A Novel GPU-Based Acceleration Algorithm for a Longwave Radiative Transfer Model

Wang, Yuzhu; Zhao, Yuan; Jiang, Junguang; Zhang, He

doi:10.3390/app10020649

Cited by 9 publications

(2 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These architectures include the residual blocks proposed in ResNet (He et al, 2015b) and the depth-wise separable convolution used in MobileNets (Howard et al, 2017) and Xception (Chollet, 2016), among others. Furthermore, EfficientNet (Tan and Le, 2019) has shown that an efficient balancing of network depth, width, and resolution can lead to better performance in terms of prediction accuracy and speed. However, any such reduction of model parameters in CNNs or exploring newer architectures must be accompanied with a rigorous validation procedure, which could be similar to the workflow proposed above.…”

Section: Discussionmentioning

confidence: 99%

“…Recent advances in NNs have led to rapid progress in the accuracy of pattern and image recognition tasks. In particular, convolutional neural networks (CNNs) (Krizhevsky et al, 2012a) have achieved impressive results for image classification (Krizhevsky et al, 2012b), while recurrent neural networks (RNNs) have made breakthroughs in sequenceto-sequence learning tasks such as machine translation (Wu et al, 2016). Efforts to use machine learning techniques to model actual physical processes in a climate model have increased recently (Schneider et al, 2017;Gentine et al, 2018;Rasp et al, 2018;O'Gorman and Dwyer, 2018;Scher, 2018;Bretherton, 2018, 2019;San and Maulik, 2018;Yuval and O'Gorman, 2020).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

RadNet 1.0: exploring deep learning architectures for longwave radiative transfer

2020

View full text Add to dashboard Cite

Abstract. Simulating global and regional climate at high resolution is essential to study the effects of climate change and capture extreme events affecting human populations. To achieve this goal, the scalability of climate models and efficiency of individual model components are both important. Radiative transfer is among the most computationally expensive components in a typical climate model. Here we attempt to model this component using a neural network. We aim to study the feasibility of replacing an explicit, physics-based computation of longwave radiative transfer by a neural network emulator and assessing the resultant performance gains. We compare multiple neural-network architectures, including a convolutional neural network, and our results suggest that the performance loss from the use of conventional convolutional networks is not offset by gains in accuracy. We train the networks with and without noise added to the input profiles and find that adding noise improves the ability of the networks to generalise beyond the training set. Prediction of radiative heating rates using our neural network models achieve up to 370× speedup on a GTX 1080 GPU setup and 11× speedup on a Xeon CPU setup compared to the a state-of-the-art radiative transfer library running on the same Xeon CPU. Furthermore, our neural network models yield less than 0.1 K d−1 mean squared error across all pressure levels. Upon introducing this component into a single-column model, we find that the time evolution of the temperature and humidity profiles is physically reasonable, though the model is conservative in its prediction of heating rates in regions where the optical depth changes quickly. Differences exist in the equilibrium climate simulated when using the neural network, which are attributed to small systematic errors that accumulate over time. Thus, we find that the accuracy of the neural network in the “offline” mode does not reflect its performance when coupled with other components.

show abstract

Section: Discussionmentioning

confidence: 99%