Forestry An International Journal of Forest Research

Forestry 2017; 90, 359–366, doi:10.1093/forestry/cpw044

Advance Access publication 9 September 2016

Form factor functions for nine commercial tree species in Bhutan

Jigme Tenzin1,2*, Tenzin Wangchuk3 and Hubert Hasenauer1

1
Institute of Silviculture, University of Natural Resources & Life Sciences, Peter-Jordan-Str. 82, A-1190 Vienna, Austria
2
Watershed Management Division, Department of Forests & Park Services (DoFPS), Ministry of Agriculture & Forests (MoAF), 11002

Downloaded from https://academic.oup.com/forestry/article-abstract/90/3/359/2605859 by guest on 06 September 2019

Thimphu, Bhutan
3
Royal Manas National Park, DoFPS, MoAF, 31101 Gelephu, Bhutan

*Corresponding author. Tel: +43 147654 4073; Fax: +43 147654 4092; E-mail: [email protected]

Received 19 May 2016

Standing timber volume in combination with the expected volume increment rates derived from volume func-
tions are essential for developing sustainable forest management plans. Tree volume cannot be measured dir-
ectly. It is derived from the diameter at breast height, tree height and a so-called form factor, which reduces
the volume of a cylinder to the actual tree form. In this paper, we test four different types of form factor func-
tions (Pollanchütz, Short Swedish, Meyer and F. Evert) for estimating total merchantable timber volume of
nine commercial tree species in Bhutan: Abies densa, Picea spinulosa, Pinus wallichiana, Tsuga dumosa, Pinus
roxburghii, Castanopsis tribuloides, Quercus glauca, Quercus lanata and Quercus lamellosa. The data for fitting
the form factor functions come from 395 felled trees. The resulting functions are evaluated using independent
validation data. Fitted statistics for evaluation include: root mean square error and mean absolute deviation.
Although all form factor functions performed similarly, we suggest that the Pollanschütz function because of
its consistency in the estimated form factors for all tree species. The evaluation of the calibrated form factor
functions by species exhibited consistent and unbiased predictions.

characterizes the shape of the tree (Burkhart and Tomé, 2012),

Introduction
whereas the stem taper is the relative rate of change in stem
Forests provide services that are important for sustaining rural diameter with increasing height (Larson, 1963; West, 2009;
livelihoods and national development, including timber produc- Burkhart and Tomé, 2012).
tion (Führer, 2000). With the increasing demand for timber, the According to Philip (1994), there are three different form fac-
correct and accurate assessment of forest growing stock (stem tor definitions: (1) the absolute form factor (based on cross-
volume) in combination with volume increment predictions sectional area at the ground level), (2) the normal or Pressler’s
derived from growth models are essential to ensure sustainable form factor (based on cross-sectional area at 0.9 of the total
forest management (Martin, 1981; Tarp-Johansen et al., 1997; height measured from the tip) and (3) the artificial form factor
Brooks et al., 2008; Hasenauer et al., 2012). The development of (based on cross-sectional area at breast height) which is consid-
the United Nations Framework Convention on Climate Change’s ered the most useful one. Different methods, such as form fac-
Reducing Emissions from Deforestation and Forest Degradation tor, form quotient, form point, taper tables, taper equation,
(REDD+) also requires accurate volume predictions for deriving taper curve and formula, exist to express the stem form to cor-
carbon stocks from our forests (United Nations Framework rectly determine the volume (Husch et al., 2002). If the form of
Convention on Climate Change, 2014). a tree can be accurately determined, then the volume can also
The volume of trees is commonly derived from the diameter at be correctly estimated (West, 2009). The standing volume can
breast height (d.b.h.), height (h) and a form factor (f); which may be calculated within a 0.2 per cent error range if calculated with
be seen as a reduction factor of a cylinder (with d.b.h. and h) to a form factor (Eastaugh, 2014) while an overestimation is evi-
the actual form of the tree (Husch et al., 2002; Akindele and dent if only the d.b.h. and h are used in a volume
LeMay, 2006; Adekunle, 2007). Stem form is an important com- equation (Hoyer, 1985; Socha and Kulej, 2007).
ponent for volume estimation (Pollanschütz, 1965; Colgan et al., In an effort to increase mitigation options against climate
2014) as trees differ in shape due to different forest manage- change and to meet the increasing demand for sustainable for-
ment practices (Larson, 1963; Ikonen et al., 2006), climatic/gen- est resources, Bhutan plans to bring large forest areas under
etic factors (Socha and Kulej, 2005; Socha and Kulej, 2007), sustainable forest management. This will require accurate esti-
species, age and d.b.h. (Avery and Burkhart, 2002) or by species mation methods for assessing forest resources, such as stand-
composition and size (Adekunle et al., 2013). Stem form ing timber volume and volume growth (Forest Resources

© Institute of Chartered Foresters, 2016. All rights reserved. For Permissions, please e-mail: [email protected].

359
Forestry

Management Division, 2013). In Bhutan, the demand for timber The purpose of this study is to develop form factor functions
has increased due to the rapid urbanization and construction for improving the volume predictions for five coniferous species,
activities (Forest Resources Management Division, 2013; i.e. A. densa (fir), Picea spinulosa (spruce), Pinus wallichiana
Department of Forests & Park Services, 2014). Approximately 9 (bluepine), T. dumosa (hemlock), P. roxburghii (chirpine) and four
per cent of the 2.71 million hectares of total forested area are broadleaved species, i.e. C. tribuloides, Quercus glauca, Quercus
under sustainable management (Department of Forests & Park lanata and Quercus lamellosa, the primary commercial tree spe-
Services, 2014). cies in Bhutan. We (1) compare the performances of different
Currently, tree volume is estimated with a general volume form factor functions and (2) select the most suitable one for
equation using only d.b.h. and h. The results by species are avail- estimating stem volume.

Downloaded from https://academic.oup.com/forestry/article-abstract/90/3/359/2605859 by guest on 06 September 2019

able as so-called local volume tables. The general volume
equations were developed during the forest inventory carried
out in 1976 (Laumans, 1994) and the local volume tables were
developed by the Forest Resources Development Division,
Material and methods
Department of Forests (Forest Resources Development Division, Data
2005). The local volume tables are currently being used in other
The research material consisted of five conifers and four broadleaved
areas of the country. However, it is speculated that the existing
species along different elevation gradients from different regions of
volume equations may overestimate/underestimate the vol- Bhutan. Data from 395 trees were used for model calibration. The
ume, which suggests that the need for a systematic review of data were collected through destructive sampling that was carried
the results and the equations that derived them (Rosset, 1999; out in timber harvesting areas as well as outside when the required
Whitfield, 2001; Forest Resources Development Division, 2004). species were not available on the harvesting sites. For each sample
For instance, the existing general volume equations for tree, the d.b.h. at 1.37 m and the total tree height after felling was
Castanopsis tribuloides and Pinus roxburghii estimate a negative recorded. Diameter measurements were recorded at 1, 3, 5 and 7 m
volume for the smaller trees; the Tsuga dumosa and Larix grif- and then at intervals of 6 m along the rest of the stem until a 5 cm
fithiana are two different species which share the same volume diameter was reached at the top. For the hardwoods, the main ter-
equation; there is also only one volume equation for all minal leader was chosen for the diameter measurements along the
Castanopsis and Quercus species; in the local volume table, stem. The measurements were recorded up until 5 cm top diameter to
address the fact that in Bhutan, small round wood is also harvested.
Castanopsis is grouped under the category of other broadleaves.
Additional information, such as slope, topography and aspect were
As an alternative to the existing volume equations, Schieler
recorded for each tree.
(1991) calculated the volume of Abies densa in Central Bhutan We obtained separate datasets for the four tree species (A. densa,
by including the form factor in a volume equation in addition to P. spinulosa, P. wallichiana and T. dumosa) from the biomass
d.b.h. and h. Whitfield (2001) suggested that adopting the equation development project of the Department of Forests & Park
approach by Schieler (1991) to all tree species in Bhutan Services, Ministry of Agriculture & Forests. This additional dataset
because volume calculations derived from form factors enhance (Table 1) was used for validation so that based on these validation
the accuracy of the volume predictions (Adekunle et al., 2013). results, the best form factor function can be selected.

Table 1 Summary of the data used in the study showing the sample size and the mean, range and standard deviation of d.b.h., height, elevation
and slope for all species

Species n d.b.h. (cm) h (m) Elevation (m.a.s.l) Slope (°)

Mean Range SD Mean Range SD Mean Range SD Mean Range SD

Data for calibration

Abies densa 30 32.8 5.0–60.3 16.15 21.7 5.8–33.3 8.52 3597 3555–3600 11.42 29.0 20–30 2.42
Picea spinulosa 32 32.9 5.6–75.3 18.83 23.5 6.1–39.7 10.23 3604 3600–3670 17.22 30.6 30–37 1.90
Pinus wallichiana 34 37.0 4.6–95.8 25.15 24.0 6.2–51.5 11.93 – – – – – –
Tsuga dumosa 46 34.6 6.4–107.4 22.05 20.9 5.8–43.3 9.10 3154 2760–3429 195.86 21.9 2–48 13.68
Pinus roxburghii 31 43.4 8.5–77.8 21.29 26.7 5.6–41.9 10.34 – – – – – –
Castanopsis tribuloides 31 42.3 6.0–89.0 23.78 20.4 5.6–34.4 8.37 2268 1970–2590 202.45 37.0 5–60 15.59
Quercus glauca 31 39.1 4.2–77.2 22.95 18.9 5.9–36.0 7.63 2279 1950–2694 242.45 35.7 15–65 13.52
Quercus lanata 31 42.4 6.6–77.5 21.84 16.0 7.2–27.9 5.18 2091 1790–2506 142.81 32.0 0–60 12.91
Quercus lamellosa 20 46.0 24.3–72.5 13.07 20.9 15–29.0 3.60 1909 1895–1935 9.51 18.5 15–44 6.64
Data for validation
A. densa 15 28.6 7.2–64.0 18.65 19.1 6.5–37.6 10.85 3231 2780–3430 156.69 29.6 14–47 14.86
P. spinulosa 26 44.9 7.1–95.0 29.72 27.7 6.8–46.0 14.02 3154 2920–3340 146.19 23.8 3–46 12.1
T. dumosa 43 42.7 8.3–86.5 21.28 26.8 7.9–46.4 9.42 2954 2820–3141 84.68 38.3 5–72 19.35
P. wallichiana 25 32.6 5.1–79.1 23.28 23.2 6.1–45.3 13.40 3013 2770–3860 255.74 15.6 3–22 3.72

n = number of trees; d.b.h. = diameter at breast height in centimetres; h = tree height in metres; – = information not available; SD, standard
deviation.

360
 Form factor functions

Form factor functions Analysis

Standing tree volume (V) can be derived from the d.b.h., h and form
factor f:
Calculation of form factor
We started our analysis by calculating the volume of the log
⎛ d. b .h.2. π ⎞ sections Vi by tree using Huber’s formula:
V=⎜ . h . f⎟ (1)
⎝ 4 ⎠ n ⎛ D12. π ⎞
VTree = ∑ Vi = ⎜ . Li ⎟ (6 )
where d.b.h. and h are in metres. i =1
⎝ 4 ⎠
The form factor refers to the characteristic shape of the tree and is

Downloaded from https://academic.oup.com/forestry/article-abstract/90/3/359/2605859 by guest on 06 September 2019

the reduction factor of the cylinder volume to the actual tree volume. where Vi is the volume of the section i, Di is the mid-diameter of
Different form factor functions are available for predicting the form fac- the section in metres and Li is the length of the section in
tor based on d.b.h. and h. For our study, we consider the form factor metres. The sum of volume of the sections provided the true
functions of Pollanschütz, Short Swedish, Meyer (equations 2–4) volume of an individual tree (equation 6).
(Pollanschütz, 1965) and F. Evert’s Australian Function (equation 5) Next we calculated for this tree the volume of a cylinder
(Evert, 1969):
(VCylinder) with its d.b.h. in metres and h of the tree in metres:
⎛ 1 1 1 ⎛ d. b. h.2 . π ⎞
f = ⎜ a + b1. ln2 (d . b. h .) + b2. + b3. + b4 . VCylinder = ⎜ . h⎟ (7 )
⎝ h d . b. h . d . b. h .2 ⎝ 4 ⎠
1 1 ⎞
+ b5. + b6. ⎟ (2)
d . b. h . . h d . b. h .2. h ⎠ Thus, for each tree in our dataset, we have the total ‘true’
tree volume from equation (6) and the volume of the cylinder
from equation (7). The corresponding ratio provides the required
⎛ 1 h h ⎞
f = ⎜ a + b1. + b2. + b3. ⎟ (3) reduction or form factor by tree:
⎝ h d. b. h. d. b .h.2 ⎠
⎛ V ⎞
fObserved = ⎜ Tree ⎟ (8 )
f = a + b1.
1
+ b2.
1
+ b3.
1 ⎝ VCylinder ⎠
d. b. h.2. h d. b. h. . h d. b. h.
1 1
+ b4 . + b5. (4)
h d. b. h.2
Selection of the form factor function
⎛ 1 1 1 ⎞ We used the calibration dataset (Table 1) of the four species
f = ⎜ a + b1. + b2. + b3. ⎟ (5)
⎝ d. b. h.2. h h d. b. h.2 ⎠ (A. densa, P. spinulosa, P. wallichiana and T. dumosa) to start the
process of model calibration for the four selected form factor
where f is the form factor; in equation (2), d.b.h. and h are in decimetres functions (see equations 2–5). Parameters were estimated using
and in equations (3–5) d.b.h. is in centimetres and h is in decimetres. the non-linear (nls) function of the R statistical software package

Table 2 Model fitting results for the validation data (only for the four species with validation data) for choosing the best form factor function

Species Obs. f Form factor function RMSE Rank MAD Rank Sum (rank)

A. densa 0.5590 Pollanschütz’s function 0.1147 3 0.0903 3 6 (3)

Short Swedish’s function 0.1112 2 0.0861 2 4 (2)
Meyer’s function 0.1225 4 0.0982 4 8 (4)
F. Evert’s function 0.0958 1 0.0828 1 2 (1)
P. spinulosa 0.5054 Pollanschütz’s function 0.0591 3 0.0405 2 5 (2)
Short Swedish’s function 0.0521 1 0.0390 1 2 (1)
Meyer’s function 0.0566 2 0.0416 3 5 (2)
F. Evert’s function 0.0566 2 0.0416 3 5 (2)
P. wallichiana 0.4986 Pollanschütz’s function 0.0499 1 0.0392 1 2 (1)
Short Swedish’s function 0.0506 2 0.0402 2 4 (2)
Meyer’s function 0.0522 3 0.0428 3 6 (3)
F. Evert’s function 0.0548 4 0.0450 4 8 (4)
T. dumosa 0.4820 Pollanschütz’s function 0.0487 1 0.0382 1 2 (1)
Short Swedish’s function 0.0490 2 0.0385 2 4 (2)
Meyer’s function 0.0516 3 0.0412 3 6 (3)
F. Evert’s function 0.0541 4 0.0439 4 8 (4)

Obs. f = mean of observed form factor for the validation data; MAD = mean absolute deviation. Values in parentheses indicate average ranking of
the models for each species.

361
Forestry

(R Core Team, 2015), requiring the parameters to have a signifi- Results and discussion
cance level of α = 0.05 to be included in the final species model.
Next, we obtained the parameter estimates by species and Model selection and validation
form factor function and predicted the form factors for the The results of the two fit statistics criteria (1) RMSE and (2) MAD
independent validation data to evaluate the performances suggest that the Pollanschütz function ranked highest for the
according to the form factor functions. As a selection criteria, species P. wallichiana and T. dumosa, the Swedish function for
we employed (1) root mean square error (RMSE) and (2) mean P. spinulosa and Evert’s function for A. densa (Table 2). However,
absolute deviation (MAD) (Weiskittel et al., 2011) defined as the results also indicate that the Pollanschütz function provides
consistent and unbiased form factor predictions for all four spe-
n

Downloaded from https://academic.oup.com/forestry/article-abstract/90/3/359/2605859 by guest on 06 September 2019

∑i = 1 (Yi − Yˆ )2 cies. Thus, we decided to select the Pollanschütz function (equa-
RMSE = (9)
n tion 2) for our further analysis.
We employed (1) MB, (2) SDR and (3) Student’s t test for fur-
n ther assessment of the selected Pollanschütz function (Table 3).
∑i = 1 Yi − Yˆ
MAD = (10) The MB (equation 11) for A. densa, P. spinulosa, P. wallichiana
n and T. dumosa ranged between 0.0207, 0.0159, 0.0126 and
where Yi is the observed form factor, Ŷ is the predicted form fac- −0.0077, respectively (Table 3). The SDR between predicted vs
tor and n is the number of observations. observed form factors exhibited a value of 0.0568 or 11 per
Model validation examines the usefulness or behaviour of the cent of the observed form factor for P. spinulosa, 0.1168 or 21
model by comparing predicted vs observed values using the per cent for A. densa, 0.0493 or 10 per cent for P. wallichiana
independent dataset (Rykiel, 1996; Hasenauer, 2006). Mean bias and 0.0487 or 10 per cent for T. dumosa. The paired Student’s t
(MB), standard deviation of residuals (SDR), R2 values (correl- test shows P-values greater than 0.05, indicating no significant
ation coefficient) (equations 11–13) and Student’s t test were differences between predicted and observed form factor values
used for further assessment and evaluation of the selected (Table 3). The MB and SDR values are small with no significant
function.
Table 3 Validation results for the Pollanschütz function for predicting
n the form factors with Student’s t test for the four species with separate
∑i = 1 (Yi − Yˆ )
MB = (11) calibration and validation dataset
n
Species Statistical indices Student’s t test
2

SDR =
n
( n
(
∑i = 1 (Yi − Yˆ )− ∑i = 1 (Yi − Yˆ )/n )) (12) Obs. f Bias SDR t-Stat P-value
n−1
A. densa 0.5590 0.0207 0.1168 (21%) 0.6882 0.5026
⎛ ∑n (Y − Yˆ )2 ⎞ P. spinulosa 0.5054 0.0159 0.0568 (11%) 1.4363 0.1633
i
R2 = 1 − ⎜⎜ ni = 1 2⎟
⎟ (13) P. wallichiana 0.4986 0.0126 0.0493 (10%) 1.2728 0.2153
⎝ i =1 i − Y ) ⎠
∑ ( Y T. dumosa 0.4820 −0.0077 0.0487 (10%) −1.0402 0.3042

where Yi is the observed form factor, Ŷ is the predicted form fac- Notes: results displayed only for those species where separate model
tor, Y is the mean of the observed form factor and n is the num- calibration and validation data were available. The t-stat and P-values
ber of observations. for Student’s t test were generated using the R statistical package.
Graphical analysis by plotting the residuals vs d.b.h. and h Values in parentheses are per cent of the observed value. Obs. f = mean
and the Anderson–Darling test for assessing the normal distri- observed form factor for the validation dataset for the four species;
bution of the residuals (R Core Team, 2015) were carried out. Bias = mean of difference between observed and predicted values.

Table 4 Final parameter estimates for the Pollanschütz function for all species based on all sample data

Species n Coefficient for the parameters Statistical indices

a b1 b2 b3 b4 b5 b6 R2 SDR

A. densa 45 0.5367 −0.0161 ns ns −0.0966 14.752 ns 0.45 0.0676 (12%)

P. spinulosa 58 0.4505 ns −16.412 0.2808 −0.1530 14.270 ns 0.57 0.0385 (8%)
P. wallichiana 59 0.6384 −0.0237 −39.308 ns −0.0861 43.190 −7.639 0.43 0.0446 (9%)
T. dumosa 89 0.5425 −0.0194 −5.847 ns ns 4.994 ns 0.32 0.0427 (9%)
P. roxburghii 31 0.5677 −0.0350 ns −0.1749 0.1534 ns −2.559 0.73 0.0398 (8%)
C. tribuloides 31 0.3899 ns −4.394 0.2967 ns −7.239 ns 0.60 0.0449 (10%)
Q. glauca 31 0.3397 ns −5.563 ns 0.1422 12.466 −11.026 0.57 0.0571 (16%)
Q. lanata 31 0.5112 −0.0340 −21.385 ns ns 14.032 −1.515 0.84 0.0302 (9%)
Q. lamellosa 20 0.2883 ns ns ns 0.7755 199.370 −434.582 0.32 0.0447 (10%)

n = number of trees; ns = non-significant parameters; values in parentheses are in per cent of the observed values.

362
 Form factor functions

Downloaded from https://academic.oup.com/forestry/article-abstract/90/3/359/2605859 by guest on 06 September 2019

Figure 1 Trend analysis of residuals vs d.b.h. (cm) and h (m) for all the species to examine the model fitting of Pollanschütz function.

differences between predicted and observed form factors. This Final model calibration and evaluation
confirms the selection of the Pollanschütz function for predicting
Based on the results of Tables 2 and 3, we refitted the selected
the form factors.
Pollanschütz function by obtaining all the available data

363
Forestry

Downloaded from https://academic.oup.com/forestry/article-abstract/90/3/359/2605859 by guest on 06 September 2019

Figure 2 Evaluation of the selected Pollanschütz function by plotting the observed and predicted form factor against tree parameters of (1) d.b.h.
(cm) and (2) h (m) for all the species.

(calibration and validation) for P. spinulosa, A. densa, P. wallichiana lamellosa) to estimate the final parameters. We combine both
and T. dumosa as well as the data available of the remaining five datasets to ensure that we cover the full range of possible varia-
species (P. roxburghii, C. tribuloides, Q. glauca, Q. lanata and Q. tions. The results for our nine commercial tree species analysed

364
 Form factor functions

in this study are presented in Table 4. R2 and SDR (see Table 4)

are provided to assess the final form factor functions for each Acknowledgements
species (Zhang, 1997; Brooks et al., 2008). The BC-CAP project was executed by the Department of Forests & Park
The SDR ranged between 0.0302 and 0.0676 and the R² Services (DoFPS), Ministry of Agriculture & Forests, Royal Government of
explained ~32 per cent to 84 per cent of the variation, Bhutan (RGoB) and the University of Natural Resources and Life
depending on the tree species (Table 4). The lowest SDR and Sciences, Austria. We thank Georg Gratzer, Lobzang Dorji, Pema Wangda,
highest R2 confirm the precision and accuracy of the Purna B. Chettri and Andras Darabant for the managerial support and
Pollanschütz function in predicting form factors (Teshome, Mr Yograj Chhetri for coordinating and arranging of the validation data.
2005; Brooks et al., 2008; Weiskittel et al., 2011) (Table 4). SDR Gratitude to Yograj Chhetri, Kezang Yangden, Madan Lama and Harka
Bdr for their support during the field work. We thank Younten Phuntsho,

Downloaded from https://academic.oup.com/forestry/article-abstract/90/3/359/2605859 by guest on 06 September 2019

is a common measure for predictive values of a non-linear
Kinley Dem and Saran Pradhan for providing information on forest
regression where low values indicate better fit (Akindele and
management, and Adam Moreno for English editing. We also thank the
LeMay, 2006). editor and the anonymous reviewers for their helpful suggestions.
Graphical presentation of the residuals exhibited no system-
atic trend or bias in the resulting form factor estimates
(Figure 1). The Anderson–Darling test revealed P-values of more
than 0.05, confirming normal distribution of the residuals. Conflict of interest statement
None declared.

Form factor predictions

The predicted form factors gradually decrease with increasing d.
Funding
b.h. and h and remain constant at a d.b.h. > 60 cm (Figure 2). This work was supported by the project ‘Climate change adaptation
potentials of forests in Bhutan-building human capacities and knowledge
These trends are a typical behaviour and were also observed in
base (BC-CAP)’ with funding from the Austrian Government through the
previous studies as given by Pollanschütz (1965) using more
Austrian Federal Ministry of Agriculture, Forestry, Environment and Water
than 10 000 trees to provide form factors for all main tree Management (BMLFUW-UW.1.3.2/0124-V/4/2013). Calibration data (Pinus
species in Austria or by Sendi et al. (2014) for form factors of roxburghii, Quercus glauca and Quercus lanata) and validation data (Picea
Tilia begonifolia as well as by Gezahgn (2015) for form factors spinulosa, Abies densa, Pinus wallichiana and Tsuga dumosa) were
of 20 broadleaved tree species in Ethiopia. The predicted obtained from the Biomass equation development project of Department
mean form factor values derived from the Pollanschütz func- of Forests & Park Services, Ministry of Agriculture & Forests, Royal
tion and the species-specific parameter estimates in Table 4 Government of Bhutan (RGoB), the part which was funded by the Bhutan
are 0.490 (P. spinulosa), 0.493 (A. densa), 0.548 (P. wallichiana), Trust Fund for Environmental Conservation and RGoB.
0.491 (T. dumosa), 0.479 (P. roxburghii), 0.458 (C. tribuloides),
0.436 (Q. lamellosa), 0.354 (Q. glauca) and 0.336 for Q. lanata
(Figure 2).
For none of these tree species, form factor estimates are References
available for comparison. The only exception is a study by Adekunle, V. 2007 Non-linear regression models for timber volume esti-
Schieler (1991) for A. densa in Central Bhutan. In this study, mation in natural forest ecosystem, Southwest Nigeria. Res. J. For. 1,
Schieler (1991) used a different dataset for this species and 40–54.
calibrated a form factor function with an R2 of 0.25 and a Adekunle, V., Nair, K., Srivastava, A. and Singh, N. 2013 Models and form
mean form factor of 0.591, while our study revealed an R2 factors for stand volume estimation in natural forest ecosystems: a case
value of 0.45 and an estimated mean form factor of 0.548 for study of Katarniaghat Wildlife Sanctuary (KGWS), Bahraich District,
A. densa. Our form factor predictions using the Pollanschütz India. J. For. Res. 24, 217–226.
approach ranged from 0.479 to 0.548 for the conifer species Akindele, S. and LeMay, V. 2006 Development of tree volume
and from 0.348 to 0.458 for the broadleaved species and thus equations for common timber species in the tropical rain forest area of
is within the range of 0.25 and 0.50 according to West Nigeria. For. Ecol. Manage 226, 41–48.
(2009). Avery, T.E. and Burkhart, H.E. 2002 Forest Measurements. McGraw Hill
Higher Education.
Brooks, J.R., Jiang, L. and Ozçelik, R. 2008 Compatible stem volume and
taper equations for Brutian pine, Cedar of Lebanon, and Cilicica fir in
Conclusions Turkey. For. Ecol. Manage 256, 147–151.
We developed form factor functions for nine commercial tree Burkhart, H.E. and Tomé, M. 2012 Modeling Forest Trees and Stands.
species in Bhutan to allow accurate and consistent volume and Springer Science & Business Media.
volume increment predictions, key information needs within Colgan, M.S., Swemmer, T. and Asner, G.P. 2014 Structural relationships
sustainable forest management. The form factor functions for between form factor, wood density, and biomass in African savanna
the nine tree species presented in this study are the first in the woodlands. Trees 28, 91–102.
region and allow for a species-specific volume assessment. The Department of Forests & Park Services 2014 Forestry Statistics 2014.
selected Pollanschütz form factor approach provides unbiased Ministry of Agriculture & Forests, Royal Government of Bhutan.
and consistent estimates and can be easily recalibrated and/or Eastaugh, C.S. 2014 Relationships between the mean trees by basal area
extended to any other tree species if new calibration data are and by volume: reconciling form factors in the classic Bavarian yield and
available. volume tables for Norway spruce. Eur. J. For. Res 133, 871–877.

365
Forestry

Evert, F. 1969 Estimating stand volume by measuring form class without the International Union of Forest Research Organizations, Section 25,
measuring diameters. For. Sci. 15, 145–148. Stockholm, pp. 7–17.
Forest Resources Development Division 2004 Forest Management Code R Core Team 2015 R: A Language and Environment for Statistical
of Bhutan. Department of Forests, Ministry of Agriculture, Royal Computing. R Foundation for Statistical Computing.
Government of Bhutan. Rosset, J. 1999 Temperate conifer forests of Bhutan: a review of forestry
Forest Resources Development Division 2005 Volume Table of Bhutan. research sctivities until June, 1998. In RNR-RC Jakar. Special Publication
Department of Forests, Ministry of Agriculture, Royal Government of No. 3, 1999. Research Extension and Irrigation Division, Ministry of
Bhutan. Agriculture, Royal Government of Bhutan, 95 pp.
Forest Resources Management Division 2013 Forest Resources Potential Rykiel, E.J. 1996 Testing ecological models: the meaning of validation.

Downloaded from https://academic.oup.com/forestry/article-abstract/90/3/359/2605859 by guest on 06 September 2019

Assessment of Bhutan 2013. Department of Forest & Park Services, Ecol. Model 90, 229–244.
Ministry of Agriculture & Forests, Royal Government of Bhutan. Schieler, K. 1991 Local Forest Inventory, Bumthang, Wangtha-La
Führer, E. 2000 Forest functions, ecosystem stability and management. Thrumshing-La Volume 1, Integrated Forest Management Project. Ministry
For. Ecol. Manage 132, 29–38. of Agriculture, Royal Government of Bhutan.
Gezahgn, T.T. 2015 Development of Form Factor and Height-Diameter Sendi, M.R.N., Navroodi, I.H., Poorbabaei, H., Milan, M.S. and Bakhshandeh,
Functions for Selected Tree Species in the Amhara Region, Ethiopia. B. 2014 Determination of lime tree (Tilia begonifolia Stev.) stems form
University of Natural Resources and Life Sciences. based on quantitative parameters (Study area: Shafaroud forests of Guilan
Hasenauer, H. 2006 Sustainable Forest Management: Growth Models for province, Iran). Folia Forestalia Polonica, series A 56 (4), 165–170.
Europe. Springer. Socha, J. and Kulej, M. 2005 Provenance-dependent variability of Abies
Hasenauer, H., Petritsch, R., Zhao, M., Boisvenue, C. and Running, S.W. grandis stem form under mountain conditions of Beskid Sądecki (south-
2012 Reconciling satellite with ground data to estimate forest productiv- ern Poland). Can. J. For. Res. 35, 2539–2552.
ity at national scales. For. Ecol. Manage 276, 196–208. Socha, J. and Kulej, M. 2007 Variation of the tree form factor and taper in
Hoyer, G.E. 1985 Tree Form Quotients as Variables in Volume Estimation. European larch of Polish provenances tested under conditions of the
U.S. Department of Agriculture, Forest Service, Pacific Northwest Forest Beskid Sądecki mountain range (southern Poland). J. For. Sci. 53, 538–547.
and Range Experiment Station, 16 pp. Tarp-Johansen, M., Skovsgaard, J., Madsen, S., Johannsen, V. and
Husch, B., Beers, T.W. and Kershaw, J.A.Jr. 2002 Forest Mensuration. Wiley. Skovgaard, I. 1997 Compatible stem taper and stem volume functions
for oak (Quercus robur L and Q petraea (Matt) Liebl) in Denmark. Ann.
Ikonen, V.-P., Kellomäki, S., Väisänen, H. and Peltola, H. 2006 Modelling
Sci. For. 54, 577–595.
the distribution of diameter growth along the stem in Scots pine. Trees
20, 391–402. Teshome, T. 2005 A ratio method for predicting stem merchantable vol-
ume and associated taper equations for Cupressus lusitanica, Ethiopia.
Larson, P.R. 1963 Stem form development of forest trees. For. Sci. Monogr.
For. Ecol. Manage 204, 171–179.
Society of American Foresters.
United Nations Framework Convention on Climate Change 2014 Key
Laumans, P. 1994 Height-diameter functions from PIS for country level
Decisions Relevant for Reducing Emissions from Deforestation and Forest
site classification and local volume table selection. Working Document
Degradation in Developing Countries (REDD+). United Nations Framework
No.11 (FO: DP/BHU/911002). Thimphu (Bhutan): UNDP/FAO Forest
Convention on Climate Change.
Resources Management and Institutional Development Project,
Department of Forestry, Ministry of Agriculture, Royal Government of Weiskittel, A.R., Hann, D.W., Kershaw, J.A.Jr and Vanclay, J.K. 2011 Forest
Bhutan. Growth and Yield Modeling. John Wiley & Sons, Oxford.
Martin, A.J. 1981 Taper and Volume Equations for Selected Appalachian West, P.W. 2009 Tree and Forest Measurement. Springer, Australia.
Hardwood Species. US Department of Agriculture, Forest Service, West Whitfield, P. 2001 Planning and Implementation for Sustainable FMU
Virginia. Management. Wang Watershed Management Project ALA/97/50.
Philip, M.S. 1994 Measuring Trees and Forests. CAB international. Ministry of Agriculture, Royal Government of Bhutan.
Pollanschütz, J. 1965 A new method of determining stem form factor of Zhang, L. 1997 Cross-validation of non-linear growth functions for mod-
standing trees. Presented at the Advisory Group of Forest Statisticians of elling tree height–diameter relationships. Ann. Bot. 79, 251–257.

366