Global Positioning System Reference
In-Depth Information
environmental modeling scenarios such as wetland ecosystems services
and watershed runoff prediction, respectively.
In contrast to the previous examples where models are exposed as
pure WPS-based services, Jones et al. (2012) proposed a WSDL-based
implementation for publishing models on the Web built internally on the
WPS service interface which can be later integrated with scientifi c workfl ow
systems such as Taverna and Kepler. In this sense, this work closely relates
with the approaches for geo-enabled scientifi c workfl ows described in the
previous section.
These examples are a fi rst attempt towards the Model Web vision,
where open networks of models exposed as web services are ready to be
shared and to support IM by allowing these services to talk to each other
directly using standard services interfaces. Yet, the above examples are still
far from the ideal vision of Model Web because integrated models have
some intrinsic characteristics that should be taken into account in practical
terms. In particular, the following issues are barriers to realizing the vision
of a useful Model Web and require further research: (i) the variability of
types of models; (ii) the increasing complexity of chained models; (iii) the
existence of multiple services interfaces for exposing models; and (iv) the
wide range of infrastructures and implementations towards realizing the
Model Web.
The idea of augmenting the Model Web with the notion of reproducibility
is a logical next step to modelers and decision makers. Essentially, IM is
seen as an example of scientifi c practice that seeks to reproduce modeling
results to be compared with historical data and verify their validness.
Therefore, reproducibility implicitly requires sustainable development
practices about enduring existing infrastructures, platforms and tools,
and re-using existing models and resources (Granell and Schade In press).
Indeed, models are reused every time they are reproduced. All of this poses
an ambitious challenge: one thing is to reuse a model, and the other is to (v)
reproduce the execution of such a model again by others. The latter implies
a large set of requirements that must be taken into account. For example,
the access to the same input data sets used in that model, the access to
notes or documentation regarding the execution of that model, access to
historical results so that new results may be assessed, as well as technical
notes and confi gurations to recreate the same execution environment. In
the rest of the section, we delve into these fi ve open issues and sketch some
possible solutions.
Not One but Multiple Types of Models
WSDL-based services and WPS-based services are designed to perform the
same computation every time they are invoked by a client application. This
Search WWH ::
Custom Search