Incremental modeling of the post-Soviet urban growth.

    MARIUS ( Modélisation des Agglomérations en Russie Impériale et ex-Union Sociétique) is a ABM model targetting the specific case of the post Soviet urban growth in the last decades.

MARIUS takes place in the direct continuation of the Simpop family of models, but features the recent advances in terms of methodology and experimentation in a massive grid computing platform (see Simprocess pages) .

In order to distinguish between generic urban mechanisms - comparable to what happened in other systems of cities in the world - and local specificity - necessary to understand the trajectory of the particular city system of Russia, we adopted an incremental modeling method.

A double hierarchy of explanatory factors 

    We suggest a methodology proposing a step wise modeling of the system of post Soviet cities that progressively complexifies the model by exploring incrementally two hierarchies of explanatory factors: The first is linked to a more and more precise description of the functional structure of the system of cities, whereas the second one corresponds to a more and more precise description of its spatio-temporal definition.
The suggested method relies on two axes of complexification, and a progression of models along those axes.

Increasing interactions complexity

    On the first axis, urban growth mechanisms are classified  according to growing levels of complexity The first level is a Gibrat’s model of growth, that is, a stochastic process with no interaction between cities. It becomes more complex when spatial interactions are added, through trade, then the creation of specialized new towns and a radical inflexion of the territorial governance. 

Increasing space-time resolution 

    On the second axis, the environment is progressively described with more and more details. This doesn’t mean the intervention of new mechanisms, but levels of constraint that are applied to mechanisms of the first axis. The representation of space and time (mainly space in our case) may affect the results of the mechanisms, but not the main logic of interaction.
    The minimum resolution corresponds to a random distribution of cities in a homogeneous geometric space. The progression in space-time resolution goes then through the introduction of localized resources, politico-demographic differentiation of territory, the exact location of cities, the physical networks and evolutive boundaries.

    Following this methodology, we believe that the progressive conception of models, along with empirical data confrontation, allows a better control of the modeling process, as it clarifies both the goals and the means of each step ot the modeling process. The application of the method should help to clarify the underlying mechanisms of the post Soviet cities’ growth, but also to develop some other models, such as the China-India Model. [TODO  lien ]

Parameter space exploration and Calibration

MARIUS is developed following the precepts of the Simprocess methodology, and benefits from its grid-computing environment, OpenMOLE, to replicate the simulations experiments , conduct sensitivity analysis, and be a benchmark for some automatic calibration methods.

/!\ This page is still under construction /!\