a hybrid model for intensively managed douglas-fir...
TRANSCRIPT
ORIGINAL PAPER
A hybrid model for intensively managed Douglas-fir plantationsin the Pacific Northwest, USA
Aaron R. Weiskittel • Douglas A. Maguire •
Robert A. Monserud • Gregory P. Johnson
Received: 6 July 2009 / Revised: 28 October 2009 / Accepted: 30 November 2009 / Published online: 30 December 2009
� Springer-Verlag 2009
Abstract Recent advances in traditional forest growth
models have been achieved by linking growth predictions
to key ecophysiological processes in a hybrid approach that
combines the strengths of both empirical and process-based
models. A hybrid model was constructed for intensively
managed Douglas-fir plantations in the Pacific Northwest,
USA, by embedding components representing fundamental
physiological processes and detailed tree allometrics into
an empirical growth model for projecting individual tree
and stand development. The simulated processes operated
at a variety of scales ranging from individual branches to
trees and stands. The canopy structure submodel improved
predictions of leaf area index at the stand level when
compared to allometric and other empirical approaches
(reducing mean square error by 30–42%). In addition, the
hybrid model achieved accuracy in short-term volume
growth prediction comparable to an empirical model.
Biases in 4-year stand growth predictions from the hybrid
model were similar to those from the empirical model
under thinning, fertilization, and the combination of these
treatments; however, volume growth predictions in
unmanaged plantations averaged approximately 36% less
bias. These improvements were attributed to detailed
information on crown structure (i.e. size, location, and
foliage mass of primary branches), simple representation of
key physiological processes, and improved site character-
ization. Soil moisture, temperature, and nitrogen mineral-
ization predicted by the hybrid model also agreed closely
with observed values from several previous studies.
Overall, the model framework will be helpful for future
analyses as it can lend insight into the influence of weather
and site edaphic factors on growth, help identify mecha-
nisms of response to silvicultural treatments, and facilitate
the design of sound management regimes for Douglas-fir
plantations across the Pacific Northwest region.
Keywords Hybrid model � Thinning � Fertilization �Growth and yield � Oregon �Washington � Leaf area index �Soil temperature � Soil moisture � Soil nitrogen
mineralization � Net primary production
Introduction
The effects of intensive management on tree growth and
stand yield are difficult to quantify empirically because
natural variability necessitates extensive replication of
silvicultural treatments across the full range of sites. Many
years must also pass before growth trajectories and impli-
cations for final yields become evident. Empirical growth
and yield models have been very effective for summarizing
Communicated by Uta Berger.
Electronic supplementary material The online version of thisarticle (doi:10.1007/s10342-009-0339-6) contains supplementarymaterial, which is available to authorized users.
A. R. Weiskittel (&)
School of Forest Resources, The University of Maine, Orono,
ME 04469, USA
e-mail: [email protected]
D. A. Maguire
Department of Forest Engineering, Resources & Management,
Oregon State University, Corvallis, OR 97331, USA
R. A. Monserud
PNW Research Station (retired), USDA Forest Service,
620 SW Main, Suite 400, Portland, OR 97205, USA
G. P. Johnson
Weyerhaeuser Company, P.O. Box 9777, Federal Way,
WA 98063, USA
123
Eur J Forest Res (2010) 129:325–338
DOI 10.1007/s10342-009-0339-6
large amounts of data from silvicultural trials or opera-
tional inventories in a way that allows the user to explore
the consequences of alternative treatments and regimes
through simulation (Pretzsch et al. 2006). However,
empirical models are more or less limited to the range of
conditions represented in available data and generally
consider site quality invariant. For this reason, they are
inappropriate for addressing many questions facing forest
managers currently, such as growth responses to changes in
water availability imposed by new cultural techniques,
likely effects of climate change, or potential growth
declines from depletion of soil nutrients under repeated
harvesting.
Models based on fundamental ecophysiological pro-
cesses have found increasing application in designing
optimal forest management regimes (Makela et al. 2000).
These models offer the advantage of simulating growth
responses to silvicultural treatments by incorporating key
ecophysiological mechanisms driving growth, and thereby
potentially yielding more robust extrapolations to untested
conditions and management regimes. However, a strong
theoretical basis does not currently exist for predicting all
aspects of tree growth from fundamental principles, and
proper parameterization for most process models is a dif-
ficult task because they often have multiple parameters,
extensive physiological data are not available for all spe-
cies, and the information necessary to initialize the models
is expensive to collect. The hybridization of empirical and
process-based models offers some promise for producing
models that draw on the strengths, and exceed the capa-
bilities, of both approaches (Landsberg 2003). However,
hybrid models have received less attention in some regions
because they have not achieved sufficient reliability or
flexibility as management tools (Pretzsch et al. 2008).
Hybrid models have been shown to improve stand-level
growth predictions in some cases relative to strictly
empirical models (Woollons et al. 1997; Snowdon 2001;
Pinjuv et al. 2006). However, the level of improvement has
varied significantly, in part because the breadth of condi-
tions covered by evaluation data relative to the modeling
data varies tremendously. Snowdon (2001) found a 23%
improvement for predicting Pinus radiata basal area
growth, while Pinjuv et al. (2006) reported only a 4%
improvement for the same species. Stand growth has gen-
erally been the metric for verifying hybrid model predic-
tions, but a more thorough evaluation of hybrid models
would include assessment of component mechanisms, for
example, by comparison of predicted and observed leaf
area index (LAI), soil moisture, and nutrient availability.
Although hybrid models generally reduce the number of
necessary parameters in comparison with purely process-
based models, the cost of requiring fewer parameters
results in some loss in site specificity. Hybrid models often
require attributes such as LAI and stand biomass compo-
nents that must be estimated from other tree dimensions
routinely measured in forest inventories or silvicultural
research trials. Consequently, these models usually involve
the use of regional allometric equations, despite recogni-
tion that they can be significantly biased when applied to
stands managed under widely varying intensities (e.g. Grier
et al. 1984). An effective hybrid model must therefore
account for silvicultural effects on allometric relationships
and the implications for related physiological processes.
The primary goal of this research project was to develop
a hybrid model for intensively managed Douglas-fir stands
and to test its performance against several long-term sil-
vicultural research trials in the northwestern United States
(hereafter referred to as the Pacific Northwest, or PNW).
The challenge was to identify those process features that
appear robust among widely differing biomes, while
adapting or adding others that seem uniquely important for
predicting the growth of Douglas-fir in intensively man-
aged PNW plantations. Output was required to include the
quantity and quality of wood at both the stand and indi-
vidual-tree levels as this information is required by forest
managers. Of particular interest was the strong control on
productivity exerted by moisture availability in the PNW,
and known structural relationships between wood quality
and responses of crown size, branching structure, and foliar
mass to silvicultural treatments (Maguire et al. 1991). This
paper describes the resulting model architecture and its
performance in simulating growth and several key pro-
cesses influencing site resource availability. Specific
objectives were to: (1) develop and assess an approach for
estimating biomass components and LAI that makes use of
branch-level detail needed for simulating silvicultural
effects on wood quality; (2) compare alternative approa-
ches for modeling key ecophysiological processes; (3) test
stand-level performance of the hybrid model across a wide
range of intensively managed plantations relative to an
empirical model developed in parallel, and (4) validate
simulations of daily soil temperature, soil moisture, and
nitrogen mineralization. Although the model was designed
to be applicable to Douglas-fir in all regions, the most
appropriate use of the model is in the PNW region of USA,
particularly in areas west of the Cascades Mountain region.
Methods
Model development
The model architecture emphasized a hierarchy of multiple
processes, with some predictions made at the branch level,
some at the tree level, and others at the stand level. The
desired temporal scale for growth predictions was annual,
326 Eur J Forest Res (2010) 129:325–338
123
but some key processes depended on diurnal cycles of
resource availability and tree physiology. The architecture
of DF.HGS (Douglas-fir Hybrid Growth System) was
intended to take advantage of the current state of knowl-
edge at each scale appropriate for representing inherent site
productivity and simulating response to silvicultural treat-
ment. It also needed the flexibility to accommodate future
modifications as better physiological and morphological
data, parameters, and submodels become available. Three
primary components comprise DF.HGS, one to character-
ize branch, crown, and canopy structure (BCACS), one to
estimate annual net primary production (NPP), and one to
allocate carbon and growth to different stand and tree
compartments (ALOGRO) (Fig. 1).
Branch, crown, and canopy simulator
Branch, crown, and canopy simulator (BCACS) contains a
set of empirical equations that predict the location, diam-
eter, and biomass of every branch for each tree measured
on a sample plot. Model input includes stand age and site
index and the following attributes for each tree: diameter at
breast height (DBH), total height (HT), height to crown
base (HCB; lowest live branch), and expansion factor
(EXPF, in trees per ha). The inputted site index is only used
to reconstruct past height growth to determine the location
of annual whorls of branches and the overall model is
relatively insensitive to it. Branch diameter, length, and
angle are predicted from basic tree dimensions using the
equations developed by Weiskittel et al. (2007b). Branch
location within an annual segment of the bole was assumed
to follow the empirical distribution described by Maguire
et al. (1994). Woody biomass, foliage biomass by age
class, and projected foliage area by age class of all bran-
ches were predicted from basal diameter and relative height
in the crown (Weiskittel et al. 2006). Woody surface area
(all-sided) of all branches was estimated at the tree level
from DBH, crown length, crown ratio, and sapwood area
(Weiskittel and Maguire 2006). These attributes were then
summed to estimate biomass and surface area at the tree-
and stand-levels.
Total stem and sapwood volume were estimated at the
tree-level from existing stem volume (Walters et al. 1985)
and sapwood taper equations (Maguire and Batista 1996).
Stem volumes were converted to biomass by applying
regional values for total and sapwood density (500 and
455 kg m-3, respectively). Belowground biomass at the
stand level was estimated from total aboveground biomass
(Ranger and Gelhaye 2001). Crown projection areas were
calculated from individual-tree crown widths computed
from branch lengths and angles, and these areas were then
summed and corrected for overlap to determine percent
canopy cover using the approach of Crookston and Stage
(1999).
Fig. 1 Components and flow of
the simulation model DF.HGS
for Douglas-fir in the Pacific
Northwest
Eur J Forest Res (2010) 129:325–338 327
123
Net primary production
Previous reviews of process-based models and their com-
ponents (Schwalm and Ek 2001; Landsberg 2003) have
suggested a common list of attributes that future models
should possess to accurately capture the effects of silvi-
cultural treatments and climate on NPP. These attributes
include: (1) daily time-step; (2) separation of direct and
diffuse radiation; (3) light extinction coefficients dependent
on solar geometry, canopy structure, and type of radiation;
(4) separation of sunlit and shaded leaf area; (5) photo-
synthesis estimates based on the Farquhar et al. (1980)
biochemical equation, (6) modification of diurnal distri-
bution of radiation and fraction of sunlit leaf area by slope
and aspect, particularly in areas with complex terrain; (7)
estimates of stomatal conductance that are sensitive to
vapor pressure deficit and soil water availability; and (8)
canopy gas exchange linked to soil water and nutrient
availability.
All of these desired attributes were incorporated into the
NPP submodel. Inputs were kept as simple as possible to
capture the fundamental processes (Table 1), so included:
(1) daily weather data (precipitation, temperature, relative
humidity, solar radiation) obtained from DAYMET (http://
www.daymet.org); (2) site physiographic features (eleva-
tion, slope, aspect); (3) stand structural attributes from
BCACS (Table 2); (4) basic soils information (rooting
depth, texture, rock content), (5) foliar nitrogen concen-
tration; and (6) physiological parameters obtained from the
literature (Appendix 1 in supplementary material). The
model runs on a daily time-step but estimates gross primary
production (GPP) five times daily based on Gaussian
integration. The development of the current-year foliage
was assumed to occur linearly for 30 days following bud
burst, with date of bud burst predicted from growing degree
days. Annual foliage litterfall and its seasonal distribution
were predicted from stand density, age, foliage retention,
crown density, slope and aspect (Weiskittel and Maguire
Table 1 Input variables
required to run Douglas-fir
hybrid growth system
(DF.HGS) for intensively
managed Douglas-fir plantations
Variable Units Definition
Tree
DBH cm Diameter at breast height (1.3 m)
HT m Total height
HCB m Height to crown base
EXPF Trees ha-1 Expansion factor
Site
SI m Site index; expected height of dominant trees
at 50 years breast height age (Bruce 1981)
SLOPE % Topographic slope
ASPECT Degrees Topographic aspect
ELEV m Elevation
Soil
SOIL_CLAY % Soil clay content by weight
SOIL_SILT % Soil silt content by weight
SOIL_SAND % Soil sand content by weight
SOIL_C % Soil carbon content by weight
SOIL_DEP m Soil root depth
SOIL_ROCK % Rock content by volume
Weather
Tmax �C Daily maximum air temperature
Tmin �C Daily minimum air temperature
Tmean �C Daily average air temperature
PRCP mm Daily total precipitation
VPD Pa Daily average vapor pressure deficit
SW.RADday W m-2 Daily shortwave solar radiation
DAYLEN Hours Day length
Geographic
LAT Degrees Latitude (to the nearest hundredths)
LONG Degrees Longitude (to the nearest hundredths)
328 Eur J Forest Res (2010) 129:325–338
123
2007). Details of NPP components follow and are also
described further in Appendix 2 in supplementary material.
Light interception Although highly detailed light inter-
ception models have been developed, they still remain
quite computer intensive and too complex for application
in management-oriented hybrid models. Estimating light
interception for an individual-tree crown based on canopy-
level principles does not fully account for the competitive
position of the tree within the stand. These limitations lead
to light interception being estimated by applying the Beer–
Lambert law after accounting for the effects of incomplete
canopy closure on the sunlit and shaded portions of the
canopy. Mean daily radiation was separated into direct and
diffuse components following the approach outlined by
Bristow and Campbell (1985), which is based on the ratio
of actual incoming shortwave radiation to potential short-
wave radiation. The amount of direct and diffuse radiation
at a given time of the day was estimated from cosine-
diurnal distribution following the procedure outlined by
Wang et al. (2002), in contrast to the more common
approach of assuming a sinusoidal pattern to diurnal radi-
ation. Light extinction coefficients for direct and diffuse
radiation were based on the relationships described by
Smith (1993) and Campbell and Norman (1998), respec-
tively. The direct radiation extinction coefficient was a
function of solar zenith angle, LAI, and stand relative
density, while the diffuse radiation extinction coefficient
was simply a linear function of LAI. Clumpiness was
accounted for by an equation presented in Kucharik et al.
(1999) that included the ratio of stand mean crown depth to
crown diameter and the solar zenith angle. The effects of
slope and aspect on the amount of incoming radiation were
assumed to be sufficiently accounted for by DAYMET.
Photosynthesis Net photosynthetic rates were estimated
with the Farquhar et al. (1980) model for both sunlit
(foliage receiving direct, diffuse, and scattered radiation)
and shaded (foliage receiving only diffuse and scattered
radiation) current-year foliage. Maximum Rubisco activity
(Vcmax), electron transport capacity (Jmax), and dark respi-
ration rate (Rdark) were estimated from their relationship to
mean canopy foliar nitrogen concentrations (Manter et al.
2005). These rates were assumed to be the same for both
sunlit and shaded foliage. Temperature modifiers for Vcmax
and Jmax were obtained from Manter et al. (2003). Other
standard parameters such as CO2 compensation point in the
absence of mitochondrial respiration and the Michaelis–
Menten constant of Rubisco were estimated from equations
given by Schwalm and Ek (2004).
To account for the effects of complex terrain on canopy
photosynthesis, the canopy was split into sunlit and shaded
leaf area based on total leaf area, slope and aspect of the
site, and solar azimuth and zenith angle (Wang et al. 2002).
Foliage age structure is also known to influence total
photosynthesis in conifer canopies, but actual Vcmax, Jmax,
and Rdark measurements for age classes older than the
current year were not available. Therefore, the net photo-
synthetic rates for the 1-, 2-, 3-, and C4-year-old foliage
age classes were assumed to be 75, 50, 30, and 10%,
respectively, of the net photosynthetic rates of current year
foliage calculated from the Farquhar et al. (1980) equation
and were inferred from information available in the liter-
ature (Ethier et al. 2006). The estimated net photosynthesis
rate for sunlit and shaded foliage was multiplied by the
amount of leaf area in each class to obtain an estimate of
stand-level GPP.
Water stress effects on canopy photosynthesis are par-
ticularly important in the PNW because of high evaporative
demand during the warm, dry summer. In a recent review
of models for simulating the effect of water stress on
canopy photosynthesis, Grant et al. (2006) found approa-
ches ranging from simple scaling factors to complex iter-
ative solutions. Proper model behavior requires the
Table 2 Stand attributes generated by submodel branch, canopy, and crown simulator (BCACS) and passed to submodel net primary production
(NPP) as the basis for estimating growth on the 67 Stand Management Cooperative plots of intensively managed Douglas-fir
Attribute Mean Standard deviation Minimum Maximum
Leaf area index (m2 m-2) 4.14 2.08 0.87 9.10
Stand-level belowground biomass (kg ha-1) 13.53 7.79 2.46 33.68
Stand-level aboveground biomass (kg ha-1) 104.72 60.37 21.04 276.33
Branch area index (m2 m-2) 0.35 0.16 0.09 0.70
Canopy cover (%) 0.89 0.11 0.54 0.99
% of leaf area in age class 1 0.28 0.008 0.25 0.30
% of leaf area in age class 2 0.26 0.004 0.24 0.27
% of leaf area in age class 3 0.18 0.021 0.15 0.25
% of leaf area in age class 4 0.16 0.009 0.15 0.20
% of leaf area in age class 5? 0.09 0.005 0.07 0.10
Eur J Forest Res (2010) 129:325–338 329
123
interactive effects of soil and atmospheric water deficits on
CO2 fixation (Grant et al. 2006). The simplest approach
was to predict stomatal conductance from both soil and
atmospheric water deficits with the equation developed by
Bond and Kavanagh (1999). Stomatal conductance was
also modified by the amount of intercepted radiation and
time since sunrise. Intercellular CO2 concentration was
then estimated from the model presented in Katul et al.
(2000) with the critical stomatal conductance set at
50 lmol m-2 s-1, and the maximum ratio of ambient to
intercellular CO2 was 0.66. One limitation of this approach
is the assumption that water deficit effects on CO2 fixation
are entirely stomatal, which is not entirely true. However, a
better understanding of the relationship between stomatal
conductance and leaf internal transfer conductance, and the
response of both to environment conditions, is required
before they can be better integrated into models (Warren
and Adams 2006).
Respiration One key simplification in the hybrid model
3-PG (Landsberg and Waring 1997) is the assumption that
NPP is a constant fraction of GPP (*0.5; Waring et al.
1998). This assumption has generated considerable debate
in the literature (e.g. Medlyn and Dewar 1999), but results
of several studies have supported the generality of this
NPP/GPP fraction across a wide range of forest types (e.g.
Litton et al. 2007). Three alternatives to this simplification
were also tested, including: (1) daily respiration as a
function of biomass and temperature (Schwalm and Ek
2004); (2) daily respiration as a function of nitrogen con-
tent and both a short- and long-term response to tempera-
ture (Kirschbaum 1999); and (3) annual respiration as a
function of nitrogen content and temperature (Battaglia
et al. 2004). However, estimating NPP as a constant frac-
tion of GPP yielded the strongest correlation with observed
stem volume growth and was therefore the approach taken
in DF.HGS.
Soil water and nutrients Daily soil water balance was
simulated in a manner similar to that described by Sch-
walm and Ek (2004). Soil water holding capacity was
estimated from rooting depth, texture, and rock content
using equations and data available in the literature (Saxton
and Rawls 2006). Daily canopy transpiration and soil
evaporation were simulated with an approach outlined by
Watt et al. (2003). In the canopy transpiration equation,
mean daily stomatal conductance was estimated from leaf
and soil water potential, leaf to air vapor pressure gradient,
and leaf specific conductance (Bond and Kavanagh 1999).
Daily soil water balance was determined by subtracting
canopy transpiration and soil evaporation from soil water
and adding daily net precipitation after subtracting inter-
ception by canopy leaf and branch surface area. The
amount of intercepted precipitation was based on empirical
rainfall attenuation in Douglas-fir stands (Keim 2004).
Although fog is an important source of moisture in some
ecosystems (Dawson 1998), prediction of its occurrence is
very difficult and it also commonly only occurs on a very
narrow belt along the Pacific Coast, so it was not directly
incorporated in the model.
For simplicity, daily nitrogen mineralization was the
only aspect of nutrient availability that was simulated in
the model. Driving variables included soil temperature,
water content, texture, bulk density, and water holding
capacity, as well as daily air temperature, rainfall, and solar
radiation (Paul et al. 2002). The mass and height of the
litter layer, the stand leaf area index, and site slope must
also be known or estimated. Simulation of daily nitrogen
mineralization is achieved by considering the influence of
all these factors on the optimum nitrogen mineralization
rate estimated from published data (Chappell et al. 1999).
Daily soil temperature was calculated from air temperature
as modified by the effects of the tree canopy, understory
vegetation, litter mass, and depth of soil layer (Paul et al.
2004). Daily nitrogen uptake was estimated as a function of
fine root biomass, root nitrogen concentration, soil tem-
perature, and the amount of available soil nitrogen
(Thornley 1991).
Allocation and growth (ALOGRO)
Allocation Several reviews on carbon allocation have
outlined the range of approaches to quantifying the sources
of its variability. However, no single solution is yet widely
accepted (Lacointe 2000). The 3-PG approach of empirical
allometric ratios (Landsberg and Waring 1997) is robust
and provides realistic values, but for many if not most
species these ratios are based on empirical relationships in
unmanaged stands. Similarly, Makela (1997) maintained a
constant ratio of foliage to fine root biomass to reflect an
expected functional balance between the two, regardless of
growing conditions. Carbon allocation, however, is highly
sensitive to intensive management, as well as to inherent
site resource availability, particularly belowground. Allo-
cation conditional on site resource availability was, there-
fore, initially built into the model. A variety of approaches
were examined, including those of Running and Gower
(1991), Johnson and Thornley (1987), Kirschbaum (1999),
and Landsberg and Waring (1997). Preliminary analysis,
however, suggested that a modified 3-PG approach resulted
in the highest correlation between simulated above-ground
NPP and observed stem volume growth. The modifications
to the 3-PG allocation approach involved estimation of the
site fertility rating from canopy N percentage, similar to
Swenson et al. (2005), and applying a correction for
inherent bias when estimating stand-level biomass (or its
330 Eur J Forest Res (2010) 129:325–338
123
allocation in this case) from stand-level mean tree size
(Duursma and Robinson 2003).
Leaf area index
Data from 14 Stand Management Cooperative (SMC;
University of Washington) installations in Oregon and
Washington, USA, comprised the dataset for validating
estimates of stand LAI (Maguire and Bennett 1996). The
installations covered a range of stand ages and site indices.
At each installation, 5–12 plots with varying silvicultural
treatment regimes (e.g. thinning intensities, fertilization)
were sampled. Multiple trees per plot were destructively
sampled and used to construct plot-specific tree leaf area
equations (Maguire and Bennett 1996). The mean LAI in
the verification dataset was 5.12 ± 3.23 (mean ± standard
deviation) with a range between 0.33 and 13.25.
LAI was estimated with BCACS and three other meth-
ods, including application of: (1) an allometric function of
DBH (Gholz et al. 1979); (2) a constant ratio of leaf area to
predicted sapwood area at crown base estimated with the
equation of Maguire and Batista (1996) (0.58 m2 cm-1;
CBSAP); and (3) an empirical function of crown length,
DBH, and height (Maguire and Bennett 1996).
Growth
Growth data from nine SMC installations in Oregon and
Washington, USA, comprised the dataset for verification of
growth (Table 3). Twenty-year mean annual rainfall for
these locations ranged from 150 to 315 cm, and January
mean minimum and July mean maximum temperatures
ranged from -2.2 to 3.5�C and 21.3 to 26.3�C, respec-
tively. Elevation ranged from 25 to 300 m above sea level
and all aspects were represented. Soils varied from a
moderately deep sandy loam to a very deep clay loam. The
plantations were established between 1971 and 1982 at a
range in tree density (905–1,575 stems per ha) and under
varying levels of vegetation control.
A set of 0.2-ha permanent plots was established by the
SMC at each installation between 1986 and 1992. Since
establishment, the 67 plots have received a variety of sil-
vicultural regimes that have involved thinning (n = 21),
nitrogen fertilization (224 kg ha-1 of urea every 4 years;
Table 3 Attributes of research plots in Oregon and Washington simulated during verification of Douglas-fir hybrid growth system (DF.HGS)
Attribute Mean Standard deviation Minimum Maximum
Leaf area index (nplot = 36)
Basal area (m2 ha-1) 37.9 15.4 10.5 76.8
Stems per ha 663.3 372.7 175.0 1,790.0
Total age (yrs) 28.8 13.9 11.0 61.0
Site index (m at 50 yrs) 39.4 4.2 26.6 46.2
Soil water holding capacity (mm) 175.2 40.2 105.2 301.5
Volume growth (nplot = 67)
Basal area (m2 ha-1) 16.5 8.8 3.3 40.5
Stems per ha 720.7 452.2 242.2 1,818.7
Total age (yrs) 17.3 3.7 12.0 24.0
Site index (m at 50 yrs) 45.9 5.1 28.1 56.9
Soil water holding capacity (mm) 246.2 93.73 55.7 381.9
Soil temperature and moisture (nplot = 1, measured 3 different years)
Mean seedling stem diameter (mm) 17.9 0.9 2.0 34.0
Mean seedling stem height (cm) 93.4 5.2 17.0 158
Total age (yrs) 2.0 1.0 1.0 3.0
Site index (m at 50 yrs) 42.0a – – –
Soil water holding capacity (mm) 527.3 – – –
Soil nitrogen mineralization (nplot = 10)
Basal area (m2 ha-1) 27.8 8.8 11.6 40.0
Stems per ha 616.3 233.8 382.9 1,112.0
Total age (yrs) 22.1 3.0 16.0 27.0
Site index (m at 50 yrs) 42.3 3.0 37.6 45.7
Soil water holding capacity (mm) 146.3 40.3 110.5 298.1
a Determined from the previous stand
Eur J Forest Res (2010) 129:325–338 331
123
n = 9), and a combination of thinning and fertilization
(n = 22) with the remaining plots serving as controls
(n = 15). Canopy mean plot foliage N concentration ran-
ged from 1.26 to 2.98%. All trees on each plot were
remeasured for DBH every 4 years at the end of the
growing season, and at each remeasurement a subsample of
at least 40 Douglas-fir trees was also measured for total
height and height to crown base. Four to eight years after
plot establishment, current-year foliage was sampled from
the upper crown of 12 trees in each plot, and analyzed for
nutrient content. Climate information for the growth peri-
ods represented on the SMC plots was obtained from
DAYMET.
LAI and growth were simulated with DF.HGS (BCACS,
NPP, and ALOGRO) on each of the 67 SMC plots. The
same plots were also simulated with an empirical growth
and yield model assembled from individual-tree diameter
and height growth equations that project on an annual cycle
(Weiskittel et al. 2007a). Relative performance of the
models was assessed by comparing predicted 4-year
growth to observed growth.
Soil temperature and moisture
Soil temperature and moisture on a set of plots at the Fall
River Long-Term Site Productivity study in coastal
Washington (Ares et al. 2007) were predicted from the
NPP portion of DF.HGS. The site is located approximately
56 km from the Pacific Ocean, has a mean elevation of
330 m, and receives an average of 226 cm of annual pre-
cipitation. The soil at Fall River is deep, well-drained,
mostly stone-free silty clay loam that has a low bulk den-
sity and high organic carbon content. The site is highly
productive with a site index of 41–43 m at base age 50. In
1999, the site was harvested and planted with Douglas-fir,
and various tree and soil responses have been intensively
monitored. Soil temperature and volumetric moisture
content were measured daily at multiple depths with
HOBO H8 data loggers and hydrosense CS620 moisture
probe, respectively. The daily mean measurements at
10 cm in the bole-only harvest with complete vegetation
control treatment were utilized for verification in this
analysis. Actual daily climate and stand biomass data from
Fall River were input to DF.HGS to drive the predictions.
Soil nitrogen mineralization
Simulation of annual soil nitrogen mineralization was
verified against data collected by Sinkhorn (2007) from ten
approximately 30-year old plantations in the northern
Oregon Coast Range. All the sites were located within
25 km of the Pacific Ocean and had soils derived from
sedimentary parent material. The sites covered a range of
Swiss needle cast (SNC) severity with mean ocular foliage
retention rating ranging from 1.4 (very severe SNC) to 4.1
(low SNC). The sites were specifically chosen to represent
a gradient in soil nitrogen ranging from 9.2 to
28.7 Mg N ha-1 (Sinkhorn 2007). In 2003, net minerali-
zation and nitrification rates in the top 10 cm of mineral
soil were assessed monthly with six buried bags per site
(Sinkhorn 2007). LAI was estimated with BCACS, climate
information for 2003 was obtained from DAYMET, and
nitrogen mineralization was simulated for the same time
period with the NPP portion of DF.HGS.
Results
Leaf area index
All four methods for estimating LAI were significantly
correlated with observed LAI (P \ 0.0001; Fig. 2). The
relative strengths of these correlations, in descending order,
were: (1) CBSAP (r = 0.93); (2) BCACS (r = 0.92); (3)
Gholz et al. (1979) (r = 0.81); and (4) Maguire and Bennett
(1996) (r = 0.59). Graphs between observed and predicted
LAI indicate a clear bias in all approaches, except BCACS.
Mean absolute error of the BCACS predictions was
1.56 ± 1.23, which was 42, 34, and 30% lower than that
from Gholz et al. (1979), CBSAP, and Maguire and
Bennett (1996) approaches, respectively.
Growth
Simulated stem volume increment was closely correlated
with observed volume growth (P \ 0.0001; r = 0.73;
Fig. 3). Mean bias in 4-year stand volume growth from
DF.HGS was 1.61 ± 12.53 m3 ha-1, while the mean
absolute error was 9.84 ± 7.82 m3 ha-1, relative to a
range in observed 4-year growth of approximately 15–
110 m3 ha-1. In comparison, mean bias and absolute error
in 4-year stand volume growth from the empirical model
were 1.49 ± 13.74 and 9.91 ± 9.56 m3 ha-1, respectively.
Although the mean error for both models was similar, the
coefficient of variation of absolute error was 23.3% lower
for hybrid model. The hybrid approach in DF.HGS sub-
stantially improved predictions of volume growth in the
control and fertilized ? thinned stands in comparison with
the empirical model, and the two approaches performed
similarly for plots that were thinned only or fertilized only
(Fig. 4). Improvements from the hybrid model were most
significant for the control stands, amounting to an
approximate 36% reduction in mean absolute error.
At the individual tree-level, MSE was reduced by 6.1 and
12.4% for 4-year diameter and height growth, respectively,
when compared to a purely empirical approach.
332 Eur J Forest Res (2010) 129:325–338
123
Fig. 2 Relationship between
predicted and observed leaf area
index (LAI) for 113 Stand
Management Cooperative
research installations based on
the data of Maguire and Bennett
(1996). LAI was predicted using
four different methods:
(1) constant leaf area per unit of
predicted sapwood area at
crown base (0.58 m2 cm-2);
(2) Gholz et al. (1979)
allometric equation as a
function of DBH; (3) Maguire
and Bennett (1996) allometric
equation as a function of crown
length, tree height, and DBH;
and (4) Branch, crown, and
canopy simulator (BCACS)
Fig. 3 Results from simulation
of 67 Stand Management
Cooperative research plots in
the Pacific Northwest, USA,
with both the hybrid model
DF.HGS and a set of empirical
diameter and height growth
equations for individual trees
(Weiskittel et al. 2007a): arelationship between observed
and predicted 4-year stand
volume growth (left; m3 ha-1);
and b prediction bias
(predicted–observed) over
predicted volume growth
Eur J Forest Res (2010) 129:325–338 333
123
Soil temperature and moisture
Simulated daily mean soil temperature and volumetric
water content were both closely correlated with observed
temperature and water (P \ 0.0001; Fig. 5). However, the
strength of the correlation was much higher for soil tem-
perature (r = 0.93 vs. 0.49 for soil moisture). DF.HGS
generally underpredicted soil temperature by approxi-
mately 0.5–1.0�C, particularly in the mid-summer. Soil
moisture also was overpredicted on average, particularly at
volumetric moisture contents above 0.25. However, during
the growing season, predictions of soil water dynamics
were closely aligned with observed values.
Soil nitrogen mineralization
Observed annual soil nitrogen mineralization rates aver-
aged 66.9 ± 26.1 kg ha-1 year-1 (range of 15.9–99.1 kg
ha-1 year-1), while predicted rates averaged 60.5 ±
17.9 kg ha-1 year-1 (range of 30.1–97.3 kg ha-1 year-1).
The hybrid model DF.HGS tended to underpredict annual
soil nitrogen mineralization by 10.6 ± 0.33 kg ha-1
year-1, but observed and predicted mineralization rates
were closely correlated (r = 0.75; P \ 0.0001; Fig. 6).
Discussion
The goal of this project was to develop a hybrid model for
assessing silvicultural strategies for intensively managed
Douglas-fir plantations. Components were selected on the
basis of their potential for explaining responses to intensive
silvicultural treatments and for representing the short-term
effects of environmental factors that impose spatial and
temporal variation in growth responses to silvicultural
manipulation. Physiological submodels and parameters
were largely gleaned from the literature and combined with
detailed stand structural information to ensure that the
framework was useful to both researchers and forest
managers. Models that rely on simpler approaches such as
radiation-use efficiency and 0–1 environmental modifiers
perform well, particularly across steep gradients in envi-
ronmental conditions (e.g. Landsberg et al. 2003). How-
ever, models designed to differentiate among alternative
silvicultural regimes on a given site, or the interaction of
alternative regimes with site factors (e.g., mineralization
rates and water availability), require more explicit treat-
ment of processes. The framework developed here for
Douglas-fir resembles many first-generation process-based
models (reviewed by Ryan et al. 1996), but with some
important refinements for simulating responses to intensive
Fig. 4 Mean absolute error in
predicted 4-year stand volume
growth (m3 ha-1) for 67 Stand
Management Cooperative
research plots in the Pacific
Northwest, USA. Plot growth
was simulated with both the
hybrid model DF.HGS and a set
of empirical, individual-tree
diameter and height growth
equations (Weiskittel et al.
2007a)
334 Eur J Forest Res (2010) 129:325–338
123
silvicultural treatments. These refinements include: (1)
hierarchical treatment of growth at branch, tree and stand
levels; (2) explicit representation of crown structure
influence on growth processes; (3) incorporation of slope
and aspect effects on canopy processes; (4) clear connec-
tion between water stress and reductions in photosynthesis;
(5) patterns in light extinction that depend on stand struc-
ture and site factors; (6) net photosynthesis rates dependent
on foliage age class distribution; and (7) initiation from
conventional inventory data.
The hybrid model presented here was designed specifi-
cally to give crown dynamics a central role in driving total
production and controlling responses to silvicultural treat-
ment. Unbiased estimation of LAI and stand biomass is
critical to any process-based and hybrid model; however,
accurately estimating these components can be challeng-
ing, particularly under the wide variety of treatment com-
binations typically imposed by intensive silviculture.
Allometric equations based only on DBH are particularly
biased in young fertilized and thinned Douglas-fir planta-
tions (Grier et al. 1984) and those with varying intensities
of competing vegetation control (Petersen et al. 2008) but
similar equations are still generally used to initiate hybrid
models. In this study, BCACS performed significantly
better than several alternative approaches across a range of
observed LAIs because it directly accounts for crown size
and the fact that it can vary widely for trees of a given
DBH, depending on silvicultural regime. However, the
Fig. 5 Relationship between
observed and predicted daily
mean soil volumetric water
content (left panel) and soil
temperature at 10 cm (rightpanel) for 2001, 2002, and 2003
at the Fall River Long-Term
Site Productivity study in
coastal Washington (Ares et al.
2007)
Fig. 6 Relationship between observed and predicted annual soil
nitrogen mineralization (kg ha-1) in Douglas-fir plantations from the
Oregon Coast Ranges. Simulations were started with initial stand and
site data from Sinkhorn (2007) and daily weather data from DAYMET
Eur J Forest Res (2010) 129:325–338 335
123
CBSAP method showed the highest correlation with
observed LAI and would suggest that determining the
optimal leaf area to sapwood area ratio is the only limita-
tion of this particular approach. On the other hand, there
can be large within-stand variation in the leaf area to
sapwood area ratio (Dean and Long 1986) and its deter-
mination would require destructive sampling, which may
limit its usefulness. Furthermore, in PNW Douglas-fir
plantations, wood quality can also vary tremendously
among alternative management regimes, primarily due to
resulting crown length and branch size. Representation of
crown structure is therefore highly desirable not only from
an ecophysiological perspective, but also in the context of
predicting consequences for wood product attributes and
performance.
All models will likely struggle to predict long-term
growth trends accurately, but short-term predictions (i.e.
5–10 years) can also be difficult in fast-growing species,
particularly under intensive management regimes and
across regions with widely varying growing conditions
like the PNW. Thus, an effective tool for assessing
intensive management of Douglas-fir plantations must
recognize that appropriate silvicultural regimes vary by
site and current stand structure. Developing a hybrid
model for intensively managed Douglas-fir plantations of
the PNW presents challenges beyond those typically
experienced in other regions, including: (1) complex ter-
rain with extremely steep slopes; (2) closed-canopy
plantations with relatively high LAIs (i.e. 7–10); (3)
severe moisture limitations imposed by extremely dry
summers; and (4) a wide variety of soil parent materials
and rock content. These factors drove many of the deci-
sions on the approaches taken in the hybrid model pre-
sented. For example, application of the radiation use
efficiency (RUE) may be questionable for Douglas-fir
stands because nearly 95% of the incoming radiation is
absorbed at a LAI of 6, which is a level exceeded by
many plantations in the PNW. Consequently, photosyn-
thesis in DF.HGS was predicted under mechanistic con-
straints imposed by daily soil moisture availability and
vapor pressure deficit. Scalars that vary from zero to one
are often insufficiently flexible to capture net effects of
daily water relations (Grant et al. 2006). Simulation
results from DF.HGS demonstrated its efficacy for cap-
turing the short-term variation in growth.
The relative performance of process-based, purely
empirical, and hybrid models depends on many factors
besides the extent to which mechanisms are represented
in their architecture, for example, conditions covered by
modeling versus evaluation datasets. A hybrid model
developed by Pinjuv et al. (2006) that accounted for
seasonal soil water availability reduced mean square error
of predicted stand basal area by *4% relative to an
empirical model. Dzierzon and Mason (2006) achieved a
reduction of 14 and 8% with a hybrid model for stand
basal area and top height growth, respectively, across
Pinus radiata plantations in New Zealand when com-
pared to regional empirical equations that relied on stand
age and site index. In contrast, DF.HGS achieved a level
of mean bias similar to a comparable set of individual
tree empirical growth equations across the full range of
tested silvicultural treatments. However, DF.HGS signif-
icantly reduced variance and bias of growth predictions
in unmanaged plantations in comparison with the
empirical model. This relative performance was attribut-
able only in part to incorporation of physiological pro-
cesses (total carbon assimilation, potential dry matter
production, and carbon allocation to volume production);
in reality, site-specific soils and climate information
localized the hybrid model to some degree. The empirical
approach relied solely on site index and therefore on a
single component of productivity. Improving growth
predictions in managed plantations will require future
refinements such as a better representation of stand
structural influences on radiation interception and improved
prediction of carrying capacity and nutrient dynamics
following fertilization. The challenge, however, is finding
the right balance between complexity and functionality
appropriate for accurately simulating stand and tree
responses to alternative management regimes. Several
existing models address many complex soil processes,
but the level of required input on soil nitrogen and car-
bon stores limits wide application of these models by
forest managers. Results from DF.HGS simulations sug-
gested that a simplified approach may be adequate for
predicting soil temperature, soil moisture, and nitrogen
mineralization.
Many shortcomings of the Douglas-fir hybrid model are
similar to those found by Ryan et al. (1996) in his
assessment of several first-generation process-based mod-
els. Missing or poorly represented components include
long-term controls on nitrogen mineralization, the uptake
and allocation of nitrogen by plants, allocation of carbon to
different tree components and processes (e.g., respiration)
and the response of all to changes in resource availability.
Satisfactory understanding of these key processes is still
lacking, so the necessary assumptions erode confidence in
model predictions. As one example, many process-based
models are built on the assumption that priorities exist for
carbon allocation. However, recent empirical evidence
does not support the hypothesized priorities (Litton et al.
2007). Resolving these differences through focused
research will be necessary before improvements in model
predictions can be realized.
336 Eur J Forest Res (2010) 129:325–338
123
Conclusions
Hybrid models that combine the features of empirical and
process-based models have been suggested as the next step
forward for improving growth projections in managed
stands (e.g. Landsberg 2003). A hybrid model constructed
for intensively managed Douglas-fir plantations in the
PNW significantly improved predictions of canopy attri-
butes, while the accuracy of volume growth predictions
were comparable to other modeling approaches. The
model’s robust performance resulted directly from: (1)
more detailed representation of individual tree crown
structure; (2) inclusion of major physiological mechanisms
at an appropriate spatial and temporal resolution; and (3)
the use of site-specific soils and climate information. Based
on this analysis, facets of hybrid models requiring future
improvement include: (1) effects of soil water and nutrient
availability on physiological processes, particularly respi-
ration and carbon allocation; (2) effects of silviculture and
environmental conditions on foliage age class dynamics,
seasonal variation in LAI, and net photosynthesis; and (3)
rigorous estimation and evaluation of key physiological
parameters.
Acknowledgments Data were generously made available by Adrian
Ares, Andreas Brunner, Hiroshii Ishii, Nathan McDowell, Cindy
Prescott, Emily Sinkhorn, Lars Vesterdal, and the Stand Management
Cooperative. Thanks to Uta Berger, David Marshall, David Hyink,
Jenny Grace, and Hans Pretzsch for reviewing a previous version of
this manuscript. This project was funded by the USDA Forest Service
PNW Research Station.
References
Ares A, Terry TA, Piatek KB, Harrison RB, Miller RE, Flaming BL,
Licata CW, Strahm BD, Harrington CA, Meade R, Anderson
HW, Brodie LC, Kraft JM (2007) The Fall River Long-Term Site
Productivity study in coastal Washington: site characteristics,
methods, and biomass and carbon and nitrogen stores before and
after harvest. In, GTR-PNW-691. USDA Forest Service. Pacific
Northwest Research Station, Portland, OR, p 85
Battaglia M, Sands P, White D, Mummery D (2004) CABALA: a
linked carbon, water and nitrogen model of forest growth for
silvicultural decision support. For Ecol Manag 193:251–282
Bond BJ, Kavanagh KL (1999) Stomatal behavior of four woody
species in relation to leaf-specific hydraulic conductance and
threshold water potential. Tree Physiol 19:503–510
Bristow KL, Campbell GS (1985) An equation for separating daily
solar irradiation into direct and diffuse components. Agric For
Meteorol 35:123–131
Bruce D (1981) Consistent height-growth and growth-rate estimates
for remeasured plots. For Sci 4:711–725
Campbell GS, Norman JM (1998) An introduction to environmental
biophysics. Springer, New York
Chappell HN, Prescott CE, Vesterdal L (1999) Long-term effects of
nitrogen fertilization on nitrogen availability in coastal Douglas-
fir forest floors. Soil Sci Soc Am J 63:1448–1454
Crookston NL, Stage AR (1999) Percent canopy cover and stand
structure statistics from the Forest Vegetation Simulator. In:
RMRS-GTR-24. USDA Forest Service, Rocky Mountain
Research Station, Ogden, UT, p 11
Dawson TE (1998) Fog in the California redwood forest: ecosystem
inputs and use by plants. Oecologia 117:476–485
Dean TJ, Long JN (1986) Variation in sapwood area-leaf area within
two stands of lodgepole pine. For Sci 32:749–758
Duursma RA, Robinson AP (2003) Bias in the mean tree model as a
consequence of Jensens’ inequality. For Ecol Manag 186:373–380
Dzierzon H, Mason EG (2006) Towards a nationwide growth and
yield model for radiata pine plantations in New Zealand. Can J
For Res 36:2533–2543
Ethier GJ, Livingston NJ, Harrison DL, Black TA, Morgan JA (2006)
Low stomatal and internal conductance to CO2 versus Rubisco
deactivation as determinants of the photosynthetic decline of
ageing evergreen leaves. Plant Cell Environ 29:2168–2184
Farquhar GD, von Caemmerer S, Berry JA (1980) A biochemical
model of photosynthetic CO2 assimilation in leaves of C3 leaves.
Planta 149:78–90
Gholz HL, Grier CC, Campbell AG, Brown AT (1979) Equations for
estimating biomass and leaf area of plants in the Pacific
Northwest. Oregon State University Forest Research Laboratory,
Corvallis
Grant RF, Zhang Y, Yuan F, Wang S, Hanson PJ, Gaumont-Guay D,
Chen J, Black TA, Barr A, Baldocchi DD, Arain A (2006)
Intercomparison of techniques to model water stress effects on
CO2 and energy exchange in temperate and boreal deciduous
forests. Ecol Model 196:289–312
Grier CC, Lee KM, Archibald RM (1984) Effects of urea fertilization
on allometric relations in young Douglas-fir trees. Can J For Res
14:900–904
Johnson IR, Thornley JHM (1987) A model of shoot:root partitioning
with optimum growth. Ann Bot 60:133–142
Katul GG, Ellsworth DS, Lai CT (2000) Modelling assimilation and
intercellular CO2 from measured conductance: a synthesis of
approaches. Plant Cell Environ 23:1313–1328
Keim RF (2004) Attenuation of rainfall by forest canopies. PhD
dissertation, College of Forestry, Oregon State University,
Corvallis, OR, p 150
Kirschbaum MUF (1999) CenW, a forest growth model with linked
carbon, energy, nutrient and water cycles. Ecol Model 181:17–59
Kucharik CJ, Norman JM, Gower ST (1999) Characterization of
radiation regimes in nonrandom forest canopies: theory, mea-
surement, and a simplified modeling approach. Tree Physiol
19:695–706
Lacointe A (2000) Carbon allocation among tree organs: a review of
basis processes and representation in functional-structural tree
models. Ann For Sci 57:521–533
Landsberg JJ (2003) Physiology in forest models: history and the
future. For Biom Model Inf Sci 1:49–63
Landsberg JJ, Waring RH (1997) A generalized model of forest
productivity using simplified concepts of radiation use effi-
ciency, carbon balance and partitioning. For Ecol Manag
95:209–228
Landsberg JJ, Waring RH, Coops NC (2003) Performance of the
forest productivity model 3-PG applied to a wide range of forest
types. For Ecol Manag 172:199–214
Litton CM, Raich JW, Ryan MG (2007) Carbon allocation in forest
ecosystems. Glob Chang Biol 13:2089–2109
Maguire DA, Batista JLF (1996) Sapwood taper models and implied
sapwood volume and foliage profiles for Coastal Douglas-fir.
Can J For Res 26:849–863
Maguire DA, Bennett WS (1996) Patterns in the vertical distribution
of foliage in young coastal Douglas-fir. Can J For Res 26:1991–
2005
Eur J Forest Res (2010) 129:325–338 337
123
Maguire DA, Kershaw JA, Hann DW (1991) Predicting the effects of
silvicultural regime on branch size and crown wood core in
Douglas-fir. For Sci 37:1409–1428
Maguire DA, Moeur M, Bennett WS (1994) Models for describing
basal diameter and vertical distribution of primary branches in
young Douglas-fir. For Ecol Manag 63:23–55
Makela A (1997) A carbon balance model of growth and self-pruning
in trees based on structural relationships. For Sci 42:7–24
Makela A, Landsberg J, Ek AR, Burk TE, Ter-Mikaelian M, Agren
GI, Olver CD, Puttonen P (2000) Process-based models for
forest ecosystem management: current state of the art and
challenges for practical implementation. Tree Physiol 20:
289–298
Manter DK, Bond BJ, Kavanagh KL, Stone JK, Filip GM (2003)
Modelling the impacts of the foliar pathogen, Phaeocryptopusgaeumannii, on Douglas-fir physiology: net canopy carbon
assimilation, needle abscission, and growth. Ecol Model
164:211–226
Manter DK, Kavanagh K, Rose CL (2005) Growth response of
Douglas-fir seedlings to nitrogen fertilization: importance of
Rubisco activation state and respiration rates. Tree Physiol
25:1015–1021
Medlyn BE, Dewar RC (1999) Comment on the article by R.H.War-
ing, J.J. Landsberg and M. Williams relating net primary
production to gross primary production. Tree Physiol 19:
137–138
Paul KI, Polglase PJ, O’Connell AM, Carlyle JC, Smethurst PJ,
Khanna PK (2002) Soil nitrogen availability predictor (SNAP): a
simple model for predicting mineralisation of nitrogen in forest
soils. Aust J Soil Res 40:1011–1026
Paul KI, Polglase PJ, Smethurst PJ, O’Connell AM, Carlyle JC,
Khanna PK (2004) Soil temperature under forests: a simple
model for predicting soil temperature under a range of forest
types. Agric For Meterol 121:167–182
Petersen KS, Ares A, Terry TA, Harrison RB (2008) Vegetation
competition effects on aboveground biomass and macronutri-
ents, leaf area, and crown structure in 5-year old Douglas-fir.
New For 35:299–311
Pinjuv G, Mason EG, Watt M (2006) Quantitative validation and
comparison of a range of forest growth model types. For Ecol
Manag 236:37–46
Pretzsch H, Biber P, Dursky J, von Gadow K, Hasenauer H, Kandler
G, Kenk G, Kublin E, Nagel J, Pukkala T, Skovsgaard JP, Sodtke
R, Sterba H (2006) Standaridizing and categorizing tree growth
models. In: Hasenauer H (ed) Sustainable forest management.
Springer, New York, pp 39–57
Pretzsch H, Grote R, Reineking B, Rotzer TH, Seifert ST (2008)
Models for forest ecosystem management: a European perspec-
tive. Ann Bot 101:1065–1087
Ranger J, Gelhaye D (2001) Belowground biomass and nutrient
content in a 47-year-old Douglas-fir plantation. Ann For Sci
58:423–430
Running SW, Gower ST (1991) FOREST-BGC, a general model of
forest ecosystem processes for regional applications. II. Dynamic
carbon allocation and nitrogen budgets. Tree Physiol 9:147–160
Ryan MG, Hunt ER, McMurtrie RE, A´
gren GI, Aber JD, Friend AD,
Rastetter EB, Pulliam WM, Raison RJ, Linder S (1996)
Comparing models of ecosystem function for temperate conifer
forests. I. Model description and validation. In: Breymeyer AI,
Hall DO, Melillo JM, A´
gren GI (eds) Global change: effects on
coniferous forests and grasslands. Wiley, New York, pp 313–362
Saxton KE, Rawls WJ (2006) Soil water characteristics estimates by
texture and organic matter for hydrologic solutions. Soil Sci Soc
Am J 70:1569–1578
Schwalm CR, Ek AR (2001) Climate change and site: relevant
mechanisms and modeling techniques. For Ecol Manag
150:241–257
Schwalm CR, Ek AR (2004) A process-based model of forest
ecosystems driven by meteorology. Ecol Model 179:317–348
Sinkhorn E (2007) Non-linear nitrogen dynamics and calcium
depletion along a temperate forest soil nitrogen gradient. M.S.
Thesis, College of Forestry, Oregon State University, p 207
Smith NJ (1993) Estimating leaf area index and light extinction
coefficients in stands of Douglas-fir (Pseudotsuga menziesii).Can J For Res 23:317–321
Snowdon P (2001) Short-term prediction of growth of Pinus radiatawith models incorporating indexes of annual climatic variation.
For Ecol Manag 152:1–11
Swenson JJ, Waring RH, Fan W, Coops NC (2005) Predicting site
index with a physiologically based growth model across Oregon,
USA. Can J For Res 35:1697–1707
Thornley JHM (1991) A transport-resistance model of forest growth
and partitioning. Ann Bot 68:211–226
Walters DK, Hann DW, Clyde MA (1985) Equations and tables
predicting gross total stem volumes in cubic feet for six major
conifers of southwest Oregon. Research Bulletin 50. Oregon State
University. Forest Research Laboratory, Corvallis, OR, p 36
Wang S, Chen W, Cihlar J (2002) New calculation methods of diurnal
distribution of solar radiation and its interception by canopy over
complex terrain. Ecol Model 155:191–204
Waring RH, Landsberg J, Williams M (1998) Net primary production
of forests: a constant fraction of gross primary production? Tree
Physiol 18:120–134
Warren CR, Adams MA (2006) Internal conductance does not scale
with photosynthetic capacity: implications for carbon isotopediscrimination and the economics of water and nitrogen use in
photosynthesis. Plant Cell Environ 29:192–201
Watt MS, Whitehead D, Richardson B, Mason EG, Leckie AC (2003)
Modelling the influence of weed competition on the growth of
young Pinus radiata at a dryland site. For Ecol Manag 178:
271–286
Weiskittel AR, Maguire DA (2006) Branch surface area and its
vertical distribution in coastal Douglas-fir. Trees: Struct Funct
20:657–667
Weiskittel AR, Maguire DA (2007) Response of Douglas-fir leaf area
index and litterfall dynamics to Swiss needle cast in north coastal
Oregon, USA. Ann For Sci 64:121–132
Weiskittel AR, Maguire DA, Garber SM, Kanaskie A (2006)
Influence of Swiss needle cast on foliage age class structure
and vertical distribution in Douglas-fir plantations of north
coastal Oregon. Can J For Res 36:1497–1508
Weiskittel AR, Garber SM, Johnson GP, Maguire DA, Monserud RA
(2007a) Annualized diameter and height growth equations for
Pacific Northwest plantation-grown Douglas-fir, western hem-
lock, and red alder. For Ecol Manag 250:266–278
Weiskittel AR, Maguire DA, Monserud RA (2007b) Modeling crown
structural responses to competing vegetation control, thinning,
fertilization, and Swiss needle cast in coastal Douglas-fir of the
Pacific Northwest, USA. For Ecol Manag 245:96–109
Woollons RC, Snowdon P, Mitchell ND (1997) Augmenting empir-
ical stand projection equations with edaphic and climate
variables. For Ecol Manag 98:267–274
338 Eur J Forest Res (2010) 129:325–338
123