A deep learning-based geostatistical framework for geospatial data analysis and modeling

Keywords: Deep learning, geostatistics, spatial uncertainty, GeoAI
Session Type: Virtual Paper Abstract
Day: Tuesday
Session Start / End Time: 3/1/2022 08:00 AM (Eastern Time (US & Canada)) - 3/1/2022 09:20 AM (Eastern Time (US & Canada))
Room: Virtual 35

Abstract

In the past several years, the deep neural network-based methods have dramatically improved the state-of-the-art in pattern recognition and applications. With a deep neural network with multiple levels of processing layers, the deep learning-based methods have been shown to excel at discovering intricate high-level of patterns from high-dimensional data. Recent studies have demonstrated the initial success of the deep neural network in modeling and analyzing geospatial data, particularly in remote sensing imagery analysis and understanding. Few has been done to exploit the power of the deep learning for general geospatial data analysis and modeling. Geostatistics represents a conventional approach to characterize and model spatial patterns and uncertainty in geospatial observations. In this paper, we explored approaches to integrate the geostatistics and deep learning to take advantages of both sides, and presented a deep learning-based geostatistical framework for statistical analysis and modeling of geospatial data. The performances of the presented framework are demonstrated with real and synthetic cases.