Premium
Assessment of convection‐permitting versions of the Unified Model over the Lake Victoria basin region
Author(s) -
Hanley Kirsty E.,
Pirret Jennifer S. R.,
Bain Caroline L.,
Hartley Andrew J.,
Lean Humphrey W.,
Webster Stuart,
Woodhams Beth J.
Publication year - 2021
Publication title -
quarterly journal of the royal meteorological society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.744
H-Index - 143
eISSN - 1477-870X
pISSN - 0035-9009
DOI - 10.1002/qj.3988
Subject(s) - precipitation , storm , climatology , convection , environmental science , tropical cyclone , meteorology , rainband , structural basin , atmospheric sciences , geology , geography , paleontology
The Met Office runs an operational high‐resolution convection‐permitting model over Tropical Africa to support National Meteorological Services across the continent. This article evaluates the prediction of convective storms in two convection‐permitting versions of the Met Office Unified Model over the Lake Victoria basin region in East Africa. Two model configurations are compared for a case of severe convection: the first was tuned for tropical regions and the second was based on the operational UK model configuration. The tropical configuration compares better with satellite‐derived rainfall observations in terms of domain‐average rain‐rate and the distribution of rain‐rates is better, particularly for the more intense rain‐rates. However, both configurations generally produce too much rain, too many small storms, and a lack of light rain, which is compensated for by heavy rain in the cores of the storms. The tropical configuration was then assessed over two rainy seasons. In both seasons, the onset time of night‐time precipitation over Lake Victoria is delayed and the amount of precipitation is underestimated, which suggests the model may miss night‐time storms that occur over the lake. During the first 12–24 hr of the simulations, the precipitation field is still spinning up from the lower‐resolution global model fields used to initialise the convection‐permitting model. This leads to large overestimates in the domain‐average precipitation during this period and highlights the need to run convection‐permitting models for at least two days to reduce the impact of cold‐starting from a non‐convective‐permitting analysis. Despite the spin‐up issues, fractions skill score analysis shows that forecast skill decreases with increasing lead time. This implies that there is inherent skill in the early stages of the model forecast and underlines that short‐term forecast (0–48 hr) improvements in the model are likely to be achieved when using an improved analysis for initialisation.