|
|
|
|
Making Models Easier to UseA significant problem with creating and maintaining complex model systems is the large number of equations, constants and parameters required for the model. A number of difficulties are presented. Firstly experiments often need to be conducted by the user of the model in order to parameterise the model for a given situation. Few current models, even process based ones, are useful in a new situation without careful thought in parameterisation. The user must then be able to find the model parameters and change their value in a straightforward manner. A secondary problem relates to the openness and understand ability of how a model is constructed. A model user has to be able to understand the form and function the equations in a given model in order to be able to establish sensible experimentation and parameterisation for their system understudy. Traditionally vegetation models have failed to fully explain their parameters and the function of internal equations and have relied instead upon the model writers working in close collaboration with experimentalists. In many cases this may not be a problem, and indeed may be desirable, but model creation is a time consuming process leaving little freedom to collaborate with the broad range of experimental research groups required to construct and test a robust mechanistically rich model. An easy to use open modelling system would allow experimentalists, who do not necessarily have a direct interest in model construction, to use standard model modules with their experimental data. Facilitating a broad audience of experimentalists for a model in this manner has many advantages: i). Experimentalists can make general use of the model to aid interpretation of results. ii) Formulation of new theories based upon the synergy between experimental observations and model predictions becomes possible iii). The approach leads to much wider testing of the general applicability of the model and if this includes feedback to the modellers improvements to the model. The availability of easy to use standard modules, that have already been coded and tested extensively, also allows experimentalists and modellers to create new unique models without the need to ‘re-invent the wheel’. Similarly one of the current difficulties with comparing the properties of existing models is the impenetrability of their model interface, parameters and controls. Open, easy to use models facilitate model testing, comparison and collaboration between modellers. Fortunately recent advances in computer software and in particular with object orientated programming practices have made it possible to simplify the construction and presentation of complex systems. Use of commercially built component objects makes it possible to construct user friendly systems without considerable expenditure of time writing user interface code. In order to comply with the openness criteria good modelling systems should be well documented or preferably self-documenting with as much of the information connected in a context sensitive fashion to specific activities within the model. User interaction with the model itself should be clear and simple. Modelling systems should support commonly available hardware and software platforms wherever possible in order to be available to a wide range of interested parties. Windows 3.11Ô , Windows 95Ô and Windows NTÔ running on desktop PC systems meet these criteria and have made the development of simple interactive model systems possible. The core algorithm of a good modelling system should also be transportable to other hardware platforms in order to facilitate comparative evaluation with other modelling systems. Installation of the modelling system onto host computer systems should be either automatic or as simple as possible. The model itself should be open to scrutiny and criticism of peers not just in the scientific literature but in form of a dynamic interactive program that can be used by others to test the limitations of the proposed system. To help users understand model results in a timely fashion modelling systems should: i) Support automatic graphing of simulation results. ii) Include at least a limited ability to analyze simulations results and make comparisons to experimentally determined results as a minimum requirement. iii) Use modern programming techniques to guide the user through complex tasks such as parameterising the model. |
Send mail to humph@essex.ac.uk with questions or comments about this web site.
|
