Times Series Analysis is Essential to Ground Water Modeling

Numerical models are generally considered unreliable by parties fearing measures that affect their groundwater. This leads to intense debates often in the press. Given the increasing number of loggers, time series analysis is more and more used to relate stresses to effects. However, results from time series analysis and numerical models seem to often conflict with each other. Worse, hydrologists often fail to explain what causes the differences. Many times, regional models fail to satisfatorily match a measured time series, even after extended calibration. A key issue to point forward is that time series analysis is a must ahead of any other modeling, because its failure or success proves whether or not the known stresses, such as precipitation, evaporation, surface water levels and pumping, can actually explain the available groundwater time series. If the time series analysis can’t, neither can a numerical model. It is essential to know before the expensive modeling exercise. We have analyzed 145 recorded daily time series observation wells in the city of Delft, The Netherlands. This revealed that only about half of them could be satisfactorily explained by the known input series, while the other half shows non-linear behavior that has not yet been explained. Constructing and calibrating a spatial groundwater model would be useless given this knowledge. Therefore, we should put effort in determining the causes of our  failure to explain the odd results instead of proceeding in the usual way with the construction of the a spatial model, when we know beforehand that it would fail, as in our Delft case.