Optimising oceanic rainfall estimates for increased lead time of stream level forecasting: A case study of GPM IMERG estimates application in the UK

Among the three main rainfall data sources (rain gauge stations, rainfall radar stations and weather satellites), satellites are often the most appropriate for longer lead times in real-time flood forecasting [1]. This is particularly relevant in the UK, where severe rainfall events often originate over the Atlantic Ocean, distant from land-based instruments although it can also limit the effectiveness of satellite data for long-term predictions [2]. The Global Precipitation Measurement (GPM) Integrated Multi-satellitE Retrievals for GPM (IMERG) estimates can be used as an alternative source for rainfall information in real-time flood forecasting models. However, the challenge lies in monitoring the vast oceanic region around the UK and integrating this extensive data into hydrological or data-driven models, which presents computational and time constraints. Identifying key monitoring area for obtaining these estimates is essential to address these challenges and to effectively use this use for water level forecasting in urban drainage systems (UDS).

This study introduced an optimised data-driven model for streamline the collection and use of GPM IMERG rainfall estimates for water level forecasting in UDS. The model’s effectiveness was demonstrated using a 20-year satellite data set from the Atlantic Ocean, west of the UK, focusing on water level forecasting for a specific UDS point in London. This data helped identify the most probable path of rainfall from the Atlantic that impacts UDS water levels. We conducted a cross-correlation analysis between the water level records and each IMERG data pixel within the selected oceanic area.

The analysis successfully pinpointed the most influential rainfall points/pixels along the Atlantic path and their respective lag times between rainfall occurrence and water level changes at any satellite-monitored point until it reaches the mainland and joins the river system. This research enhances understanding of long-distance rainfall patterns while optimising the use of GPM IMERG data. It also aids in reducing data volume and processing time for stream-level forecasting models, aiming for longer lead times.