The Canadian Terrestrial Ecosystem Model (CTEM) is the interactive
vegetation component in the Earth system model of the Canadian
Centre for Climate Modelling and Analysis. CTEM models
land–atmosphere exchange of
Dynamic global vegetation models (DGVMs) are now considered an integral
component of Earth system models (ESMs). DGVMs model vegetation as a dynamic
component of the Earth system, allowing simulation of the atmosphere–land
flux of
DGVMs typically represent vegetation in terms of a modest number of plant
functional types (PFTs). This simplification is justified by the ability to
categorize plant species on the basis of their form and interaction with the
environment
Parameters used in the competition module of CTEM v. 2.0.
In its simplest form, the dynamic behaviour of vegetation can be described by two aspects: changes in plant structure and changes in areal coverage. First, as vegetation responds to changes and variability in climate, its structure changes affecting its height, LAI and rooting depth, typically on seasonal to decadal timescales. Second, vegetation also adapts by changing its areal extent. Areal changes are slower than the structural changes, typically occurring on decadal to centennial timescales for woody vegetation and sub-decadal to decadal timescales for herbaceous vegetation. Together, these changes in structure and areal extent capture vegetation's dynamic behaviour in response to changes and variability in climate at different timescales.
DGVMs may incorporate only the structural, or both the structural and areal,
aspects of vegetation dynamics. The Canadian Terrestrial Ecosystem Model
(CTEM)
Removing the constraint of specified fractional coverage of PFTs and simulating the areal vegetation dynamics adds another degree of freedom to DGVMs. Simulating both structural and areal vegetation dynamics in a realistic manner is a more stringent test of a DGVM's abilities. An incorrect simulation of the geographical distribution of PFTs will lead to a similarly flawed distribution of primary terrestrial carbon pools and fluxes, regardless if the model correctly simulates the structural vegetation dynamics.
Modelling areal dynamics of vegetation requires simulating competition for
available space and resources between PFTs. In the real world, plants compete
for space to acquire both aboveground (light) and belowground (water and
nutrients) resources. Attempts to capture competition between plants to allow
these interactions have generally been accomplished using three different
kind of models: (i) theoretical models, (ii)
Theoretical models (e.g.
Here, we evaluate CTEM v. 2.0, which explicitly models the competition for
space between PFTs using a modified version of the L–V equations
Version 1 of the CTEM is the terrestrial carbon cycle component of the second
generation Canadian Earth System Model (CanESM2)
Version 1.0 of CTEM is described in a collection of papers detailing
parametrization of photosynthesis, autotrophic and heterotrophic respiration
A parametrization for competition between PFTs in an earlier version of CTEM
is described by
Competition between PFTs in CTEM is based upon modified L–V equations
The change in fractional coverage (
The value of parameter
For the usual form of the L–V equations with
For
In the standard L–V equations for predator–prey interactions coexistence is
possible because the predator depends on prey for its food and so the
predator population suffers as the prey population declines. This is in
contrast to the application of the equations for competition between PFTs
where the dominant PFT does not depend on sub-dominant PFTs for its existence
and is thus able to exclude them completely. The PFTs interact with each
other through the invasion term
The PFT-dependent colonization rate (
The value of
The PFT-dependent mortality rate (
Intrinsic- or age-related mortality uses a PFT-specific maximum age,
Bioclimatic limits used in the competition parametrization of CTEM
v. 2.0 as described in Sect.
Mortality associated with bioclimatic criteria,
The annual mortality rates for
The mortality associated with bioclimatic criteria,
We perform three equilibrium simulations for pre-industrial conditions on
a Gaussian
The simulations are driven with the Climate Research Unit – National Centers for Environmental Prediction (CRU-NCEP) v. 4 climate data
The simulations with prescribed PFT fractions (PRES) use a reconstruction of
land cover for the year 1861 based on the approach described in
A Moderate-Resolution Imaging Spectroradiometer (MODIS)-derived land cover
product
The GLC2000 land cover product comprising of 22 land cover types is mapped on
to the nine PFTs represented in CTEM by W2006 (their Table 2). W2006 split
the broadleaf deciduous PFT into cold deciduous and drought deciduous
versions. They made the simple but reasonably realistic assumption that at
high latitudes (above 34
The MODIS land cover product contains 17 land cover types and we map it to
CTEM PFTs following the mapping scheme of W2006, as closely as possible (see
Table S1 in the Supplement). The MODIS product is averaged across the years 2001–2011
and interpolated to the Gaussian
Both the MODIS-derived product (17 land cover types) and the modified
W2006 land cover (based on GLC2000's 22 land cover types) product are
subject to errors in categorizing remotely sensed vegetation into
broad-scale vegetation types and then their further mapping onto
CTEM's nine PFTs. The latter mapping requires assumptions about which
of the land cover types contribute to which of the CTEM PFTs and in
what proportion, including the bare fraction. For instance, W2006
assign the GLC2000
We also compare the latitudinal distribution of simulated gross primary
productivity (GPP), vegetation biomass and soil carbon with observation-based
estimates of these quantities. The observation-based estimate of GPP is from
The performance of the original L–V-based competition parametrization
(LVCOMP) and its modified version as implemented in CTEM (CTCOMP) is
evaluated at the most basic level in Fig.
Comparison of observation-based and simulated global areal
extent of grass, treed, bare ground and total vegetated area (sum of
grass, treed and crop areas) from pre-industrial simulations that
use the CTEM competition scheme (modified L–V; CTCOMP) and the
unmodified L–V equations (LVCOMP)
Global tree
Overall, the modelled global coverage of trees and grasses, and the global
bare and vegetated areas in the CTCOMP and LVCOMP simulations compare
reasonably with their observation-based estimates. The global tree cover in
the CTCOMP simulation is about 6 % lower than the W2006 data set but 20 %
more than the MODIS-derived estimates, which are for the modern period.
Simulated coverage by CTCOMP lies in-between the observation-based estimates
based on the modified W2006 and the MODIS-derived products, except for the
area covered by grasses, which is lower than both observation-based estimates
(13 and 24 %, respectively). LVCOMP simulated global area covered by trees
is the highest while coverage of grasses and vegetated area is the lowest compared to
the CTCOMP simulation and observation-based estimates. Total vegetation cover
in the CTCOMP (90.7
Figure
The global areas of individual PFTs in Fig.
Comparison of simulated fractional coverage of tree PFTs in the
LVCOMP (based on the original L–V equations, left column) and CTCOMP (based
on modified L–V equations, middle column) simulations with the
observation-based estimate from the modified
As with Fig.
Figure
The broad spatial distribution of tree cover is simulated reasonably
well in the CTCOMP simulation, including the boreal and tropical
forests, although the model tends to simulate less tree cover in arid
regions, especially the south-western United States. The spatial correlation between
the modelled tree fraction in the CTCOMP simulation and the modified
W2006 data set is 0.79 with a RMSD of 22.8 %. Three main factors
make simulating tree cover challenging in our modelling
framework. First, the sub-grid scale climatic niches are unresolved at
the
Comparison of simulated fractional coverage of grass PFTs in the
LVCOMP (based on the original L–V equations, left column) and CTCOMP (based
on modified L–V equations, middle column) simulations with the
observation-based estimate from the modified
The spatial distribution of trees in the LVCOMP simulation is similar
to that in the CTCOMP simulation but generally shows less tree cover
in arid regions and higher cover in predominately forested regions
compared to both the W2006 data set and the CTCOMP simulation. The
LVCOMP simulation demonstrates similar spatial correlation (0.76) but
poorer RMSD (27.5 %) against the modified W2006 data set than the
CTCOMP simulation (Fig.
The spatial coverage of grass in the CTCOMP simulation shows a fairly
reasonable agreement compared to modified W2006 data set, especially
on the African continent (Fig.
Grass cover in CTEM, regardless of the competition parametrization (CTCOMP or LVCOMP), is influenced by three main factors: (i) tree coverage, as trees are considered superior to grasses because of their ability to invade them, (ii) moisture availability and (iii) the disturbance regime. Higher disturbance rates act to reduce tree cover and increase grass cover because grasses colonize faster than trees after a disturbance. Frequent disturbance thus reduces the superiority of trees over grasses.
Grass coverage is overestimated in the northern high latitude regions
in the CTCOMP simulation (Fig.
The grassland extent in the parts of the US Plains and Canadian Prairies is also underestimated in the CTCOMP simulation. The modified
W2006 data set shows around 60 % grass cover in this region while
CTEM estimates about 30 % grass cover. This underestimate could be
due to the following three reasons. First, fire is simulated
interactively in CTEM and biases in simulated area burned will
influence the PFT dynamics as fire tends to reduce tree cover and
increase grass cover. The fire parametrization includes both human
(as a function of population density) and natural influences but
cultural differences in the human influence are not represented (see
Appendix
Grass cover in the LVCOMP simulation is relegated to areas where trees
are not able to effectively colonize, due to conditions being overly
arid, overly cold and/or with high disturbance regimes. In the LVCOMP
simulation, even more extensive grass cover is simulated in the high
latitudes than in the CTCOMP simulation, approaching 90 % in many
regions. The LVCOMP simulation's low grass cover in many other regions
reflects the high tree cover estimated in the LVCOMP simulation due to
its inability to allow appropriate PFT coexistence (as mentioned
earlier) acting to exclude grasses. As a result the grass cover in the
LVCOMP simulation is low and patchy and differs greatly from the
modified W2006 data set, as is also indicated by a low spatial
correlation and higher RMSD in Fig.
Scatter plot of the model simulated vegetation cover root mean
square difference and correlation coefficient as compared to the
observation-based estimate derived from the modified
Bare ground in the modified W2006 data set occurs primarily in arid,
mountainous and arctic regions (Fig.
The comparison of geographical distributions of simulated tree, grass
cover and bare ground cover with observation-based estimates shows
that, while some limitations remain, the competition parametrization
of CTEM performs reasonably realistically. A more stringent test of
the model performance is the comparison of geographical distributions
of individual tree and grass PFTs against observation-based estimates,
as shown in Fig.
The modelled global total areal extent of the needleleaf evergreen (NDL-EVG) tree PFT
in the CTCOMP simulation is around 5 % less than the modified
W2006 estimate and about 21 % more than the modern-day MODIS-derived
estimate (Fig.
The simulated distribution of NDL-EVG trees in the LVCOMP simulation differs
greatly from the modified W2006 estimate with much lower coverage of NDL-EVG
trees except in certain regions where the simulated coverage of NDL-EVG trees
is high (
The observation-based distribution of needleleaf deciduous (NDL-DCD) trees is
primarily limited to areas in eastern Siberia with some additional areas of
low coverage in North America (
While CTEM parametrizes the broadleaved evergreen (BDL-EVG) PFT as tropical
trees (through bioclimatic limits; see Sect.
The western margin of the Amazon River basin shows the limitations of the
In CTEM, broadleaf deciduous (BDL-DCD) trees are divided into cold
(BDL-DCD-COLD) and drought/dry (BDL-DCD-DRY) sub-types. The
bioclimatic indices for BDL-DCD-COLD and BDL-DCD-DRY trees are assigned so
that their geographical distributions are broadly limited to regions where
their deciduousness is primarily controlled by temperature and soil moisture,
respectively. For the observation-based data set, this distinction is more
arbitrary. W2006 used latitudinal limits of
The simulated geographical distribution of BDL-DCD-COLD in the CTCOMP
simulation is broadly similar to the modified W2006 data set, although some
differences remain. Compared to the modified W2006 data set, the model
simulates larger BDL-DCD-COLD extent in western Canada (and lower NDL-EVG as
a result) and smaller BDL-DCD-COLD coverage in south-eastern Europe (with
comparably more NDL-EVG). These areas demonstrate that the competition
between the NDL-EVG and BDL-DCD-COLD PFTs is not perfectly modelled in these
regions. The arbitrary latitudinal limits used by W2006 imply that
BDL-DCD-COLD trees exist as far north as Australia's Gold Coast in the
eastern state of Queensland, which is certainly not realistic given its
tropical climate. The model does not simulate any broad-scale existence of
BDL-DCD-COLD trees between 24 and 34
The simulated geographical distribution of BDL-DCD-DRY trees in the CTCOMP
simulation is also broadly similar to the modified W2006 data set with some
differences over India, and in southern Africa near Botswana and Zambia. CTEM
coverage in these regions is lower by approximately 20 % than the
modified W2006 estimate. As with other tree PFTs, the simulated areal extent
of BDL-DCD-DRY trees in the LVCOMP simulation is more limited compared to the
CTCOMP simulation and the modified W2006 data set, but of a higher percent
cover in the grid cells where BDL-DCD-DRY trees do exist. The global total
areal coverage of BDL-DCD-DRY trees in the CTCOMP and LVCOMP simulations are
The total BDL-DCD coverage, consisting of both the cold and dry
deciduous sub-types, in the CTCOMP simulation is similar to the
estimates based on MODIS and modified W2006 products but much higher
in the LVCOMP simulation (Fig.
Both
Zonal mean GPP (top), vegetation biomass (middle) and soil carbon
(bottom) for CTEM simulations using (i) prescribed PFT fractional cover from
the modified
Figure
Reasons for overly extensive bare ground include either the inadequacies of
using a single grass PFT globally for each photosynthetic pathway (
The simulated distribution of
The differences between the simulations with prescribed fractional coverage
of PFTs (PRES) and the simulations in which fractional coverage of PFTs are
dynamically simulated (CTCOMP and LVCOMP) are also the result of somewhat
different parameter values between the two. The simulations with dynamically
determined fractional coverage of PFTs provide an additional constraint,
which the model must meet, i.e. the observation-based estimates of the
fractional coverage of PFTs. We found that the default parameter values used
for simulations with prescribed fractional coverage of PFTs did not yield
optimum comparison with observation-based estimates of fractional coverage
of PFTs. Changes in parameter values were required because the simulations
with dynamically determined fractional coverage of PFTs include additional
processes such as mortality generating bare ground (see Sect.
Comparison of primary model quantities from the three CLASS–CTEM
simulations with observation-based estimates and other model estimates. The
CTCOMP simulation uses a modified form of the Lotka–Volterra equations for
simulating competition between PFTs. The LVCOMP simulation uses the
unmodified Lotka–Volterra equations. The PRES simulation uses prescribed
fractional coverage of PFTs based on the modified
Table
GPP is slightly higher in the CTCOMP and LVCOMP simulations compared to the
PRES simulation. The primary reason for this increase is the changes in PFT
specific parameters, rather than the spatial distribution of PFTs itself.
Performing the PRES simulation, with the same parameter set as in the CTCOMP
and LVCOMP simulations, yields a global GPP value of
140.8
Similar to GPP, NPP in the CTCOMP and LVCOMP simulations is 9.1 and 6.2 %
higher, respectively, compared to the PRES simulation
(Table
The simulated vegetation biomass and soil carbon mass are the lowest in the CTCOMP simulation amongst the three simulations. The LVCOMP simulation conversely yields the highest vegetation biomass and soil carbon mass. The PRES simulation has intermediate values of vegetation biomass and soil carbon. Simulated litter mass is similar across all the three simulations. Global values of vegetation, soil carbon and litter mass from all simulations compare reasonably with other estimates.
Compared to the PRES simulation, the zonally averaged vegetation biomass in
the CTCOMP and LVCOMP simulations is higher in the high southern latitudes
(
The zonal distribution of soil carbon (Fig.
Annual fire emissions are the highest in the CTCOMP simulation even though it has
the lowest annual burned area (24.1 % less than in the PRES simulation)
while the LVCOMP has the largest burned area (7.8 % more than in the PRES
simulation) but the same amount of emissions as the PRES simulation
(Table
Geographical distribution of modelled GPP
The geographical distribution of GPP, vegetation biomass, soil carbon
mass and burned area from the PRES, CTCOMP and LVCOMP simulations are
compared in Figs.
The geographical distribution of GPP (Fig.
The geographical distribution of vegetation biomass between the three
simulations (Fig.
As with the zonally averaged values, the geographical distribution of soil
carbon (Fig.
Geographical distribution of modelled soil carbon mass
The geographical distribution of annual burned area is broadly similar in the
PRES, CTCOMP and LVCOMP simulations. All three simulations show higher area
burned in savanna regions of the tropics (Fig.
Overall the impact of dynamically modelling fractional coverage of PFTs in the CTCOMP simulation yields the largest differences for simulated vegetation biomass and annual area burned by fire, while other simulated primary carbon pools and fluxes remain similar to those in the PRES simulation and to observation-based estimates.
Modelling vegetation spatial dynamics explicitly in DGVMs overcomes the
limitations inherent in prescribing a vegetation cover that is unable to
respond to changes in climate, atmospheric
Plot showing the maximum number of PFTs in the pre-industrial equilibrium CTCOMP and LVCOMP simulations that can compete for space within a grid cell based upon each PFT's bioclimatic limits. The total number of natural non-crop PFTs represented in CTEM is seven.
The strength of bioclimatic limits imposed within models varies greatly. At
one end of the spectrum there are biogeographic models where present-day
geographical distributions of PFTs are used to derive climate envelopes
within which PFTs can exist. While this approach gives information about
a PFT's areal extent under present-day conditions, the information is
essentially binary, i.e. within a grid cell the PFTs exist or they do not.
This biogeography approach neither provides any information about the
fraction of a grid cell that the PFTs occupy nor how the PFTs' competitive
success might evolve under changing environmental conditions. At the other
end of the spectrum are parametrizations that require no bioclimatic
information to limit species extent, i.e. each PFT's geographic extent is
solely derived from its physiological responses and competitive interactions
with other PFTs arising from modelled processes. We are not aware of any
competition parametrizations used within a global-scale DGVM that does not
require the use of bioclimatic constraints. There are, however, recent
attempts at developing models without bioclimatic limitations at the
regional-scale.
We believe the use of bioclimatic constraints is reasonably moderate in CTEM.
With the exception of the needleleaf deciduous PFT, the model uses fairly
relaxed bioclimatic constraints for its tree PFTs. We do not use any
bioclimatic limits for the
Equilibrium offline simulations corresponding to the year 1861, using the
original and modified L–V equations show that the model is able to capture
the broad geographical distributions of tree and grass cover, and the bare
fraction, when compared to the observation-based modified W2006 data set,
especially when using the modified L–V equations. The global areas covered
by tree and grass PFTs, the bare fraction, and the individual PFTs also
compare reasonably to the observation-based estimates of the modified W2006
data set and those derived from a MODIS product
Some limitations are evident in our simulations. Overall, the simulated tree and grass cover is low in hot arid regions and the simulated grass cover is overly high in cold arid regions. As a result the model generates more bare fraction in Australia, the Andes mountains of South America and the Kalahari desert of Africa, and too little bare fraction in the high Siberian Arctic. The simulated geographical distribution of individual PFTs appears reasonable although limitations remain here too. The model simulates a small fractional coverage of NDL-EVG trees in warm regions of southern Africa, Australia and South America that is not consistent with the observation-based estimate based on the modified W2006 data set. The gradation in the fractional coverage of PFTs from regions of high to low coverage is simulated much better when the modified version of the L–V equations is used, but the simulated fractional coverage is still not as graded as in the modified W2006 data set. These limitations are due primarily to (i) the coarse resolution used in our study, which does not allow resolution of climatic niches, (ii) the absence of shrub and moss/lichen PFTs and (iii) the small number of natural PFTs (seven) that are represented in CTEM. These constraints limit the model's ability to capture distributions of plants within the same broad functional group but that exist in geographically and climatically distinct regions, and will be the focus of future model development.
Modelling competition between PFTs in CTEM required adjusting some of the
model parameters in order to realistically simulate the fractional coverage
of its seven natural, non-crop PFTs. These parameter changes removed the
positive bias in simulated vegetation biomass in certain high-latitude
regions but yield too high vegetation biomass in the tropics. Overall,
simulated global values of GPP, NPP, vegetation biomass, soil carbon and area
burned in the CTCOMP simulation compare reasonably with observation-based and
other model estimates (Table
Despite its limitations, the behaviour of the competition module that uses
the modified L–V equations, in the CLASS–CTEM modelling framework, is
sufficiently realistic to yield a tool with which to study the impact of
changes in climate and atmospheric
The basic model structure of CTEM includes three live vegetation components
(leaf (L), stem (S) and root (R)) and two dead carbon pools (litter or
detritus (D) and soil carbon (H)). The amount of carbon in these pools
(
All biogeochemical processes in CTEM are simulated at a daily time step
except gross photosynthetic uptake and associated calculation of canopy
conductance, which are simulated on a half hour time step with CLASS. The
photosynthesis module of CTEM calculates the net canopy photosynthesis rate,
which, together with atmospheric
The photosynthesis parametrization is based upon the approach of
The gross leaf photosynthesis rate,
The Rubisco enzyme limited photosynthesis rate,
The transport capacity (
Parameters used in the photosynthesis module of CTEM v. 2.0.
The influence of soil moisture stress is simulated via
The
The current version of CTEM does not include nutrient constraints on
photosynthesis and, as a result, increasing atmospheric
Finally, the leaf-level gross photosynthesis rate,
The net canopy photosynthetic rate,
When using the
Canopy (
Version 2.0 of CTEM calculates autotrophic respiratory fluxes
(
Autotrophic respiration (
Maintenance respiration from the stem and root components is estimated based
on PFT-specific base respiration rates (
PFT-specific base respiration rates
(
Growth respiration,
Heterotrophic respiration,
PFT-specific base respiration rates
(
The effect of soil moisture is accounted for via dependence on soil matric
potential (
The temperature of the litter pool is a weighted average of the temperature
of the top soil layer (
The response of heterotrophic respiration to soil moisture is formulated
through soil matric potential (
for
The amount of humidified litter, which is transferred from the litter to the
soil carbon pool (
With heterotrophic respiration known, net ecosystem productivity (NEP)
is calculated as
Positive NPP is allocated daily to the leaf, stem and root components, which
generally causes their respective biomass to increase, although the biomass
may also decrease depending on the autotrophic respiration flux of
a component. Negative NPP generally causes net carbon loss from the
components. While CTEM offers the ability to use both specified constant or
dynamically calculated allocation fractions for leaves, stems and roots, in
practice the dynamic allocation fractions are primarily used. The formulation
used in CTEM v. 2.0 differs from that for CTEM v. 1.0 as described in
The dynamic allocation to the live plant tissues is based on the light, water and leaf phenological status of vegetation. The preferential allocation of carbon to the different tissue pools is based on three assumptions: (i) if soil moisture is limiting, carbon should be preferentially allocated to roots for greater access to water, (ii) if LAI is low, carbon should be allocated to leaves for enhanced photosynthesis and finally (iii) carbon is allocated to the stem to increase vegetation height and lateral spread of vegetation when the increase in LAI results in a decrease in light penetration.
The vegetation water status,
The light status,
PFT-specific base allocation fractions for leaf
(
Grasses do not have a stem component (i.e.
The dynamic allocation fractions are superseded under three conditions.
First, during the leaf onset for crops and deciduous trees, all carbon must
be allocated to leaves (
The leaf phenology parametrization used in CTEM v. 1.0 is described in
detail by
The leaf phenological state transitions are dependent upon environmental
conditions. In particular, the transition from no leaves/dormant state to the
maximum growth state is based on the carbon-gain approach. CTEM uses
PFT-specific parameters used in the phenology module of CTEM v. 2.0:
maximum cold (
The transition from the normal growth state to the leaf fall state is
triggered by unfavourable environmental conditions and shorter day length.
Broadleaf deciduous trees transition to the leaf fall state when either: (i) day
length is less than 11 h and the rooting zone temperature drops below
11.15
The model vegetation is able to transition between the different leaf phenological states in response to changing conditions. For example, a leaf out in spring for broadleaf cold deciduous trees can be interrupted by a cold event when the vegetation goes into a leaf fall state until the return of more favourable conditions.
Leaf litter generation is caused by normal turnover of leaves
(
The conversion of leaf carbon to leaf litter (
The turnover of stem and root components is modelled via their PFT-dependent
specified lifetimes. The litter generation (
With gross canopy photosynthesis rate (
When the daily NPP (
If the daily NPP is negative (
The rate change equations for the litter and soil carbon pools are given by
The time-varying biomass in the leaves (
Leaf biomass is converted to LAI using specific leaf area (SLA,
The vegetation height (
CTEM dynamically simulates root distribution and depth in soil following
PFT-specific leaf lifespans (
The rooting depth
CTEM v. 2.0 represents disturbance as both natural and
human-influenced fires. The original fire parametrization
corresponding to CTEM v. 1.0 is described in
Fire in CTEM is simulated using a process-based scheme of intermediate
complexity that accounts for all elements of the fire triangle: fuel load,
combustibility of fuel, and an ignition source. CTEM represents the
probability of a fire occurrence (
The
The probability of fire based on the presence of ignition sources
(
The probability of fire due to the combustibility of the fuel,
The probability of fire based on the moisture content in the
The area burned (
The fire spread rate in the downwind direction (
PFT-specific parameter values used in the fire module of CTEM
v. 2.0. The maximum fire spread rate,
The wind speed (
The dependence of fire spread rate on the rooting zone and duff soil
wetness,
With fire spread rate determined, and the geometry of the burned area
defined, the area burned in 1 day,
The fire extinguishing probability,
The dot product of
Litter generated by fire is based on similar mortality factors, which reflect
a PFT's susceptibility to damage due to fire
When CTEM is run with prescribed PFT fractional cover, the area of PFTs does
not change and the fire-related emissions of
CTEM explicitly simulates C
The land use change (LUC) module of CTEM is based on
A decrease in the area of natural non-crop PFTs, associated with an increase
in area of crop PFTs, results in deforested biomass (while the term
When croplands are abandoned, the area of natural PFTs increases. In
simulations with prescribed fractional coverage of PFTs this results in
a decreased carbon density for all model pools as the same amount of carbon
is spread over a larger fraction of the grid cell. This reduced density
implies that natural vegetation is able to take up carbon as it comes into
equilibrium with the driving climate and atmospheric
Fortan code for CLASS–CTEM modelling framework is available on request and
upon agreeing to Environment Canada's licensing agreement available at
The works published in this journal are distributed under the Creative Commons Attribution 3.0 License. This license does not affect the Crown copyright work, which is re-usable under the Open Government Licence (OGL). The Creative Commons Attribution 3.0 License and the OGL are interoperable and do not conflict with, reduce or limit each other. © Crown copyright 2015.
J. R. Melton was supported by a National Scientific and Engineering Research Council of Canada (NSERC) Visiting Postdoctoral Fellowship. We thank Ben Poulter for sharing his version of the Saatchi tropical vegetation biomass data set. Rudra Shrestha wrote the code that calculates the dry season length. We thank two anonymous reviewers and the editor for their comments, which helped to improve this manuscript. Edited by: G. Munhoven