No abstract

No abstract

Abstract. Groundwater head and stream discharge is assimilated using the ensemble transform Kalman filter in an integrated hydrological model with the aim of studying the relationship between the filter performance and the ensemble size. In an attempt to reduce the required number of ensemble members, an adaptive localization method is used. The performance of the adaptive localization method is compared to the more common distance-based localization. The relationship between filter performance in terms of hydraulic head and discharge error and the number of ensemble members is investigated for varying numbers and spatial distributions of groundwater head observations and with or without discharge assimilation and parameter estimation. The study shows that (1) more ensemble members are needed when fewer groundwater head observations are assimilated, and (2) assimilating discharge observations and estimating parameters requires a much larger ensemble size than just assimilating groundwater head observations. However, the required ensemble size can be greatly reduced with the use of adaptive localization, which by far outperforms distance-based localization. The study is conducted using synthetic data only.

Observed 3-D temperature distributions within a streambed were used to analyze the effects on exchange fluxes between groundwater and the stream during rainfall-runoff events. By combining a dense vertical and lateral monitoring network of streambed temperatures with coupled surface/subsurface 3-D flow and heat transport modeling, we demonstrate how temperature can be used directly as a calibration target. Three model setups with different hydraulic conductivity distributions were evaluated in an optimization approach using temperature and hydraulic head data. The hydraulic conductivity distributions were based on slug test surveys within the streambed and aquifer. A detailed characterization of the hydraulic conductivity of the streambed and aquifer is needed to accurately simulate observed temperatures. Hence, the most sensitive parameters, the vertical hydraulic conductivity and the thermal conductivity, were calibrated within different conductivity zones of the heterogeneous model. Simulated exchange fluxes across the streambed showed variations up to a factor of four within just a meter. Such differences may not have been correctly predicted using 1-D heat transport models due to lateral conduction amongst the different flow paths. During the rainfall-runoff event, fluxes decreased substantially (250%) due to a decrease in the hydraulic gradient with increased stream stage. Although no flow reversals were observed during the studied conditions, it is possible that these can occur during larger rainfall-runoff events. We show that with the current sampling and modeling techniques, 3-D temperature data can be used to estimate dynamic exchange in heterogeneous flow fields encountered in the field.

Abstract. Groundwater head and stream discharge is assimilated using the Ensemble Transform Kalman Filter in an integrated hydrological model with the aim of studying the relationship between the filter performance and the ensemble size. In an attempt to reduce the required number of ensemble members, an adaptive localization method is used. The performance of the adaptive localization method is compared to the more common local analysis localization. The relationship between filter performance in terms of hydraulic head and discharge error and the number of ensemble members is investigated for varying numbers and spatial distributions of groundwater head observations and with or without discharge assimilation and parameter estimation. The study shows that (1) more ensemble members are needed when fewer groundwater head observations are assimilated, and (2) assimilating discharge observations and estimating parameters requires a much larger ensemble size than just assimilating groundwater head observations. However, the required ensemble size can be greatly reduced with the use of adaptive localization, which by far outperforms local analysis localization.