Environmental Engineering Reference
In-Depth Information
Addressing Uncertainty
A resampling routine, Ecoranger, has been included to accept input probability
distributions for the biomasses, consumption and production rates, ecotrophic
efficiencies, catch rates, and diet compositions. Using a Monte Carlo approach, a
set of random input variables is drawn from user selected frequency distributions
and the resulting model is evaluated based on user-defined criteria and physiologi-
cal and mass balance constraints.
To facilitate this task of describing probability distributions for all input
parameters (including the diet compositions matrices) and to make the process
more transparent a “Pedigree” routine was implemented (Pauly et al.
2000
)that
allows the user to mark the data origin using a pre-defined table for each type of
input parameters (from in-situ sampling, taken from other models etc.). The
confidence interval around the input parameter is smallest (
10%) for data
derived from sampling the same system and largest for those derived from other
models (
80%).
The Pedigree index values for input data scale from 0 for data that is not rooted
in local data up to a value of 1 for data that are fully rooted in local data. Based on
the individual index value an overall “pedigree index”,
, is calculated as the
t
average of the individual pedigree value based on
X
n
t
i
;
p
n
;
t
¼
(5.5)
i¼
1
where
t
i
,
p
is the pedigree index value for group
i
and input parameter
p
for each of
the
n
living groups in the ecosystem;
p
can represent either
B
,
P
/
B
,
Q
/
B
,
Y
or the diet
composition, DC. To scale based on the number of living groups in the system (n),
an overall measure of fit,
t
* is calculated (using an equation based on how the
t
-value for a regression is calculated) as
p
1
n
2
t
¼
t
p
(5.6)
t
2
This measure of fit describes how well rooted a given model is in local data.
Mass-balancing an ECOPATH model is usually achieved by manually adjusting
biomasses, mortality rates, diets, etc., searching for data inconsistencies and gradu-
ally obtaining a balanced model. An iterative method for obtaining mass-balance
has been added to EwE, offering a well defined, reproducible approach, while also
allowing exploration of alternative solutions based on parameter confidence inter-
vals as explained above. This routine, called
automated mass balance
, is further
explained in Christensen and Walters (
2004
).
