Environmental Engineering Reference
In-Depth Information
therefore requires updates if the model is updated, which is difficult for maintenance.
Such wrappers are not required for the FSSIM-AM models, as these have been
developed in Java tm , which is also the programming language of SeamFrame.
The Open Modelling Interface and Environment (OpenMI; Moore and Tindall
2005) was used in the SEAMLESS modelling framework to link the models at run
time into a model chain. OpenMI is based on a pull-approach in which the last model
in the chain pulls its outputs from other models in the chain by calling “getValues()”-
methods, which means requesting outputs from a model or data source. The model
can set its outputs as inputs to other models through a “setValues()”-method.
The model components (Fig. 5. 4 ) are developed as OpenMI-components. If a model
is wrapped, then the wrapper needs to be developed as an OpenMI-component,
which implies that the models are not aware of OpenMI or affected by OpenMI.
The definition of data exchanged in setValues() (e.g. inputs) and getValues()
(e.g. outputs) forced modellers to be specific about the inputs and outputs of a
model and facilitated linking of the models in a model chain. Wrapping the model
as an (OpenMI) component facilitated the definition of models independently of
each other, of data sources and of the graphical user interface.
Model assumptions, interfaces and available data sources need to be clearly and
explicitly specified, so that the models can be linked to each other. The FSSIM model
interfaces in terms of inputs, outputs, states and parameters have been defined
explicitly in an ontology, which is a collection of all concepts and relationships
between concepts relevant to the domain (Antoniou and Van Harmelen 2004) and
which functions as a dictionary. It sets up clear definitions for loosely integrating
models in an open environment, facilitated by the knowledge manger for knowledge
processing (such as reasoning and consistency checking) and by the domain manager
(such as automatic generation of code templates for models and domain classes,
accessing an instance of a domain class at runtime to supply the model component
with the appropriate data).
FSSIM Application: Detailed and Simple Applications
Through the first application of FSSIM to a few regions it appeared that the data
requirements of the models (FSSIM/APES) are too high (i.e. good data on farm
management are extremely scarce). For this reason, it was decided to allow two
variants of FSSIM: one that uses detailed data an agro-management and one that
uses less detailed (simple or summarized) data on agro-management. The simple
version can be more easily used for a larger number of applications necessary for
up-scaling to the EU level (cf. Pérez Domínguez et al. 2009) , whereas the detailed
application is useful for application to specific regions. The principal differences
between the two versions of FSSIM are summarised in Table 5.4 .
The purpose of this section is to describe the results of a detailed and simple
application of FSSIM to explain the followed procedure for running the model and
to assess its capacity to reproduce the current situation and forecast the future.
Search WWH ::




Custom Search