Abstract
Abstract. In this study, we explore a new approach based on machine learning (ML) for deriving aerosol extinction coefficient profiles, single-scattering albedo and asymmetry parameter at 360 nm from a single multi-axis differential optical absorption spectroscopy (MAX-DOAS) sky scan. Our method relies on a multi-output sequence-to-sequence model combining convolutional neural networks (CNNs) for feature extraction and long short-term memory networks (LSTMs) for profile prediction. The model was trained and evaluated using data simulated by Vector Linearized Discrete Ordinate Radiative Transfer (VLIDORT) v2.7, which contains 1 459 200 unique mappings. From the simulations, 75 % were randomly selected for training and the remaining 25 % for validation. The overall error of estimated aerosol properties (1) for total aerosol optical depth (AOD) is -1.4±10.1 %, (2) for the single-scattering albedo is 0.1±3.6 %, and (3) for the asymmetry factor is -0.1±2.1 %. The resulting model is capable of retrieving aerosol extinction coefficient profiles with degrading accuracy as a function of height. The uncertainty due to the randomness in ML training is also discussed.
Highlights
Aerosols play an important role in the Earth–atmosphere system by modifying the global energy balance, participating in cloud formation and atmospheric chemistry, and fertilizing land and ocean
This paper investigates the potential of using advances in machine learning to invert aerosol properties from a hyperspectral remote-sensing technique called multi-axis differential optical absorption spectroscopy (MAX-DOAS)
For the branch corresponding to sequencebased outputs, the features extracted from 1D convolutional neural networks (CNNs) layers are fed to a long short-term memory network (LSTM; Hochreiter and Schmidhuber, 1997) to produce a sequence of partial aerosol optical depth (AOD) values at varying atmospheric layers
Summary
Aerosols play an important role in the Earth–atmosphere system by modifying the global energy balance, participating in cloud formation and atmospheric chemistry, and fertilizing land and ocean. The MAX-DOAS technique has been widely used to derive vertical aerosol extinction coefficient profiles in the lower troposphere This is typically done from ground-based measurements of oxygen collision complex (O2O2) absorption This study describes and evaluates a fast novel machine learning (ML) approach for retrieving aerosol extinction coefficient profiles, asymmetry factor and single-scattering albedo at 360 nm from SCD(O2O2) observations within a single MAX-DOAS sky scan. The basic idea of our approach is as follows: (1) develop an “inverse model” by one-time offline training of a supervised ML algorithm on simulated MAX-DOAS data and corresponding atmospheric aerosol conditions and (2) use the relationships derived in the first step to estimate the aerosol extinction profile, asymmetry factor and single-scattering albedo from the MAXDOAS SCD(O2O2) measurements.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have