Source author record

Menno A. Veerman

Menno A. Veerman appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

physics.ao-ph Machine Learning physics.flu-dyn

Catalog footprint

What is connected

3works

3topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2022arXiv

Shallow cumulus convection over land in cloud-resolving simulations with a coupled ray tracer

We present simulations of cumulus convection over land with shortwave radiation computed by a Monte Carlo ray tracer coupled to the flow solver. Ray tracing delivers very realistic in-cloud heating rates and global horizontal irradiance fields. The necessary performance of the ray tracer has been enabled by the raw power of GPU computing and from techniques for accelerating data lookup and ray tracer convergence. We used a case study over a grassland in the Netherlands to compare simulations with a coupled ray tracer to those with a conventional two-stream solver, and to test ray tracer convergence. We demonstrate that the simulated cloud evolution is insensitive across a wide range of samples sizes in the ray tracer. Furthermore, simulations with a coupled ray tracer produce surface irradiance patterns that resemble observations and that strongly feed back to the evolution of clouds via locally enhanced surface heat fluxes.

preprint2021arXiv

Record high solar irradiance in Western Europe during first COVID-19 lockdown largely due to unusual weather

Spring 2020 broke sunshine duration records across western Europe. The Netherlands recorded the highest surface irradiance since 1928, exceeding the previous extreme of 2011 by 13 %, and the diffuse fraction of the irradiance measured a record low percentage (38 %). The coinciding irradiance extreme and a reduction in anthropogenic pollution due to COVID-19 measures triggered the hypothesis that cleaner-than-usual air contributed to the record. Based on analyses of ground-based and satellite observations and experiments with a radiative transfer model, we estimate a 1.3 % (2.3 W m$^{-2}$) increase in surface irradiance with respect to the 2010-2019 mean due to a low median aerosol optical depth, and a 17.6 % (30.7 W m$^{-2}$) increase due to several exceptionally dry days and a very low cloud fraction overall. Our analyses show that the reduced aerosols and contrails due to the COVID-19 measures are far less important in the irradiance record than the dry and particularly cloud-free weather.

preprint2020arXiv

Predicting atmospheric optical properties for radiative transfer computations using neural networks

The radiative transfer equations are well-known, but radiation parametrizations in atmospheric models are computationally expensive. A promising tool for accelerating parametrizations is the use of machine learning techniques. In this study, we develop a machine learning-based parametrization for the gaseous optical properties by training neural networks to emulate a modern radiation parameterization (RRTMGP). To minimize computational costs, we reduce the range of atmospheric conditions for which the neural networks are applicable and use machine-specific optimised BLAS functions to accelerate matrix computations. To generate training data, we use a set of randomly perturbed atmospheric profiles and calculate optical properties using RRTMGP. Predicted optical properties are highly accurate and the resulting radiative fluxes have average errors within \SI{0.5}{\flux} compared to RRTMGP. Our neural network-based gas optics parametrization is up to 4 times faster than RRTMGP, depending on the size of the neural networks. We further test the trade-off between speed and accuracy by training neural networks for the narrow range of atmospheric conditions of a single large-eddy simulation, so smaller and therefore faster networks can achieve a desired accuracy. We conclude that our machine learning-based parametrization can speed-up radiative transfer computations whilst retaining high accuracy.