New estimation models for determining the Q347

Abstract

In Switzerland, low flows are described by the five percent quantile denoted by Q347. This threshold value not only has consequences for the planning, but also necessitates authorities to adjust the operation of pertinent infrastructure to mitigate ecological impacts on watercourses. Given a discharge timeseries spanning at least a ten-year period, determination of the Q347 can be done using the duration curve. Typically, said timeseries are not available for smaller catchments necessitating the estimation of the threshold value Q347. In Switzerland, the utilization of multiple linear regression has been established to estimate the area-specific discharge q347.

The primary objective of these investigations is to estimate the Q347 value for 383 ungauged catchments in the Canton of Solothurn, each covering an area less than 100 km². Daily discharge, precipitation and temperature timeseries ranging from 1990 to 2020 were collected from 56 gauged catchments smaller than 500 km² surrounding the target area. 30 “static” parameters delineating geometry, topography, geology, land use, and drainage along with nine “climatic” parameters describing temperatures, precipitation distributions, and potential evapotranspiration were defined and computed to characterize gauged and ungauged catchments. Alongside comparing three regression methods, coupled with two adjustment techniques supplementing truncated discharge timeseries, three parameter selection methods are evaluated. The validation of the proposed models shows reduced errors and increased linear correlations between estimated and observed values compared to currently applied models. Notably, a spatially more homogeneous yet catchment-specific distribution of estimated values is observable. Particularly when timeseries remain unadjusted or adjustment is done using the Antecedent Precipitation Index (API) and the flow duration curve from a donor basin (Ridolfi, E.; Kumar, H.; Bárdossy, A., 2020), the proposed models yield promising results.

Furthermore, the temporal variability of low flow events for the glacier-free catchments in the study area has been analysed. The frequency of low flow events below the threshold systematically increased over the last 30 years, while the 10-year Q347 value of said catchments has systematically decreased in the same period. The increase in low flow days leads to large errors in the estimation of the Q347 value, especially when its estimation is based on truncated timeseries. As further changes in runoff behaviour are to be expected due to climate change, extending the definition of "low flow" to include event duration and intensity alongside a fixed threshold value could offer a more suitable description.