Large-scale seasonal forecasts of river discharge by coupling local and global datasets with a stacked neural network: Case for the Loire River system.

Accurate prediction of river discharge is critical for a wide range of sectors, from human activities to environmental hazard management, especially in the face of increasing demand for water resources and climate change. To address this need, a multivariate model that incorporates both local and global data sources, including river and piezometer gauges, sea level, and climate parameters. By employing phase shift analysis, the model optimizes correlations between the target discharge and 12 parameters related to hydrologic and climatic systems, all sampled daily. In addition, a stacked LSTM - a more complex neural network architecture - is used to improve information extraction ability. Exploring river dynamics in the Loire-Bretagne basin and its surroundings, the investigation delves into predictions in daily time steps for one, three, and six months ahead. The resulting forecast features high accuracy and efficiency in predicting river discharge fluctuations, showcasing superior performance in forecasting drought periods over flood peaks. A detailed examination on data used highlights the significance of both local and global datasets in predicting river discharge, where the former dictates short-term predictions, while the latter drives long-range forecasts. Seasonally extended forecasting confirms a strong connection between the forecast leading time and the shift in data correlation, with lower correlation at a lag of 3 months due to seasonal changes affecting forecast quality, compensated by a higher correlation at a longer lag of 6 months. Such mutual effect in this multi-time-step forecasting improves the predictive quality of a six-month horizon, thus encourages progress in long-term prediction to a seasonal scale. The research establishes a practical foundation for effectively utilizing big data to leverage long-term forecasting of environmental dynamics.

Long-run forecasting surface and groundwater dynamics from intermittent observation data: An evaluation for 50 years.

The accurate prediction of water dynamics is critical for operational water resource management. In this study, we propose a novel approach to perform long-term forecasts of daily water dynamics, including river levels, river discharges, and groundwater levels, with a lead time of 7-30 days. The approach is based on the state-of-the-art neural network, bidirectional long short-term memory (BiLSTM), to enhance the accuracy and consistency of dynamic predictions. The operation of this forecasting system relies on an in-situ database observed for over 50 years with records gauging in 19 rivers, the karst aquifer, the English Channel, and the meteorological network in Normandy, France. To address the problem of missing measurements and gauge installations over time, we developed an adaptive scheme in which the neural network is regularly adjusted and re-trained in response to changing inputs during a long operation. Advances in BiLSTM with extensive learning past-to-future and future-to-past further help to avoid time-lag calibration that simplifies data processing. The proposed approach provides high accuracy and consistent prediction for the three water dynamics within a similar accuracy range as an on-site observation, with approximately 3 % error in the measurement range for the 7 day-ahead predictions and 6 % error for the 30 d-ahead predictions. The system also effectively fills the gap in actual measurements and detects anomalies at gauges that can last for years. Working with multiple dynamics not only proves that the data-driven model is a unified approach but also reveals the impact of the physical background of the dynamics on the performance of their predictions. Groundwater undergoes a slow filtration process following a low-frequency fluctuation, favoring long-term prediction, which differs from other higher-frequency river dynamics. The physical nature drives the predictive performance even when using a data-driven model.

Dramatic changes in a phytoplankton community in response to local and global pressures: a 24-year survey of the river Loire (France).

The impact of climate change and of other anthropogenic pressures on the structure and composition of phytoplankton communities of large European rivers remains poorly documented. Here we report the findings of a study of the changes in the phytoplankton community of the middle segment of the river Loire over the past 24 years. An attempt is made to distinguish between the impact of changes acting at the local scale and that of those acting more globally. A dramatic reduction in phytoplankton abundance was observed, particularly in the mid -1990s; this was concomitant with an increase in the relative proportion of cyanobacteria. At the same time, the phytoplankton community displayed increasing richness and diversity, and little change in its size structure. All these changes seem to be related to local changes, in particular to the reduction in phosphorus concentrations, as well as to changes in climate, throughout modifications in the river discharge and water temperature. Interestingly, herbicide contamination also appeared to be of particular importance in explaining the unexpected increase in the proportion of cyanobacteria in the phytoplankton community after the 1990s. These findings suggest that combinations of numerous anthropogenic pressures acting at different spatial and temporal scales have led to a mix of predictable and unpredictable changes occurring in the phytoplankton community of the river Loire, with probable consequences for the trophic networks in this river.

Investigation of transport processes inside karst aquifer by means of STATIS.

Environmental data sets are often multidimensional and consequently display complex structure. This article shows the limitations of principal component analysis (PCA) for the study of such three-dimensional (3D) data sets. These limitations can be resolved by the use of the statistical tool STATIS. The inlet (a swallow hole) and the outlet (a spring) of a karst system of the Western Paris basin were sampled during three rain events of various intensities. These 3D geochemical data sets (variables x sites x dates) for a karst system were analyzed by STATIS method to identify hydrological processes. STATIS proceeds in three steps (interstructure, compromise, and intrastructure), which allows us to focus the analysis of hydrologic systems at different temporal and spatial scales. Compromise plane shows that suspended matter and flood are not simultaneous and highlights a rapid flow, characterized by turbidity and phosphate, which represents a point source contamination, and a ground water flow contaminated by nitrate. Intrastructure plane allows us to compare hydrochemical variations between the swallow hole and the spring lead. By this way, hydrological processes such as direct transfer and resuspension of intrakarstic sediments before and after the flood were identified what cannot be realized by comparison of inlet and outlet breakthrough curves. Finally, results obtained from the same data set by STATIS and a coupled study using PCA and normalized hysteresis curves were compared. This comparison shows the efficiency of STATIS at the identification of transport processes and vulnerability of karst system and its potential for hydrological applications.

Nutrient dynamics as indicators of karst processes: comparison of the Chalk aquifer (Normandy, France) and the Edwards aquifer (Texas, U.S.A.).

Karst aquifers display a range of geologic and geomorphic characteristics in a wide range of climatic and land-use settings; identification of transport dynamics representative of karst aquifers in general could help advance our understanding of these complex systems. To this end, nutrient, turbidity, and major ion dynamics in response to storms were compared at multiple sites in two karst aquifers with contrasting characteristics and settings: the Chalk aquifer (Eure Department, Normandy, France) and the Barton Springs segment of the Edwards Aquifer (Texas, U.S.A.). The Chalk aquifer is typified by high matrix porosity, thick surficial deposits (up to 30 m thick), and agricultural land use; the Barton Springs segment is typified by low matrix porosity, outcropping limestone, and urban land use. Following one to three storms, from 5 to 16 samples from springs and wells were analyzed for major ions, and specific conductance and turbidity were monitored continuously. Comparison of the chemographs indicated some generalized responses, including an increase in turbidity and potassium concentrations and a decrease in major ion and nitrate concentrations with infiltrating storm runoff. Factor analysis of major ions and turbidity revealed strikingly similar behavior of the chemical variables for the two aquifers: The first two factors, explaining more than 75% of the variability, illustrate that dynamics of most major ions (including nitrate) are opposed to those of turbidity and of potassium. The results demonstrate that potassium and nitrate are effective tracers of infiltrating storm runoff and resident ground water, respectively, and the similar results for these two highly contrasting aquifers suggest that the dynamics identified might be applicable to karst systems in general.

Quantitative interpretation of specific conductance frequency distributions in karst.

Use of the coefficient of variation (CV) of specific conductance has been a simple and popular approach to classifying karst aquifers; however, problems with this approach arise because specific conductance frequency distributions (CFDs) are usually multimodal and the use of the CV sometimes erroneously classifies aquifers in terms of their dominant flow type or recharge type. Here, we demonstrate a more rigorous analysis of the CFD, which gives insight into the water types contributing to spring flow. The CFD for a water year is separated into an additive series of normal distributions, each related to a hydrogeochemical population. For each water type, its mean, variance, and contribution to the overall CFD can be quantified and compared between water types and water years. We applied this method to 4 years of data collected at Barton Springs, Austin, Texas. Although the overall shape of the CFD changed from year to year, it could consistently be separated into the same set of normally distributed populations. We suggest that each population represents a water type resulting from a particular mode of aquifer functioning. Changes in the parameters describing the curves reflect aquifer response to climatic variations. The results suggest that no single parameter of specific conductance can be used to describe the degree of karst behavior of an aquifer but that the degree of karst behavior itself varies from year to year depending on hydrologic conditions.