In the Luangwa basin in Zambia, long-term total water storage variations were observed with Gravity Recovery and Climate Experiment, but not reproduced by a standard conceptual hydrological model that encapsulates our current understanding of the dominant regional hydrological processes. The objective of this study was to identify potential processes underlying these low-frequency variations through combined data analysis and model hypothesis testing. First, we analyzed the effect of data uncertainty by contrasting observed storage variations with multi-annual estimates of precipitation and evaporation from multiple data sources. Second, we analyzed four different combinations of model forcing and evaluated their skill to reproduce the observed long-term storage variations. Third, we formulated alternative model hypotheses for groundwater export to potentially explain low-frequency storage variations. Overall, the results suggest that the initial model's inability to reproduce the observed low-frequency storage variations was partly due to the forcing data used and partly due to the missing representation of regional groundwater export. More specifically, the choice of data source affected the model's ability to reproduce annual maximum storage fluctuations, whereas the annual minima improved by adapting the model structure to allow for groundwater export from a deeper groundwater layer. This suggests that, in contrast to previous research, conceptual models can reproduce long-term storage fluctuations if a suitable model structure is used. Overall, the results highlight the value of alternative data sources and iterative testing of model structural hypotheses to improve runoff predictions in a poorly gauged basin leading to enhanced understanding of its hydrological processes.

Satellite observations can provide valuable information for a better understanding of hydrological processes and thus serve as valuable tools for model structure development and improvement. While model calibration and evaluation have in recent years started to make increasing use of spatial, mostly remotely sensed information, model structural development largely remains to rely on discharge observations at basin outlets only. Due to the ill-posed inverse nature and the related equifinality issues in the modelling process, this frequently results in poor representations of the spatio-Temporal heterogeneity of system-internal processes, in particular for large river basins. The objective of this study is thus to explore the value of remotely sensed, gridded data to improve our understanding of the processes underlying this heterogeneity and, as a consequence, their quantitative representation in models through a stepwise adaptation of model structures and parameters. For this purpose, a distributed, process-based hydrological model was developed for the study region, the poorly gauged Luangwa River basin. As a first step, this benchmark model was calibrated to discharge data only and, in a post-calibration evaluation procedure, tested for its ability to simultaneously reproduce (1) the basin-Average temporal dynamics of remotely sensed evaporation and total water storage anomalies and (2) their temporally averaged spatial patterns. This allowed for the diagnosis of model structural deficiencies in reproducing these temporal dynamics and spatial patterns. Subsequently, the model structure was adapted in a stepwise procedure, testing five additional alternative process hypotheses that could potentially better describe the observed dynamics and pattern. These included, on the one hand, the addition and testing of alternative formulations of groundwater upwelling into wetlands as a function of the water storage and, on the other hand, alternative spatial discretizations of the groundwater reservoir. Similar to the benchmark, each alternative model hypothesis was, in a next step, calibrated to discharge only and tested against its ability to reproduce the observed spatio-Temporal pattern in evaporation and water storage anomalies. In a final step, all models were re-calibrated to discharge, evaporation and water storage anomalies simultaneously. The results indicated that (1) the benchmark model (Model A) could reproduce the time series of observed discharge, basin-Average evaporation and total water storage reasonably well. In contrast, it poorly represented time series of evaporation in wetland-dominated areas as well as the spatial pattern of evaporation and total water storage. (2) Stepwise adjustment of the model structure (Models B-F) suggested that Model F, allowing for upwelling groundwater from a distributed representation of the groundwater reservoir and (3) simultaneously calibrating the model with respect to multiple variables, i.e. discharge, evaporation and total water storage anomalies, provided the best representation of all these variables with respect to their temporal dynamics and spatial patterns, except for the basin-Average temporal dynamics in the total water storage anomalies. It was shown that satellite-based evaporation and total water storage anomaly data are not only valuable for multi-criteria calibration, but can also play an important role in improving our understanding of hydrological processes through the diagnosis of model deficiencies and stepwise model structural improvement.

Study region: The study area is the river basin upstream of the Kariba dam located in the Zambezi River at the border of Zambia and Zimbabwe. Study focus: During the dry season of 2019 in Sub-Saharan Africa, extremely low water levels occurred in the Zambezi. According to news media, locals perceived this drought as the worst in several decades. We analyzed the 2019 drought in the Zambezi River Basin upstream of the Kariba dam to determine whether it indeed was the longest, most intense, and severe drought, in terms of precipitation, total water storage and reservoir water level observations over recent decades. New hydrological insights for the region: Data analysis indicates that the 2019 drought indeed had the lowest basin-averaged annual rainfall, most severe local rainfall deficit in the north of the basin, and lowest reservoir level since 1995. However, the rainfall deficit was more severe in 2002, both basin-wide and locally in the south of the basin. The total storage deficit was more severe in 2004, both basin-wide and locally in the central part of the basin. However, as the available storage data did not cover the entire deficit for 2019, its final duration and severity remain unknown. Therefore, it depends on the drought characteristic, hydrological variable, and location within the basin, whether the 2019 drought was indeed the most extreme over recent decades.

Limited availability of ground measurements in the vast majority of river basins world-wide increases the value of alternative data sources such as satellite observations in hydrological modelling. This study investigates the potential of using remotely sensed river water levels, i.e. altimetry observations, from multiple satellite missions to identify parameter sets for a hydrological model in the semi-arid Luangwa River basin in Zambia. A distributed process-based rainfall-runoff model with sub-grid process heterogeneity was developed and run on a daily timescale for the time period 2002 to 2016. As a benchmark, feasible model parameter sets were identified using traditional model calibration with observed river discharge data. For the parameter identification using remote sensing, data from the Gravity Recovery and Climate Experiment (GRACE) were used in a first step to restrict the feasible parameter sets based on the seasonal fluctuations in total water storage. Next, three alternative ways of further restricting feasible model parameter sets using satellite altimetry time series from 18 different locations along the river were compared. In the calibrated benchmark case, daily river flows were reproduced relatively well with an optimum Nash-Sutcliffe efficiency of ENS,Q = 0.78 (5/95th percentiles of all feasible solutions ENS,Q,5/95 = 0.61-0.75). When using only GRACE observations to restrict the parameter space, assuming no discharge observations are available, an optimum of ENS,Q =-1.4 (ENS,Q,5/95 =-2.3-0.38) with respect to discharge was obtained. The direct use of altimetry-based river levels frequently led to overestimated flows and poorly identified feasible parameter sets (ENS,Q,5/95 =-2.9-0.10). Similarly, converting modelled discharge into water levels using rating curves in the form of power relationships with two additional free calibration parameters per virtual station resulted in an overestimation of the discharge and poorly identified feasible parameter sets (ENS,Q,5/95 =-2.6-0.25). However, accounting for river geometry proved to be highly effective. This included using river cross-section and gradient information extracted from global high-resolution terrain data available on Google Earth and applying the Strickler-Manning equation to convert modelled discharge into water levels. Many parameter sets identified with this method reproduced the hydrograph and multiple other signatures of discharge reasonably well, with an optimum of ENS,Q = 0.60 (ENS,Q,5/95 =-0.31-0.50). It was further shown that more accurate river cross-section data improved the water-level simulations, modelled rating curve, and discharge simulations during intermediate and low flows at the basin outlet where detailed on-site cross-section information was available. Also, increasing the number of virtual stations used for parameter selection in the calibration period considerably improved the model performance in a spatial split-sample validation. The results provide robust evidence that in the absence of directly observed discharge data for larger rivers in data-scarce regions, altimetry data from multiple virtual stations combined with GRACE observations have the potential to fill this gap when combined with readily available estimates of river geometry, thereby allowing a step towards more reliable hydrological modelling in poorly gauged or ungauged basins.