Machine Learning for Hydrological Forecasting in Ungauged Upland Catchments: A Case Study from County Mayo, Ireland

Understanding the relationships between rainfall and river flows, especially as climate shifts and rainfall patterns are modified, and in the context of human intervention, requires studying rivers within their catchments. This is especially important for ungauged, remote catchments, which are often understudied due to the challenges in data collection. In line with the EU Water Framework Directive, studying such catchments can offer valuable insights into broader hydrological processes. This study focuses on two instrumented, small, upland rivers near Newport, County Mayo, in the west of Ireland. The lack of data from most small, upland catchments highlights the importance of using innovative approaches like Machine Learning (ML) for hydrological forecasting. ML, a branch of artificial intelligence, enables the development of predictive models that identify patterns in meteorological and hydrological data, even in the absence of direct measurements. Using at least 48 months of meteorological and hydrological data, this research aims to model catchment behaviour and improve the understanding of hydrological responses to climate variability and change in remote areas. This work seeks to enhance our ability to predict the physical drivers of hydrological change and contribute to more accurate forecasting in ungauged catchments.