Responsible AI for Earth Observation

TL;DR

This paper surveys responsible AI in Earth Observation (EO), integrating five guiding pillars—bias mitigation, AI security, geo-privacy, ethical principles, and AI4EO for social good—into a cohesive framework that spans academia and industry. It emphasizes concrete practices such as bias auditing, adversarial defense, uncertainty quantification, and explainable AI, while addressing privacy risks from UAVs and high-resolution imagery. The authors discuss regulatory and ethical contexts, propose actionable guidelines, and highlight opportunities in social good applications like early warning systems and climate teleconnections, as well as business sustainability through governance and open data ecosystems. By linking SDGs with transparent, reproducible EO AI workflows, the work argues for responsible, open, and inclusive AI4EO development that can inform policy, foster trust, and reduce harms across environmental and societal domains. $P(y|x^{*},\mathcal{D})=\int P(y|x^{*},\theta)P(\theta|\mathcal{D})\,d\theta$ representations and other EO-specific metrics underscore the need for uncertainty-aware, transparent modeling in complex geospatial contexts.

Abstract

The convergence of artificial intelligence (AI) and Earth observation (EO) technologies has brought geoscience and remote sensing into an era of unparalleled capabilities. AI's transformative impact on data analysis, particularly derived from EO platforms, holds great promise in addressing global challenges such as environmental monitoring, disaster response and climate change analysis. However, the rapid integration of AI necessitates a careful examination of the responsible dimensions inherent in its application within these domains. In this paper, we represent a pioneering effort to systematically define the intersection of AI and EO, with a central focus on responsible AI practices. Specifically, we identify several critical components guiding this exploration from both academia and industry perspectives within the EO field: AI and EO for social good, mitigating unfair biases, AI security in EO, geo-privacy and privacy-preserving measures, as well as maintaining scientific excellence, open data, and guiding AI usage based on ethical principles. Furthermore, the paper explores potential opportunities and emerging trends, providing valuable insights for future research endeavors.

Figures (7)

• Figure 1: Number of AI4EO-related studies published in IEEE TGRS from 2014 to 2023. Data is collected from Google Scholar advanced search: keywords: ("machine learning" or "deep learning") and "remote sensing."
• Figure 2: Overview of the development of AI algorithms over the last decades. The focus of AI algorithms has shifted multiple times between being model-centric and data-centric. The blue part illustrates the historical development of AI concerning data and models, while the orange part reflects our beliefs in the EO community regarding current and future developments. The concept of responsible AI gained significant attention in the late 2010s and early 2020s. During this period, concerns about AI biases, fairness, transparency, and accountability began to receive increased attention.
• Figure 3: Overview of the main building blocks of Responsible AI in EO: mitigating (unfair) biases, securing AI (defenses, uncertainty modeling, and explainability), preserving (geo)privacy, and addressing ethical concerns outline the considerations necessary for implementing responsible AI methodologies within the fields of EO. Social good presents the opportunities and goals related to how a responsible AI system can effectively be utilized to make a positive difference in people's lives.
• Figure 4: A schematic overview of the different biases that play a role during various stages of a machine learning workflow. Inspired by suresh2021framework.
• Figure 5: Illustration of adversarial attack: a minor perturbation can fool the classifier into wrong predictions. yu2023universal