SDI Convergence - Nederlandse Commissie voor Geodesie - KNAW

fication both of data and models. The usage of spatial data metadata for this purpose is
a fundamental service within SDI and is in wide use (van der Wel, 2005; Stout et al.,
2007; Nogueras-Iso et al., 2005). At this point there are many model registers supporting
model search and discovery (i.e., NASA Global Change Master Directory (Website
3), Freebase - Environmental Modelling Collection (Website 4), Catchment Modelling
toolkit (Website 5)).
At level 2 a deeper description of both information and models is required to assist their
utilisation. For data and information in particular this includes metadata describing the
genesis, quality and condition of the data resource. In the fields of Agriculture and NRM
this metadata is of high importance as data often exists as a collection of data instances
that have been produced by a number of projects and initiatives. These are
often spread over time and may be subject to an evolving standard or improving collection
methods and associated technologies. Unless there has been strong adherence to
a data collection standard the specific instances in the collection may differ in subtle
ways that can have a profound influence on their utility. At this level these differences
are captured in the metadata which is often of a more technical nature. At present this
type of metadata does not get much attention in metadata schemas (NIOS, 2004) and
can be costly to collect. An exception is the emerging Numerical Model Metadata standard
associated with climate modelling (Steenman-Clark et al., 2004). Stout et al.
(2007) provide an example where the US Federal Geographic Data Committee (FGDC)
metadata standard is being extended to support some of this type of metadata. At level
2 the use paradigm for model metadata is changed. Rather than generally describing
the model itself emphasis is placed on storing contextual information about the modelling
activity and how to implement. To support effective model use not only are technical
descriptions of the inputs and description of how to execute the model important but
understanding where and who implemented the model is extremely useful. This can
benefit to prevent duplication in modelling effort. In contrast to level 1 there are not
many tools that effectively address and use model metadata at this level. Most approaches
to date are associated with theme based modelling groups and a good example
is the ‘Earth System Curator’ and its prototype database used by climate modellers.
In this context the record of model activity is confined to the community of interest.
This prototype has the stated objective to “store metadata related to model runs and
datasets” but is still under development. The authors have not yet found published
model metadata systems recording details of model instances spatially (other than the
MIKE system described herein). Practical support for this function requires user-friendly
and efficient approaches to the spatial registration of these instances. At level 2 there
are opportunities to link model and data metadata repositories (for those models with
stable or fixed data inputs) and support queries informing understanding of spatial data
availability for modelling. These approaches require a common referencing scheme for
datasets in both registers. This is a key component within the case study described in
the following section.
At the third level the emphasis for metadata shifts to record details that can support associated
processes both for models and data. In the data area examples include metadata
to support publication and transformation of data such as in web based mapping
or even workflows associated with data such as managing collection by third parties
(Stout et al., 2007). In relation to models this metadata serves as a base to underpin
applications for automated model processing or pre-processing of data for models.
Current development in this area is largely confined to process models in remotely
sensed data (Jianto et al., 2003) or in climate modelling. The development of metadata
in this area is a crucial part of an SDI to support automated publication of data and
models. Where the metadata is extended to store processing instructions, algorithms or
140