Nonstationary frequency pairing reveals a highly sensitive peak flow regime to harvesting across a wide range of return periods

Abstract

Decades of pairing peak flows of forested and harvested watersheds by equal chronology when evaluating the effects of forest harvesting on peak flows in paired watershed studies lately came under criticisms; and suggestions have been made to abandon the practice and pair by equal frequency instead. However, the stationarity assumption imposed by conventional frequency analyses complicates the use of frequency pairing because peak flows contain change-point and trend-shifting nonstationarities caused by continuous harvesting and forest regrowth. Here we introduce a new nonstationary frequency pairing method for evaluating harvesting effects by allowing the parameters of peak flow frequency distributions to change in time using physically based covariates. The method’s demonstration on a paired watershed study in the snow environment of British Columbia, Canada, reveals how both small (return periods < 10 years) and large (return periods > 10 years) peak flows are highly sensitive to harvesting within the critical mid-elevation south-exposed slopes of the watershed. Such outcomes challenge an age-old deterministic wisdom, but are consistent with the emerging probabilistic understanding of how forests affect peak flows. Advantages of the new method include: (i) bypassing the need for the calibration equation traditionally used in paired watershed studies, thereby eliminating some associated sources of uncertainty; (ii) making use of longer peak flow records by explicitly accounting for the physical causes of the nonstationarities, and thus more explicit inferences about the effects of harvesting on the larger peak flows; and (iii) allowing for the estimation of harvesting effects on peak flows at different points in time during the disturbance history of a watershed, thus providing a direct evaluation of the hydrologic recovery. This study falls within the newly emerging field of “attribution science” which uses a combination of observations and models to identify separately the factors that contribute to extremes. It paves the way to a wider range of applications, not just in the wider hydrology but also other geosciences.