Environmental Engineering Reference
In-Depth Information
Table 5.4 FSSIM application to regions with detailed or summarized data availability
FSSIM with detailed data
FSSIM with summarized data
Use APES with observed input data
Use APES with generated input data
Use detailed survey
Use simplified survey
Includes current and alternatives activities
Includes current and alternatives activities
Use all FSSIM-MP modules
Use only some FSSIM-MP modules
Use semi-automatic procedure for calibration
based on risk and/or Positive Mathematical
Programming
Use automated procedure for
calibration based on risk or/and Positive
Mathematical Programming
Detailed Application of FSSIM
FSSIM was tested for a range of detailed applications with the aim to analyse the
current situation and to anticipate the impact of new, alternative scenarios and
policy changes. In this chapter, results of Midi-Pyrénées (France) are presented as
an example of the test application.
An overview on the selected components, modules and calibration procedure used in
the detailed application as well as the tested scenario is described (Fig. 5.5 ) below:
-
Components: the selected components are: (i) the farm typology; (ii) the
detailed computer-based survey for agro-management and FSSIM-AM; (iii)
the biophysical model APES; and (vi) the mathematical programming model
FSSIM-MP.
FSSIM-MP modules: the selected modules are the crops, premiums, risk, PMP,
-
perennial, policy and common modules.
Calibration procedure: the calibration procedure is based on two steps: in the
-
first step, we apply the risk approach in order to calibrate the model, as precisely
as possible. The model assigns automatically a value to the risk aversion
coefficient 1 which gives the best fit between the model's predicted crop pattern and
the observed values. The difference between both values is assessed statistically
by using the Percent Absolute Deviation 2 (PAD). The aim of this step is to ensure
that the model produces acceptable results before going to the second step.
1 The chosen value can vary from 0 to 1.65, as suggested by the literature.
2 Percent absolute deviation (%):
n
ˆ
XX
i
i
(%)
=
i
=
1
.100
n
ˆ
X
i
i
=
1
where ˆ i is the observed value of the variable i and X i is the simulated value. The best calibration
is reached when PAD is close to zero.

Search WWH ::

Custom Search