Information Technology Reference
In-Depth Information
stakeholders provides the scale of spatialisation of flow representation. Hydrological
signal needs to be realistic at this spatial scale since it might be used for feedback in
the water assessment component. Additionally, model needs to provide a dual
representation of levels and flows. These both requirements are rather new for
hydrological modeling. We now present how we adapted previous hydrological
modeling frameworks to cope with them.
4
Existing Approaches for Hydrological Modeling
Hydrological modelling is currently divided into two main categories: conceptual
models and physically based models. Conceptual models provide a simplified
representation of the general behaviour of the catchment based on the continuity
equation as well as additional mathematical relationships to simulate the links
between rainfall and surface runoff. In this category GR models [9] for example are
conceptual models based on relations between data series of inputs and outputs of
water and calibrated on past series. Physically based models which try to represent the
rainfall-runoff transformation based on the understanding of hydrological
mechanisms which control the response through physically based equations. They aim
at being explicit on water flows between surface, soil and ground water compartments
which ends up in impacting on hill slope flows and river discharge [10]. Most of them
are spatially distributed models accounting for the variability in the input variables as
well as in the properties which influences the processes across the catchment. These
distributed physically based models, such as in [11-12] among others, represent the
dynamics of water flows on a landscape represented as a computing grid, from inputs
due to rain to outputs including evaporation. Both categories of models are able to
tackle connections between surface and groundwater. When dealing with whole river
basins or territories equivalent to the size of a county, distributed models are still at a
scale rather too large to cope with farm level, in order to represent the hydrology and
the discharge at the outlet of the basin in an acute way.
More recently a few scholars have attempted to represent hydrological process in a
fully distributed way, based on techniques such as cellular automata or agent based
modeling. Delahaye and colleagues [13] have represented interactions between land
use and flows with a cellular automaton featuring a topological graph with the only
coded characteristics being the topography. Water then flows on this simulated
landscape. A more extreme attempt has agentified “water bowls” in the RIVAGE
model. In this model elementary particles of water move according to basic physical
laws on a given landscape, they can meet up and aggregate in various form of water
bodies [14]. These both innovative approaches have inspired our work. However they
handle much more local scale than our need.
4.1
Integrated Hydrological Models and Agent Based Models
Agent based modeling is currently a common approach to represent the dynamic
relations between a hydrological model and water uses. Le Page and colleagues
