Applying recession models for low-flow prediction: a comparison of regression and matching strip approaches
Abstract. Low flows in the Swiss Plateau are expected to occur more often, to last longer and, hence, to be more severe under climate change. To predict and manage such periods of water scarcity effectively, more precise information on the drainage behavior of catchments is required. The drainage behavior of a catchment can be characterized by recession analysis methods (RAMs; e.g., recession curves) of which several have been developed in the last decades. Their recession parameters have been related to different aquifer characteristics or more general catchment characteristics like lithology, topography, or climatology. Such parameters vary widely, and the effects of uncertainties on the model’s outcomes are diverse and complex. Despite the obvious potential of recession curves for prediction, they have so far not been used for operational low flow prediction and guidance for hazard mitigation. In addition, recession curves of slowly draining catchment states are hardly represented by current RAMs.
To fill the gap of RAMs representing slow draining catchment states we developed two novel RAMs, one fully automated and based on the matching strip method (MRC_slow), the other one (SDSC) relying on a careful expert-based selection of few recession segments with the slowest recession behavior. Alongside we used three established RAMs from the literature (one further matching strip model, linear regression and lower envelope in the discharge decay – discharge recession diagram). We applied the five RAMs on previously extracted low flow segments of 33 catchments in the Swiss Plateau and compared them on their recession curvatures, durations, and volumes. We designed a procedure that evaluates which of any selected RAMs best matches the recession behavior of individual low flow segments of a hydrograph. Applying this in a simulated prediction situation, we evaluated in retrospect, which of the five specifically selected RAMs predicted the low flow hydrographs between 2021 and 2022 most accurately.
We found the variability of recession durations and volumes between catchments to be higher than between the five RAMs. Within 30 of the 33 catchments, the order of recession durations and recession volumes was the same. Hence the different recession behaviors of the RAMs could be related to different catchment states. Upon evaluating the low flow predictions, we found that the MRC_slow approach overall performed best followed by linear regression and SDSC. However, for operational low flow prediction we recommend using four of the five RAMs. This allows for changing the recession model(s) at every timestep if the recession behavior changes. It is also possible to present predictions with a model ensemble, indicating a range of uncertainties if several models perform similarly well. The described data-driven approach and the newly developed models are, therefore, very promising for improving low flow predictions in gauged catchments.
This study aims to address the estimation of low-flow hydrograph recession using five approaches, including two methods proposed by the authors. Analyses were conducted with daily streamflow data from 33 catchments in the Swiss Plateau. The authors present the findings from three perspectives of identified low-flow recession: 1) the duration of the recession, 2) the specific discharge volume embedded in the recession, and 3) the overall prediction performance.
The issues raised by the authors are significant; however, the manuscript is hindered by several factors: 1) illogical or incorrect interpretations with insufficient or no supporting evidence, 2) subjective descriptions based on visualization without quantification, and 3) redundant and colloquial writing, a relatively large number of grammatical issues, and unclear figure legends. These factors diminish the clarity and rigor of their findings, which are essential for a high-quality paper. Major comments regarding these key concerns are outlined below, followed by additional comments related to the analyses:
Other comments:
Bart, R., & Hope, A. (2014). Inter-seasonal variability in baseflow recession rates: The role of aquifer antecedent storage in central California watersheds. Journal of Hydrology, 519(PA), 205–213. https://doi.org/10.1016/j.jhydrol.2014.07.020
Biswal, B., & Marani, M. (2010). Geomorphological origin of recession curves. Geophysical Research Letters, 37(24), 1–5. https://doi.org/10.1029/2010GL045415
Botter, G., Porporato, A., Rodriguez-Iturbe, I., & Rinaldo, A. (2009). Nonlinear storage-discharge relations and catchment streamflow regimes. Water Resources Research, 45(10), 1–16. https://doi.org/10.1029/2008WR007658
Dralle, D. N., Karst, N. J., Charalampous, K., Veenstra, A., & Thompson, S. E. (2017). Event-scale power law recession analysis: Quantifying methodological uncertainty. Hydrology and Earth System Sciences, 21(1), 65–81. https://doi.org/10.5194/hess-21-65-2017
Jachens, E. R., Rupp, D. E., Roques, C., & Selker, J. S. (2020). Recession analysis revisited: Impacts of climate on parameter estimation. Hydrology and Earth System Sciences, 24(3), 1159–1170. https://doi.org/10.5194/hess-24-1159-2020
Roques, C., Rupp, D. E., & Selker, J. S. (2017). Improved streamflow recession parameter estimation with attention to calculation of − dQ/dt. Advances in Water Resources, 108, 29–43. https://doi.org/10.1016/j.advwatres.2017.07.013
Shaw, S. B., & Riha, S. J. (2012). Examining individual recession events instead of a data cloud: Using a modified interpretation of dQ/dt-Q streamflow recession in glaciated watersheds to better inform models of low flow. Journal of Hydrology, 434–435, 46–54. https://doi.org/10.1016/j.jhydrol.2012.02.034
Tashie, A., Pavelsky, T., & Band, L. E. (2020). An Empirical Reevaluation of Streamflow Recession Analysis at the Continental Scale. Water Resources Research, 56(1), 1–18. https://doi.org/10.1029/2019WR025448