Information theory for sustainable management of water resources

点击翻译按钮获取中文摘要

Hydrology and water resources management are key disciplines for outlining a sustainable future, since water interacts, among others, with food, energy, health and ecosystem functioning. The main difference between the environmental science and fundamental physics, is that environmental systems, although governed by the same laws of physics, are generally complex systems that are largely unobservable. The behaviour at temporal and spatial scales of interest is often emergent from complex interactions and feedbacks, which make it hard to predict. Uncertainty is therefore an inevitable part of predictions and this should be acknowledged in science and engineering. Information -the opposite of uncertainty- then becomes a key common currency in environmental science and engineering problems, where it is often far from trivial what to measure at what scale and what model complexity is warranted by the data. In fact, all challenges in water resources management can be seen as decisions made under some form of uncertainty Those decisions improve when more information becomes available. The question then is: How does information get to a decision? In principle, all information we have about our environment ultimately stems from observation. Part of those observations have been condensed into physical laws. This is why we don't need access to data from experiments performed centuries ago, but can use the laws of e.g. mass, energy and momentum conservation in our hydrological models. This going from data to laws is a form of data compression, where patterns in the data allow a shorter description. Information theory, originally found to describe limits of optimal information transmission and storage therefore also describes the way we derive laws from data, how we distill information from numbers. A growing number of physicists sees information as a fundamental physical concept. The research program that I intend to set up will be mainly focused on using information theory to formalize how information flows from the environment into observations, into models, where they are combined with previously distilled laws to yield predictions that subsequently improve decisions and designs. One of the application areas is operation of hydro-electrical systems, investigating the role of hydrological and market information at time scales from minute to seasonal. This topic poses both a large number of science questions from an information perspective, as well as practical applications with large direct benefits. Due to this large interest there are many possible collaborations with industry and Canadian academia. To summarize, the proposed program aims to address practical questions in water resources management by putting the quantity "information" in a central role. This will serve to streamline the information flow from source to decisions relevant for society.