Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Natural Language-Driven Global Mapping of Martian Landforms

Yiran Wang, Shuoyuan Wang, Zhaoran Wei, Jiannan Zhao, Zhonghua Yao, Zejian Xie, Songxin Zhang, Jun Huang, Bingyi Jing, Hongxin Wei

TL;DR

MarScope is presented, a planetary-scale vision-language framework enabling natural language-driven, label-free mapping of Martian landforms, and establishes a new paradigm where natural language serves as a direct interface for scientific discovery over massive geospatial datasets.

Abstract

Planetary surfaces are typically analyzed using high-level semantic concepts in natural language, yet vast orbital image archives remain organized at the pixel level. This mismatch limits scalable, open-ended exploration of planetary surfaces. Here we present MarScope, a planetary-scale vision-language framework enabling natural language-driven, label-free mapping of Martian landforms. MarScope aligns planetary images and text in a shared semantic space, trained on over 200,000 curated image-text pairs. This framework transforms global geomorphic mapping on Mars by replacing pre-defined classifications with flexible semantic retrieval, enabling arbitrary user queries across the entire planet in 5 seconds with F1 scores up to 0.978. Applications further show that it extends beyond morphological classification to facilitate process-oriented analysis and similarity-based geomorphological mapping at a planetary scale. MarScope establishes a new paradigm where natural language serves as a direct interface for scientific discovery over massive geospatial datasets.

Natural Language-Driven Global Mapping of Martian Landforms

TL;DR

MarScope is presented, a planetary-scale vision-language framework enabling natural language-driven, label-free mapping of Martian landforms, and establishes a new paradigm where natural language serves as a direct interface for scientific discovery over massive geospatial datasets.

Abstract

Planetary surfaces are typically analyzed using high-level semantic concepts in natural language, yet vast orbital image archives remain organized at the pixel level. This mismatch limits scalable, open-ended exploration of planetary surfaces. Here we present MarScope, a planetary-scale vision-language framework enabling natural language-driven, label-free mapping of Martian landforms. MarScope aligns planetary images and text in a shared semantic space, trained on over 200,000 curated image-text pairs. This framework transforms global geomorphic mapping on Mars by replacing pre-defined classifications with flexible semantic retrieval, enabling arbitrary user queries across the entire planet in 5 seconds with F1 scores up to 0.978. Applications further show that it extends beyond morphological classification to facilitate process-oriented analysis and similarity-based geomorphological mapping at a planetary scale. MarScope establishes a new paradigm where natural language serves as a direct interface for scientific discovery over massive geospatial datasets.
Paper Structure (17 sections, 3 equations, 5 figures)

This paper contains 17 sections, 3 equations, 5 figures.

Table of Contents

  1. Introduction
  2. Results
  3. Acknowledgments

Figures (5)

  • Figure 1: Workflow of the MarScope platform. MarScope enables three query modes---text-based, image-based, and image-text---to retrieve Martian landforms. The system outputs global distribution maps, heat maps, detailed records, and curated image sets, supporting both mapping and dataset construction.
  • Figure 2: Validation of MarScope outputs. Left column: Published global distributions for six representative Martian landforms: alluvial fans morgan2022global, glacier-like forms souness2012inventory, landslides roback2021controls, pitted cones mills2024global, yardangs liu2020mapping, and dark slope streaks bickel2025streaks. Middle column: corresponding MarScope retrievals generated using text-mode queries, with the Hi-Res mode applied to dark slope streaks and the default mode used for the other five landforms. Right column: dynamic F1 scores as a function of the top-K retrieved tiles for image (green), text (red), and multi-modal (blue) query modes. Peak F1 values and the corresponding K are shown for each landform type.
  • Figure 3: Landform distributions retrieved by MarScope. Global maps of aeolian (top), glacial and periglacial (middle), and volcanic/tectonic (bottom) landforms on Mars. Colors indicate landform classes, with representative CTX patches on the right showing characteristic morphologies.
  • Figure 4: Process-based geomorphological mapping enabled by MarScope. Examples retrieved using formation-mechanism--oriented queries, including shallow ground ice, regional stress fields, and catastrophic flooding.
  • Figure 5: Visual discovery of rare Martian landforms using MarScope. Retrieval of (A) doublet craters and (B) inverted craters, showing reference examples (left) and morphologically similar features identified across global CTX data (right).