China is setting an end-of-2028 target for a geological map of Mars, a proposed research project that would bring together spacecraft observations, planetary geology and large-model techniques. The goal described in late-September reporting is not a finished map or a demonstrated AI product. It is a planned scientific workflow intended to organize scattered measurements into a common picture of the planet’s surface and geological history.
The project was presented by Hou Zengqian, a Chinese Academy of Sciences academician and chief scientist for Tianwen-3, according to South China Morning Post. Its proposed deadline gives the effort a clear deliverable, but the available accounts do not establish that the schedule is guaranteed or that the resulting system has already been built.
Accounts from Xinhua and China Daily likewise describe the work as a planned mapping effort. That attribution matters: the public material sets out intended methods and products, while leaving future execution and validation open.
A planetary map built from several measurement layers
The proposed map would start with data from China’s Tianwen-1 mission and add observations from probes operated by the United States and the European Space Agency. The source accounts describe a combination of imagery, spectral measurements, radar, gravity and magnetic data. Each layer records a different aspect of Mars, so the planned work would require researchers to align observations before interpreting surface features or geological relationships.
That data-integration task matters because a global map cannot rely on one type of image alone. Visual imagery can show terrain and surface patterns, while spectral, radar, gravity and magnetic measurements provide other clues about composition and subsurface structure. The sources do not claim that the proposed method has solved those integration challenges. Instead, they describe a framework for assembling the measurements into map products at different levels of detail.
China’s Zhurong rover, which landed on Mars in 2021, is part of the scientific context for the effort. Findings from Tianwen-1 are intended to inform the map, giving the project a connection to observations gathered on the Martian surface as well as orbital data. The plan therefore links an existing Chinese Mars mission to a larger cartographic objective, rather than presenting the map as a standalone software release.
The reported deliverables include a 1:5 million global geological map, a much more detailed 1:50,000 map for landing areas and a set of thematic atlases. Those scales indicate that the proposal is aimed at multiple users and decisions. A global product can support broad geological comparisons, while landing-area maps can organize local information around sites where spacecraft may operate. The thematic atlases would separate particular kinds of geological information rather than compressing every observation into one image.
The proposal also includes an international scientific objective: seeking acceptance of a Chinese Martian geologic-time framework. That is an ambition for how researchers might describe and organize Mars’s geological history, not evidence that such a framework has already gained international acceptance. The distinction is important because scientific recognition would depend on review, comparison with other interpretations and broader use by researchers.
China’s space-computing push provides a useful adjacent context for why large-scale planetary data work is receiving attention. But the fact cards do not establish a dedicated computing architecture, a processing benchmark or a performance advantage for this Mars project. The map remains a proposed research program with stated outputs and methods.
How the proposed AI workflow would use geological expertise
The reported AI approach is designed to put geologists’ expertise into large models through rule engines. That description points to a system in which domain knowledge would shape how the model handles geological features, rather than leaving interpretation entirely to statistical pattern matching. The cards also describe multi-scale attention for feature extraction, suggesting that the planned method would examine structures at different spatial levels.
Those elements are significant because geological mapping involves both visible patterns and interpretive rules. A model may help identify features across very large datasets, while rules supplied by specialists can guide how those features are classified or related. The source material, however, does not provide accuracy figures, processing times or results from a completed demonstration. It would be premature to treat the proposed architecture as proven scientific performance.
The plan’s use of international probe data also raises a practical data-management question. Combining Tianwen-1 observations with U.S. and ESA measurements would require compatible formats, comparable interpretations and careful handling of differences among instruments and missions. The available reporting says those sources would be used together, but it does not spell out a finalized data standard or collaboration agreement.
A related EastFrontier report on Xuannv’s Earth geospatial AI shows why the proposed Mars project fits within a wider interest in geospatial intelligence. The comparison is limited: Xuannv concerns Earth-focused data, while the Mars proposal concerns planetary geology. Still, both examples place AI alongside mapping and structured geographic information rather than treating models as general-purpose chat systems.
Why the 2028 deadline remains an open research question
The deadline gives scientists and mission planners a way to measure progress, but it does not settle the scientific or operational questions between presentation and delivery. The project must still turn multiple measurement types into consistent map layers, encode geological expertise in usable rules and test whether multi-scale feature extraction produces interpretations that specialists can examine.
The intended landing-area map makes the schedule especially consequential. At 1:50,000, local products would demand more detail than the 1:5 million global map. The sources do not say which landing areas would receive priority, how many sites would be covered or how the team would validate local classifications. Those omissions do not invalidate the proposal, but they leave the implementation plan less defined than the headline deadline might suggest.
The international dimension is similarly aspirational. U.S. and ESA probe data are described as inputs, and international acceptance of a Chinese geologic-time framework is identified as an objective. Neither point demonstrates a completed multinational mapping effort or consensus around the framework. Acceptance would be a later outcome of scientific scrutiny, not an automatic result of combining datasets.
For now, the significance lies in the proposed combination of Mars observations and AI-assisted geological reasoning. China is framing a map as a research product with global, landing-area and thematic layers, while assigning large models a role in extracting features under expert-guided rules. Whether the work reaches its end-2028 target, and whether researchers outside China adopt its geological framework, will depend on evidence that the planned pipeline can turn diverse measurements into reproducible interpretations.
A third EastFrontier link to China’s orbital AI plans underscores the broader setting without changing the facts of this project. The Mars map is still a stated goal. Its next test will be whether the proposed data fusion, expert rules and multi-scale analysis produce the detailed atlases and geological framework described by its planners.