“Nowcasting” provides weather forecasting central to issuing alerts and warnings of high-impact weather on timescales typically shorter than 6 hours. On these timescales numerical weather prediction (NWP) models cannot be utilised due to their slow runtime, and therefore forecasters generally rely on directly interpreting observations to make predictions about the evolution of weather conditions. By utilising recent advances in machine learning this challenge presents an opportunity to learn directly from data the near-time evolution of weather and through this gain new insights into atmospheric processes of storm formation. Accurate nowcasting tools are extremely useful in providing warning of imminent flash flood events, potentially providing a few hours’ notice in which to take action to prevent risk to life or property.
The aim of the project is to provide precipitation nowcasting using real-time remote sensing observations (ground-based radar and satellite) to predict precipitation for the next hours. Satellite observations have the advantage of near global coverage, particularly valuable in data sparse parts of the world such as Africa, while ground based radar offer higher spatial and temporal resolution observations. The student will develop a novel approach to nowcasting by coupling a deep learning model and a simple 2D fluid dynamics model, placing the research within the realms of emerging field of physics-informed neural networks. A precipitation nowcasting system must predict both the movement and evolution of a storm. Current machine learning approaches to nowcasting tend to use neural network as a black box and learn both the processes together, but this can often lead to solutions which are unrealistically smeared out. In practice we already have considerable physical insight into the motion of storms and physics-informed neural networks provide an exciting new approach to integrating this prior knowledge to improve predictions. By utilising a 2D fluid dynamics model to take care of the horizontal advection of storms, the neural network is tasked with only predicting the formation and evolution of storms. Based on existing research it is hypothesised that utilising a separate advection model will alleviate the numerical diffusion (smearing) seen in contemporary neural network-only approaches to nowcasting. The decoupling of transport and evolution will also allow for the introduction of physics-based constraints on the neural network solution, for example constraining the horizontal wind field to be approximately divergence free.
While clouds and precipitation are routinely observed from existing remote sensing platforms, it is much harder to observe wind fields over a large area at a high enough frequency for nowcasting. Initially the horizontal (synoptic) flow field will initially be given by a forecast model with the possibility to drive the project towards diagnosing or correcting the flow-field with machine learning applied to observations. Older model runs can be used for this as forecasts of large-scale synoptic winds generally are driven by large pressure systems and are well-resolved by NWP models. Some newer Doppler weather radars (for example the NCAS C-band radar operated from Leeds) are capable of directly observing winds as well as precipitation when there is rain present. Storm motion can also be inferred by tracking storms over multiple observations. The student could also investigate the feasibility of either using entirely observational wind data or using the observational data to nudge the model data towards a more realistic wind field in order to maximise the short-range accuracy of the nowcasts.
Physics-based constraints need not be just related to the winds. Observations of surface temperature / humidity or profiles of temperature / humidity, either in-situ or remotely sensed, also contain valuable information which can help predict the likely evolution of a storm and so the student could also investigate the value of adding such observations in to a neural network nowcasting system.
Depending on the interest of the student the project will utilise high-resolution (both spatially and temporally) radar and satellite observations from the EU/UK (Nimrod radars and MSG Seviri satellite), USA (Nexrad radars and GOES-16/GOES-17 satellite) or South East Asia (NZ MetService radars and Himawari-8 satellite). The synoptic wind fields will be given by either the Met-Office UKV (4km) model or the ECMWF ERA5 reanalysis (0.1deg). Through the Met Office CASE award, the student will have access to state-of-the-art computing, modelling and observational data as well as staff expertise in atmospheric science, and additional funding provided by a CASE award.