基础模型时代的地球科学:AlphaEarth如何重塑定量地学

doi:10.13745/j.esf.sf.2025.9.68

摘要/Abstract

摘要：

进入21世纪以来,随着大数据和人工智能技术的快速发展与深度应用,地球科学正在经历一场深刻的变革,其特征是研究范式由定性描述转向定量分析,由对自然现象的观察转向对内在机理的深入揭示,由局部区域的研究扩展至全球尺度的综合视角,由基于经验的推断发展为依托数据与模型的智能预测。AlphaEarth Foundations(AEF)作为新一代地球空间(Geo)智能平台,通过构建统一的64维共享嵌入空间,首次实现了光学、雷达、激光等12类对地观测数据的标准化表征与无缝融合,其数据同化效率显著提高,解决了长期困扰地球科学研究的“数据孤岛”问题,同时也正促进重塑地球科学研究的范式与方法论体系,特别是在定量地球科学领域实现突破性进展。本文系统论述AEF如何通过其创新的多源数据融合架构、高维特征标注学习和可扩展计算框架,对推动数据驱动的地球科学研究进入智能化、精准化和实时化发挥作用,同时,通过资源环境领域的应用实例表明AEF的潜在应用前景和创新需求。研究表明,AEF不仅显著提升了传统地球科学问题的解决效率,更催生了多个全新的地学研究方向和方法论体系。

关键词: 大模型, 人工智能, 矿产预测, AlphaEarth, 知识图谱, 深部与覆盖区找矿

Abstract:

Since the beginning of the 21st century, advances in big data and artificial intelligence have driven a paradigm shift in the geosciences, moving the field from qualitative descriptions toward quantitative analysis, from observing phenomena to uncovering underlying mechanisms, from regional-scale investigations to global perspectives, and from experience-based inference toward data- and model-enabled intelligent prediction. AlphaEarth Foundations (AEF) is a next-generation geospatial intelligence platform that addresses these changes by introducing a unified 64-dimensional shared embedding space, enabling—for the first time—standardized representation and seamless integration of 12 distinct types of Earth observation data, including optical, radar, and lidar. This framework significantly improves data assimilation efficiency and resolves the persistent problem of “data silos” in geoscience research. AEF is helping redefine research methodologies and fostering breakthroughs, particularly in quantitative Earth system science. This paper systematically examines how AEF’s innovative architecture—featuring multi-source data fusion, high-dimensional feature representation learning, and a scalable computational framework—facilitates intelligent, precise, and real-time data-driven geoscientific research. Using case studies from resource and environmental applications, we demonstrate AEF’s broad potential and identify emerging innovation needs. Our findings show that AEF not only enhances the efficiency of solving traditional geoscientific problems but also stimulates novel research directions and methodological approaches.

Key words: large-scale models, artificial intelligence, mineral prospectivity mapping, AlphaEarth, knowledge graphs, deep and covered mineral exploration

中图分类号:

P628
TP18

成秋明, 杨一琳, 周远志, 张渊智. 基础模型时代的地球科学:AlphaEarth如何重塑定量地学[J]. 地学前缘, 2025, 32(5): 1-11.

CHENG Qiuming, YANG Yilin, ZHOU Yuanzhi, ZHANG Yuanzhi. Earth science in the era of foundation models: How AlphaEarth is reshaping quantitative geoscience[J]. Earth Science Frontiers, 2025, 32(5): 1-11.

图/表 7

图1 AEF以及开展AEF+示意图 (A)—AEF嵌入向量矩阵;(B)—AEF+拓展地质(G)、地球物理(Gy)、地球化学(Gc)嵌入。

Fig.1 Schematic diagram of AEF and the proposed AEF+ framework. (A) AEF embedding vector matrix; (B) AEF+ extension incorporating geological (G), geophysical (Gy), and geochemical (Gc) embeddings.

图2 集宁浅覆盖区岩性分类图及预测结果 (A)—实际测量的岩性分类图;(B)—基于AEF数据集和野外验证岩性记录点,用随机森林方法进行集宁浅覆盖区岩性分类结果。(A)和(B)中的图例颜色相同。(Q:第四纪;N1h:汉诺坝玄武岩;N2b:新近纪宝格达乌拉组;Ewl:古近纪乌兰戈楚组;γ:燕山期花岗岩;Pt1j:古元古代集宁群;Ar3x:新太古代兴和群)(据文献[25]修改)

Fig.2 Lithologic classification map and prediction results for the Jining Area. (A) A geology map; (B) Lithologic classification results for the Jining area obtained using the random forest method based on the AEF dataset and field-validated lithology points. The legend colors are consistent between (A) and (B). Modified after [25].

图3 融合地球化学数据与AEF数据预测岩性分布 (A)—1∶5万比例尺水系沉积物地球化学数据,利用主成分分析方法计算的元素组合预测得到的玄武岩分布(据文献[25]修改);(B)—融合地球化学数据与AEF数据的岩性识别结果。

Fig.3 Lithology distribution prediction through integration of geochemical and AEF data. (A) Stream sediment geochemical data at 1∶50000 scale, showing element associations identified by principal component analysis. (B) Lithology identification results from fused geochemical and AEF data.

图4 西藏多龙矿集区圈定矿化蚀变带分布(据文献[28]修改) (A)—ASTER影像提取剥蚀强度指数分布图;(B)—AEF数据随机森林回归剥蚀强度指数分布图。

Fig.4 Delineation of mineralized alteration zones in the Duolong ore district, Tibet. Modified after [28]. (A) Erosion Index distribution map extracted from ASTER imagery; (B) Erosion Index distribution map derived from random forest regression using AEF data.

图5 ASTER影像提取蚀变分类分布图(A)和基于AEF数据的随机森林多回归模型所提取的蚀变矿物信息分布图(B)

Fig.5 Alteration mineral classification map derived from ASTER imagery (A) and alteration mineral information distribution map extracted using a random forest multi-regression model based on AEF data (B).

图6 在菲律宾Mayon火山周围采用AEF数据计算的2018年相对于2017年的地表综合变化强度图(Δθ,单位°)

Fig.6 Comprehensive surface change intensity map (Δθ, unit: degrees) around Mayon Volcano, Philippines, for 2018 relative to 2017, calculated using AEF data.

图7 分别采用Landsat 8和AEF数据估算珠江三角洲河流泥沙浓度 (A)—Landsat 8 数据计算结果;(B)—AEF数据计算结果。

Fig.7 Estimation of river suspended sediment concentration in the Pearl River Delta using Landsat 8 and AEF data, respectively. (A) Results derived from Landsat 8 data; (B) Results derived from AEF data.

参考文献 46

[1]	成秋明. 什么是数学地球科学及其前沿领域?[J]. 地学前缘, 2021, 28(3): 6-25.
[2]	CHENG Q M. Mathematical geosciences: local singularity analysis of nonlinear Earth systems[C]// DAYA SAGAR B, CHENG Q M, AGTERBERG F. Handbook of mathematical geosciences. Cham: Springer, 2018: 179-208.
[3]	JORDAN M I, MITCHELL T M. Machine learning: trends, perspectives, and prospects[J]. Science, 2015, 349(6245): 255-260.
[4]	KARPATNE A, ATLURI G, FAGHMOUS J H, et al. Theory-guided data science: a new paradigm for scientific discovery[J]. IEEE TKDE, 2017, 29(10): 2318-2331.
[5]	HINZE W J, VON FRESE R R B, SAAD A H. Gravity and magnetic exploration:principles, practices, and applications[M]. Cambridge: Cambridge University Press, 2013.
[6]	DENTITH M, MUDGE S T. Geophysics for the mineral exploration geoscientist[M]. Cambridge: Cambridge University. Press, 2014.
[7]	ANDERSON S, GISLASON S. Geochemistry of the Earth’s surface[J]. Applied Geochemistry, 2011, 26: S1-S2.
[8]	SMITH D B, CANNON W F, WOODRUFF L G, et al. Geochemical and mineralogical data for soils of the conterminous United States[R]. Reston: US Geological Survey Data Series, 2013, 801: 19.
[9]	WULDER M A, MASEK J G, COHEN W B, et al. Opening the archive: How free data has enabled the science and monitoring promise of Landsat[J]. Remote Sensing of Environment, 2012, 122: 2-10.
[10]	DRUSCH M, DEL BELLO U, CARLIER S, et al. Sentinel-2: ESA’s optical high-resolution mission for GMES operational services[J]. Remote Sensing of Environment, 2012, 120(15): 25-36.
[11]	GONG J, CHENG P, WANG Y. Three-dimensional modeling and application in geological exploration engineering[J]. Computers & Geosciences, 2004, 30(4): 391-404.
[12]	BROWN C F, KAZMIERSKI M R, PASQUARELLA V J, et al. AlphaEarth Foundations: an embedding field model for accurate and efficient global mapping from sparse label data[EB/OL]. 2025. [2025-09-17]. https://doi.org/10.48550/arXiv.2507.22291.
[13]	REICHSTEIN M, CAMPS-VALLS G, STEVENS B, et al. Deep learning and process understanding for data-driven Earth system science[J]. Nature, 2019, 566(7743): 195-204.
[14]	CHENG Q, AGTERBERG F P. Fuzzy weights of evidence method and its application in mineral potential mapping[J]. Natural Resources Research, 1999, 8(1): 27-35.
[15]	RODRIGUEZ-GALIANO V, SANCHEZ-CASTILLO M, CHICA-OLMO M, et al. Machine learning predictive models for mineral prospectivity: An evaluation of neural networks, random forest, regression trees and support vector machines[J]. Ore Geology Reviews, 2015, 71: 804-818.
[16]	KIM Y, NAKATA N. Geophysical inversion versus machine learning in inverse problems[J]. The Leading Edge, 2018, 37: 894-901.
[17]	BANERJEE S, CARLIN B P, GELFAND A E. Hierarchical Modeling and analysis for spatial data[M]. Boca Raton: CRC Press, 2014.
[18]	LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. Nature, 2015, 521(7553): 436-444.
[19]	GOODCHILD M F, LI L. Replication across space and time must be weak in the social and environmental sciences[J]. PNAS, 2021, 118(35): e2015759118.
[20]	CRESSIE N, WIKLE C K. Statistics for spatio-temporal data[M]. New York: Wiley, 2011.
[21]	ZHAO T, WANG S, OUYANG C J, et al. Artificial intelligence for geoscience: progress, challenges, and perspectives[J]. The Innovation, 2024, 5(5): 100691.
[22]	WILKINSON M D, DUMONTIER M, AALBERSBERG I J J, et al. The FAIR guiding principles for scientific data management and stewardship[J]. Scientific Data, 2016, 3(1): 1-9.
[23]	DRURY S A. Image interpretation in geology[M]. Sydney: Allen & Unwin, 1987.
[24]	HUNT G R. Spectral signatures of particulate minerals in the visible and near infrared[J]. Geophysics, 1977, 42(3): 501-513.
[25]	GE Y, ZHANG Z, CHENG Q, et al. Geological mapping of basalt using stream sediment geochemical data: case study of covered areas in Jining, Inner Mongolia, China[J]. Journal of Geochemical Exploration, 2022, 232: 106888.
[26]	AGTERBERG F P. Geomathematics[M]. Amsterdam: Elsevier, 1974.
[27]	CARRANZA E J M. Geochemical anomaly and mineral prospectivity mapping in GIS[M]//Handbook of exploration and environmental geochemistry. Amsterdam: Elsevier, 2008.
[28]	FU H Z, CHENG Q M, JING L H, et al. Deep learning-based hydrothermal alteration mapping using GaoFen-5 hyperspectral data in the Duolong Ore District, Western Tibet, China[J]. Journal of Applied Remote Sensing, 2021, 15(4): 044512.
[29]	SINGH A. Digital change detection techniques using remotely-sensed data[J]. International Journal of Remote Sensing, 1989, 10(6): 989-1003.
[30]	JOYCE K E, BELLISS S E, SAMSONOV S V, et al. A review of the status of satellite remote sensing and image processing techniques for mapping natural hazards and disasters[J]. Progress in Physical Geography, 2009, 33(2): 183-207.
[31]	GOETZ A F H, VANE G, SOLOMON J E, et al. Imaging spectrometry for Earth remote sensing[J]. Science, 1985, 228(4704): 1147-1153.
[32]	GILLESPIE A R, KAHLE A B, WALKER R E. Color enhancement of highly correlated images. I. Decorrelation and HSI contrast stretches[J]. Remote Sensing of Environment, 1986, 20(3): 209-235.
[33]	CRACKNELL A P. Synergy in remote sensing: what’s in a pixel?[J]. International Journal of Remote Sensing, 1998, 19(11): 2025-2047.
[34]	FU H, KERESZTURI G, CHENG Q, et al. Deciphering differential exhumation in the Gangdese orogen in southern Tibet using exposed porphyry alteration systems and geomorphic analysis[J]. Geological Society of America Bulletin, 2024, 136(9/10): 3795-3809.
[35]	DEEPMIND. AlphaEarth Foundations helps map our planet in unprecedented detail[EB/OL]. 2025a. [2025-9-15]. https://deepmind.google/discover/blog/alphaearth-foundations-helps-map-our-planet-in-unprecedented-detail/.
[36]	ASIF R. Meet AlphaEarth Foundations: Google DeepMind’s so called ‘virtual satellite’ in AI-driven planetary mapping[EB/OL]. 2025. [2025-09-15]. https://www.marktechpost.com/2025/07/31/meet-alphaearth-foundations.
[37]	DEEPMIND. Dataset release:satellite embedding V1 annual (2017-2024)[EB/OL]. Google Earth Engine. 2025. [2025-09-15]. https://developers.google.com/earth-engine/datasets/catalog/GOOGLE_SATELLITE_EMBEDDING_V1_ANNUAL.
[38]	PASQUARELLA V, SCHECHTER E. AI-powered pixels: introducing Google’s satellite embedding dataset[EB/OL]. (2025-07-31)[2025-09-18]. https://medium.com/google-earth/aipowered-pixels-introducing-googles-atellite-embeddingdataset-31744c1f4650
[39]	LUCÍA G D. From imagery to insight: Google AlphaEarth Foundations in CARTO[EB/OL]. 2025. [2025-09-01]. https://carto.com/blog/google-alphaearth-foundations-in-carto.
[40]	BOONE A. Google’s newest AI acts like a satellite to track climate change[EB/OL]. 2025. [2025-09-17]. https://www.wired.com/story/googles-newest-ai-model-acts-like-a-satellite-to-track-climate-change.
[41]	MICHAEL N. Google DeepMind says its new AI can map the entire planet with unprecedented accuracy[EB/OL]. 2025. [2025-09-17]. https://venturebeat.com/ai/google-deepmind-says-its-new-ai-can-map-the-entire-planet-with-unprecedented-accuracy.
[42]	HOURIEZ L, PILARSKI S, VAHEDI B, et al. Scalable geospatial data generation using AlphaEarth Foundations model[EB/OL]. 2025. [2025-09-10]. https://arxiv.org/abs/2508.11739.
[43]	CHENG Q, OBERHÄNSLI R, ZHAO M. A new international initiative for facilitating data-driven Earth science transformation[J]. Geological Society, London, Special Publications, 2020, 499(1): 225-240.
[44]	CHENG Q. IUGS’ initiative on data-driven geoscience discovery[J]. Journal of Earth Science, 2021, 32(2): 468-470.
[45]	OBERHNSLI R. Deep-time Digital Earth (DDE) the First IUGS Big Science Program[J]. Journal of the Geological Society of India, 2020, 95(3): 223-226.
[46]	STEPHENSON M H, CHENG Q, WANG C, et al. Progress towards the establishment of the IUGS Deep-time Digital Earth (DDE) programme[J]. Episodes, 2020, 43: 1057-1062.