Abstract:Understanding the influence of the Antarctic on the global climate is crucial for the prediction of global warming. However, due to very few observation sites, it is difficult to reconstruct the rational spatial pattern by filling in the missing values from the limited site observations. To tackle this challenge, regional spatial gap-filling methods, such as Kriging and inverse distance weighted (IDW), are regularly used in geoscience. Nevertheless, the reconstructing credibility of these methods is undesirable when the spatial structure has massive missing pieces. Inspired by image inpainting, we propose a novel deep learning method that demonstrates a good effect by embedding the physics-aware initialization of deep learning methods for rapid learning and capturing the spatial dependence for the high-fidelity imputation of missing areas. We create the benchmark dataset that artificially masks the Antarctic region with ratios of 30%, 50% and 70%. The reconstructing monthly mean surface temperature using the deep learning image inpainting method RFR (Recurrent Feature Reasoning) exhibits an average of 63% and 71% improvement of accuracy over Kriging and IDW under different missing rates. With regard to wind speed, there are still 36% and 50% improvements. In particular, the achieved improvement is even better for the larger missing ratio, such as under the 70% missing rate, where the accuracy of RFR is 68% and 74% higher than Kriging and IDW for temperature and also 38% and 46% higher for wind speed. In addition, the PI-RFR (Physics-Informed Recurrent Feature Reasoning) method we proposed is initialized using the spatial pattern data simulated by the numerical climate model instead of the unified average. Compared with RFR, PI-RFR has an average accuracy improvement of 10% for temperature and 9% for wind speed. When applied to reconstruct the spatial pattern based on the Antarctic site observations, where the missing rate is over 90%, the proposed method exhibits more spatial characteristics than Kriging and IDW.

Phlst with Adaptive Tiling and Its Application to Antarctic Remote Sensing Image Approximation

Estimation of Sea Ice Movement Using Locally Parallel Matching Model.

Polyharmonic Wavelet Transform and Image Approximation

Adaptively Tiled Image Mosaics Utilizing Measures of Color and Region Entropy

High-Dimensional Data Compression Via Phlct

Multilevel T-spline Approximation for Scattered Observations with Application to Land Remote Sensing

Harmonic Wavelet Transform and Image Approximation

Automatic Local Phenology Simulation for Landsat TM Image

Robust multi-scale image matching for deriving ice surface velocity field from sequential satellite images

A Novel Image Registration Algorithm for Remote Sensing under Affine Transformation.

Physics-Informed Deep Learning for Reconstruction of Spatial Missing Climate Information in the Antarctic

A Fast and Adaptive Surface Reconstruction Algorithm Based on Implicit Tensor-Product B-Spline Surfaces ∗

A Bayesian Approach for Interpolating Clear-Sky MODIS Land Surface Temperatures on Areas With Extensive Missing Data

A Generalized and Accelerated Approach of Ambiguity-Free Imaging for Sub-Nyquist SAR

Adaptive multiresolution analysis based on anisotropic triangulations

Compensated Convexity Methods for Approximations and Interpolations of Sampled Functions in Euclidean Spaces: Applications to Contour Lines, Sparse Data and Inpainting

Efficient Analysis-Suitable T-spline Fitting for Freeform Surface Reconstruction and Intelligent Sampling

An Adaptive Angular Aomain Compression Scheme For Solving Multiscale Radiative Transfer Equation

Simultaneous Seismic Data Interpolation and Denoising Based on Nonsubsampled Contourlet Transform Integrating With Two-Step Iterative Log Thresholding Algorithm

An Adaptive Phase Optimization Algorithm for Distributed Scatterer Phase History Retrieval

Quasi-P-Wave Reverse Time Migration in TTI Media with a Generalized Fractional Convolution Stencil