arXiv subjects

Robin J. Hogan

Publications and source records attributed to Robin J. Hogan.

3 recordsLinked to original sources

arXiv2022

Machine Learning Emulation of 3D Cloud Radiative Effects

The treatment of cloud structure in numerical weather and climate models is often greatly simplified to make them computationally affordable. Here we propose to correct the European Centre for Medium-Range Weather Forecasts 1D radiation scheme ecRad for 3D cloud effects using computationally cheap neural networks. 3D cloud effects are learned as the difference between ecRad's fast 1D Tripleclouds solver that neglects them and its 3D SPARTACUS (SPeedy Algorithm for Radiative TrAnsfer through CloUd Sides) solver that includes them but is about five times more computationally expensive. With typical errors between 20 % and 30 % of the 3D signal, neural networks improve Tripleclouds' accuracy for about 1 % increase in runtime. Thus, rather than emulating the whole of SPARTACUS, we keep Tripleclouds unchanged for cloud-free parts of the atmosphere and 3D-correct it elsewhere. The focus on the comparably small 3D correction instead of the entire signal allows us to improve predictions significantly if we assume a similar signal-to-noise ratio for both.

cs.LG↗

arXiv2021

Copula-based synthetic data augmentation for machine-learning emulators

Can we improve machine-learning (ML) emulators with synthetic data? If data are scarce or expensive to source and a physical model is available, statistically generated data may be useful for augmenting training sets cheaply. Here we explore the use of copula-based models for generating synthetically augmented datasets in weather and climate by testing the method on a toy physical model of downwelling longwave radiation and corresponding neural network emulator. Results show that for copula-augmented datasets, predictions are improved by up to 62 % for the mean absolute error (from 1.17 to 0.44 W m$^{-2}$).

cs.LG↗

arXiv2018

Rainfall nowcasting by combining radars, microwave links and rain gauges

The objective of this work is to provide high-resolution rain rate maps at short lead-time forecasts (nowcasts) necessary to anticipate flooding and properly manage sewage systems in urban areas by combining radars, rain gauges, and operational microwave links, and taking into account their respective uncertainties. A variational approach (3D-Var) is used to find the best estimate for the rain rate, and its error covariance, from the different rain sensors. Short-term rain rate forecasts are then produced by assuming Lagrangian persistence. A velocity field is obtained from the operational radar-derived rain fields, and the rain rate field is advected using the Total Variance Diminishing (TVD) scheme. The error covariance associated to the estimated rain rate is also propagated, and we use these two in the 3D-Var at the next observation time step. This approach can be seen as a Variational Kalman Filter (VKF), in which the covariance of the prior is not constant but dependent on time. The proposed approach has been tested using data from 14 rain gauges, 14 microwave links and the operational radar rain product from MeteoSwiss in the area of Zurich (Switzerland). During the applications the assumption of the Lagrangian persistence appears to be valid up to 20 min (a bit longer for stratiform events). During convective events, the algorithm is less powerful and shorter lead times should be considered (i.e., 15 min). Although such lead times are short, they are still useful to various hydrological and outdoor applications.

physics.ao-ph↗