I led the initial development of Simile, and am still closely involved with the development team. I’m delighted to see your two questions: they are exactly the sort of question that people should be asking of a modelling environment which claims to be suitable for implementing the serious models needed for addressing real-world problems. Our aim in developing Simile was to allow modellers to implement the models they wanted, without programming, and with as little restriction as possible, and we are very happy to be challenged in the way you have. (We are even happy when someone comes up with something Simile can’t do, since that gives us something to aim for!)

Roel has addressed your second point, and it is indeed the case that one model can include processes operating at different spatial and temporal scales. One example is a landscape vegetation model running over years, which has occasional fire outbreaks in which the fire’s spatial spread is modelled on an hourly basis - but only for the duration of the fire.

Your first point - on having different spatial units modelled in diffierent ways - is particularly interesting. We do this by having the concept of a ‘conditional submodel’. Thus, to take your example, the Cell submodel (with multiple instances, representing the cells in the grid) would contain two submodels, one for each method of modelling fish population dynamics. Each would be conditional (it would contain the Simile ‘condition’ element, a question-mark). Each condition element would contain a Boolean (true/false) expression, which is true in those cells using one model and false in the others. The model can even switch these dynamically during a simulation run.

An example of this mechanism in operation is provided by a model of landuse change (http://www.simulistics.com/examples/landuse/landuse.htm). In this case, the two conditional submodels are for Crop and Forest. Which one is switched on depends on what landuse state the spatial unit happens to be in.

Please do feel free to contact me if you would like to discuss these issues further, or if you have more questions. robertm -at- ed -dot- ac -dot -uk.

Robert

I think you bring up an extremely important issue. It is on my mind a lot. With some of the landuse models we just started to experiment with strategies to estimate uncertainties and do the calibrations, and I expect to become much more sophisticated as our project moves along. Thank you for the NUSAP link. It seems like a good start for documenting (meta dating) the components entering the larger model.

I am working on a species model within Mimes that does both, it looks at habitat conditions and species interactions within a cell, while allowing the populations to move across the landscape based on extractors (factors that causes individuals to move away) and contractors (factors that cause individuals to congregate.
Also the Simile/MIMES framework has no problems with handling multiple scales in time and space during the same model runs. Having said that I would advise against using these features to prevent the models from becoming to complex and less reliable. But when needed, we will use it.