Quantifying hydrologic modeling uncertainties originating from meteorological uncertainties is crucial yet unexplored on the global scale. This study integrates a novel ensemble meteorological dataset with process-based hydrologic models, assessing the impact of precipitation and temperature uncertainties across approximately three million subbasins globally. We introduce two metrics to identify uncertainty hotspots: one tracing the uncertainty propagation from inputs to model outputs and the other measuring the uncertainty magnitude relative to hydrologic climatology (i.e., ratio between uncertainty and climate average). Our findings reveal different uncertainty responses across hydrologic variables to the combined precipitation and temperature uncertainties. For routed river streamflow, uncertainty propagation is strong in tropical rain forests and Europe (except the Scandinavian Peninsula) but weak in the deserts, which is partly attributed to the regional differences in baseflow ratios. In contrast, both uncertainty metrics indicate low streamflow uncertainties in cryosphere areas and downstream areas of major rivers. The substantial modeling uncertainties observed, particularly in the Southern Hemisphere and less developed regions, underscore the need to improve global spatial meteorological datasets.

Global sensitivity analysis (GSA) enhances our understanding of computational models and simplifies model parameter estimation. VISCOUS (VarIance-based Sensitivity analysis using COpUlaS) is a variance-based GSA framework. The advantage of VISCOUS is that it can use existing model input-output data (e.g., water model parameters-responses) to estimate the first- and total-order Sobol’ sensitivity indices. This study improves VISCOUS by refining its handling of marginal densities of the Gaussian mixture copula model (GMCM). We then evaluate VISCOUS using three types of generic functions relevant to water system models. We observe that its performance depends on function dimension, input-output data size, and non-identifiability. Function dimension refers to the number of uncertain input factors analyzed in GSA, and non-identifiability is the inability to estimate GMCM parameters. VISCOUS proves powerful in estimating first-order sensitivity with a small amount of input-output data (e.g., 200 in this study), regardless of function dimension. It always ranks input factors correctly in both first- and total-order terms. For estimating total-order sensitivity, it is recommended to use VISCOUS when the function dimension is not very high (e.g., less than 20) due to the challenge of producing sufficient input-output data for accurate GMCM inferences (e.g., more than 10000). In cases where all input factors are equally important, a rarity in practice, VISCOUS faces non-identifiability issues that impact its performance. We provide a didactic example and an open-source Python code, pyVISCOUS, for broader user adoption.

Reservoirs are ubiquitous water infrastructure, providing functional capability to manage, and often mitigate, hydrological variability across space and time. The presence and operation of a reservoir control the downstream flow regime, such that in many locations understanding reservoir operations is crucial to understanding the hydrological functioning of a catchment. Despite many advances in modelling reservoir operations, inclusion of reservoirs in large-scale hydrological modelling remains challenging, particularly when the number of reservoirs is large and data access is limited. Here we design a set of simple reservoir operating rules (with only two calibrated parameters) focused on simulating small water supply reservoirs across large scales with various types of open-access data (i.e. catchment attributes and flows at downstream gauges). We integrate our rules into a national-scale hydrological model of Great Britain and compare hydrological simulations with and without the new reservoir component. Our simple reservoir operating rules significantly increase model performance in reservoir-impacted catchments, particularly when the rules are calibrated individually at each downstream gauge. We also test the feasibility of using transfer functions (which transform reservoir and catchment attributes into operating rule parameters) to identify a nationally consistent calibration. This works well in ∼ 50 % of the catchments, while nuances in individual reservoir operations limit performance in others. We suggest that our approach should provide a lower benchmark for simulations in catchments containing water supply reservoirs and that more complex methods should only be considered where they outperform our simple approach.

Accurate discharge values form the foundation of effective water resource planning and management. Unfortunately, these data are often perceived as absolute and deterministic by users, modelers, and decision-makers, despite the inherent subjectivity and uncertainty in the data preparation processes. This study is undertaken to examine the discharge estimation methods used by the Water Survey of Canada (WSC) and their impacts on reported discharge values. First, we explain the hydrometric station network, essential terminologies, and fundamental concepts of rating curves. Subsequently, we examine WSC's standard operating procedures (SOPs), including shift, temporary shift, and override, in discharge estimation. Based on WSC's records of similar to 1800 active hydrometric stations for discharge monitoring, we evaluated sample rating curves and their correlation with stage and discharge measurement. We investigate under-ice measurements, ice condition periods and frequency, and extreme values in contrast to rating curves. Employing an independent workflow, we demonstrate that 69 % of existing records align with the rating curve and temporary shift concept, while the remaining 31 % follow alternative discharge estimation methods (override). Selected example stations illustrate discharge estimation methods over time. We also demonstrate the impact of override and temporary shifts on commonly assumed uncertainty models. Given the practices of override and temporary shifts within WSC, there is a need to explore innovative methods for discharge uncertainty estimation. We hope our research helps in the critical challenge of estimating and communicating uncertainty in published discharge values.

Monitoring programmes worldwide use biota to assess the "health" of water bodies. Indices based on biota are used to describe the change in status of sites over time, to identify progress against management targets and to diagnose the causes of biological degradation. A variety of numerical stressor-specific biotic indices have been developed based on the response of biota to differences in stressors among sites. Yet, it is not clear how variation in pressures within sites, over what time period, and in what combination has the greatest impact on different biotic groups. An understanding of how temporal variation in pressures influences biological assessment indices would assist in setting achievable targets and help focus catchment-scale mitigation strategies to ensure that they deliver the desired improvements in biological condition. Hydrochemical data provided by a network of high-frequency (15 or 30 min) automated monitoring stations over 3 years were matched to replicated biological data to understand the influence of spatio-temporal variation in pollution pressures on biological indices. Hydrochemical data were summarised in various ways to reflect central tendency, peaks, troughs and variation over 1-90 days before the collection of each biological sample. An objective model selection procedure was used to determine which hydrochemical determinand, and over what time period, best explained variation in the biological indices. Stressor-specific indices derived from macroinvertebrates which purportedly assess stress from low flows, excess fine sediment, nutrient enrichment, pesticides and organic pollution were significantly inter-correlated and reflected periods of low oxygen concentration, even though only one index (ASPT(WHPT), average score per taxon) was designed for this purpose. Changes in community composition resulting from one stressor frequently lead to confounding effects on stressor-specific indices. Variation in ASPT(WHPT) was best described by dissolved oxygen calculated as Q(5) over 10 days, suggesting that low oxygen events had most influence over this period. Longer-term effects were apparent, but were masked by recovery. Macroinvertebrate abundance was best described by Q(95) of stream velocity over 60 days, suggesting a slower recovery in numbers than in the community trait reflected by ASPT(WHPT). Although use of ASPT(WHPT) was supported, we recommend that additional independent evidence should be used to corroborate any conclusions regarding the causes of degradation drawn from the other stressor-specific indices. The use of such stressor-specific indices alone risks the mistargeting of management strategies if the putative stressor-index approach is taken to be more reliable than the results herein suggest.

As the adverse impacts of hydrological extremes increase in many regions of the world, a better understanding of the drivers of changes in risk and impacts is essential for effective flood and drought risk management and climate adaptation. However, there is currently a lack of comprehensive, empirical data about the processes, interactions, and feedbacks in complex human-water systems leading to flood and drought impacts. Here we present a benchmark dataset containing socio-hydrological data of paired events, i.e. two floods or two droughts that occurred in the same area. The 45 paired events occurred in 42 different study areas and cover a wide range of socio-economic and hydro-climatic conditions. The dataset is unique in covering both floods and droughts, in the number of cases assessed and in the quantity of socio-hydrological data. The benchmark dataset comprises (1) detailed review-style reports about the events and key processes between the two events of a pair; (2) the key data table containing variables that assess the indicators which characterize management shortcomings, hazard, exposure, vulnerability, and impacts of all events; and (3) a table of the indicators of change that indicate the differences between the first and second event of a pair. The advantages of the dataset are that it enables comparative analyses across all the paired events based on the indicators of change and allows for detailed context- and location-specific assessments based on the extensive data and reports of the individual study areas. The dataset can be used by the scientific community for exploratory data analyses, e.g. focused on causal links between risk management; changes in hazard, exposure and vulnerability; and flood or drought impacts. The data can also be used for the development, calibration, and validation of socio-hydrological models. The dataset is available to the public through the GFZ Data Services (Kreibich et al., 2023,https://doi.org/10.5880/GFZ.4.4.2023.001).

<p>Reliable and accurate river streamflow or discharge measurement and reporting are essential for engineering, economic, and social decision-making. Discharge values are often perceived as true and deterministic by users, modelers, and decision-makers. In this study, the processes of discharge estimation by the Water Survey of Canada, WSC, are presented. The process of inferring the discharge (water volume over time) based on stage (water level) through stage-discharge relationships or “rating curves” including related terminologies is described. Multiple practices of rating curve construction and discharge estimation across WSC hydrometric stations are explored. Major processes of "override" and "temporary shift" which significantly affect the discharge estimation are elaborated. The reproducibility of the published discharge data using data from the production process for approximately 1750 active hydrometric stations operated by WSC is examined. Other impacts of temporary shift and override have been evaluated on the properties such as discharge residuals or performance metrics. Recommendations are made for wider access to metadata and measurements that are essential to quantify the reproducibility and uncertainty of reported discharge values. Open science, particularly Earth system modeling, demands clear communication of reproducibility, and uncertainty of published discharge.</p>

In this paper we implement a degree day snowmelt and glacier melt model in the Dynamic fluxEs and ConnectIvity for Predictions of HydRology (DECIPHeR) model. The purpose is to develop a hydrological model that can be applied to large glaciated and snow-fed catchments yet is computationally efficient enough to include model uncertainty in streamflow predictions. The model is evaluated by simulating monthly discharge at six gauging stations in the Naryn River catchment (57 833 km2) in central Asia over the period 1951 to a variable end date between 1980 and 1995 depending on the availability of discharge observations. The spatial distribution of simulated snow cover is validated against MODIS weekly snow extent for the years 2001–2007. Discharge is calibrated by selecting parameter sets using Latin hypercube sampling and assessing the model performance using six evaluation metrics. The model shows good performance in simulating monthly discharge for the calibration period (NSE is 0.74<NSE<0.87) and validation period (0.7<NSE<0.9), where the range of NSE values represents the 5th–95th percentile prediction limits across the gauging stations. The exception is the Uch-Kurgan station, which exhibits a reduction in model performance during the validation period attributed to commissioning of the Toktogul reservoir in 1975 which impacted the observations. The model reproduces the spatial extent in seasonal snow cover well when evaluated against MODIS snow extent; 86 % of the snow extent is captured (mean 2001–2007) for the median ensemble member of the best 0.5 % calibration simulations. We establish the present-day contributions of glacier melt, snowmelt and rainfall to the total annual runoff and the timing of when these components dominate river flow. The model predicts well the observed increase in discharge during the spring (April–May) associated with the onset of snow melting and peak discharge during the summer (June, July and August) associated with glacier melting. Snow melting is the largest component of the annual runoff (89 %), followed by the rainfall (9 %) and the glacier melt component (2 %), where the values refer to the 50th percentile estimates at the catchment outlet gauging station Uch-Kurgan. In August, glacier melting can contribute up to 66 % of the total runoff at the highly glacierized Naryn headwater sub-catchment. The glaciated area predicted by the best 0.5 % calibration simulations overlaps the Landsat observations for the late 1990s and mid-2000s. Despite good predictions for discharge, the model produces a large range of estimates for the glaciated area (680–1196 km2) (5th–95th percentile limits) at the end of the simulation period. To constrain these estimates further, additional observations such as glacier mass balance, snow depth or snow extent should be used directly to constrain model simulations.

There is a general trend toward the increasing inclusion of uncertainty estimation in the environmental modelling domain. We present the Consortium on Risk in the Environment: Diagnostics, Integration, Benchmarking, Learning and Elicitation (CREDIBLE) Uncertainty Estimation (CURE) toolbox, an open-source MATLAB (TM) toolbox for uncertainty estimation aimed at scientists and practitioners who are not necessarily experts in uncertainty estimation. The toolbox focusses on environmental simulation models and, hence, employs a range of different Monte Carlo methods for forward and conditioned uncertainty estimation. The methods included span both formal statistical and informal approaches, which are demonstrated using a range of modelling applications set up as workflow scripts. The workflow scripts provide examples of how to utilize toolbox functions for a variety of modelling applications and, hence, aid the user in defining their own workflow; additional help is provided by extensively commented code. The toolbox implementation aims to increase the uptake of uncertainty estimation methods within a framework designed to be open and explicit in a way that tries to represent best practice with respect to applying the methods included. Best practice with respect to the evaluation of modelling assumptions and choices, specifically including epistemic uncertainties, is also included by the incorporation of a condition tree that allows users to record assumptions and choices made as an audit trail log.

Abstract. Groundwater-dominated catchments are often critical for nationally important water resources. Many conceptual rainfall–runoff models tend to degrade in their model performance in groundwater-dominated catchments, as they are rarely designed to simulate spatial groundwater behaviours or interactions with surface waters. Intercatchment groundwater flow is one such neglected variable. Efforts have been made to incorporate this process into existing models, but there is a need for improving our perceptual models of groundwater–surface water interactions prior to any model modifications. In this study, national meteorological, hydrological, hydrogeological, geological and artificial influence (characterising abstractions and return flows) datasets are used to infer a perceptual model of intercatchment groundwater flow (IGF) and how it varies across the river Thames, United Kingdom (UK). We characterise the water balance, presence of gaining/losing river reaches and intra-annual dynamics in 80 subcatchments of the river Thames, taking advantage of its wealth of data, densely gauged river network and geological variability. We show the prevalence of non-conservative river reaches across the study area, with heterogeneity both between, and within, geological units, giving rise to a complex distribution of recharge and discharge points along the river network. We infer where non-conservative reaches can be attributed to IGF and where other processes (e.g. surface water abstractions) are the likely cause. Through analysis of recorded reach length water balance data and hydrogeological perceptualisation, we conclude that outcrops of carbonate fractured aquifers (Chalk and Jurassic limestone) show evidence of IGF both from headwater to downstream reaches, and out-of-catchment via spring lines. We found variability across the study area, with more seasonality and variability in river catchments on Jurassic limestone outcrops compared to Chalk and Lower Greensand outcrops. Our results demonstrate the need for local investigation and hydrogeological perceptualisation within regional analysis, which we show to be achievable given relatively simple geological interpretation and data requirements. We support the inclusion of IGF fluxes within existing models to enable calibration improvements in groundwater-dominated catchments, but with geologically specific characteristics, and (when perceptually appropriate) connectivity between catchments.

Despite the recent advances, the identification of influential hydrologic processes and parameters of the process-based hydrologic model is still challenging. Part of the reason is the uncertain and interacting hydrologic process and the high dimensional parameter space. The motivation for this work is to effectively select an appropriate set of hydrologic processes and parameters for each basin on the globe, which is not necessarily the same everywhere. Here we evaluate the applications of the Structure for Unifying Multiple Modeling Alternatives (SUMMA) model to a large number of representative areas on the globe. Our objective is to identify the dominant hydrologic processes and sensitive model parameters for each representative area. First, sensitivity indices of the SUMMA parameters are computed using the VISCOUS global sensitivity analysis method. VISCOUS is the abbreviation of Variance-based Sensitivity Analysis using Copulas. Second, the sensitivity values are summarized per hydrologic process (e.g., snowmelt, surface runoff, infiltration, soil moisture, evapotranspiration, baseflow, and runoff) and per simulation statistic (e.g., mean, coefficient of variance, and autoregressive lag 1). The summarized sensitivity indices enable modelers to identify the most dominant hydrologic processes in each representative area. The results of this study will provide a foundation to estimate parameters in large-domain applications of process-based hydrologic models.

Sensitivity analysis is used to increase our understanding of the evaluated model and ease model parameter estimation. VISCOUS (VarIance-based Sensitivity analysis using COpUlaS) is a given-data, computationally frugal variance-based global sensitivity analysis framework. Grounded in Copula theory, VISCOUS computes the Sobol sensitivity indices using a probability model that describes the relationship between model inputs (e.g., the perturbations in the model parameters) and outputs (e.g., the model responses given a parameter perturbation). In this technical note, we make three contributions to make the VISCOUS framework easier to understand and apply. First, we provide additional derivations of VISCOUS to connect the VISCOUS framework to recent developments in the data science community. We provide didactic examples with simple test functions in order to help a wider group of modelers understand the underpinnings of the VISCOUS framework. Second, we evaluate the VISCOUS framework using three types of Sobol functions and provide a cautionary note on using VISCOUS to approximate Sobol' sensitivity indices for applications where model inputs are of similar importance. Third, we provide an open-source code of VISCOUS in Python, namely, pyVISCOUS. pyVISCOUS is model-independent and can be applied with user-provided input-output data.

Many hydrologic modelling groups face similar challenges, with untapped opportunities to share code and concepts across different model development groups. An active community of practice is emerging, where the focus is not so much on developing a community hydrologic model, but more on advancing the science and practice of community hydrologic modeling. This presentation will summarize our recent efforts to develop open-source models, methods, and datasets to enable process-based hydrologic prediction across North America (and beyond). The contributions include (1) developing ensemble meteorological datasets for North America and the globe; (2) developing modular approaches to hydrologic modeling through a hierarchal approach that separates different model sub-domains (vegetation, snow, soil, groundwater) and separates the physical representations from the numerical solution; (3) implementing third-party numerical solvers (sundials) to improve the robustness and efficiency of the numerical solutions; (4) developing agile parallelization methods capable of handling heterogeneous computing loads and bottlenecks in the downstream reaches of large river networks; (5) implementing flexible model configuration toolbox to accelerate the implementation of large-domain hydrologic models; (6) advancing methods for river lake routing, including development of integrated river-lake hydrography datasets and development of large-domain reservoir management models; (7) advancing methods for large-domain parameter estimation; (8) advancing methods for ensemble data assimilation; and (9) advancing methods for probabilistic hydrologic prediction on time scales from seconds to seasons. We will discuss some of the major challenges encountered and the high-priority research that is necessary to advance capabilities in large-domain hydrologic prediction.

Risk management has reduced vulnerability to floods and droughts globally 1 , 2 , yet their impacts are still increasing 3 . An improved understanding of the causes of changing impacts is therefore needed, but has been hampered by a lack of empirical data 4 , 5 . On the basis of a global dataset of 45 pairs of events that occurred within the same area, we show that risk management generally reduces the impacts of floods and droughts but faces difficulties in reducing the impacts of unprecedented events of a magnitude not previously experienced. If the second event was much more hazardous than the first, its impact was almost always higher. This is because management was not designed to deal with such extreme events: for example, they exceeded the design levels of levees and reservoirs. In two success stories, the impact of the second, more hazardous, event was lower, as a result of improved risk management governance and high investment in integrated management. The observed difficulty of managing unprecedented events is alarming, given that more extreme hydrological events are projected owing to climate change 3 .

Climate change may significantly increase flood risk globally, but there are large uncertainties in both future climatic changes and how these propagate into changing river flows. Here, the impact of climate change on the magnitude and frequency of high flows is analysed for Great Britain (GB) to provide the first spatially consistent GB projections to include both climate ensembles and hydrological model parameter uncertainties. We use the latest high-resolution (12 km) regional climate model ensemble from the UK Climate Projections (UKCP18). These projections are based on a perturbed-physics ensemble of 12 regional climate model simulations and allow exploration of climate model uncertainty beyond the variability caused by the use of different models. We model 346 larger (>144 km2) catchments across GB using the DECIPHeR hydrological modelling framework. Generally, results indicated an increase in the magnitude and frequency of high flows (Q10, Q1, and annual maximum) along the western coast of GB in the future (2050–2075), with increases in annual maximum flows of up to 65 % for western Scotland. In contrast, median flows (Q50) were projected to decrease across GB. Even when using an ensemble based on a single regional climate model (RCM) structure, all flow projections contained large uncertainties. While the RCM parameters were the largest source of uncertainty overall, hydrological modelling uncertainties were considerable in eastern and south-eastern England. Regional variations in flow projections were found to relate to (i) differences in climatic change and (ii) catchment conditions during the baseline period as characterised by the runoff coefficient (mean discharge divided by mean precipitation). Importantly, increased heavy-precipitation events (defined by an increase in 99th percentile precipitation) did not always result in increased flood flows for catchments with low runoff coefficients, highlighting the varying factors leading to changes in high flows. These results provide a national overview of climate change impacts on high flows across GB, which will inform climate change adaptation, and highlight the impact of hydrological model parameter uncertainties when modelling climate change impact on high flows.

Land use changes, landscape modifications and changing climate have resulted in local to regional flood risk increases in recent decades. As an alternative to traditional engineering approaches, there has been a movement towards catchment-based flood risk management, a subset of which is natural flood management (NFM). NFM aims to enhance flood resilience through the slowing and storing of runoff and flow, based on the restoration and augmentation of hydrological and morphological catchment features. However, despite research highlighting their potential benefits, there is a limited quantity of robust and science-based empirical evidence on how these structures function in the landscape and their efficacy in reducing flood hazards. To address this knowledge gap and contribute to the growing NFM evidence base, this study evaluates the efficacy of offline water storage ponds for flow attenuation. Two contrasting pond sites in the Tone and Parrett catchment headwaters (SW-England) were monitored from April 2018 to December 2020 for channel flow, pond volume, and rainfall. Field-scale, high resolution (1.61 cm spatial support) digital elevation models from Structure-from-Motion and manual surveys were collected to quantify pond dynamics and connectivity. A comprehensive, storm definition methodology was developed for event separation, to enable quantification of the impact of pond behaviour on stream flow, across storm hydrographs. Results show that where offline ponds function as designed, with direct pond filling from the channel, peak flow attenuation can reach 7% (maximum event AEP of 83%). Smaller events, where pond filling occurs directly from rainfall or runoff display only a maximum of 3% peak flow reduction (event AEP from 83% ->99%). The ability of ponds to attenuate flow is heavily reliant on sufficient structural conditions for ponds to fill directly from the channel and to drain slowly following event peaks. This study provides an empirical database for future NFM applications, including key criteria for future design and for use as observational data in modelling applications, by upscaling the efficacy of ponds to reduce flood impacts at larger scales.

Spatial parameter fields are required to model hydrological processes across diverse landscapes. Transfer functions are often used to relate parameters to spatial catchment attributes, introducing large uncertainties. Quantifying these uncertainties remains a key challenge for large‐scale modeling. This paper extends the multiscale parameter regionalization (MPR) technique to consider parameter uncertainties. We evaluate this method of producing nationally consistent parameter fields, which maintain a constant relationship between model parameters and catchment attributes, across 437 catchments in Great Britain (GB). By sampling multiple transfer function parameters, we produce thousands of possible model parameter fields which are constrained within an uncertainty framework. This is compared to spatially homogeneous parameter sets constrained for individual catchments. The nationally consistent MPR parameter fields perform well (KGE* > 0.75) across 60% of catchments. Performance is similar or better than catchment‐constrained parameters (KGE* drop < 0.1) across 82% of catchments. Advantages of our national parameter fields include (a) improved representation of flows within catchments, (b) more robust performance between calibration and evaluation periods, and (c) spatial parameter fields reflecting hydrologically meaningful variation in catchment characteristics. By including uncertainties, we show that hydrographs produced using MPR have smaller uncertainty bounds which are better able to encompass flows. As the first application of MPR to both the DECIPHeR modeling framework and GB, we developed transfer functions and identified key catchment attributes to constrain model parameters, which are transferrable to other models alongside the addition of uncertainty. Methodologies presented here are informative for future regionalization efforts in GB and elsewhere.

Landscape evolution models (LEMs) have the capability to characterize key aspects of geomorphological and hydrological processes. However, their usefulness is hindered by model equifinality and paucity of available calibration data. Estimating uncertainty in the parameter space and resultant model predictions is rarely achieved as this is computationally intensive and the uncertainties inherent in the observed data are large. Therefore, a limits‐of‐acceptability (LoA) uncertainty analysis approach was adopted in this study to assess the value of uncertain hydrological and geomorphic data. These were used to constrain simulations of catchment responses and to explore the parameter uncertainty in model predictions. We applied this approach to the River Derwent and Cocker catchments in the UK using a LEM CAESAR‐Lisflood. Results show that the model was generally able to produce behavioural simulations within the uncertainty limits of the streamflow. Reliability metrics ranged from 24.4% to 41.2% and captured the high‐magnitude low‐frequency sediment events. Since different sets of behavioural simulations were found across different parts of the catchment, evaluating LEM performance, in quantifying and assessing both at‐a‐point behaviour and spatial catchment response, remains a challenge. Our results show that evaluating LEMs within uncertainty analyses framework while taking into account the varying quality of different observations constrains behavioural simulations and parameter distributions and is a step towards a full‐ensemble uncertainty evaluation of such models. We believe that this approach will have benefits for reflecting uncertainties in flooding events where channel morphological changes are occurring and various diverse (and yet often sparse) data have been collected over such events.

Rainfall-runoff models based on conceptual "buckets" are frequently used in climate change impact studies to provide runoff projections. When these buckets approach empty, the simulated evapotranspiration approaches zero, which places an implicit limit on the soil moisture deficit that can accrue within the model. Such models may cease to properly track the moisture deficit accumulating in reality as dry conditions continue, leading to overestimation of subsequent runoff and possible long-term bias under drying climate. Here, we suggest that model realism may be improved through alternatives which remove the upper limit on simulated soil moisture deficit, such as "bottomless" buckets or deficit-based soil moisture accounting. While some existing models incorporate such measures, no study until now has systematically assessed their impact on model realism under drying climate. Here, we alter a common bucket model by changing the soil moisture storage to a deficit accounting system in such a way as to remove the upper limit on simulated soil moisture deficit. Tested on 38 Australian catchments, the altered model is better able to track the decline in soil moisture at the end of seasonal dry periods, which leads to superior performance over varied historic climate, including the 13-year "Millennium" drought. However, groundwater and GRACE data reveal long-term trends that are not matched in simulations, indicating that further changes may be required. Nonetheless, the results suggest that a broader adoption of bottomless buckets and/or deficit accounting within conceptual rainfall runoff models may improve the realism of runoff projections under drying climate.