| 19 Jun 2026
Divergent responses of streamflow reanalysis errors to precipitation reanalysis errors modulated by catchment heterogeneity
Abstract. Streamflow reanalysis is vital for water resources management and climate impact assessment; however, the extent to which it is affected by precipitation forcing errors remains poorly understood. Focusing on the reanalysis dataset of Global Flood Awareness System driven by the European Centre for Medium-Range Weather Forecasts Reanalysis v5 (GloFAS-ERA5), this paper details how streamflow reanalysis errors respond to precipitation errors. Specifically, the root mean square errors (RMSEs) are calculated by hydrological year for reanalysis products across 671 catchments in the Catchment Attributes and Meteorology for Large-sample Studies (CAMELS) dataset; and by combining catchment-specific linear regression with global panel regression, the effects of precipitation errors on streamflow errors are quantified. The results demonstrate an improved performance from GloFAS-ERA5 v2.1 to v4.0, with the median RMSE decreasing from 2.16 mm to 1.81 mm. For GloFAS-ERA5 v4.0, the panel regression indicates that for every 1 mm increase in precipitation RMSE, the corresponding streamflow RMSE increases by an average of 0.51 mm–reflecting the buffering capacity of catchment storage. In the meantime, the corresponding catchment-specific increase of streamflow RMSE reaches up to 2.5 mm in humid catchments but remains below 0.7 mm in arid catchments. These divergent responses reflect that the saturation-excess mechanism makes the precipitation errors immediately affect the streamflow error while soil moisture deficits dampen their effects. Furthermore, incorporating interaction terms into panel regression increases the coefficient of determination (R2) from 0.16 to 0.36, indicating that error responses are modulated by catchment heterogeneity. This modulation is further confirmed by targeted case studies, indicating that the temperature controls the storage and release of snow water, thereby dampening and delaying the responses of streamflow errors to precipitation errors in snow-dominated catchments. These findings provide a valuable diagnostic method and practical guidance for applications of global streamflow reanalysis to complex, heterogeneous catchments.
This paper precisely addresses a long-standing issue that has been ambiguously treated in the field of global hydrological reanalysis, namely, whether precipitation input errors are amplified or attenuated during the runoff concentration process. It does not merely provide a global average number; rather, for the first time, it systematically quantifies the distinctly different patterns of this error propagation in humid, arid, and snow-covered regions. Using two specific case studies, which include a rain-fed basin and a snow-fed basin, the paper interprets statistical regression coefficients into two physical processes, namely saturation-excess runoff and snowmelt delay, making the conclusions highly convincing.
Nevertheless, the paper still has the following areas that require improvement:
(1) Lines 18-19: The physical interpretation of the buffering coefficient of 0.51 is somewhat thin. Although it is mentioned as "the buffering capacity of catchment storage," the storage capacities of the 671 basins (e.g., baseflow index, groundwater recharge rate) vary considerably. The paper does not further cross-validate the average coefficient of 0.51 with basin-specific storage capacity curves or soil moisture memory. Does this 0.51 predominantly reflect soil infiltration, or is it dominated by evapotranspiration (ET) consumption? The current explanation is somewhat vague.
(2) The paper does not address the issue of water balance closure. The explanation that the error in arid regions is <0.7 is attributed to soil moisture deficit buffering, but this implies a premise—that the runoff error caused by precipitation error is absorbed. Where does the absorbed error go? Does this imply that the evapotranspiration (ET) error in arid regions is amplified?
(3) Although the paper states that VIF < 5, indicating weak collinearity between precipitation and temperature, in snow seasons, higher precipitation is often accompanied by lower temperatures, and the actual impacts of the two are highly temporally coupled. While the panel regression passes statistical tests, the physical simultaneity (i.e., winter low temperatures cause snowfall, and the quality of both low-temperature and precipitation data often deteriorates simultaneously) has not been sufficiently disentangled and discussed.
(4) Logical confusion: Section 4.2 indicates that basin-specific regressions perform better than panel regression. Why then continue to stubbornly focus on improving the panel regression? Although the authors explain that panel regression provides a stable average, they do not explicitly answer: given the substantial individual differences (ranging from 0 to 2.5), does the forced use of a global panel regression (even with interaction terms) statistically obscure the extreme physical mechanisms of the extremes? The logic would be clearer if a sentence were added, such as: "Panel regression is primarily used to reveal cross-basin universal laws, rather than to precisely predict specific values for a given basin."
(5) The limitation of the temporal scale is not discussed: all error calculations are based on annual RMSE. The annual-scale RMSE smooths out intra-seasonal phase errors.