In the last decades, advancements in computational science have greatly expanded the use of artificial neural networks (ANNs) in hydrogeology, including applications on groundwater forecast, variable selection, extended lead-times, and regime-specific analysis. However, ANN-model performance often omits the sensitivity to observational uncertainties in hydroclimate forcings. The goal of this paper is to implement a data-driven modeling framework for assessing the sensitivity of ANN-based groundwater forecasts to the uncertainties in observational inputs across space, time, and hydrological regimes. The objectives are two-folded. The first objective is to couple an ANN model with the PAWN sensitivity analysis (SA). The second objective is to evaluate the scale- and process-dependent sensitivities of groundwater forecasts to hydroclimate inputs, computing the sensitivity index in groundwater wells (1) across the whole time-series (for the global sensitivity analysis); (2) across the output sub-regions with conditions of water deficit and water surplus (for the ‘regional’ sensitivity analysis); and (3) at each time step (for the time-varying sensitivity analysis). The implementation of the ANN-PAWN occurs in 68 wells across the Northern High Plains aquifer, USA, with pre-time-step rainfall, evapotranspiration, snowmelt, streamflow, and groundwater measurements as inputs. Results show that evapotranspiration and rainfall are the major sources of uncertainty, with the latter being particularly relevant in water surplus conditions and the former in water deficit conditions. The time-varying sensitivity analysis leads to the identification of localized sensitivities to other sources of uncertainty, as snowmelt in spring or river flow during the annual peak period at the groundwater level.

@en

The aim of this paper is to improve semiseasonal forecast of groundwater availability in response to climate variables, surface water availability, groundwater level variations, and human water management using a two-step data-driven modeling approach. First, we implement an ensemble of artificial neural networks (ANNs) for the 300 wells across the High Plains aquifer (USA). The modeling framework includes a method to choose the most relevant input variables and time lags; an assessment of the effect of exogenous variables on the predictive capabilities of models; and the estimation of the forecast skill based on the Nash-Sutcliffe efficiency (NSE) index, the normalized root mean square error, and the coefficient of determination (R²). Then, for the ANNs with low- accuracy, a MultiModel Combination (MuMoC) based on a hybrid of ANN and an instance-based learning method is applied. MuMoC uses forecasts from neighboring wells to improve the accuracy of ANNs. An exhaustive-search optimization algorithm is employed to select the best neighboring wells based on the cross correlation and predictive accuracy criteria. The results show high average ANN forecasting skills across the aquifer (average NSE > 0.9). Spatially distributed metrics of performance showed also higher error in areas of strong interaction between hydrometeorological forcings, irrigation intensity, and the aquifer. In those areas, the integration of the spatial information into MuMoC leads to an improvement of the model accuracy (NSE increased by 0.12), with peaks higher than 0.3 when the optimization objectives for selecting the neighbors were maximized.tT.

@en

This study aims at proposing novel approaches for integrating qualitative flow observations in a lumped hydrologic routing model and assessing their usefulness for improving flood estimation. Routing is based on a three-parameter Muskingum model used to propagate streamflow in five different rivers in the United States. Qualitative flow observations, synthetically generated from observed flow, are converted into fuzzy observations using flow characteristic for defining fuzzy classes. A model states updating method and a model output correction technique are implemented. An innovative application of Interacting Multiple Models, which use was previously demonstrated on tracking in ballistic missile applications, is proposed as state updating method, together with the traditional Kalman filter. The output corrector approach is based on the fuzzy error corrector, which was previously used for robots navigation. This study demonstrates the usefulness of integrating qualitative flow observations for improving flood estimation. In particular, state updating methods outperform the output correction approach in terms of average improvement of model performances, while the latter is found to be less sensitive to biased observations and to the definition of fuzzy sets used to represent qualitative observations.@en

In agricultural areas where groundwater is the main water supply for irrigation, forecasts of the water table are key to optimal water management. However, water management can be constrained by semiseasonal to seasonal forecast. The objective is to create an ensemble of water table one- to fivemonth lead-time forecasts based on multiple data-driven models (DDMs). We hypothesize that datadriven modeling capabilities (e.g., random forests, support vector machines, artificial neural networks (ANNs), extreme learning machines, and genetic programming) will improve naïve and autoregressive simulations of groundwater tables. An input variable selection method was used to carry the maximum information in the input (precipitation, crop water demand, changes in groundwater table, snowmelt, and evapotranspiration) and output relationship for the DDMs. Five DDMs were implemented to forecast groundwater tables in an unconfined aquifer in the Northern High Plains (Nebraska, USA). Root mean squared error (RMSE) and Nash-Sutcliffe efficiency index were used to evaluate the skill of the model and three hydrologic regimes were determined statistically as high, mid, and low groundwater table levels. Additionally, varying storage regimes were used to construct rising and falling limbs in the tested well. Results show that all models outperform the baseline for all the lead times, ANNs being the best of all.

@en

This paper comparatively assesses the performance of five data assimilation techniques for three-parameter Muskingum routing with a spatially lumped or distributed model structure. The assimilation techniques used include direct insertion (DI), nudging scheme (NS), Kalman filter (KF), ensemble Kalman filter (EnKF) and asynchronous ensemble Kalman filter (AEnKF), which are applied to river reaches in Texas and Louisiana, USA. For both lumped and distributed routing, results from KF, EnKF and AEnKF are sensitive to the error specification. As expected, DI outperformed the other models in the case of lumped modelling, while in distributed routing, KF approaches, particularly AEnKF and EnKF, performed better than DI or nudging, reflecting the benefit of updating distributed states through error covariance modelling in KF approaches. The results of this work would be useful in setting up data assimilation systems that employ increasingly abundant real-time observations using distributed hydrological routing models.

@en