Regional regression models for estimating monthly streamflows

Abstract

Environmental flow science has attracted significant attention in recent years and has become more widely implemented by various agencies responsible for managing water resources. In addition to annual environmental flows, monthly environmental flows have been increasingly used to protect aquatic ecosystems. Regional regression analysis is commonly used to estimate streamflows when long-term continuous streamflow records are not available. While literature related to regional regression models for estimating annual flows is relatively rich, studies focused on regional regression analysis for estimating monthly flows are rare. This study contributes a comprehensive assessment of regional regression models for estimating monthly flows. A comprehensive database of watershed characteristics was developed, and a suite of monthly and annual flows were estimated using long-term continuous streamflow records from 72 watersheds within or adjacent to the Susquehanna River Basin. Regional regression models were developed for 104 flows, including 96 monthly flow statistics and 8 annual flow statistics. The performance of regional regression models for estimating various monthly flows, as well as which watershed characteristics were most important for estimating them, was investigated. The results showed regional regression analysis performed better for wet months than dry months, and better for high and medium flows than low flows. Drainage area and precipitation were the most important watershed characteristics for estimating flows. The performance of regional regression models for estimating annual flow was often better for estimating monthly flows of dry months but worse than for estimating monthly flows of wet months. The study also provides guidance for water resources managers regarding where to focus streamflow monitoring efforts with limited resources, considering flow estimation error is greatest for low flow statistics in dry weather months.