Pointers for the Future - Environmental Modelling: Finding Simplicity in Complexity - page 443

Modeling System 21 brings together 189 models of terres-
trial, coastal, marine, hydrological, carbonate and climate
systems, which can be downloaded, as well as providing
discussion fora and educational materials. Such progress
was never so rapid under the proprietary model of soft-
ware development. Open-sourcing software has also led
to more open sourcing of models with positive implica-
tions for more rapid model development, more dynamic
interoperation of models and greater transparency of
models (see, for example, the OpenMI initiative). 22 Still,
only two of the 17 General Circulation Models used for
the IPCC Fourth Assessment are open source at the time
of writing, 23

networked sensor networks, other web models or even
with models running on the desktop. A web climate
model might interact with a hydrological impacts
model with either one way to two way communica-
tion. Examples of inter-operating web-based models
include the CIAS 24 (Community Integrated Assess-
ment System) that links climate model simulations to
a number of models including the WaterWorld web
based hydrological model (see Chapter 20) and the link
between the desktop WEAP 25 model (Water Evaluation
and Planning) and the web-based WaterWorld system.
In the latter, a WEAP user requests a baseline or sce-
nario simulation for a set of WEAP coordinates and the
software calls the web-based WaterWorld which then
runs a simulation and provides WEAP with the required
information for further analysis. The development of
open standards for web processing services, 26 so-called
'systems of systems' such as the Global Earth Obser-
vation System of Systems, 27 and other frameworks for
interoperation, promise to deliver interoperating suites
of web-based models that enable models to be indepen-
dently developed, maintained and delivered by a range
of modellers yet appear as a single model to the end user.
This is the vision of the so-called Model Web (Geller
and Melton, 2008). One issue that increasingly needs
to be addressed in putting together increasingly com-
plicated, interoperating models is to ensure that the
code is still performing in a verifiable way. Test-driven
development (TDD) is commonly used in producing
large software products, often where several developers
are modifying parts of the code at the same time (Mar-
tin, 2008; Freeman and Pryce, 2009). For examples,
initiatives have been developed to produce 'self-testing'
programs, such as the JUnit library 28 for Java (although
in reality, self-testing means that the programmer builds
in the tests to ensure that expected behaviour occurs).

so much progress remains to be made in

that regard.
In addition to the hardware and software developments
outlined above, improvements in the ubiquity, speed,
availability and sophistication of computer networking
have also yielded significant improvements for modelling.
Perhaps the most valuable of these have been as follows:

The availability of the indexed and searchable world
wide web (WWW) which has brought a significant
increase in access to research including downloadable
models for all, and has provided model developers
with a much longer reach towards a wider range of
users for their modelling tools. In recent years, there
has also been movement away from downloadable,
installable models, which are invariably out of date soon
after they are installed and subject to the vagaries of
users' differing computer configurations, to web-based
models (see Chapter 20) and models which make use
of 'crowdsourced' computing power available through
a user community's computing resources (for example
climateprediction.net which is a distributed computing
climate model). Web-based models can make use of the
client's computing power (by running in the browser
using Javascript or Java) or can run on the modeller's
server but be operated from the user's browser.
There are a number of benefits of having models run
online as a so-called 'web service'. In addition to the
benefits of end-user responsive and adaptive model
development outlined in Chapter 20, running models
as web services holds the potential for enabling the inter-
operation of models with web based datasets, wireless or

•

Despite the tremendous growth in computing power,
there are still some modelling problems that are limited
by computing power. This is particularly the case for
computational fluid dynamics models (see Chapter 6)
or climate models (see Chapter 9), both of which
need to solve complex equations at finer and finer

•

24 See www.tyndall.ac.uk/research/cias (accessed 6 April 2012).
25 See www.weap21.org/ (accessed 6 April 2012).
26 See www.opengeospatial.org/standards/wps (accessed 6 April
2012).
27 See www.earthobservations.org/geoss.shtml (accessed 6 April
2012).
28 See www.junit.org/ (accessed 6 April 2012).

21 See http://csdms.colorado.edu/wiki/Main_Page (accessed
6 April 2012).
22 See www.openmi.org/.
23 See www.easterbrook.ca/steve/?p

= 667 (accessed 6 April 2012).