Appendix A Caveats, impacts on results and how each was addressed.
Type of Caveat
Caveat
Impact
How addressed
Ecosim models
Structure of EwE models
dependent upon user aims,
expertise, and philosophy
Possible designer bias in
EwE results
Used a suite of EwE models, performed
comparative analysis of Ecopath models
Not all Ecosim models have
been calibrated to timeseries data
Some models lack realism
for effects of biological
interactions
Investigate how predictions change for
different settings of the key Ecosim
vulnerability parameters
No changes in distribution
of species
Changes in ecological
interactions that occur due
distribution changes not
considered
Because functional groups (and generally not
species) are used in the models, changes in
species distribution are considered indirectly
if range changes substitute ecologically
equivalent species
Effects of global warming
on consumer species not
modelled
Global warming may impact
physiology and ultimately
food web dynamics
Currently technically difficult to address this
effect in EwE and parameterise relationships
for a large number of functional groups
Human activities such as
fishing and nutrient run-off
are assumed to remain
constant
Human activities will
change in the future and
these will affect marine
ecosystems
We are interested in the effect of climate
change, so we have isolated it from other
effects
Uncertainty in NPZD model
predictions
Contributes to uncertainty in
predictions of ecosystem
response to climate change
The sign of production and nutrient change is
robust to uncertainty. Magnitude of change
less certain, but predictions comparable to
other NPZD models (Sarmiento et al. 2004)
Spatial resolution of NPZD
output too coarse
Fine and meso-scale
oceanographic processes not
resolved
Used the finest scale future productivity
available
Effects of extreme events
(e.g. cyclones) on primary
productivity not considered
Extreme events contribute to
variability in ecosystem
dynamics
These process are not well predicted by
current climate models, so we chose to focus
on climate trends
NPZD does not incorporate
land-based nutrient inputs to
coastal waters
Land-based nutrient inputs
can be considerable in bays
and estuaries
Nutrient availability inshore dependent on
offshore nutrients in Australia. Further
modelling of climate impacts on nutrient
inputs required in future.
No feedbacks from the
ecosystem to primary
production rate
Feedbacks probably weak in
open ocean ecosystems, but
may be important in
enclosed bay ecosystems
This subject is the focus of ongoing work
(Travers et al. 2007, Fulton in press)
Uncertainty in GCM
predictions
Contributes to uncertainty in
predictions of ecosystem
response to climate change
Sensitivity of Ecosim models to primary
production change cover the range of
plausible scenarios
Uncertainty in future greenhouse gas emissions
Contributes to uncertainty in
predictions of ecosystem
response to climate change
Used IPCC emission scenario and sensitivity
analyses
Primary
production
models
Climate models
References
Fulton EA (2009) Approaches to end to end ecosystem models. Ices Journal of Marine Science,
In press
Sarmiento JL, Slater R, Barber R, et al. (2004) Response of ocean ecosystems to climate
warming. Global Biogeochemical Cycles, 18, GB3003, doi:10.1029/2003GB002134
Travers M, Shin YJ, Jennings S, Cury P (2007) Towards end-to-end models for investigating the
effects of climate and fishing in marine ecosystems. Progress in Oceanography, 75, 751-770.
Appendix B Predicting benthic primary production rate
Benthic production models, for seagrasses, benthic micro-algae and macro-algae, were based on
those by Murray & Parslow (1999), which have also been used to model primary production for
other food web models (e.g. Fulton et al. 2004). Similar to the NPZD, production rate of these
groups was also dependent upon nitrogen availability, light and temperature. Nitrogen
availability was obtained from the NPZD model, and temperature and light level projections
were obtained from the GCM and were averaged across the Ecosim model regions, as for
phytoplankton production. Benthic production rates were calculated at monthly time intervals,
using mean monthly light, temperature and nutrients. Production rate (grams nitrogen per month
per gram of primary producer) in the primary production models is estimated independent of
population biomass.
For seagrass, macro-algae and benthic micro-algae, the production rate equation for each primary
producer type i at month t, was:
Pi,t = μmax,i Ci,t min[hNi,t , hIi,t]
(1)
Where Pi,t is production rate (grams of nitrogen per month per gram of producer biomass); μmax,i
is the maximum production rate per month; Ci,t is the temperature rate multiplication factor;
hNi,t is the nitrogen multiplier; and hIi,t is the light multiplier. At elevated nutrient levels, growth
of epiphytic algae hinders seagrass production (Short & Neckles 1999). Therefore, the
production rate for seagrass was reduced by an amount proportional to the predicted nitrogen
concentration for each region. The proportional reduction in seagrass production rate was also
dependent on the temperature rate multiplier, because epiphytic algal growth is temperature
dependent (Murray & Parslow 1999).
We used Liebig’s law of the minimum for light and nutrient limitation, where only the factor in
shortest supply at a time is limiting, because multiplicative models can predict unrealistically low
production rates for moderate levels of light and nutrients (Murray & Parslow 1999). Light and
nutrient limitation multipliers are calculated using the Michalis-Menton equation:
hNi,t = DINt /( KN,i + DINt)
(2)
Where DINt is dissolved inorganic nitrogen (N m-3); and KN,i is the nitrogen half saturation
constant (N m-3). We assumed that DINt is correlated with nitrogen in the sediments, thus,
seagrasses would respond similarly to changes in water nitrogen content.
Light attenuation is affected by temperature, therefore:
hIi,t = It / (KI,i Ci,t + It)
(3)
Where It is irradiance in watts per m2; and KI,i is the light half saturation constant in watts per m2.
Irradiance at depth was determined by the light extinction parameter. We assumed the same light
extinction parameter for all regions, because changing this parameter did not qualitatively affect
results. The temperature rate multiplication factor was constant for all parameters and is
calculated using the equation:
Ci,t = Q10,i (Tt-TE) / 10
(4)
Where Q10,i is the factor by which rate processes increase for a 10oC increase in temperature; Tt
is the temperature in the environment in oC; and TE is the standardisation temperature. The
standardisation temperature was set to the mean for each model region (Fig. 1) from the GCM in
the years 2000-2010. Parameters for each primary producer group are shown in Table B1.
Table B1 Parameter value ranges for benthic production models
Primary
producer
group
Parameter
Units
Benthic microalgae
Maximum
production rate
mgN/day/
gram of
algae
Light half
saturation
constant
W/m2
Nutrient half
saturation
constant
mgN/m3
Maximum
production rate
mgN/d/gra
m of algae
Light half
saturation
constant
W/m2
5 - 20
Nutrient half
saturation
constant
mgN/m3
1 – 30
Maximum
production rate
mgN/d/
gram of
seagrass
0.035 – 0.1
Light half
saturation
constant
W/m2
30 – 60
Nutrient half
saturation
constant
mgN/m3
5 – 50
Epiphytic algal
multiplier
-
0.005 – 0.03
Q10
-
2
Macro-algae
Seagrass
All groups
Range
0.1 – 0.75
2.5 - 12
100 - 300
0.05 - 2
References
Fulton EA, Smith ADM (2004) Lessons learnt from a comparison of three ecosystem models for
Port Phillip Bay, Australia. African Journal of Marine Science, 26, 219-243.
Murray AG, Parslow JS (1999) Modelling of nutrient impacts in Port Phillip Bay - a semienclosed marine Australian ecosystem. Marine and Freshwater Research, 50, 597-611.
Short FT, Neckles HA (1999) The effects of global climate change on seagrasses. Aquatic
Botany, 63, 169-196.
Appendix C Habitat mediation
Non-trophic interactions may be important mediators of the response of functional group
biomasses to primary production change. We contrasted two scenarios of consumer habitat
dependence. The first scenario has no effect of the abundance of benthic-macro producers
(seagrass, mangroves and macro-algae) on the survival of consumer species and was used in the
main text. The second considers these habitats as important nursery grounds for functional
groups with split juvenile and adult life history stages included in their Ecosim models. These
groups were: juvenile tiger and banana prawns in the Gulf of Carpentaria model, juvenile
snapper, flatfish, whiting, piscivores and mullet in the Port Phillip Bay model and juvenile
banded morwong in the east coast Tasmania model. For this second scenario, vulnerability of
juvenile groups to predation was linearly dependent on the abundance of habitat (Table 1) and
predation vulnerability could at most double (Okey et al. 2004).
The Gulf of Carpentaria, Tasmania, and Port Phillip Bay models were the only models with
juveniles and adults distinguished for some habitat dependent functional groups. Comparison of
model simulations with and without juvenile habitat dependency showed habitat dependency
affected habitat dependent functional groups, by amplifying responses to primary production
change (Fig. C1). However, the effect of habitat dependency on system level indices including
total fishery landings, community abundance evenness, mean system longevity and the mean
trophic level was small. Thus, to simplify analyses we excluded habitat dependency effects in the
main analyses.
There are two caveats to the habitat mediation analyses. First, only a small number of the total
functional groups in our 12 Ecosim models have split life history stages specified. This limits our
ability to specify juvenile habitat dependencies. Second, little information is currently available
on the functional form of habitat dependencies. Future research should consider habitat
dependencies across a broad range of species and seek ways to parameterise relationships
between habitat availability and juvenile survival.
Fig.C1 Change in biomass for change in primary production rates of all primary producers (%)
over 50 years for habitat dependent groups with unfitted (solid line), fitted (starred line) and
trophic level (dashed line) vulnerability formulations and formulations with juvenile dependence
on habitat (grey lines) and without (black lines) juvenile dependence on habitat. Banded
morwong are from the Tasmania model, snapper, flatfish, whiting, piscivores and mullet are
from the Port Phillip Bay model and the prawn species are from the Gulf of Carpentaria model.
Results are relative to simulations with no climate change.
Reference
Okey TA, Vargo GA, Mackinson S, Vasconcellos M, Mahmoudi B, Meyer CA (2004)
Simulating community effects of sea floor shading by plankton blooms over the West Florida
Shelf. Ecological modeling, 172, 339-359.