Use this URL to cite or link to this record in EThOS:
Title: Information gain that convective-scale models bring to probabilistic weather forecasts
Author: Cafaro, Carlo
ISNI:       0000 0004 7971 8962
Awarding Body: University of Reading
Current Institution: University of Reading
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
Ensemble prediction systems, run nowadays at convective-scale by several operational and research forecasting centres, undoubtedly provide a large amount of data. Whether and how the utilisation of these data can lead to additional valuable information for the probabilistic prediction of specific weather phenomena is an open challenge. Here, the information gain of the Met Office convective-scale ensemble relative to its lower resolution global model counterpart (33 km) is quantified for the sea breeze phenomenon. It occurs on small spatio-temporal scale but influenced by the large-scale conditions, well represented by the global ensembles. Sea breeze is also an atmospheric counterpart of the so called density currents and could lead to high impact weather, especially when colliding with other sea breezes or mesoscale flows. In the first part of the thesis, a new set of numerical simulations of colliding density currents is presented, with the aim of understanding the dynamics of collision. A novel parametric formula for predicting the collision angle is shown to agree well with numerical data. This can be useful for future parametrizations of these processes. Probabilistic forecasts of the occurrence of sea breezes have been then produced, compared and verified over an extended period. This motivated the development of a novel method for the automatic identification of the sea breeze from convective-scale ensembles. This method can be in principle applied to every coastline. A Bayesian approach is instead used to extract information from the coarser resolution ensemble. It is trained on paired high/low resolution ensemble member, but it can be trained also on observations. This method creates a statistical forecast of the high-resolution ensemble member, given the knowledge of the global ensemble predictors alone. In the last part of the thesis, the same methodology has been applied to the prediction of wind gusts. Comparison of the two forecasting methods, using a variety of well established verification metrics all lead to the same conclusion: although the Bayesian forecasts have potential skill for the prediction of event occurrence, the convective-scale ensemble is shown to be more skillful.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral