World Water Congress 2015 Edinburgh Scotland
Special session 34: Habitat Restoration And Natural Flood Management
E Marian Scott
E Marian Scott
School of Mathematics and Statistics
With Adrian Bowman, Claire Miller, Kelly Gallacher, Ruth O’Donnell, Mengyi Gong, Craig Wilkie, Alistair Rushworth
Networks in space and through time
A set of monitoring sites- which may be physically connected Sampling through time and over space, often sparse in space and not always representative. Water-Quality Watch displays real-time water quality data collected remotely by sensors installed in rivers, and lakes. Readings taken every 5 to 60 minutes are transmitted via satellite to the USGS National Water Information System (NWIS). Data include water temperature, pH, specific conductance, turbidity, dissolved oxygen, and (or) nitrate depending on the site.
Dealing with the quantity of data in terms of the number of pixels, combined with the sparsity of the data, in terms of the time series observed will be a huge challenge.
Modern functional data approaches are suitable for large numbers of time series of potentially noisy data and enable clusters of curves to be identified which are coherent in terms of temporal dynamics.
Opportunities and challenges
From a network we can learn:
whether there are seasonal patterns and trend
Whether the patterns are different at different locations
How to make predictions at any point on the network, regardless of whether it is near a monitoring site
Data characteristics- quantity and quality and relatedness
Non stationary, complex nature of the relationships
For networks, how to build fast and efficient spatio-temporal models,
Designing the network and the resulting power to detect change