Deep learning framework for true amplitude imaging: Effect of conditioners and initial models

Abstract

We propose a workflow to correct migration amplitudes by estimating the inverse Hessian operator weights using a neural network–based framework. We train the network such that it learns the transformation between the migration output and true amplitude reflectivity constrained by different conditioners. We analyse the network output with a velocity model and with source illumination as a conditioner. Compared to the velocity model, source illumination as a conditioner performs better because source illumination encodes the geometrical spreading information and accounts for non‐stationarity. We further use the output of the deep neural network as a starting model for accelerating the convergence of iterative least‐squares reverse time migration. Using a deep learning framework, the proposed method combines the model domain and data domain least‐squares migration approaches to recover images with interpretable amplitudes, attenuated migration artefacts, better signal‐to‐noise ratio and improved resolution. We compare the output of the proposed algorithm with conventional least‐squares output and show that the proposed workflow is more robust, especially in the areas with weak illumination.

Diffusion Model-based Probabilistic Downscaling for 180-year East Asian Climate Reconstruction

“…Statistical downscaling methods leverage statistical relationships between low-resolution and high-resolution climate data. These statistical downscaling methods rely on mathematical techniques, including deep learning and traditional statistical approaches, to establish statistical relationships between lowresolution and high-resolution climate data, enabling the derivation of detailed localscale information [17][18][19][20]22 . Currently, both dynamical and statistical downscaling techniques find widespread use in studies related to climate change, climate variability, hydro-climate extremes, and impact assessments at regional scales, particularly within sectors such as agriculture, energy, and water resources [23][24][25][26] .…”

Diffusion model-based probabilistic downscaling for 180-year East Asian climate reconstruction

“…These downscaling techniques, including both dynamical and statistical methods, have been developed to generate high-resolution climate data [13][14][15][16][17][18][19][20] . Among these techniques, Regional Climate Models (RCMs) stand out as dynamical models that utilize topography and circulation conditions from GCMs to generate the regional climate information [13][14][15] .…”

“…Statistical downscaling methods leverage statistical relationships between lowresolution and high-resolution climate data. These statistical downscaling methods rely on mathematical techniques, including deep learning and traditional statistical approaches, to establish statistical relationships between low-resolution and high-resolution climate data, enabling the derivation of detailed local-scale information [17][18][19][20]22 . Currently, both dynamical and statistical downscaling techniques find widespread use in studies related to climate change, climate variability, hydro-climate extremes, and impact assessments at regional scales, particularly within sectors such as agriculture, energy, and water resources [23][24][25][26] .…”

Deep generative model super-resolves spatially correlated multiregional climate data

“…In this paper, we propose a machine learning approach that super-resolves the GCM outputs and reproduces both the local statistics and the instantaneous spatial correlations between distant regions. Among several options for improving the resolution of geophysical or climatological data [29][30][31][32] , our method is based on the generative adversarial network (GAN) approach, which has been proven to be a very powerful downscaling tool through several previous studies 33,34 . To accurately reproduce the physical nature, we use auxiliary but climatologically important data, sea-level pressure distribution and topographic information, in addition to the target variables, temperature and precipitation distributions (see Fig.…”