There is an increasing need for accurate groundwater-level (GWL) prediction to support effective seasonal water management. It is desirable for forecasting tools to be not only accurate but also accessible for decision-makers. We test the Prophet forecasting procedure, an open-source code released by Facebook, to address these challenges. It is based on an additive model considering non-periodic changes and periodic components in a Bayesian framework with easily-interpretable parameters. Predictions of daily GWL data in an area affected by pumping near a tourist complex in the Ramsar wetland area of Doñana (Spain) are compared to other forecasting methods. Prophet outperforms most methods in predicting GWL making it a fast and flexible forecasting tool for hydrologists and water managers. Furthermore, it allows gaining insight into the influence of each component of the forecast separately, helping to assess the hydrodynamic response to external drivers such as groundwater pumping.

Accurate estimation of missing daily precipitation data remains a difficult task. A wide variety of methods exists for infilling missing values, but the percentage of gaps is one of the main factors limiting their applicability. The present study compares three techniques for filling in large amounts of missing daily precipitation data: spatio-temporal kriging (STK), multiple imputation by chained equations through predictive mean matching (PMM), and the random forest (RF) machine learning algorithm. To our knowledge, this is the first time that extreme missingness (>90%) has been considered. Different percentages of missing data and missing patterns are tested in a large dataset drawn from 112 rain gauges in the period 1975–2017. The results show that both STK and RF can handle extreme missingness, while PMM requires larger observed sample sizes. STK is the most robust method, suitable for chronological missing patterns. RF is efficient under random missing patterns. Model evaluation is usually based on performance and error measures. However, this study outlines the risk of just relying on these measures without checking for consistency. The RF algorithm overestimated daily precipitation outside the validation period in some cases due to the overdetection of rainy days under time-dependent missing patterns.

Groundwater resources are regularly the principal water supply in semiarid and arid climate areas. However, groundwater levels (GWL) in semiarid aquifers are suffering a general decrease because of anthropic exploitation of aquifers and the repercussions of climate change. Effective groundwater management strategies require a deep characterization of GWL fluctuations, in order to identify individual behaviors and triggering factors. In September 2019, the Guadalquivir River Basin Authority (CHG) declared that there was over-exploitation in three of the five groundwater bodies of the Almonte-Marismas aquifer, Southwest Spain. For that reason, it is critical to understand GWL dynamics in this aquifer before the new Spanish Water Resources Management Plans (2021–2027) are developed. The application of GWL series clustering in hydrogeology has grown over the past few years, as it is an extraordinary tool that promptly provides a GWL classification; each group can be related to different responses of a complex aquifer under any external change. In this work, GWL time series from 160 piezometers were analyzed for the period 1975 to 2016 and, after data pre-processing, 24 piezometers were selected for clustering with k-means (static) and time series (dynamic) clustering techniques. Six and seven groups (k) were chosen to apply k-means. Six characterized types of hydrodynamic behaviors were obtained with time series clustering (TSC). Number of clusters were related to diverse affections of water exploitation depending on soil uses and hydrogeological spatial distribution parameters. TSC enabled us to distinguish local areas with high hydrodynamic disturbance and to highlight a quantitative drop of GWL during the studied period.