Verification of probabilistic seasonal rainfall forecasts for Fiji Islands

Accurate and skillful seasonal rainfall forecasting is crucial for various socioeconomic activities, particularly providing essential climate information for agricultural planning and decision-making, prepare and respond to disasters such as droughts, landslides and floods and provide support to water resources management. In Fiji, where the economy is highly reliant on rainfall, reliable forecasts can significantly mitigate the adverse impacts of hydroclimate variability. Although, Fiji Meteorological Services (FMS) issues seasonal rainfall forecast for the country, validation of forecasts has not been done before. This study aims to evaluate the seasonal forecasts for the Fiji Islands using probabilistic verification methods to assess their accuracy and skill. The forecasts, produced by the FMS, are analyzed through probabilistic categories: below-normal (BN), normal (N), and above-normal (AN), across five regions of Fiji (Western, Central, Eastern, and Northern Divisions, and Rotuma). To assess forecast performance, the study compares the regional seasonal forecasts with observations from the FMS. Additionally, the study delineates the climatological zone for Fiji to ensure consistency in the forecast approach. The results show that the percentage correct exceeds 50%, and the Probability of Detection indicates that more than half of the forecasted categories are accurate. However, the False Alarm Ratio shows that 60% of the forecasted events are false alarms. In terms of the Critical Success Index for N category, all Divisions and Rotuma (except Eastern Division) show that more than half of N rainfall events are correctly predicted based on observations. The Heidke skill score ranges from 0.23 (Western Division) to 0.003 (Rotuma), indicating varying degrees of forecast skill across the regions. The study also identifies that some stations belong to different climatological zones than initially assumed. Therefore, there is a need to improve the seasonal forecasts, and the use of consistent and homogeneous climatological zones is recommended to enhance forecast accuracy.