Data assimilation in groundwater modelling: ensemble Kalman filter versus ensemble smoothers

Groundwater modelling calls for an effective and robust data integrating method to fill the gap between the model and observation data. The ensemble Kalman filter (EnKF), a real‐time data assimilation method, has been increasingly applied in multiple disciplines such as petroleum engineering and hydrogeology. In this approach, a groundwater model is updated sequentially with measured data such as hydraulic head and concentration. As an alternative to the EnKF, the ensemble smoother (ES) has been proposed for updating groundwater models using all the data together, with much less computational cost. To further improve the performance of the ES, an iterative ES has been proposed for continuously updating the model by assimilating measurements together. In this work, we compare the performance of the EnKF, the ES, and the iterative ES using a synthetic example in groundwater modelling. Hydraulic head data modelled on the basis of the reference conductivity field are used to inversely estimate conductivities at unsampled locations. Results are evaluated in terms of the characterization of conductivity and groundwater flow predictions. It is concluded that (a) the iterative ES works better than the standard ES because of its continuous updating and (b) the iterative ES could achieve results comparable with those of the EnKF, with less computational cost. These findings show that the iterative ES should be paid much more attention for data assimilation in groundwater modelling.     

Using data assimilation method to calibrate a heterogeneous conductivity field conditioning on transient flow test data

A data assimilation method is developed to calibrate a heterogeneous hydraulic conductivity field conditioning on transient pumping test data. The ensemble Kalman filter (EnKF) approach is used to update model parameters such as hydraulic conductivity and model variables such as hydraulic head using available data. A synthetical two-dimensional flow case is used to assess the capability of the EnKF method to calibrate a heterogeneous conductivity field by assimilating transient flow data from observation wells under different hydraulic boundary conditions. The study results indicate that the EnKF method will significantly improve the estimation of the hydraulic conductivity field by assimilating continuous hydraulic head measurements and the hydraulic boundary condition will significantly affect the simulation results. For our cases, after a few data assimilation steps, the assimilated conductivity field with four Neumann boundaries matches the real field well while the assimilated conductivity field with mixed Dirichlet and Neumann boundaries does not. We found in our cases that the ensemble size should be 300 or larger for the numerical simulation. The number and the locations of the observation wells will significantly affect the hydraulic conductivity field calibration.     

An approach to handling non-Gaussianity of parameters and state variables in ensemble Kalman filtering

The ensemble Kalman filter (EnKF) is a commonly used real-time data assimilation algorithm in various disciplines. Here, the EnKF is applied, in a hydrogeological context, to condition log-conductivity realizations on log-conductivity and transient piezometric head data. In this case, the state vector is made up of log-conductivities and piezometric heads over a discretized aquifer domain, the forecast model is a groundwater flow numerical model, and the transient piezometric head data are sequentially assimilated to update the state vector. It is well known that all Kalman filters perform optimally for linear forecast models and a multiGaussian-distributed state vector. Of the different Kalman filters, the EnKF provides a robust solution to address non-linearities; however, it does not handle well non-Gaussian state-vector distributions. In the standard EnKF, as time passes and more state observations are assimilated, the distributions become closer to Gaussian, even if the initial ones are clearly non-Gaussian. A new method is proposed that transforms the original state vector into a new vector that is univariate Gaussian at all times. Back transforming the vector after the filtering ensures that the initial non-Gaussian univariate distributions of the state-vector components are preserved throughout. The proposed method is based in normal-score transforming each variable for all locations and all time steps. This new method, termed the normal-score ensemble Kalman filter (NS-EnKF), is demonstrated in a synthetic bimodal aquifer resembling a fluvial deposit, and it is compared to the standard EnKF. The proposed method performs better than the standard EnKF in all aspects analyzed (log-conductivity characterization and flow and transport predictions).     

Use of Machine Learning Methods to Reduce Predictive Error of Groundwater Models

Quantitative analyses of groundwater flow and transport typically rely on a physically‐based model, which is inherently subject to error. Errors in model structure, parameter and data lead to both random and systematic error even in the output of a calibrated model. We develop complementary data‐driven models (DDMs) to reduce the predictive error of physically‐based groundwater models. Two machine learning techniques, the instance‐based weighting and support vector regression, are used to build the DDMs. This approach is illustrated using two real‐world case studies of the Republican River Compact Administration model and the Spokane Valley‐Rathdrum Prairie model. The two groundwater models have different hydrogeologic settings, parameterization, and calibration methods. In the first case study, cluster analysis is introduced for data preprocessing to make the DDMs more robust and computationally efficient. The DDMs reduce the root‐mean‐square error (RMSE) of the temporal, spatial, and spatiotemporal prediction of piezometric head of the groundwater model by 82%, 60%, and 48%, respectively. In the second case study, the DDMs reduce the RMSE of the temporal prediction of piezometric head of the groundwater model by 77%. It is further demonstrated that the effectiveness of the DDMs depends on the existence and extent of the structure in the error of the physically‐based model.     

Using data assimilation method to calibrate a heterogeneous conductivity field and improve solute transport prediction with an unknown contamination source

Hydraulic conductivity distribution and plume initial source condition are two important factors affecting solute transport in heterogeneous media. Since hydraulic conductivity can only be measured at limited locations in a field, its spatial distribution in a complex heterogeneous medium is generally uncertain. In many groundwater contamination sites, transport initial conditions are generally unknown, as plume distributions are available only after the contaminations occurred. In this study, a data assimilation method is developed for calibrating a hydraulic conductivity field and improving solute transport prediction with unknown initial solute source condition. Ensemble Kalman filter (EnKF) is used to update the model parameter (i.e., hydraulic conductivity) and state variables (hydraulic head and solute concentration), when data are available. Two-dimensional numerical experiments are designed to assess the performance of the EnKF method on data assimilation for solute transport prediction. The study results indicate that the EnKF method can significantly improve the estimation of the hydraulic conductivity distribution and solute transport prediction by assimilating hydraulic head measurements with a known solute initial condition. When solute source is unknown, solute prediction by assimilating continuous measurements of solute concentration at a few points in the plume well captures the plume evolution downstream of the measurement points.     

Influence of calibration methodology on ground water flow predictions

We constructed a numerical model of transient ground water flow and solute transport for a portion of the Biscayne Aquifer in Florida, and calibrated the model with three different combinations of data from a 193-day period: head (h) data alone, data on h and ground water discharge to a canal (q), and data on h, q, and ground water chloride concentration (C). We used each of the three calibrated models to predict h and q during a 182-day test period separate from the calibration period. All three calibrated models predicted h equally well during the test period (r = 0.95, where r = 1 indicates perfect agreement between measured and simulated values), though the model calibrated on h alone had significantly different parameter values than the other two models. Predictions of q during the test period depended on calibration methodology; models calibrated with multiple targets simulated q more accurately than the model calibrated on h alone (r = 0.79 compared to r = 0.49). Based on the results of these simulations, we conclude: (1) Post-calibration prediction is important in assessing the value of different data types in automated calibration; (2) inverse-solution uniqueness is not a requirement for accurate h predictions; (3) relatively simple models can predict with reasonable accuracy transient ground water flow in a complex aquifer, and parameters governing this prediction can be estimated by nonlinear regression methods that incorporate both h and q data; (4) addition of C data to the calibration did not improve model predictive capacity because the information in the C data was similar to that in the q data, from the perspective of model calibration (the subsurface chemical signal in question was controlled mainly by seepage of high-chloride canal water into the low-chloride ground water system).     

Comparison of Ensemble Kalman Filter groundwater-data assimilation methods based on stochastic moment equations and Monte Carlo simulation

Traditional Ensemble Kalman Filter (EnKF) data assimilation requires computationally intensive Monte Carlo (MC) sampling, which suffers from filter inbreeding unless the number of simulations is large. Recently we proposed an alternative EnKF groundwater-data assimilation method that obviates the need for sampling and is free of inbreeding issues. In our new approach, theoretical ensemble moments are approximated directly by solving a system of corresponding stochastic groundwater flow equations. Like MC-based EnKF, our moment equations (ME) approach allows Bayesian updating of system states and parameters in real-time as new data become available. Here we compare the performances and accuracies of the two approaches on two-dimensional transient groundwater flow toward a well pumping water in a synthetic, randomly heterogeneous confined aquifer subject to prescribed head and flux boundary conditions.     

The Ensemble Kalman Filter for Groundwater Plume Characterization: A Case Study

The Kalman filter is an efficient data assimilation tool to refine an estimate of a state variable using measured data and the variable's correlations in space and/or time. The ensemble Kalman filter (EnKF) (Evensen 2004, 2009) is a Kalman filter variant that employs Monte Carlo analysis to define the correlations that help to refine the updated state. While use of EnKF in hydrology is somewhat limited, it has been successfully applied in other fields of engineering (e.g., oil reservoir modeling, weather forecasting). Here, EnKF is used to refine a simulated groundwater tetrachloroethylene (TCE) plume that underlies the Tooele Army Depot‐North (TEAD‐N) in Utah, based on observations of TCE in the aquifer. The resulting EnKF‐based assimilated plume is simulated forward in time to predict future plume migration. The correlations that underpin EnKF updating implicitly contain information about how the plume developed over time under the influence of complex site hydrology and variable source history, as they are predicated on multiple realizations of a well‐calibrated numerical groundwater flow and transport model. The EnKF methodology is compared to an ordinary kriging‐based assimilation method with respect to the accurate representation of plume concentrations in order to determine the relative efficacy of EnKF for water quality data assimilation.     

Assimilating transient groundwater flow data via a localized ensemble Kalman filter to calibrate a heterogeneous conductivity field

A localized ensemble Kalman filter (EnKF) method is developed to assimilate transient flow data to calibrate a heterogeneous conductivity field. To update conductivity value at a point in a study domain, instead of assimilating all the measurements in the study domain, only limited measurement data in an area around the point are used for the conductivity updating in the localized EnKF method. The localized EnKF is proposed to solve the problems of the filter divergence usually existing in a data assimilation method without localization. The developed method is applied, in a synthetical two dimensional case, to calibrate a heterogeneous conductivity field by assimilating transient hydraulic head data. The simulations by the data assimilation with and without localized EnKF are compared. The study results indicate that the hydraulic conductivity field can be updated efficiently by the localized EnKF, while it cannot be by the EnKF. The covariance inflation and localization are found to solve the problem of the filter divergence efficiently. In comparison with the EnKF method without localization, the localized EnKF method needs smaller ensemble size to achieve stabilized results. The simulation results by the localized EnKF method are much more sensitive to conductivity correlation length than to the localization radius. The developed localized EnKF method provides an approach to improve EnKF method in conductivity calibration.     

Assimilation of SLA and SST data into an OGCM for the Indian Ocean

 Remotely sensed observations of sea-level anomaly and sea-surface temperature have been assimilated into an implementation of the Miami Isopycnic Coordinate Ocean Model (MICOM) for the Indian Ocean using the Ensemble Kalman Filter (EnKF). The system has been applied in a hindcast validation experiment to examine the properties of the assimilation scheme when used with a full ocean general circulation model and real observations. This work is considered as a first step towards an operational ocean monitoring and forecasting system for the Indian Ocean. The assimilation of real data has demonstrated that the sequential EnKF can efficiently control the model evolution in time. The use of data assimilation requires a significant amount of additional processing and computational resources. However, we have tried to justify the cost of using a sophisticated assimilation scheme by demonstrating strong regional and temporal dependencies of the covariance statistics, which include highly anisotropic and flow-dependent correlation functions. In particular, we observed a marked difference between error statistics in the equatorial region and at off-equatorial latitudes. We have also demonstrated how the assimilation of SLA and SST improves the model fields with respect to real observations. Independent in situ temperature profiles have been used to examine the impact of assimilating the remotely sensed observations. These intercomparisons have shown that the model temperature and salinity fields better resemble in situ observations in the assimilation experiment than in a model free-run case. On the other hand, it is also expected that assimilation of in situ profiles is needed to properly control the deep ocean circulation. Received: 8 January 2002 / Accepted: 8 April 2002     

Calibration of a Land Subsidence Model Using InSAR Data via the Ensemble Kalman Filter

The application of interferometric synthetic aperture radar (InSAR) has been increasingly used to improve capabilities to model land subsidence in hydrogeologic studies. A number of investigations over the last decade show how spatially detailed time‐lapse images of ground displacements could be utilized to advance our understanding for better predictions. In this work, we use simulated land subsidences as observed measurements, mimicking InSAR data to inversely infer inelastic specific storage in a stochastic framework. The inelastic specific storage is assumed as a random variable and modeled using a geostatistical method such that the detailed variations in space could be represented and also that the uncertainties of both characterization of specific storage and prediction of land subsidence can be assessed. The ensemble Kalman filter (EnKF), a real‐time data assimilation algorithm, is used to inversely calibrate a land subsidence model by matching simulated subsidences with InSAR data. The performance of the EnKF is demonstrated in a synthetic example in which simulated surface deformations using a reference field are assumed as InSAR data for inverse modeling. The results indicate: (1) the EnKF can be used successfully to calibrate a land subsidence model with InSAR data; the estimation of inelastic specific storage is improved, and uncertainty of prediction is reduced, when all the data are accounted for; and (2) if the same ensemble is used to estimate Kalman gain, the analysis errors could cause filter divergence; thus, it is essential to include localization in the EnKF for InSAR data assimilation.     

Improving hydrologic predictions of a catchment model via assimilation of surface soil moisture

This paper examines the potential for improving Soil and Water Assessment Tool (SWAT) hydrologic predictions of root-zone soil moisture, evapotranspiration, and stream flow within the 341 km² Cobb Creek Watershed in southwestern Oklahoma through the assimilation of surface soil moisture observations using an Ensemble Kalman filter (EnKF). In a series of synthetic twin experiments assimilating surface soil moisture is shown to effectively update SWAT upper-layer soil moisture predictions and provide moderate improvement to lower layer soil moisture and evapotranspiration estimates. However, insufficient SWAT-predicted vertical coupling results in limited updating of deep soil moisture, regardless of the SWAT parameterization chosen for root-water extraction. Likewise, a real data assimilation experiment using ground-based soil moisture observations has only limited success in updating upper-layer soil moisture and is generally unsuccessful in enhancing SWAT stream flow predictions. Comparisons against ground-based observations suggest that SWAT significantly under-predicts the magnitude of vertical soil water coupling at the site, and this lack of coupling impedes the ability of the EnKF to effectively update deep soil moisture, groundwater flow and surface runoff. The failed attempt to improve stream flow prediction is also attributed to the inability of the EnKF to correct for existing biases in SWAT-predicted stream flow components.     

Comparison between gradient based UCODE_2005 and the ensemble Kalman Filter for transient groundwater flow inverse modeling

Gradient based UCODE_2005 and data assimilation based on the Ensemble Kalman Filter(EnKF) are two different inverse methods. A synthetic two-dimensional flow case with four no-flow boundaries is used to compare the UCODE_2005 with the Ensemble Kalman Filter(EnKF) for their efficiency to inversely calculate and calibrate a hydraulic conductivity field based on hydraulic head data. A zonal, random heterogeneous conductivity field is calibrated by assimilating the time series of heads observed in monitoring wells. The study results indicate that the two inverse methods, UCODE_2005 and EnKF, could be used to calibrate the hydraulic conductivity field to a certain degree. More available observations and information about the conductivity field, more accurate inverse results will be obtained for the UCODE_2005. On the other hand, for a realistic zonal heterogeneous hydraulic conductivity field, EnKF can only efficiently determine the hydraulic conductivity field at the first several assimilated time steps. The results obtained by the UCODE_2005 look better than those by the EnKF. This is possibly due to the fact that the UCODE_2005 uses observed head data at every time step, while EnKF can only use observed heads at first several steps due to the filter divergence problem.     

AN EXERCISE IN GROUND-WATER MODEL CALIBRATION AND PREDICTION

Abstract. For a classroom exercise, nine groups of graduate students calibrated a numerical ground-water flow model to a set of perfectly observed hydraulic head data for a hypothetical phreatic aquifer. All groups used exactly the same numerical model and identical sets of observed data. After calibration, the students predicted the hydraulic head distribution in the aquifer resulting from a modification in one boundary condition. A quantitative analysis of the results of this calibration-prediction exercise vividly demonstrates some of the difficulties in parameter identification for ground-water flow models. Group predictions differed significantly. Successful prediction was strongly correlated with successful estimation of conductivity values, and was essentially unrelated to successful estimation of aquifer bottom elevations or with the number of trial-and-error simulations required for calibration. Most importantly, success in prediction was unrelated to success in matching observed heads under premodification conditions. In this sense, good calibration did not lead to good prediction.     

Transient Recharge Estimability Through Field‐Scale Groundwater Model Calibration

The estimation of recharge through groundwater model calibration is hampered by the nonuniqueness of recharge and aquifer parameter values. It has been shown recently that the estimability of spatially distributed recharge through calibration of steady‐state models for practical situations (i.e., real‐world, field‐scale aquifer settings) is limited by the need for excessive amounts of hydraulic‐parameter and groundwater‐level data. However, the extent to which temporal recharge variability can be informed through transient model calibration, which involves larger water‐level datasets, but requires the additional consideration of storage parameters, is presently unknown for practical situations. In this study, time‐varying recharge estimates, inferred through calibration of a field‐scale highly parameterized groundwater model, are systematically investigated subject to changes in (1) the degree to which hydraulic parameters including hydraulic conductivity (K) and specific yield (S_y) are constrained, (2) the number of water‐level calibration targets, and (3) the temporal resolution (up to monthly time steps) at which recharge is estimated. The analysis involves the use of a synthetic reality (a reference model) based on a groundwater model of Uley South Basin, South Australia. Identifiability statistics are used to evaluate the ability of recharge and hydraulic parameters to be estimated uniquely. Results show that reasonable estimates of monthly recharge (<30% recharge root‐mean‐squared error) require a considerable amount of transient water‐level data, and that the spatial distribution of K is known. Joint estimation of recharge, S_y and K, however, precludes reasonable inference of recharge and hydraulic parameter values. We conclude that the estimation of temporal recharge variability through calibration may be impractical for real‐world settings.     

Real-time assimilation of streamflow observations into a hydrological routing model: effects of model structures and updating methods

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.     

Application of a data assimilation method via an ensemble Kalman filter to reactive urea hydrolysis transport modeling

With growing importance of water resources in the world, remediations of anthropogenic contaminations due to reactive solute transport become even more important. A good understanding of reactive rate parameters such as kinetic parameters is the key to accurately predicting reactive solute transport processes and designing corresponding remediation schemes. For modeling reactive solute transport, it is very difficult to estimate chemical reaction rate parameters due to complex processes of chemical reactions and limited available data. To find a method to get the reactive rate parameters for the reactive urea hydrolysis transport modeling and obtain more accurate prediction for the chemical concentrations, we developed a data assimilation method based on an ensemble Kalman filter (EnKF) method to calibrate reactive rate parameters for modeling urea hydrolysis transport in a synthetic one-dimensional column at laboratory scale and to update modeling prediction. We applied a constrained EnKF method to pose constraints to the updated reactive rate parameters and the predicted solute concentrations based on their physical meanings after the data assimilation calibration. From the study results we concluded that we could efficiently improve the chemical reactive rate parameters with the data assimilation method via the EnKF, and at the same time we could improve solute concentration prediction. The more data we assimilated, the more accurate the reactive rate parameters and concentration prediction. The filter divergence problem was also solved in this study.     

Data assimilation for distributed hydrological catchment modeling via ensemble Kalman filter

Catchment scale hydrological models are critical decision support tools for water resources management and environment remediation. However, the reliability of hydrological models is inevitably affected by limited measurements and imperfect models. Data assimilation techniques combine complementary information from measurements and models to enhance the model reliability and reduce predictive uncertainties. As a sequential data assimilation technique, the ensemble Kalman filter (EnKF) has been extensively studied in the earth sciences for assimilating in-situ measurements and remote sensing data. Although the EnKF has been demonstrated in land surface data assimilations, there are no systematic studies to investigate its performance in distributed modeling with high dimensional states and parameters. In this paper, we present an assessment on the EnKF with state augmentation for combined state-parameter estimation on the basis of a physical-based hydrological model, Soil and Water Assessment Tool (SWAT). Through synthetic simulation experiments, the capability of the EnKF is demonstrated by assimilating the runoff and other measurements, and its sensitivities are analyzed with respect to the error specification, the initial realization and the ensemble size. It is found that the EnKF provides an efficient approach for obtaining a set of acceptable model parameters and satisfactory runoff, soil water content and evapotranspiration estimations. The EnKF performance could be improved after augmenting with other complementary data, such as soil water content and evapotranspiration from remote sensing retrieval. Sensitivity studies demonstrate the importance of consistent error specification and the potential with small ensemble size in the data assimilation system.     

Hydrological data assimilation with the Ensemble Square-Root-Filter: Use of streamflow observations to update model states for real-time flash flood forecasting

The objective of the study is to evaluate the potential of a data assimilation system for real-time flash flood forecasting over small watersheds by updating model states. To this end, the Ensemble Square-Root-Filter (EnSRF) based on the Ensemble Kalman Filter (EnKF) technique was coupled to a widely used conceptual rainfall-runoff model called HyMOD. Two small watersheds susceptible to flash flooding from America and China were selected in this study. The modeling and observational errors were considered in the framework of data assimilation, followed by an ensemble size sensitivity experiment. Once the appropriate model error and ensemble size was determined, a simulation study focused on the performance of a data assimilation system, based on the correlation between streamflow observation and model states, was conducted. The EnSRF method was implemented within HyMOD and results for flash flood forecasting were analyzed, where the calibrated streamflow simulation without state updating was treated as the benchmark or nature run. Results for twenty-four flash-flood events in total from the two watersheds indicated that the data assimilation approach effectively improved the predictions of peak flows and the hydrographs in general. This study demonstrated the benefit and efficiency of implementing data assimilation into a hydrological model to improve flash flood forecasting over small, instrumented basins with potential application to real-time alert systems.     

Dual states estimation of a subsurface flow-transport coupled model using ensemble Kalman filtering

Modeling the spread of subsurface contaminants requires coupling a groundwater flow model with a contaminant transport model. Such coupling may provide accurate estimates of future subsurface hydrologic states if essential flow and contaminant data are assimilated in the model. Assuming perfect flow, an ensemble Kalman filter (EnKF) can be used for direct data assimilation into the transport model. This is, however, a crude assumption as flow models can be subject to many sources of uncertainty. If the flow is not accurately simulated, contaminant predictions will likely be inaccurate even after successive Kalman updates of the contaminant model with the data. The problem is better handled when both flow and contaminant states are concurrently estimated using the traditional joint state augmentation approach. In this paper, we introduce a dual estimation strategy for data assimilation into a one-way coupled system by treating the flow and the contaminant models separately while intertwining a pair of distinct EnKFs, one for each model. The presented strategy only deals with the estimation of state variables but it can also be used for state and parameter estimation problems. This EnKF-based dual state-state estimation procedure presents a number of novel features: (i) it allows for simultaneous estimation of both flow and contaminant states in parallel; (ii) it provides a time consistent sequential updating scheme between the two models (first flow, then transport); (iii) it simplifies the implementation of the filtering system; and (iv) it yields more stable and accurate solutions than does the standard joint approach. We conducted synthetic numerical experiments based on various time stepping and observation strategies to evaluate the dual EnKF approach and compare its performance with the joint state augmentation approach. Experimental results show that on average, the dual strategy could reduce the estimation error of the coupled states by 15% compared with the joint approach. Furthermore, the dual estimation is proven to be very effective computationally, recovering accurate estimates at a reasonable cost.