Validation of ERA5 rainfall data over the South Pacific Region: case study of Fiji Islands

Abstract

Rainfall variability has a significant impact on hydrological cycle. Understanding rainfall variability over Fiji Islands is important for decision-making in the backdrop of global warming. Reanalysis rainfall products are commonly used to overcome observed data quality challenges especially over ungauged highland areas. However, an evaluation of reanalysed datasets is important to ensure accurate and reliable climate information generated using such datasets, especially for small Island with high variable topography like Fiji. This work aims to validate the spatiotemporal performance of European Centre for Medium-Range Weather Forecasts (ECMWF) fifth-generation reanalysis rainfall (ERA5) data against ground-based station data from 19 stations for the period 1971–2020 over Fiji Islands. Correlation coefficient and difference statistics: bias, and root mean square error, are used to assess the performance of the data. Further, common Empirical Orthogonal Function (common EOFs) analysis was used to evaluate spatiotemporal performance of ERA5 datasets. The results of the station-by-station comparison shows that interpolated ERA5 annual rainfall matches the corresponding results from rain gauges remarkably well for many stations. The correlation coefficient values range from 0.5 to 0.85, while the bias spans from a negative 282 to a positive 575, and the root mean square error (RMSE) varies between 285 and 662 mm for the annual rainfall across the study area. However, there is overestimation and underestimation of the observed rainfall by ERA5 datasets. The leading common EOF principal component for annual rainfall suggests that the inter-annual variability in ERA5 dataset is generally consistent with observed station datasets, cross validation results indicated high scores (correlations of 0.82), with limited spatial variation. This work presents a reliable data assessment of the ERA5 data over Fiji Islands, indicating there is good match of the annual observed rain gauged station data and ERA5. The findings give accuracy references for further use of the ERA5 data in understanding rainfall variability and change over the region.