Evaluating the accuracy and skill of predicting rainfall in Terminal Aerodrome Forecasts at Entebbe International Airport

Abstract

Weather forecasting is a critical aspect for optimizing aerodrome operations. It allows ensuring on-ground and en-route safe and efficient air traffic management. The Terminal Aerodrome Forecast (TAF) in particular summarizes the expected atmospheric conditions at an aerodrome or airport, especially those related to the prediction of rainfall. Rainfall variability, especially bad weather in terms of heavy rainfall is a major cause of flight delays at Entebbe International Airport (EIA) as it reduces visibility and causes slippery runways, hence the need for reliable rainfall forecasts in the Terminal Aerodrome Forecast reports. It is thus important to verify rainfall forecasts in order to comprehend a general understanding of the atmosphere for provision of information for policy and decision making in aviation. This study assessed the accuracy and skill in predicting the occurrence or non-occurrence of rainfall in Terminal Aerodrome Forecasts made at Entebbe International Airport. TAFs and Meteorological Aerodrome Routine Reports (METARs) as well as Special Weather Reports (SPECIs) were used to provide the forecasted and observational data, respectively that was used for the verification process. This data was objectively categorized into hits, misses, false alarms and correct rejections in a contingency table on a monthly basis. From this, metrics for determining accuracy and skill such as Probability of Detection (POD), False Alarm Ratio (FAR), Bias, Proportional Correct (PC), Heidke Skill Score (HSS), Gilbert Skill Score (GSS) and True Skill Score (TSS) were determined. In general, the results included a POD of 0.49, FAR of 0.84 and PC of 0.8. With an average bias of 3.7, the results indicate that rainfall occurrences were over forecasted throughout all seasons of the year. In addition, there was significant difference in accuracy of TAFs across seasons although MAM season slightly had a significantly different accuracy from all the other seasons. However, the skill metrics had no significant variation across the seasons and that the probability of forecasting rainfall in TAFs was not more than just a chance. The study recommends the development and implementation of an automatic verification model that can enhance forecasters’ accuracy and skill when they fall short in regard to the method or model used during the forecasting process.