Communicating uncertainty about groundwater scenarios using stochastic simulation of water table depths time series

Abstract

Time series modeling provides an empirical stochastic method to model monitoring data from observation wells, without the complexity of physical mechanistic models. In the same direction, geostatiscal methods are used to make probabilistic statements about quantities of interest at non-measured locations. The aim of this work was to present water-table levels scenarios results of a combination of time series modeling and geostatistics to predict and discuss the communication via probability maps. The study case was held in a watershed located in an outcrop of the Guarani Aquifer System (GAS). The Onça Creek watershed has a monitoring scheme with 23 wells spatially distributed over the area. The water heads are measured with a semi-monthly frequency. First, the time series are inspected and modeled with a special type of Transfer-function noise model, the so called PIRFICT-model and then the model outputs are interpolated spatially using geostatistics. How communicate this results is discussed via the resulted maps that contain probabilistic measures about model uncertainty. Understand uncertainty and communicate it to practitioners, decision makers and stakeholders in a clear and simple form is a key element for efficient water resources planning.