Global Positioning System Reference
In-Depth Information
is true for example to run a spatial intersection process between two given
geometries and to geocode a place name. In these cases, a geoprocessing
service takes input data, uses this data to perform a computation, encodes
the result into XML, and fi nally sends it back to the client. In contrast,
environmental models often consist of several phases (e.g., initialization,
run, and fi nish) and must be capable of performing different types of
computations in each phase. Unlike environmental models, geoprocesing
services are often time-independent and therefore do not require storing
and referencing previous calculations, which implies that WPS-based
services are stateless.
Is the Model Web in conjunction with geoprocesing services and
workfl ow systems able to capture not only one type of model but multiple
types of models such as time-dependent, stochastic, empirical and so on?
To this respect, Castranova et al. (2013) recently showed that it is possible,
under certain circumstances, to adapt WPS-based services for encoding
time-dependent model computations and maintaining model state.
Nevertheless, innovative approaches are yet required to bring different
computations models together.
Not One but Multiple Perspectives of Complexity
As introduced earlier, a model basically simplifi es a natural system or a
part of it. Therefore, IM should be regarded as a simplifi cation process that
abstract from complex interactions among natural systems. Nevertheless,
most of available models cannot be easily linked with each other when
assembling IM chains due to several reasons (Granell et al. 2013b). Here,
we overview some of them.
First, the expectation that the Model Web can deliver integrated models
by simply linking existing models as web services can lead to meaningless
models. Link to the previous issue on variability of types of models,
some environmental models are regarded as time-dependent models and
empirical models that rely strongly on calibration data. That is, “as more
components are brought together, the calibration of the whole integrated
model becomes only more diffi cult” (Voinov and Shugart 2013). It is less thus
evident how current web service technologies and workfl ow systems can
handle calibration data tied to models (see Section “Not One but Multiple
Reproducible Resources”).
Second, semantics and formats of model input/output data sets (Feng
et al. 2011) along with multiple spatial resolutions and scales (Voinov and
Shugart 2013) may diffi cult IM. Again, it is less clear how to exchange data
between models with distinct level of spatial resolutions and scales.
Third, accessing models without proper supporting tools may be seen
as an obstacle at fi rst sight by non-expert users (Skøien et al. 2013) thereby
Search WWH ::
Custom Search