Spatial distribution prediction of geothermal gradient in North China Craton driven by the combination of machine learning and stratification modeling

doi:10.13745/j.esf.sf.2025.4.67

Abstract

Abstract:

As a key parameter characterizing the lithospheric thermal state, the spatial distribution of the geothermal gradient is crucial for understanding cratonic thermal evolution and guiding geothermal exploration. Previous studies on the North China Craton (NCC) were largely limited to one-dimensional analyses, failing to resolve vertical variations and limiting model accuracy. This study innovatively develops a depth-stratified prediction model for the NCC geothermal gradient, systematically elucidating its depth-dependent patterns and thermo-tectonic controls. Integrating 573 geothermal gradient measurements with depth information from global databases and literature, data were partitioned into six depth intervals: <500 m, 500-1000 m, 1000-2000 m, 2000-3000 m, 3000-4000 m, and >4000 m. Thirteen geological/geophysical predictors (e.g., Moho depth, geothermal heat flow) were used to train machine learning regression models for each interval. Key results show: (1) Model performance (R²) is depth-dependent: exceeding 0.45 in middle and shallower intervals (<3000 m), but declining significantly at greater depths due to sparse data; (2) Feature importance analysis reveals Moho depth and heat flow dominate deep predictions (weight>40%), while magnetic anomalies and geological age exert minor influence (<10%); (3) The 3D geothermal structure shows systematic depth variations: uppermost high-value zones (>35 ℃/km) align with active fault systems, while mid-deep high-value zones (500-3000 m) migrate eastward, correlating spatially with crustal thinning zones associated with Pacific Plate subduction. This work presents the first 3D geothermal gradient model of the NCC. Results provide critical data for regional geothermal assessment and new constraints on thermo-tectonic coupling during craton destruction. The depth-stratified framework offers a methodological paradigm for thermal studies of analogous tectonic units.

Key words: North China Craton, geothermal gradient, machine learning, geological and geophysical features, stratification modeling

CLC Number:

P314.3
P628

LI Jinming, ZHENG Yang, CHENG Qiuming. Spatial distribution prediction of geothermal gradient in North China Craton driven by the combination of machine learning and stratification modeling[J]. Earth Science Frontiers, 2025, 32(4): 291-302.

Figures/Tables 10

Fig.1 Distribution of sample points in the study area. The background map is the digital topographic map made with GMT6[5], the boundary of the North China Craton (black line) indicates the study area (data from [6]), and the red star indicates a drill hole (data from [7]).

Table 1 Statistical results of thermal gradient sample data in different depth segments

分组	深度/m	样本数量/个	占比
A	0~<500	256	44.68%
B	500~<1 000	231	40.31%
C	1 000~<2 000	208	36.30%
D	2 000~<3 000	150	26.18%
E	3 000~4 000	81	14.14%
F	>4 000	22	3.84%

Fig.2 Histogram of thermal gradient values for each depth segments. The red line represents the normal distribution curve.

Table 2 Geological and geophysical features used in the study and data sources

编号	特征名称	数据来源
1	地质年代	[28]
2	磁异常	[29]
3	地壳厚度	[30]
4	结晶壳厚度	[30]
5	沉积物厚度	[30]
6	岩石圈厚度	[31]
7	均衡重力异常	[32]
8	布格重力异常	[32]
9	地形	[33]
10	自由空气重力异常	[34]
11	居里点深度	[35]
12	距火山距离	[36]
13	大地热流	[12]

Fig.3 Statistical chart of the performance evaluation standard value of the optimal algorithm model corresponding to different depth Segments A: 0~<500 m; B: 500~<1 000 m; C: 1 000~<2 000 m; D: 2 000~<3 000 m; E: 3 000~4 000 m; F: >4 000 m。

Fig.4 Scatter plot of actual measured value and predicted values. The blue dots represent all the observed values of each depth segments, the red line is the straight line fitted by the least square method, and the light red band is the 95% confidence interval.

Fig.5 Statistical line chart of feature importance ranking of each depth segment. a—the variations in the importance of geothermal heat flow and Curie point depth features at different depths; b—the variations in the importance of sediment thickness, crystalline crust thickness and crustal thickness features at different depths; c—the variations in the importance of crustal thickness and isostatic gravity anomaly features at different depths; d—the variations in the importance of bouguer gravity anomaly and free air gravity anomaly features at different depths; e—the variations in the importance of the thickness of lithosphere and topography features at different depths; f—the variations in the importance of isostatic gravity anomaly and the distance to volcano features at different depths.

Fig.6 Predicted geothermal gradient distribution maps of six depth segments. The boundary of the North China Craton (black line) indicates the study area (data from [6]), and the background is the mountain shadow in the mapping software ArcGis pro.

Fig.7 Active faults and 0-500 m geothermal gradient distribution. The boundary of the North China Craton (black line) indicates the study area (data from [6]), the background is the mountain shadow in the mapping software ArcGis pro, and the red line represents the distribution of active faults (data from [48]).

Fig.8 Comparison of measured temperature and predicted temperature curves for the CCSD-1 well (data from CCSDP). The well location is shown in Fig.1. The black line represents the drilling temperature curve, the blue line represents the well temperature curve after the well temperature reaches stability, and the red line represents the temperature-depth variation curve at this point calculated by predicting the geothermal gradient in this study.

References 53

[1]	袁玉松, 米立军, 张功成, 等. 沉积盆地地温梯度研究中应注意的问题[J]. 地质论评, 2009, 55(4): 531-535.
[2]	LUCAZEAU F. Analysis and mapping of an updated terrestrial heat flow data set[J]. Geochemistry, Geophysics, Geosystems, 2019, 20(8): 4001-4024. DOI
[3]	SCLATER J G, HASTEROK D, GOUTORBE B, et al. Marine heat flow[M]// HARFFJ, MESCHEDEM, PETERSENS, et al. Encyclopedia of marine geosciences. Dordrecht: Springer, 2014: 1-16.
[4]	何丽娟, 汪集旸. “大地热流”等地热学重要术语的概念与应用[J]. 中国科技术语, 2021, 23(3): 3-9. DOI
[5]	WESSEL P, LUIS J F, UIEDA L, et al. The generic mapping tools version 6[J]. Geochemistry, Geophysics, Geosystems, 2019, 20(11): 5556-5564. DOI
[6]	潘桂棠, 肖庆辉, 陆松年, 等. 中国大地构造单元划分[J]. 中国地质, 2009, 36(1): 1-16.
[7]	WANG D, ZHANG W, ZHANG X, et al. Casing and well cementation[M]// WANGD, ZHANGW, ZHANGX, et al. The China continental scientific drilling project:CCSD-1 Well drilling engineering and construction. Berlin, Heidelberg: Springer, 2015: 303-331.
[8]	CARLSON R W, PEARSON D G, JAMES D E. Physical, chemical, and chronological characteristics of continental mantle[J]. Reviews of Geophysics, 2005, 43(1): RG1001.
[9]	朱日祥, 陈凌, 吴福元, 等. 华北克拉通破坏的时间、范围与机制[J]. 中国科学: 地球科学, 2011, 41(5): 583-592.
[10]	HU S, HE L, WANG J. Heat flow in the continental area of China: a new data set[J]. Earth and Planetary Science Letters, 2000, 179(2): 407-419.
[11]	JIANG G, HU S, SHI Y, et al. Terrestrial heat flow of continental China: updated dataset and tectonic implications[J]. Tectonophysics, 2019, 753: 36-48.
[12]	ZHANG Y, CHENG Q, HONG T, et al. Ocean singularity analysis and global heat flow prediction reveal anomalous bathymetry and heat flow[J]. Geoscience Frontiers, 2025, 16(3): 102013.
[13]	王亚茹, 蔺文静, 朱传庆, 等. 基于地质年代与热流关系的中国陆区热流分布[J]. 地球科学, 2025, 50(2): 763-781.
[14]	SHAHDI A, LEE S, KARPATNE A, et al. Exploratory analysis of machine learning methods in predicting subsurface temperature and geothermal gradient of Northeastern United States[J]. Geothermal Energy, 2021, 9(1): 18.
[15]	MEJÍA-FRAGOSO J C, FLÓREZ M A, BERNAL-OLAYA R. Predicting the geothermal gradient in Colombia: a machine learning approach[J]. Geothermics, 2024, 122: 103074.
[16]	江晓雪, 朱传庆. 复杂热传递作用下冀中坳陷深、浅部地热资源差异分布[J/OL]. 地球科学, (2024-04-10)[2024-09-18]. https://link.cnki.net/urlid/42.1874.P.20240408.1659.025.
[17]	QIU N, ZUO Y, ZHOU X, et al. Geothermal regime of the Bohai offshore area, Bohai Bay Basin, North China[J]. Energy Exploration & Exploitation, 2010, 28(5): 327-350.
[18]	WANG X, MAO X, MAO X, et al. Characteristics and classification of the geothermal gradient in the Beijing-Tianjin-Hebei Plain, China[J]. Mathematical Geosciences, 2020, 52(6): 783-800.
[19]	HE J, LI K, WANG X, et al. A machine learning methodology for predicting geothermal heat flow in the Bohai Bay Basin, China[J]. Natural Resources Research, 2022, 31(1): 237-260.
[20]	陈国雄, 张越鹏, 罗磊, 等. 数据驱动斑岩型矿床时空预测模型[J]. 地学前缘, 2025, 32(4): 46-59.
[21]	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. DOI
[22]	BELL G, HEY T, SZALAY A. Beyond the data deluge[J]. Science, 2009, 323(5919): 1297-1298.
[23]	周永章, 陈烁, 张旗, 等. 大数据与数学地球科学研究进展: 大数据与数学地球科学专题代序[J]. 岩石学报, 2018, 34(2): 255-263.
[24]	张汝惠, 赵国泽. 河北邢束地区地壳上地幔的热作用[J]. 中国地震, 1993(3): 45-55.
[25]	常健, 邱楠生, 赵贤正, 等. 渤海湾盆地冀中坳陷现今地热特征[J]. 地球物理学报, 2016, 59(3): 1003-1016. DOI
[26]	Global Heat Flow Data Assessment Group. The global heat flow database: release 2024[DB/OL]. (2024-04-17)[2024-07-01]. https://doi.org/10.5880/fidgeo.2024.014.
[27]	QUINTERO W, CAMPOS-ENRÍQUEZ O, HERNÁNDEZ O. Curie point depth, thermal gradient, and heat flow in the Colombian Caribbean (northwestern South America)[J]. Geothermal Energy, 2019, 7(1): 16.
[28]	STEINSHOUER D W, QIANG J, MCCABE P J, et al. Maps showing geology, oil and gas fields, and geologic provinces of the Asia Pacific region[R]. Reston: U.S. Geological Survey, 1999.
[29]	DYMENT J, CHOI Y, LESUR V, et al. The world digital magnetic anomaly map: version 2.1[C]// Proceedings of the 22nd EGU general assembly. Vienna: EGU, 2020: 22098.
[30]	MOONEY W D, BARRERA-LOPEZ C, SUÁREZ M G, et al. Earth crustal model 1 (ECM1): a 1° x 1° global seismic and density model[J]. Earth-Science Reviews, 2023, 243: 104493.
[31]	HOGGARD M J, CZARNOTA K, RICHARDS F D, et al. Global distribution of sediment-hosted metals controlled by craton edge stability[J]. Nature Geoscience, 2020, 13(7): 504-510.
[32]	BALMINO G, VALES N, BONVALOT S, et al. Spherical harmonic modelling to ultra-high degree of bouguer and isostatic anomalies[J]. Journal of Geodesy, 2012, 86(7): 499-520.
[33]	TOZER B, SANDWELL D T, SMITH W H F, et al. Global bathymetry and topography at 15 Arc Sec: SRTM15+[J]. Earth and Space Science, 2019, 6(10): 1847-1864. DOI
[34]	SANDWELL D T, MÜLLER R D, SMITH W H F, et al. New global marine gravity model from CryoSat-2 and Jason-1 reveals buried tectonic structure[J]. Science, 2014, 346(6205): 65-67.
[35]	LI C, LU Y, WANG J. A global reference model of curie-point depths based on EMAG2[J]. Scientific Reports, 2017, 7(1): 45129.
[36]	PROGRAM VOLCANISM PROGRAM. Volcanoes of the world[DB/OL]. (2024-06-06)[2024-07-01]. https://doi.org/10.5479/si.GVP.VOTW5-2024.5.2.
[37]	王子烨, 左仁广. 基于多源数据融合的喜马拉雅淡色花岗岩识别[J]. 地学前缘, 2023, 30(5): 216-226. DOI
[38]	ARTEMIEVA I, MOONEY W. Thermal thickness and evolution of Precambrian lithosphere: a global study[J]. Journal of Geophysical Research: Solid Earth, 2001, 106: 16387-16414.
[39]	DRAPER N R, SMITH H. Applied regression analysis[M]. 3rd ed. Hoboken: John Wiley & Sons, Inc., 1998.
[40]	CORTES C, VAPNIK V. Support-vector networks[J]. Machine Learning, 1995, 20(3): 273-297.
[41]	BOUKERCHE A, TAO Y, SUN P. Artificial intelligence-based vehicular traffic flow prediction methods for supporting intelligent transportation systems[J]. Computer Networks, 2020, 182: 107484.
[42]	KRZYWINSKI M, ALTMAN N. Classification and regression trees[J]. Nature Methods, 2017, 14(8): 757-758.
[43]	SEEGER M. Gaussian processes for machine learning[J]. International Journal of Neural Systems, 2004, 14(2): 69-106. PMID
[44]	KOHAVI R. A study of cross-validation and bootstrap for accuracy estimation and model selection[C]// Proceedings of the 14th international joint conference on artificial intelligence: volume 2. Montreal: Morgan Kaufmann Publishers Inc, 1995: 1137-1143.
[45]	DERBELA M, NOUIRI I. Intelligent approach to predict future groundwater level based on artificial neural networks (ANN)[J]. Euro-Mediterranean Journal for Environmental Integration, 2020, 5(3): 51.
[46]	LÖSING M, EBBING J. Predicting geothermal heat flow in antarctica with a machine learning approach[J]. Journal of Geophysical Research: Solid Earth, 2021, 126(6): e2020JB021499.
[47]	成秋明. 洋中脊动力学与俯冲带地震-岩浆-成矿事件远程效应[J]. 地学前缘, 2024, 31(1): 1-14. DOI
[48]	STYRON R, PAGANI M. The GEM global active faults database[J]. Earthquake Spectra, 2020, 36: 875529302094418.
[49]	马刚, 常恩祥, 周瑞良, 等. 华北平原北部地温场特征及其形成条件的初步探讨[J]. 中国地质科学院562综合大队集刊, 1983: 109-126.
[50]	李彤霞, 王岩, 郭晓燕, 等. 郯庐地震带分段特征与小震序列活动分析[J]. 防灾减灾学报, 2017, 33(2): 78-84.
[51]	汪集旸, 胡圣标, 庞忠和, 等. 中国大陆干热岩地热资源潜力评估[J]. 科技导报, 2012, 30(32): 25-31.
[52]	许天福, 文冬光, 袁益龙. 干热岩地热能开发技术挑战与发展战略[J]. 地球科学, 2024, 49(6): 2131-2147.
[53]	XIONG B, XU H, FANG C, et al. Construction and application of favorable target evaluation system for hot dry rock[J]. Petroleum Exploration and Development, 2025, 52(1): 258-271. DOI