Environmental Engineering Reference
In-Depth Information
on the models, rather than on the user interfaces. Whenever flexibility of use is
important, providing domain specific views of data adds value to such a tool, both in
operational usage and in model development.
The component GDD (Graphic Data Display) is a Microsoft .NET component
developed by CRA (Di Guardo et al.,
2007
), which has the specific purpose of
retrieving a set of output variables and allowing values to be displayed either as
textual tables or as graphs. GDD can be used as a stand-alone tool or as a compo-
nent inside an application. In the former case it provides access to a file dialogue
that allows the user to select a file, whereas in the latter case it can be opened
inside a modelling framework to directly load the current dataset. GDD accepts
inputs in three different formats: XML, MS Excel, and the more compact and
faster binary form (another available component also allows I/O operations with
the binary format). Readers can however be extended by third parties implement-
ing the proper interface. Each variable can be either a table column, or an entire
table of the dataset, depending on whether it is either only time-variant or time and
one-dimensional space-variant (the latter are variables that vary down soil profiles,
such as soil temperature). GDD has seven tab pages supporting data views such as
tabular views (which can be saved using the Microsoft Excel format), scatter
graphs, time courses, histograms, soil profiles (water, temperature, nitrates, agro-
chemicals), 'Micale' graphics, frequency histograms, and probability of exce-
dence. Also, GDD allows showing reference data against simulation outputs via
configurations which can be saved. GDD can read APES GUI output files, in both
XML and binary formats.
The Simulation Output Evaluator
The Simulation Output Evaluator (SOE) is a data analysis tool developed by CRA
that provides easy access to model evaluation techniques. As the literature gives
neither a standard theory on model evaluation, nor a standard “box of tools”, the
emphasis is on statistical techniques for comparing estimates either with actual
measurements, or two series of estimates, making use of an extensible library called
IRENE (the.NET 2.0 version). Non-replicated estimates are mostly compared with
the non-replicated measurements. The program also allows comparison of indi-
vidual estimates with replicated measurements (or vice versa) and replicated esti-
mates with replicated measurements. The program provides extensive statistical
capabilities with tools for a variety of needs. Ready-to-use procedures handle a
wide range of statistical indices and test statistics. Basic statistics allow a prelimi-
nary check of data quality. The evaluation of model performance is based on either
the model residuals or on the correlation coefficient. In addition, model evaluation
by probability distributions (i.e., probability of excedence), residual analysis (i.e.,
pattern indices), and fuzzy-based aggregation statistics are allowed both for indices
produced internally by the component and for external numerical values. The fuzzy
aggregation model is saved as an XML file. Graphics are included in most analytical
