Stormflow threshold behaviour in a subtropical mountainous headwater catchment during forest recovery period

Abstract

Forest ecohydrological feedbacks complicate the threshold behaviour of stormflow response to precipitation or wetting conditions on a long-term scale (e.g. several years). In this study, the threshold behaviours in an evergreen-deciduous mixed forested headwater catchment in southern China were examined during 2009-2015, when damaged vegetation was recovering after the great 2008 Chinese ice and snowstorm. The non-uniqueness of the thresholds and the slow and rapid responses of stormflow at the outlet of the catchment in different hydro-climate datasets with different maximum values of gross precipitation (P) and sums of precipitation and antecedent soil moisture index (P + ASI) were assessed. The thresholds of P and P + ASI required to trigger stormflows (i.e. 'generation thresholds') and the transition from slow to rapid responses of stormflow (i.e. 'rise thresholds') were compared both seasonally and annually. The results indicated significant differences in the analysed datasets, highlighting the need to compare thresholds with care to avoid misinterpretation. Seasonal variations in threshold behaviours in the catchment suggested that vegetation canopy interception contributed to higher rise thresholds, and wetter conditions resulted in higher runoff sensitivity to precipitation during the growing and rainy seasons. Furthermore, the generation thresholds were higher in the dormant season, possibly due to drier soil moisture conditions in the near-channel areas. During the vegetation recovery period, the annual generation thresholds increased, however the rise thresholds did not exhibit a similar trend. The rapid stormflow response above the threshold decreased, possibly due to transpiration and interception of the recovered vegetation. However, the slow stormflow response to small rainfall events below the thresholds was higher in wetter years but lower in drier years, suggesting that the total water input dominated the stormflow response during small rainfall events. In conclusion, the seasonal and annual variations in threshold behaviours highlight that vegetation recovery and hydro-climatic conditions had a notable impact on the stormflow response. K E Y W O R D S2008 Chinese ice and snowstorm, stormflow threshold, vegetation recovery, ecohydrological feedback, rainfall-runoff processes, post-disturbance stormflow, Nanling Mountain

Evaluating the Ubiquity of Thresholds in Rainfall‐Runoff Response Across Contrasting Environments

“…This sensitivity analysis showed that threshold detection for the sites considered in this study did not change considerably across the 10%-50% minimum slope difference range that was tested (Table S1). It is also notable that in the present study and other recent ones (e.g., Scaife & Band, 2017;Wei et al, 2020), the exclusion of possible slower stormflow generation was avoided by not assuming a below-threshold slope (i.e., m 1 ) of 0. This is fundamentally different from operational definitions of threshold behavior found in other studies that only assess a below-threshold regime that is characterized by negligible runoff response (i.e., m 1 = 0).…”

“…Most runoff initiation thresholds have been attributed to response from matrix flow, macropore flow, or pipe flow (McGuire & McDonnell, 2010; Tromp‐van Meerveld & McDonnell, 2006b; Uchida et al., 2005; Wei et al., 2020). However, others have identified relationships with multiple thresholds: in these multithreshold cases, the runoff initiation threshold is followed by a secondary threshold (referred to as a “rise threshold”) when additional rainfall triggers a transition to different runoff generation processes (Wei et al., 2020). The thresholds are therefore sequential and manifest as multiple, distinct breakpoints at two or more different moments in time (Wei et al., 2020).…”

“…This may be explained by the fact that these previous studies reported critical meteorological factor values required to initiate response (i.e., runoff‐initiation thresholds), while the thresholds observed in the current study were not necessarily runoff‐initiation thresholds. Most runoff initiation thresholds have been attributed to response from matrix flow, macropore flow, or pipe flow (McGuire & McDonnell, 2010; Tromp‐van Meerveld & McDonnell, 2006b; Uchida et al., 2005; Wei et al., 2020). However, others have identified relationships with multiple thresholds: in these multithreshold cases, the runoff initiation threshold is followed by a secondary threshold (referred to as a “rise threshold”) when additional rainfall triggers a transition to different runoff generation processes (Wei et al., 2020).…”

“…The thresholds are therefore sequential and manifest as multiple, distinct breakpoints at two or more different moments in time (Wei et al., 2020). In the case of the HJA7 site, the larger threshold value that was observed in the present study could be associated with a “rise threshold”: it may be caused by the addition of rainfall onto wetted soils that results in the activation of matrix or preferential flow in hillslope soils, thereby increasing the hillslope‐riparian‐stream connectivity (McGuire & McDonnell, 2010), and leading to a nonlinear response change (Sidle et al., 2000; Wei et al., 2020). If such a hypothesis were true, however, it would raise the question of why the PRA did not identify the runoff initiation threshold but rather identified a subsequent “rise threshold.” Multibreakpoint PRA approaches or alternate threshold identification methods are warranted to better evaluate the presence of sequential initiation and rise thresholds.…”

“…This methodological choice was driven by the fact that high-frequency snowfall and snowmelt data were not available, and HydRun is only designed to delineate rainfall-runoff events (Tang & Carey, 2017). Furthermore, while some past studies did identify seasonal differences in threshold behavior (Scaife & Band, 2017;Wei et al, 2020), such differences were not explored here. Although many rainfall-runoff events were considered in the present study, the number of events per site varied significantly due to shorter measurement periods for some sites and the fact that drier sites tend to have fewer rainfall-runoff events.…”