Event-based changes in hydraulic properties of surface soils in semi-arid regions may generate surface runoff regimes at decadal timescales

Climate seasonality, characterized by alternating drought and rainfall, drives the degradation and recovery of soil properties that shape hydrological functioning. These shifts can trigger ecohydrological responses such as increased runoff, reduced soil water availability, and vegetation decline, while subsequent recovery may enable partial or full restoration of soil hydraulic properties. The objective of this study was to determine how the interaction of these mechanisms, driven by stochastic climatic inputs, shapes the soil water balance, with particular emphasis on runoff generation.

We developed a conceptual model in which soil hydraulic properties are represented as a dynamic variable subject to degradation, recovery, or no change, formulated using discrete-state logic. In a specific implementation, state-dependent thresholds determine system behavior: degradation occurs as a discrete jump triggered by abrupt wetting of the soil, characterized by rapid wetting from a dry state to an upper moisture threshold; recovery follows a time-dependent trajectory when soil moisture is maintained within this threshold for a sufficient duration; and outside these conditions, no change occurs. This variable dynamically scales infiltration capacity, allowing rainfall under degraded conditions to generate surface runoff.

Resulting runoff time series were normalized by annual evaporation and classified using agglomerative hierarchical clustering with dynamic time warping. Three characteristic runoff patterns emerged: (i) rapid recovery with sub-decadal runoff decline; (ii) an intermediate transitional pattern with slower recovery and moderate runoff persistence; and (iii) permanent degradation associated with multi-decadal runoff regimes. To identify the drivers of these patterns, we analyzed a dimensionless set of parameters using linear discriminant analysis. The dominant control leading to degradation was the temporal clustering of rainfall, defined as the relative duration of rainfall events compared to the expected interarrival time between events.

Using dimensionless parameter combinations that express state-shifting forcings associated with each regime, we investigated clusters of runoff patterns. The results indicate that temporally isolated rainfall events can trigger irreversible state shifts leading to runoff-dominated regimes, whereas sustained or closely spaced rainfall events have the potential to initiate recovery processes spanning multiple decades.