| 21 Sep 2026
Circulation-driven streamflow variability in Great Britain and its representation by UKCP18 climate models
Abstract. Interannual to decadal seasonal streamflow variability has major implications for flood risk, drought management, water resources and hydropower production across Great Britain. A substantial component of this variability is linked to large-scale atmospheric circulation, particularly over the North Atlantic, but its contribution to seasonal streamflow variability remains insufficiently quantified. Here, we use a dynamical adjustment framework based on multiple linear regression to quantify the contribution of synoptic-scale weather regimes to seasonal Standardised Streamflow Index (SSI) variability across Great Britain. The method is applied to reconstructed SSI series for 167 non-overlapping catchments, which are aggregated to the UKCP18 river regions.
The strongest dynamical contribution occurs in winter, especially across western and northern Great Britain, where weather-regime-driven SSI explains around 70% of interannual winter SSI variance in several regions. The dynamically driven contribution to SSI also explains a substantial fraction of spring and autumn variability, but the relationship is weaker in summer, consistent with the reduced influence of synoptic-scale circulation on warm-season rainfall. In one third of GB regions, including the Clyde, Solway, and North-West England, dynamically driven SSI captures much of the observed interannual and decadal variability, including several historically wet and dry winters. This strong circulation–streamflow relationship provides useful information for seasonal forecasting, climate-model evaluations and historical reconstruction.
Applying the same framework to weather-regime sequences from the UKCP18 global climate models ensemble shows that the ensemble systematically under-represents interannual variance in winter dynamically driven SSI, with an ensemble-mean bias of approximately -20% relative to observations and substantial spatial and inter-member variation. These results show that atmospheric circulation biases can propagate into climate-model-driven assessments of seasonal streamflow variability, with implications for hydrological projections, model evaluation, and process-informed bias correction.
This is an excellent paper that makes an important contribution to understanding of the role of large-scale atmospheric circulation in driving seasonal streamflow variability across Great Britain. The extension of the dynamical adjustment framework previously applied to precipitation to seasonal streamflow is innovative and well justified. The study provides valuable new insights into the regional and seasonal dependence of circulation-driven hydrological variability and offers a useful framework for evaluating the representation of these processes in climate models.
The authors demonstrate that atmospheric circulation explains a substantial proportion of winter streamflow variability, particularly across western and northern Great Britain, where explained variance reaches approximately 80% in some regions. The comparison with precipitation is especially valuable, helping to explain the spatial patterns observed and highlighting the modifying influence of catchment characteristics and hydrological processes. The finding that UKCP18 systematically underestimates the variance of circulation-driven winter streamflow by approximately 20% is an important result with implications for climate-model evaluation and hydrological impact assessments.
The paper is clearly written, logically structured and supported by convincing analyses. The discussion is particularly strong and appropriately acknowledges the limitations of the approach, including the influence of catchment characteristics, hydrological memory, non-stationarity and the limitations of applying historical circulation-streamflow relationships to future climates. I also welcome the measured discussion of potential applications and the distinction between using the framework as a process-based diagnostic and as a substitute for conventional hydrological projections.
Overall, I consider this a strong and worthwhile contribution that is suitable for publication. I have a small number of comments, principally concerning uncertainty inherited from the hydrological reconstructions and some aspects of the regression methodology, that I believe could further strengthen the paper.
Uncertainty in reconstructed SSI and interpretation of regression residuals
The analysis relies on historical SSI series derived from hydrologically reconstructed river flows, with the best-performing hydrological model selected for each catchment. While these reconstructions provide an important foundation for the analysis, I wonder whether the uncertainties associated with them, and their implications for the estimated circulation-streamflow relationships, could be considered more explicitly.
In particular, the regression residual is described as representing the component of seasonal SSI not explained by the weather regimes, with acknowledgement that it may contain other dynamical influences. However, the residual will also contain errors inherited from the hydrological reconstructions, including uncertainty associated with hydrological model structure, the combination of objective functions used, parameterisation and meteorological forcing, alongside errors associated with the statistical regression itself.
I wonder whether these uncertainties might also vary across catchments, particularly between relatively responsive upland catchments and those with more complex groundwater or storage influences. Such differences could contribute to the geographical patterns in explained variance, alongside the physical processes discussed by the authors.
The uncertainty inherited from reconstructed SSI may also have implications for the subsequent UKCP18 evaluation. Although applying a fixed regression relationship to observed and modelled weather regimes provides an effective means of isolating differences in atmospheric circulation, the estimated transfer function itself remains subject to uncertainty arising from the historical flow reconstructions.
Could the authors comment further on these issues and their potential implications? I do not consider a comprehensive uncertainty propagation analysis necessary for this paper. However, a clearer recognition of the uncertainty inherited from the hydrological reconstructions would strengthen the interpretation of the findings.
Validation and temporal stability of regression relationships
Could the authors clarify whether the reported explained variance represents performance over the period used to calibrate the regression relationships, or whether any independent validation was undertaken? Given the potential applications to seasonal forecasting and extending historical streamflow variability, it would be useful to understand the extent to which these relationships are temporally stable and retain explanatory or predictive skill outside the calibration period. Sorry if i missed this.Â
Weather-regime frequency, persistence and sequencing
The discussion highlights the importance of weather-regime frequency, persistence and sequencing in determining seasonal hydrological variability. However, the regression methodology uses the seasonal number of days assigned to each weather regime and does not appear to explicitly account for the duration or temporal ordering of individual weather-regime spells. Consequently, seasons with identical weather-regime frequencies but different sequences or persistence characteristics would produce the same dynamically driven SSI estimate. It may be useful to clarify this distinction, particularly where weather-regime persistence and sequencing are discussed as contributing to hydrological variability. This could also be acknowledged as a potential avenue for further methodological development, especially given the importance of antecedent conditions and hydrological memory.
Minor clarification of analysis periods
Section 2.2.1 refers to the application of UKCP18 weather-regime sequences over 1900–2020, whereas the caption of Figure 4 indicates a comparison period of 1901–2024. Please check and clarify the periods used in the analysis and ensure consistency throughout the manuscript.
In summary, this is a high-quality, well-executed study that makes a valuable contribution to understanding circulation-driven hydrological variability and its representation in climate models. The methodological approach is clear, the principal findings are convincing, and the discussion provides an appropriately balanced interpretation of the results and their implications.
My comments are intended primarily to improve methodological transparency and strengthen the treatment of uncertainty, rather than to challenge the central findings. In particular, I consider some further acknowledgement of the uncertainty inherited from hydrological reconstructions worthwhile.