . :b o t r
:ry qttri rt
-
\r'
n o )
-
t!'t
. l .{?t) F
+f rX
O \ J
6 *
+,
.,{
r-{qt
t
@
c)
+,
..-a
a
Ioo
o\/
o(o
e
s
o6r
e
o
F{
E
c)
o \ /@ b
+,
o
E
o . g(o l.\
G
ou
=a
6TA
cd
E
6
?a
-\
C\
6 E9 o
t 6 lq -
F l =g 8
l Q o
+.r
l ' Eg q t
l-a 3r
. ( ,
B o
- u ,
' 6 i
! O
E O
t r *
+,
.-
F{
6
=
@
o
+,
.-
m
E
o
o(,
o
CO
o
sf
t-{
o
6l
F,{
e
oF { E
c)
t-.
-r O)
= + ,w c )
F
d
a
e(0
qt
o)
tr
F{
d
TA
q,
m
et
6crt
o t r
l i -{ o
r< 6l
a i G
t t 4
:t .d
.!t o
* . 4
' e t
Q . EF l t r
i q ,
ut 9.g ( 5
id .?.
E F
O GB F
.5
d
=
@
o
+,
a
t .
o
og
a? 6-{ n
unuuB/rxo luaruoJoul JalauBIC
T r ' ' 6 : n
unuuB/uro luauaJoul JalouBlc
6? 6.r ":
unuuB/uto luaualoul JeleuBlo
"3 6: ':
unuuB/luo luoualoul JalaruBlc
r/
II
\
\
f/ /?/7'�
( { \
\
//
I
I
I
\
\
\

cl
-g
cD
o
I
o
6
E
o
.I
G
o
a
E(l,
€
o
a
- 9 -
Figure 2. (p. 8, opposite). Diameter rncrement Functions.
too
90
80
70
60
50
40
30
20
10
o
50 60 70
D&uneter (cm)
Figure 3. Logging Rule
Otameter (cm)
Figure 4. Damage Fuuction
CD
C
o
a9 e o
-s
E
o
o
o
€ 1 0
d
E
o
o
o o
a
Rub A
Rub B
I
8 i
I
I
I
I
I
I
I
I
I
I
I
!
lt"*""ro.dl

- 1 0 -
TeHc 4. Ilrrvcst groulx
Retcntlon lntcnel
Hrrvcrt ltccmrrklng
SrouP group
Cuttlng Logglng
dhmter rulc
A
Logglry 19td
rulc trccmrklng
B 3uldcllncr
1
2
3
4
5
6
7
I
9
Qld walnut
A
A
B
B
c
D1
a2
Non-cornmercial
0
20
20
20
20
20
0
0
1 0 0 0 0
80 20 15
70 20 15
70 20 15
70 20 15
60 20 ls
6 0 2 0 0
:'
Logging damage functions were expressed as a series of simple linear relationships fitted by
eye (Figure 4). Data w€re derivcd from nine rainforest sitcs (three recut and six virgin)
sampled before and after logging, using clusters of 60 point samples (BAF 10 sq m/ha) on a
30 metre grid.
It was determined from Departmental logging records that non-compulsory stems comprised
thrce percent of harvest group 1 to 3 stems, and seven percent of group 4 to 8 stems.
Yolume Equatlonr
Reliable rainforest volume equations are available in the form of two-way equations which
predict log volumc frorn tree diameter and log length (Vanclay ct al t987). However, sincc
forecasting future log lengths is unnecessarily complex and inaccurate, oDe-way equations
predicting log volume from diametcr arc required. One-way equations were developed, using
data from the 1978-81 inventory. Inventory data included cstimatcs of log lengths and
centre diametcrs (thc diameter over bark, half way along the log), and measurements of
diameter at breast height or abovc buttrcss, oyer bark. Although it is desirable to use
measured data for thc development of volumc equations, such data wcre availablc from only
a few geographic locations, which were believcd to bc unrepresentative of thc resourcc as a
whole. Thus the use of the geographically diverse inventory data was considered preferable.
Log volumes for thc inventory data were calculated using Hubcr's formulalo from the
estimated log length, and thc cstimated centrc diameter reduced by a standard (0.956) bark
thickness correction factor. Experience suggests that inventory stafr can reliably estimate
the log length on a standing tree, but that estimates of stem diameters higher up the bole
are less accurate. Thus a scaling factor was developcd to ensure that the one-way cquation
produced estimatcs consistent with the morc reliablc two-way equation. The resulting
volume equations are presentcd in Figurc 5.

- 1 1
r @OUP I SPECES. All. ZONEII
a OTHER SPECEq ZOI{E,t.
O ON€R SPECEq ON{ER ZONEII
,
o
atIoa tIoI
,
l '
,4r/ '
14lo
.5 lto 3g co 65 ?o tE c0
Obnrlr (Ercrrt Hlgh) ovr brrt (nm).
Figure 5. Volume equations
Yield Calculatlons
Existlng resource. Table 5 presents a description of the current standing loggable volume.
This volume would be removed under a tree-marked sale, and is not the iotal merchantable
volume. It is evident that the current loggable volume of about 2 million cubic metres
would be sufficient to maintain the present rate of logging for many years, even if it is
assumed that there is no growth.
Table 5 also illustrates how the loggable area may yary with the logging rule applied. A
comparison of the loggable areas given in Tablc 5 with the productivc areas given in Table
2 reveals that at present, only two thirds of the productiyc area could be logged.
Cuttlng cycle analysls. Cutting cycle analysis8 (McGrath and Carron 7966) has been widely
used for calculation of the allowable cut for irregular Australian native forests, and was
used in this study. Cutting cycle analysis requires a nett productive area estimate, inventory
data, growth and harvesting models and volume equations. It also requires a nominal
cutting cycle to be specified. For north eueensland rainforests, a cutting cycle of 40 years
was adopted for all zones.

- t 2
Trblc 5. Volumcc prercntly rvrll,rble.
Allocetlon
zonc
Estlmated Estlmated Estlmrted Averege
loggeble rversgc loggeble stem
lrel yleld volumc Yolurne
(he) (cu n/hr) (cu n) (cu m)
I-ogging rule A
I
2t
3
4
5
38 700
24400
23200
13 300
14 800
15
1 8
t 9
L7
18
599400 2.50
433 600 2.23
440500 2.43
226200 2.06
266 500 z.OL
Total 114 400 1926200 2.25t7
I-ogging rule B
1
2t
3
4
5
668 000 2.48
498 700 2.18
502000 2.38
273300 L.97
298400 t.97
38 700
25 300
23200
15 300
15 700
t 7
20
22
1E
19
Total 118 200 t9 2240400 L20
t estimates for 7nrc 2 priot to Cyclorrc Winifred-
Traditionally, cutting cycle analysis involves stratifying each allocation zone, computing the
average stand table and area within each stratum, and conducting the analysis on this
aggregsted data. As the actual time of harvest of any area cannot be determined, logging is
simulated at the mid-point of each cycle. Thus for a 40 year cycle, logging would be
simulated at 20, 60 and 100 years for the first, second and third cutting cycles respectively.
Computer software developed for thc current study allowed simulation of the growth of
each individual plot rather than the stratum average stand. The yield from each plot was
weighted by the corresponding nett productive area, stratified by soil parent material and
logging history (pre-1970, post-l970). Cutting cycle analysis was performed separately for
each allocation zone using only inventory data from plots located in that zone.