Historical Correlation Between IMERG V07 Rainfall Estimates Over the Atlantic Ocean and Stream Level Variations in Four UK Catchments

One of the major challenges in improving flood management is the low reliability and limited availability of rainfall forecasts. This challenge becomes more pronounced in areas affected by rainfall originating in oceanic regions beyond the reach of land-based instruments. In such cases, satellite precipitation products (SPPs) present an alternative source of rainfall data for use in flooding- risk forecasts. To explore this possibility, it is essential to first establish whether there is a relationship between the SPP data and flood risk, and to what extent this relationship can be detected. Therefore, this research explores the historical relationship between rainfall events over the Atlantic Ocean, as captured by IMERG V07 estimates, and stream level variations in four UK catchments. The study utilizes over 20 years of data to perform cross-correlation analyses between stream level records and precipitation estimates from each pixel of the IMERG V07 grid in a selected region of the Atlantic near the UK. The analysis revealed several key insights into these relationships.: A) Each catchment has a distinct historical path for rainfall events moving across the ocean toward the UK, that are related to stream level variations. B) It is possible to identify the regions of the Atlantic that consistently produce the most impactful rainfall events affecting catchments in specific areas of the UK. C) The strongest rainfall-stream level relationships were observed at distances of up to 650 km, which may help to compensate for the 4-hour latency of IMERG V07 early run data, enhancing its suitability for real-time flood forecasting. Such findings are significant as they allow for a more focused approach and the direction of monitoring efforts for flood risk detection toward specific regions of the Atlantic rather than monitoring vast oceanic areas that leads to the processing large amounts of irrelevant data. The next phase of this study focus on applying these findings on the development of a machine learning model able to predict stream level variations based on the long-distance relationship with rainfall events, exploring the potential of earlier risk detection for increasing lead time of flooding forecasts