Lossy Neural Compression for Geospatial Analytics: A Review

TL;DR

This paper addresses the bottlenecks of storage, bandwidth, and compute in Earth Observation and climate simulations by examining lossy neural compression (NC) as a data-driven alternative to traditional codecs. It systematically reviews NC fundamentals, including transform coding, quantization, and entropy models, and then analyzes NC applications for remote sensing and climate data, highlighting challenges and architectural adaptations. The authors connect NC to self-supervised learning and foundation models, discuss platform-level implementation considerations, and propose future directions such as feature compression, onboard processing, and standards for neurally compressed embeddings. The work emphasizes the potential for substantial reductions in data transfer and storage costs while enabling scalable, automated geospatial analytics, with careful attention to preserving scientifically relevant information and uncertainty quantification. Overall, it provides practical guidance for leveraging NC in EO/ESM pipelines and outlines concrete research directions to make NC widely adoptable in geospatial domains.

Abstract

Over the past decades, there has been an explosion in the amount of available Earth Observation (EO) data. The unprecedented coverage of the Earth's surface and atmosphere by satellite imagery has resulted in large volumes of data that must be transmitted to ground stations, stored in data centers, and distributed to end users. Modern Earth System Models (ESMs) face similar challenges, operating at high spatial and temporal resolutions, producing petabytes of data per simulated day. Data compression has gained relevance over the past decade, with neural compression (NC) emerging from deep learning and information theory, making EO data and ESM outputs ideal candidates due to their abundance of unlabeled data. In this review, we outline recent developments in NC applied to geospatial data. We introduce the fundamental concepts of NC including seminal works in its traditional applications to image and video compression domains with focus on lossy compression. We discuss the unique characteristics of EO and ESM data, contrasting them with "natural images", and explain the additional challenges and opportunities they present. Moreover, we review current applications of NC across various EO modalities and explore the limited efforts in ESM compression to date. The advent of self-supervised learning (SSL) and foundation models (FM) has advanced methods to efficiently distill representations from vast unlabeled data. We connect these developments to NC for EO, highlighting the similarities between the two fields and elaborate on the potential of transferring compressed feature representations for machine--to--machine communication. Based on insights drawn from this review, we devise future directions relevant to applications in EO and ESM.

Figures (21)

• Figure 1: Literature on neural compression (NC) summarized for the past 15 years. We plot separate bars for remote sensing (RS) and recent developments in foundation model (FM) methodology. Data Source: Queries to the Web of ScienceWoCGraphic2000-2024AICompressorHistory.
• Figure 2: Domain-specific shares in publications in NC methodologies from years 2000 through 2023 for the ten biggest (sub)categories as per Web of ScienceWoCGraphic2024AICompressor. The stacked subcategories for Computer Science have been ordered from largest to smallest bottom-up.
• Figure 3: Taxonomy of compression methods. This review focuses on the family of methods defined by following the blue nodes. We group all variations within this family into transforms, quantization strategies, entropy models, and optimization objectives.
• Figure 4: Image compression from the perspective of pixel correlations, Left to Right: constant image, correlated noise, uncorrelated noise. A detailed description is provided in the main text.
• Figure 5: Depiction of typical compression pipeline.