基于机器学习的长江溶解态有机碳输送的长期变化及驱动因素分析

doi:10.3969/j.issn.1000-5641.2026.03.002

摘要/Abstract

摘要：

解析河流溶解态有机碳 (DOC) 浓度及通量变化对于刻画全球碳循环过程和细化全球碳收支计算至关重要. 然而, 中国河流的相关研究受“长时间连续观测样本少”和“控制参数时间分布不统一”等限制, 河流DOC的季节和长期变化规律及其驱动因素的影响机制尚未得到充分认识. 该研究利用长江徐六泾站长时间尺度的逐月DOC浓度实测数据和相关流域特征数据, 对比不同机器学习方法, 基于最佳模型模拟了2001—2020年徐六泾站逐月的DOC浓度变化; 并使用SHAP (SHapley Additive exPlanations) 方法分析了流域特征对DOC浓度和通量的影响. 结果表明, 随机森林算法是最佳选择 (R²=0.72, RMSE=0.09 mg·L⁻¹). 2001—2020年间, 长江徐六泾站DOC浓度范围为1.24~2.27 mg·L⁻¹, 平均为1.67 mg·L⁻¹; 年DOC通量变化范围为0.93~2.41 Tg·a⁻¹, 平均为1.46 Tg·a⁻¹. 研究表明, 徐六泾站DOC浓度的季节性模式已从洪季低、枯季高转变为洪季和枯季均较高, 这种变化可以归因于人为水调节活动和流域生态系统格局的变化. 在长期趋势上, 徐六泾站的DOC浓度和通量均显著上升, 递增速率分别为0.026 mg·L⁻¹·a⁻¹, p<0.05; 0.0025 Tg·a⁻¹, p<0.05. 其中在近十年尺度上, 人类活动的增强解释了54.1%的DOC浓度变化, 是最主要的驱动因素. 本研究将有助于进一步了解过去十几年间长江DOC浓度和通量的变化规律及其驱动因素的影响机制, 并为河流碳循环的大数据分析研究提供新的视角.

关键词: 长江, 溶解态有机碳, 机器学习, 驱动因素

Abstract:

Analyzing variations in riverine dissolved organic carbon (DOC) concentration and flux is essential for understanding global carbon cycle processes and refining carbon budget estimations. Machine learning and big data analysis have become invaluable in this field. However, research on Chinese rivers is limited due to lack of long-term continuous observational data and non-uniform temporal distribution of key influencing factors. Consequently, mechanisms driving seasonal and long-term variations in riverine DOC and their influencing factors remain unclear. This study compared various machine learning methods using long-term, monthly DOC concentration data from the Yangtze River’s Xuliujing Station, as well as watershed characteristic data. Using the optimal model, we simulated monthly DOC concentration changes at Xuliujing Station from 2001 to 2020 and employed the SHAP (SHapley Additive exPlanations) method to analyze the impact of watershed characteristics on DOC concentration and flux. The findings demonstrated that the Random Forest algorithm yielded the highest accuracy, achieving an R² of 0.72 and an RMSE of 0.09 mg·L⁻¹. Over the study period, DOC concentrations ranged from 1.24 to 2.27 mg·L⁻¹, with a mean of 1.67 mg·L⁻¹. Annual DOC flux varied between 0.93 and 2.41 Tg·a⁻¹, averaging 1.46 Tg·a⁻¹. Notably, the seasonal pattern of DOC concentration shifted from low during the flood season and high during the dry season to high levels in both seasons. This shift was primarily due to anthropogenic water regulation activities and changes in watershed ecosystem patterns. Over the long term, both DOC concentration and flux at Xuliujing Station have significantly increased, at rates of 0.026 mg·L⁻¹·a⁻¹ and 0.0025 Tg·a⁻¹, respectively (both p<0.05). Human activities were the predominant driving factor, accounting for 54.1% of the changes in DOC concentration. This study provides valuable insights into the evolving patterns of DOC concentration and flux in the Yangtze River over recent decades and the mechanisms by which driving factors influence these changes. It also provides a novel perspective for the big data analysis of riverine carbon cycling.

Key words: Yangtze River, dissolved organic carbon, machine learning, driving factors

中图分类号:

P734

鲁兴宇, 吴莹, 樊俊宁, 叶雨萍, 王佳. 基于机器学习的长江溶解态有机碳输送的长期变化及驱动因素分析[J]. 华东师范大学学报（自然科学版）, 2026, 2026(3): 17-29, 43.

Xingyu LU, Ying WU, Junning FAN, Yuping YE, Jia WANG. Long-term variations and driving factors of dissolved organic carbon transport in the Yangtze River using machine learning[J]. J* E* C* N* U* N* S*, 2026, 2026(3): 17-29, 43.

图/表 9

图1

表1

表2

图2

图4

图5

图3

图6

图7

参考文献 51

1	FRIEDLINGSTEIN P, O’SULLIVAN M, JONES M W, et al.. Global carbon budget 2023. Earth System Science Data, 2023, 5 (12): 5301- 5369.
2	LUDWIG W, PROBST J L, KEMPE S.. Predicting the oceanic input of organic carbon by continental erosion. Global Biogeochemical Cycles, 1996, 10 (1): 23- 41.
3	REGNIER P, FRIEDLINGSTEIN P, CIAIS P, et al.. Anthropogenic perturbation of the carbon fluxes from land to ocean. Nature Geoscience, 2013, 6, 597- 607.
4	LIU D, JIANG X T, DUAN M W, et al.. Human and natural activities regulate organic matter transport in Chinese rivers. Water Research, 2023, 245, 120622.
5	刘娇娇, 刘军志, 宋超, 等.. 河流碳循环模型研究进展. 地球科学进展, 2024, 39 (2): 111- 123.
6	AITKENHEAD J A, MCDOWELL W H.. Soil C: N ratio as a predictor of annual riverine DOC flux at local and global scales. Global Biogeochemical Cycles, 2000, 14 (1): 127- 138.
7	REN W, TIAN H Q, CAI W J, et al.. Century-long increasing trend and variability of dissolved organic carbon export from the Mississippi River basin driven by natural and anthropogenic forcing. Global Biogeochemical Cycles, 2016, 30 (9): 1288- 1299.
8	BAO H Y, WU Y, ZHANG J.. Spatial and temporal variation of dissolved organic matter in the Changjiang: Fluvial transport and flux estimation. Journal of Geophysical Research: Biogeosciences, 2015, 120 (9): 1870- 1886.
9	LYU S C, YU Q B, WANG F, et al.. A synthetic model to quantify dissolved organic carbon transport in the Changjiang River system: Model structure and spatiotemporal patterns. Journal of Advances in Modeling Earth Systems, 2019, 11 (9): 3024- 3041.
10	MING Y, GAO L, GUO L D.. Dissolved and particulate organic carbon dynamics in the lower Changjiang River on timescales from seasonal to decades: Response to climate and human impacts. Journal of Marine Systems, 2023, 239, 103855.
11	谢莹, 张军, 王子腾, 等.. 1980—2019年人为干扰和绿色投资对长江干流有机碳的影响. 生态学杂志, 2024, 43 (11): 3456- 3469.
12	PARK J H, NAYNA O K, BEGUM M S, et al.. Reviews and syntheses: Anthropogenic perturbations to carbon fluxes in Asian river systems–concepts, emerging trends, and research challenges. Biogeosciences, 2018, 15 (9): 3049- 3069.
13	ZHONG J, LI S L, ZHU X T, et al.. Dynamics and fluxes of dissolved carbon under short-term climate variabilities in headwaters of the Changjiang River, draining the Qinghai-Tibet Plateau. Journal of Hydrology, 2021, 596, 126128.
14	LIU D, PAN D L, BAI Y, et al.. Variation of dissolved organic carbon transported by two Chinese rivers: The Changjiang River and Yellow River. Marine Pollution Bulletin, 2015, 100 (1): 60- 69.
15	王威, 郭庆军, 杜陈军, 等.. 长江流域水环境碳循环研究进展. 生态学杂志, 2023, 42 (3): 736- 747.
16	黄春林, 侯金亮, 李维德, 等.. 深度学习融合遥感大数据的陆地水文数据同化: 进展与关键科学问题. 地球科学进展, 2023, 38 (5): 441- 452.
17	林晶, 吴莹, 张经, 等.. 长江有机碳通量的季节变化及三峡工程对其影响. 中国环境科学, 2007, 27 (2): 246- 249.
18	ZHAO L B, GAO L, GUO L D.. Seasonal variations in molecular size of chromophoric dissolved organic matter from the lower Changjiang (Yangtze) river. Journal of Geophysical Research: Biogeosciences, 2021, 126 (8): e2020JG006160.
19	徐新良. 中国年度NDVI、EVI 1KM数据集[Z/OL]. 北京: 资源环境科学数据注册与出版系统. [2025-03-01]. https://www.resdc.cn/DOI/doi.aspx?DOIid=49.
20	PENG S Z, DING Y X, LIU W Z, et al.. 1 km monthly temperature and precipitation dataset for China from 1901 to 2017. Earth System Science Data, 2019, 11 (4): 1931- 1946.
21	SIMS K, REITH A, BRIGHT E, et al. LandScan global 2022 [Z/OL]. Oak Ridge National Laboratory, 2023. [2025-03-01]. https://doi.org/10.48690/1529167.
22	SUN H, QIAN X, ZHAO Z Y. Monthly gap-filled CCI soil moisture over region of China (Combined Product) [Z/OL]. Science Data Bank, 2023. [2025-03-01]. https://doi.org/10.57760/sciencedb.07849.
23	YANG J, HUANG X.. The 30 m annual land cover dataset and its dynamics in China from 1990 to 2019. Earth System Science Data, 2021, 13 (8): 3907- 3925.
24	GUYON I, ELISSEEFF A.. An introduction to variable and feature selection. Journal of Machine Learning Research, 2003, 3, 1157- 1182.
25	SCHÖLKOPF B, SMOLA A, MÜLLER K R.. Nonlinear component analysis as a kernel eigenvalue problem. Neural Computation, 1998, 10 (5): 1299- 1319.
26	SMOLA A J, SCHÖLKOPF B.. A tutorial on support vector regression. Statistics and Computing, 2004, 14 (3): 199- 222.
27	BREIMAN L.. Random forests. Machine Learning, 2001, 45 (1): 5- 32.
28	CHEN T, GUESTRIN C. XGBoost: A scalable tree boosting system[C]// KDD’16: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: Assoc Computing Machinery, 2016: 785-794.
29	KE G, MENG Q, FINLEY T, et al. LightGBM: A highly efficient gradient boosting decision tree[C]// Neural Information Processing Systems 30. La Jolla, California, 2017.
30	ROGERS T T, MCCLELLAND J L.. Parallel distributed processing at 25: Further explorations in the microstructure of cognition. Cognitive Science, 2014, 38 (6): 1024- 1077.
31	GHAHRAMANI Z.. Probabilistic machine learning and artificial intelligence. Nature, 2015, 521, 452- 459.
32	LUNDBERG S, LEE S I. A unified approach to interpreting model predictions [EB/OL]. (2017-05-22)[2025-02-01]. https://arxiv.org/abs/1705.07874.
33	ZHANG L J, XUE M, LIU Q Z.. Distribution and seasonal variation in the partial pressure of CO₂ during autumn and winter in Jiaozhou Bay, a region of high urbanization. Marine Pollution Bulletin, 2012, 64 (1): 56- 65.
34	YOU Y B, XIE Z H, JIA B H, et al.. Impacts of anthropogenic water regulation on global riverine dissolved organic carbon transport. Earth System Dynamics, 2023, 14 (5): 897- 914.
35	ZHOU Y Q, YAO X L, ZHOU L, et al.. How hydrology and anthropogenic activity influence the molecular composition and export of dissolved organic matter: Observations along a large river continuum. Limnology and Oceanography, 2021, 66 (5): 1730- 1742.
36	LIU Y Y, DENG B, DU J Z, et al.. Nutrient burial and environmental changes in the Yangtze Delta in response to recent river basin human activities. Environmental Pollution, 2019, 249, 225- 235.
37	孔令桥, 张路, 郑华, 等.. 长江流域生态系统格局演变及驱动力. 生态学报, 2018, 38 (3): 741- 749.
38	LAN Z L, ZHAO Y, ZHANG J G, et al.. Long-term vegetation restoration increases deep soil carbon storage in the Northern Loess Plateau. Scientific Reports, 2021, 11 (1): 13758.
39	YAN Y Z, LAUERWALD R, WANG X H, et al.. Increasing riverine export of dissolved organic carbon from China. Global Change Biology, 2023, 29 (17): 5014- 5032.
40	WANG S, FU B J, PIAO S L, et al.. Reduced sediment transport in the Yellow River due to anthropogenic changes. Nature Geoscience, 2016, 9, 38- 41.
41	ZHANG W H, ZHU G F, WAN Q Z, et al.. Influence of irrigation on vertical migration of soil organic carbon in arid area of inland river. Land, 2023, 12 (8): 1532.
42	CLARK J M, LANE S N, CHAPMAN P J, et al.. Export of dissolved organic carbon from an upland peatland during storm events: Implications for flux estimates. Journal of Hydrology, 2007, 347 (3/4): 438- 447.
43	YU Q, LI H, ZHAO Y, et al.. Long-term trends of dissolved organic carbon dynamics in a subtropical forest responding to environmental changes. Environmental Science & Technology, 2024, 58 (28): 12420- 12429.
44	LIU D, TIAN L Q, JIANG X T, et al.. Human activities changed organic carbon transport in Chinese rivers during 2004—2018. Water Research, 2022, 222, 118872.
45	LEVSHINA S I.. Dissolved and suspended organic matter in the Amur and Songhua River water. Water Resources, 2008, 35 (6): 716- 724.
46	KRAUSE A, PAPASTEFANOU P, GREGOR K, et al.. Quantifying the impacts of land cover change on gross primary productivity globally. Scientific Reports, 2022, 12 (1): 18398.
47	吴健敏, 郗敏, 孔范龙, 等.. 土壤溶解性有机碳(DOC)动态变化影响因素研究进展. 地质论评, 2013, 59 (5): 953- 961.
48	梁红霞, 陈富荣, 陈永宁, 等.. 安徽省怀远-灵璧地区表层土壤碳储量研究. 安徽大学学报(自然科学版), 2014, 38 (6): 96- 103.
49	王敏, 张龙军, 桂祖胜.. 长江干流有机碳的时空输运特征及三峡工程对其影响. 中国海洋大学学报(自然科学版), 2011, 41 (S1): 117- 124.
50	LIU D, BAI Y, HE X Q, et al.. Changes in riverine organic carbon input to the ocean from mainland China over the past 60 years. Environment International, 2020, 134, 105258.
51	ITTEKKOT V, ARAIN R.. Nature of particulate organic matter in the River Indus, Pakistan. Geochimica et Cosmochimica Acta, 1986, 50 (8): 1643- 1653.

月度变量				年度变量
核主成分	特征值	解释率/%	累计解释率/%	核主成分	特征值	解释率/%	累计解释率/%
1	31.56	50.50	50.50	1	35.65	46.40	46.40
2	14.53	23.30	73.80	2	22.80	29.60	76.00
3	4.63	7.42	81.22	3	10.38	13.50	89.50
4	4.05	6.48	87.70	4	5.16	6.70	96.20
5	3.25	5.21	92.91	5	2.92	3.80	100
6	2.58	4.13	97.00
7	1.86	2.98	100

变量		不同核主成分下的载荷
变量		1	2	3	4	5	6	7
月度变量	Temp	0.940	0.035	0.125	−0.187	−0.202	−0.052	0.057
	NDVI	0.907	0.053	0.293	0.101	−0.129	−0.066	0.050
	SM	0.901	0.075	0.326*	0.155	0.122	−0.033	−0.029
	PET	0.931	−0.045	0.079	−0.305	−0.111	−0.046	0.033
	GPP	0.958*	0.036	0.153	−0.150	−0.088	−0.041	−0.020
	Precip	0.888	−0.142	0.086	−0.258	0.293	0.028	0.040
	Q	0.847	−0.188*	0.316	−0.043	0.001	0.287	−0.067
年度变量	Grassland	−0.894	−0.037	−0.289	0.079	0.024
	Forest	0.806	0.291*	0.456	−0.136	0.023
	Cropland	−0.807	0.010	−0.563*	0.032	0.158
	PD	0.941*	0.033	0.306	0.045	−0.110
	Shrub	−0.922	−0.068	−0.130	−0.032	0.195

[1]	钟鹏权, 王维, 张嘉伟, 李徐, 周子清, 王苗勋, 许媛. 长江口滨海湿地生态修复工程对底栖纤毛虫群落的影响[J]. 华东师范大学学报（自然科学版）, 2026, 2026(3): 110-123.
[2]	裘诚. 杭州湾北岸登陆台风对长江口咸潮入侵的影响——以2024年9月台风“贝碧嘉”“普拉桑”为例[J]. 华东师范大学学报（自然科学版）, 2025, 2025(6): 141-151.
[3]	傅宇飞, 徐泽松, 冯安昊, 魏同权. 基于流量感知的移动网络授权信息在线预测方法[J]. 华东师范大学学报（自然科学版）, 2025, 2025(4): 1-14.
[4]	叶可, 田波, 王玉纯. 基于Sentinel-2卫星遥感的河口潮滩植被指数构建研究[J]. 华东师范大学学报（自然科学版）, 2025, 2025(2): 141-153.
[5]	田丰, 袁琳, 张利权, 李阳, 班璐, 陈平伢. 长江口海三棱藨草种子产量与休眠萌发特性研究[J]. 华东师范大学学报（自然科学版）, 2025, 2025(2): 68-76.
[6]	王余意, 黄晔, 杨静, 丁方方, 何天豪, 李雨珊, 黄琳, 李晔, 刘敏. 长江口及其近岸海域抗生素的时空分异特征与生态风险评估[J]. 华东师范大学学报（自然科学版）, 2024, 2024(6): 136-150.
[7]	陈灏, 何贤强, 李润, 曹芳. 基于机器学习的长江口表层水体溶解有机碳遥感反演研究[J]. 华东师范大学学报（自然科学版）, 2024, 2024(4): 123-136.
[8]	张勇, 王慧, 朱传华, 周浩, 詹宇, 李灿, 肖逸凡, 杨丽丽, 刘佳奇. 基于机器学习的卫星遥感水质富营养化评价——以合肥市环城河为例[J]. 华东师范大学学报（自然科学版）, 2024, 2024(1): 1-8, 112.
[9]	陈家明, 王世明, 杨荣荣, 陈子彦, 梁霞, 侯立军. 长江河口水体暗碳固定的温度适应性[J]. 华东师范大学学报（自然科学版）, 2024, 2024(1): 104-112.
[10]	金智, 朱建荣, 仇威. 长江流域梯级水库夏末秋初蓄水对河口盐水入侵和淡水资源的影响[J]. 华东师范大学学报（自然科学版）, 2024, 2024(1): 90-103.
[11]	高祥云, 孟丹, 罗明凯, 王俊, 张丽平, 孔超. 支持隐私保护的端云协同训练[J]. 华东师范大学学报（自然科学版）, 2023, 2023(5): 77-89.
[12]	仇威, 朱建荣. 持续强北风天气下长江口盐水入侵对径流量的响应[J]. 华东师范大学学报（自然科学版）, 2023, 2023(3): 132-146.
[13]	王晓婷, 栾华龙, 张卫国, LAMNina Siu-Ngan. 三角洲海岸土地利用侵蚀暴露度分析及其空间规划意义——以上海浦东新区为例[J]. 华东师范大学学报（自然科学版）, 2023, 2023(3): 181-193.
[14]	游智湧, 刘博林, 刘程, 高灯州. 长江口沉积物固氮过程的温度敏感性及影响因素[J]. 华东师范大学学报（自然科学版）, 2022, 2022(3): 101-108.
[15]	李志鹏, 朱建荣. 2007—2016年北支河势变化对长江口盐水入侵影响数值研究[J]. 华东师范大学学报（自然科学版）, 2022, 2022(3): 109-124.