Channelling information flows from observation to decision; or how to increase certainty

Abstract:

To make adequate decisions in an uncertain world, information needs to reach the decision problem, to enable overseeing the full consequences of each possible decision.

On its way from the physical world to a decision problem, information is transferred through the physical processes that influence the sensor, then through processes that happen in the sensor, through wires or electromagnetic waves. For the last decade, most information becomes digitized at some point. From moment of digitization, information can in principle be transferred losslessly. 

Information about the physical world is often also stored, sometimes in compressed form, such as physical laws, concepts, or models of specific hydrological systems. It is important to note, however, that all information about a physical system eventually has to originate from observation (although inevitably coloured by some prior assumptions). This colouring makes the compression lossy, but is effectively the only way to make use of similarities in time and space that enable predictions while measuring only a a few macro-states of a complex hydrological system.

Adding physical process knowledge to a hydrological model can thus be seen as a convenient way to transfer information from observations from a different time or place, to make predictions about another situation, assuming the same dynamics are at work.

The key challenge to achieve more certainty in hydrological prediction can therefore be formulated as a challenge to tap and channel information flows from the environment.

For tapping more information flows, new measurement techniques, large scale campaigns, historical data sets, and large sample hydrology and regionalization efforts can bring progress. For channelling the information flows with minimum loss, model calibration, and model formulation techniques should be critically investigated. Some experience from research in a Swiss high alpine catchment are used as an illustration.