To aid the development of models of soil-water relations we are collecting high temporal resolution datasets using new technologies. This will enable us to increasingly learn from the complexity that emerges from such patterns and thus allow us to inform current, and generate new, hypotheses. We will develop new analytical techniques for sensing biogeochemical cycles in soils and water that may feed into new mathematical and geostatistical models for analysing complex signals (see below). The research will initially focus on the collection and analysis (e.g. nitrate, P and dissolved organic C) of drainage and stream waters from the Rowden plots and the River Taw, in conjunction with the construction and maintenance of new databases, in order to feed model development.

Building on the data gathered during the high resolution monitoring, we shall test the hypothesis that time-series data on outputs and controlling variables for soil systems contain complex non-stationary features which give insight into system behaviour. Environmental systems, particularly those that include the soil, often vary markedly in time in complex ways. Standard methods for time-series analysis assume that the underlying variables have statistical properties that are uniform over time (stationarity). This will often not be realistic in environmental systems. Wavelet transforms are a powerful and relatively recent addition to the armoury of applied mathematics. In a wavelet transform a time series is decomposed into components of different temporal scale without assuming stationarity. Using wavelets we can study the distribution of variation between different scales, and how this changes over time. We can also investigate the covariation of two or more variables at different temporal scales. We will show the potential of wavelet methods for the analysis of time series data on stream-flow data from monitored systems at North Wyke (looking at multiple observations on flow and water quality) and comparing time series of flows at different scales (catchment drainage, field drainage and small plots). We will also analyse eddy correlation flux measurements of trace gas emission from managed land collected by Rothamsted and collaborators. The multivariate wavelet analysis will elucidate underlying processes, for example, measurements on nutrients that are transported in and from the catchment by related pathways will show similar multi-scale behaviour in time, and trace gas fluxes from land under different management will respond differently to rainfall events.