Neural network emulator for atmospheric chemical ODE.

Area of Science:

• Atmospheric Chemistry
• Computational Science
• Artificial Intelligence

Background:

• Atmospheric chemistry modeling is computationally intensive.
• Deep neural networks show promise in complex modeling tasks.

Purpose of the Study:

• To develop a fast and accurate neural network emulator for atmospheric chemistry.
• To model atmospheric chemistry as a time-dependent Ordinary Differential Equation (ODE) using neural networks.

Main Methods:

• Proposed ChemNNE, an Attention-based Neural Network Emulator (NNE).
• Utilized sinusoidal time embedding for temporal patterns and Fourier neural operator for ODE modeling.
• Introduced three physics-informed loss functions for training.

Main Results:

• Achieved state-of-the-art performance in modeling accuracy.
• Demonstrated significant improvements in computational speed.
• Introduced a novel large-scale dataset for benchmarking.

Conclusions:

• ChemNNE offers an efficient and accurate solution for atmospheric chemistry modeling.
• The physics-informed approach enhances model reliability.
• The developed dataset facilitates future research in the field.