Data Assimilation (DA) is the process of estimating a distribution of possible environmental model states by combining a prior or forecast distribution of environmental states with information from error prone observations of the model-related variables. Major improvements in weather forecasting over the last few decades are directly attributable to improved data assimilation. However, much work remains. Vast numbers of observations of clouds, precipitation and aerosols are currently unassimilated because current data assimilation techniques do not adequately account for the non-linear, multi-scale and non-Gaussian uncertainty distributions associated with these variables. Breakthrough research is urgently needed in this area because clouds, precipitation and aerosols remain the largest contributor to climate change and weather prediction uncertainty.
Ensemble forecasting attempts to predict the distribution of future possible states from a collection or ensemble of environmental model forecasts using information about the errors in model initial conditions and the model evolution. Over nearly all time scales the mean of an ensemble forecast can be shown to have much smaller mean square errors than any single forecast. Without ensemble forecasting, it would be extremely difficult to extract useful information from forecasts beyond seven days. Data assimilation is closely linked to ensemble forecasting because data assimilation methods are greatly improved by ensembles and, at the same time, data assimilation greatly assists attempts to accurately represent initial condition and model error uncertainty. Ensemble forecasting is currently revolutionizing the way leading organizations in electricity, agriculture, transportation, public safety, insurance and finance manage environmental risk.
Dynamics refers to studies aimed at understanding why the atmosphere appears and evolves as it does. Such understanding is fundamental to the development of innovative new ways of improving data assimilation and ensemble forecasting techniques. I am particularly interested in dynamical understanding that helps expose weaknesses in current atmospheric modelling techniques, and hence, suggest research directions that would improve atmospheric modelling methods.
Predictability refers to our ability to use an estimate of the current state of a system to predict its future state. Research in this area requires a good understanding of data assimilation, ensemble forecasting and dynamics. Central questions in predictability research include “How much more predictability could be gained by improving the observational network and/or the data assimilation system and/or our models?”
I have a range of PhD projects available in each of these areas. Please stop by my office or email me if at all interested and together we can mould a high impact PhD research topic for your thesis.