Quantification of climate-induced changes in groundwater levels using fuzzy rule-based modelling

"Quantification of climate-induced changes in groundwater levels using fuzzy rule-based modelling"

Ronja Iffland and Uwe Haberlandt
Institute of Hydrology and Water Resources Management, Leibniz University of Hanover, Germany (iffland@iww.uni-hannover.de)

The impact of climate change on hydrological processes, such as increased flooding and prolonged droughts, also affects groundwater recharge and therefore groundwater levels. To make reliable statements about possible changes in groundwater level dynamics prediction models are needed. In this study, fuzzy rule-based models are used to analyse and quantify the effects of changing climate conditions on groundwater levels.

Focussing on 114 groundwater wells in Lower Saxony, Germany, the study aims to explain groundwater dynamics using assumed relationships between climatic indices and groundwater levels. Starting from linear methods, we used fuzzy logic to capture the non-linearities in groundwater level systems. While fuzzy logic models have mostly been considered in combination with neural networks in groundwater level prediction, our approach utilises the transparency of fuzzy rule-based modelling to maintain model interpretability. To improve the forecast accuracy, we introduced moving averages and time lags to account for the persistent influence of meteorological indices. As reference we calculated multiple linear regression models. The performance of both fuzzy rule-based models and linear regression models are evaluated using split validation. To predict future changes in groundwater levels, we applied both models to climate model data based on the RCP8.5 scenario. It is expected that the non-linear fuzzy rule-based models outperform the linear regression models.