Does Antecedent Catchment Wetness Explain the Timing of Rainfall‐Runoff Relationship Shifts?

Abstract

Annual rainfall‐runoff relationships have been shown to shift during droughts and, in many catchments, not recover, resulting in less runoff per unit of precipitation than prior to the drought. While the cause of these shifts appears to be related to plant water use and subsurface water dynamics, it is still unclear which climatic conditions explain the timing of runoff shifts. The main objective of this study is to determine whether antecedent catchment wetness alone can explain the timing of shifts into and out of low runoff states. To do this, we explain the annual probability of being in a normal runoff state, as derived from Peterson et al. (2021), https://doi.org/10.1126/science.abd5085 Hidden Markov models (HMMs), across 158 catchments in Victoria, Australia. Specifically, we predict the annual HMM state probability using a convolution of historical seasonal or annual rainfall deficit, which is then input to a logistic function to estimate state probability. To examine non‐recovery, we develop two logistic functions with different thresholds for shifting into and out of the normal runoff state. Model parameters are estimated using beta‐distribution maximum likelihood and global optimisation. Overall, the results show that antecedent catchment wetness can explain the timing of rainfall‐runoff relationship shifts, that autumn wetness is important for transitions between low and normal runoff states, and that the results provide further evidence that catchments have finite resilience to disturbances. These findings suggest that further development of a non‐homogeneous HMM could help identify how antecedent catchment wetness contributes to catchment shifts and recovery.