Publication Date:
2023-08-30
Description:
Better understanding snowfall microphysics is a key challenge in atmospheric science, crucial for snowfall quantification, remote sensing, and weather forecasting in general. Using meteorological radars, we propose a novel approach to retrieve the microphysical properties of snowfall from dual-frequency Doppler spectral observations, while relaxing assumptions on beam matching and non-turbulent atmosphere. The approach relies on a two-step deep-learning framework inspired from data compression techniques: an encoder maps a high-dimensional signal to a lower-dimensional latent space, while the decoder reconstructs the original signal from this latent space. Here, dual-frequency Doppler spectrograms constitute the high-dimensional input, while the dimensions of the latent space are constrained to represent the snowfall properties of interest. The decoder neural network is first trained to emulate Doppler spectra from a set of microphysical variables, using simulations from the radiative transfer model PAMTRA as training data. In a second step, the encoder network learns the inverse mapping, from measured dual-frequency spectrograms to the microphysical latent space. The method was applied to X- and W-band data from the ICE GENESIS campaign that took place in Switzerland in January 2021. The approach was thoroughly evaluated by comparisons with collocated aircraft in situ measurements collected during three precipitation events, with an overall good agreement. The main contribution of this work is (i) the inversion framework itself, which can be applied to other remote-sensing retrieval applications ; and (ii) the seven retrieved microphysical descriptors providing relevant insights into snowfall processes.
Language:
English
Type:
info:eu-repo/semantics/conferenceObject
Permalink