Groundwater System Identification through Time Series Analysis

“As perhaps is not uncommon in research, the initial objectives differ from the final results presented in this thesis. Initially, the primary objective was to improve current methods with which the relationship between groundwater dynamics (and impact of hydrologic measures) and the species composition of groundwater dependent ecosystems is modeled. In the alternative proposed originally by [Maas, 1995], a key notion is that the spatial differences in groundwater level dynamics are mainly determined by the spatially variable propertjes of groundwater systems, white the variation through time or temporal dynamics are mainly driven by the spatially less variable meteorologic dynamics. Consequently, it was hypothesized that spatial differences in vegetation are also primarily determined by system propertjes, or in other words, that they can be modeled more accurately by ‘filtering out’ the temporal, meteorologic dynamics. To be more specific, Maas proposed the use of time series models for inferring the so-Galled impulse response function from series of groundwater level observations, which in turn can be characterized using statistica) moments (see section 2.3.4). Moments are scalars and constants in linear, time-invariant
systems. Together with the spatially less variable driving forces, they completely characterize the (deterministic part of the) dynamics at a certain point in
space. As an additional advantage, moments can also be simulated spatially using a standard distributed groundwater model. The scope of this research thus initially encompassed three fields of modeling, i.e. time series, groundwater and ecohydrologic modeling. These fields of modeling together constitute the method of impulse response moments, from which moments can be derived and between which they can be mutually exchanged. Although it was envisaged that the link between time series models and groundwater models would be a subject in its own right, at the time the existing ARMA time series models were thought to be welt developed and to be directly usable without modification. ARMA models, however, turned out to have several important limitations (see section 2.2.3). What followed was a process of improving and adapting time series analysis methods to the needs set by the initial objectives of this research (but not limited to that). As such, most of the actual contributions of this thesis are in the field of time series analysis.”

(Citaat: Asmuth, J. von – Groundwater System Identification through Time Series Analysis)