Time series modeling of Groundwater levels with Pastas

The term aquifer derives from the Latin aqua ferre, meaning “to carry water”. Embedded in this definition is the concept of movement, and therefore of time. Groundwater systems can be seen as living systems, continuously responding to external solicitations such as recharge, evapotranspiration, pumping, and surface water interactions. Groundwater level time series represent the integrated signal of these stresses, filtered by the physical properties of the aquifer. Today, large amounts of
monitoring data are routinely collected through automated measurements, yet a significant portion of these datasets remains largely unused. Groundwater levels are recorded, archived, and reported, but often not analysed in a way that extracts the information contained in the signal itself. This represents a loss of precious knowledge, essential for informed groundwater management.

Time series analysis provides a framework to study aquifer response without requiring a complete physical description of the subsurface. Through a data-driven approach, the observed signal can be decomposed into the contributions of different processes controlling groundwater dynamics, quantify response times, and support management decisions. For a more detailed overview, see: Bakker and Schaars (2019). In many cases, attempts to reproduce complex realities through highly detailed physically
based models may lead to over-complex numerical models with limited predictive capability, whereas time series models focus directly on reproducing and interpreting observable system behaviour.
This workshop introduces groundwater time series analysis using impulse response functions, with hands-on exercises using the open-source Python package Pastas. Participants will learn how to use Pastas to describe and predict groundwater level fluctuations.