Accurate quantification of carbon stocks and their dynamic changes in forest ecosystems is essential for assessing regional carbon sink potential and addressing climate change. However, conventional machine learning models often suffer from reduced extrapolation ability in long-term predictions due to the absence of ecological process constraints. To overcome this limitation, this study established a knowledge-guided machine learning (KGML) framework for natural forests in Heilongjiang Province by integrating fixed-plot observations from 1976-2015 with historical and future climate and CO₂ concentration datasets. The CO₂ fertilization effect was incorporated and optimized into the forest carbon sequestration (FCS) model, and its simulated results were used as prior features input into forest carbon sequestration model (XGBoost) to achieve the fusion of process knowledge with data-driven learning.Results showed that the optimized FCS model significantly improved prediction accuracy (R² increased from 0.85 to above 0.90), while KGML model demonstrated superior simulation accuracy of carbon density acrossvarious forest types compared to machine learning and process models, while also exhibiting greater stability and extrapolation capability in long-term predictions (≥10 years). Scenario simulations indicated that natural forests under SSP126 would reach the highest sequestration potential by 2060, with carbon stocks of 1 379.26-1 439.02 Tg C. This study demonstrates the effectiveness and potential for promotion of knowledge-guided machine learning in forest carbon prediction and provides a new path for improving the accuracy of carbon sink assessment and achieving the dual-carbon goals.
本研究使用Cheng等[34]的研究成果,2015年天然林空间分布数据集(30 m×30 m)和美国国家航空航天局(national aeronautics and space administration,NASA)[35]的MCD12Q1土地分类数据(500 m×500 m),通过掩膜获得了黑龙江省2015年针叶林、阔叶林、混交林3种林型的天然林空间分布数据。为确保遥感数据与样地数据的一致性,本研究以样地信息为准使用空间叠加的方法对两种数据进行了匹配和校验。
ZHANGY R, LIUX T, GAOW Q,et al.Dynamic changes of forest vegetation carbon storage and the characteristics of carbon sink (source) in the Natural Forest Protection Project region for the past 20 years[J].Acta Ecologica Sinica,2021,41(13):5093-5105.
WUM L, SHAOY C, HUL T,et al.Machine learning-based prediction of carbon fluxes in terrestrial ecosystems in China and its response to land use change[J].Environmental Science,2025,46(8):4733-4741.
[6]
ZHANGX M, BRANDTM, YUEY M,et al.The carbon sink potential of Southern China after two decades of afforestation[J].Earth’s Future,2022,10(12):e2022EF002674.
TIANH L, ZHUJ H, HEX,et al.Projected biomass carbon stock of arbor forest of three provinces in Northeastern China based on random forest model[J].Scientia Silvae Sinicae,2022,58(4):40-50.
[9]
ARÉVALOP, BACCINIA, WOODCOCKC E,et al.Continuous mapping of aboveground biomass using Landsat time series[J].Remote Sensing of Environment,2023,288:113483.
[10]
HANSONP C, STILLMANA B, JIAX,et al.Predicting lake surface water phosphorus dynamics using process-guided machine learning[J].Ecological Modelling,2020,430:109136.
[11]
DAW A, KARPATNEA, WATKINSW,et al.Physics-guided neural networks (PGNN):An application in lake temperature modeling[J].arXiv:2017.
[12]
OLDENJ D, LAWLERJ J, POFFN L.Machine learning methods without tears:A primer for ecologists[J].Quarterly Review of Biology,2008,83(2):171-193.
[13]
REICHSTEINM, CAMPS-VALLSG, STEVENSB,et al.Deep learning and process understanding for data-driven Earth system science[J].Nature,2019,566(7743):195-204.
[14]
XUB, YUH F, SHIZ B,et al.Knowledge-guided machine learning reveals pivotal drivers for gas-to-particle conversion of atmospheric nitrate[J].Environmental Science and Ecotechnology,2024,19:100333.
[15]
ABIMBOLAO P, MEYERG E, MITTELSTETA R,et al.Knowledge-guided machine learning for improving daily soil temperature prediction across the United States[J].Vadose Zone Journal,2021,20(5):e20151.
[16]
LIUL C, ZHOUW, GUANK Y,et al.Knowledge-guided machine learning can improve carbon cycle quantification in agroecosystems[J].Nature Communications,2024,15(1):357.
[17]
ZHENGH Y, LINB T, JIANGJ,et al.Knowledge-guided machine learning method for downhole gauge record prediction in deep water gas field[M].Offshore Technology Conference Asia,2024.
[18]
RAISSIM, PERDIKARISP, KARNIADAKISG E.Physics-informed neural networks:A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations[J].Journal of Computational Physics,2019,378:686-707.
FENGJ, SHAOP P, ZHANGJ R,et al.Research review of physics-guided data-driven methods and their application and prospects in hydrological model development[J].Journal of Hohai University(Natural Sciences),2025,53(5):1-11.
LIQ, HUANGC, LIUJ,et al.Research on passive recognition method of ocean acoustic target based on guiding learning model[J].Journal of Ocean Technology,2025,44(3):66-76.
ZHOUY J, JIANGH, CHENX,et al.Scale knowledge guided deep learning for simulating ecosystem carbon fluxes[J].Chinese Science Bulletin,2025,70(17):2791-2804.
[26]
CAIS Z, WANGZ C, WANGS F,et al.Physics-informed neural networks for heat transfer problems[J].Journal of Heat Transfer,2021,143(6):4050542.
[27]
CAIW X, HEN P, LIM X,et al.Carbon sequestration of Chinese forests from 2010 to 2060:Spatiotemporal dynamics and its regulatory strategies[J].Science Bulletin,2022,67(8):836-843.
[28]
CAIW X, XUL, WEND,et al.The carbon sequestration potential of vegetation over the Tibetan Plateau[J].Renewable and Sustainable Energy Reviews,2025,207:114937.
WEND, XUL, WANGC Y,et al.Remote sensing assessment and prediction of forest vegetation carbon sequestration in the Beijing-Tianjin-Hebei urban area[J].Progress in Geography,2024,43(12):2520-2529.
[31]
CHENC, RILEYW J, PRENTICEI C,et al.CO2 fertilization of terrestrial photosynthesis inferred from site to global scales[J].Proceedings of the National Academy of Sciences,2022,119(10):e2115627119.
[32]
WANGS H, ZHANGY G, JUW M,et al.Recent global decline of CO2 fertilization effects on vegetation photosynthesis[J].Science,2020,370(6522):1295-1300.
[33]
HEY, LIUY, LEIL,et al.CO2 fertilization contributed more than half of the observed forest biomass increase in northern extra-tropical land[J].Global Change Biology,2023,29(15):4313-4326.
[34]
YAOY T, CIAISP, JOETZJERE,et al.The impacts of elevated CO2 on forest growth,mortality,and recovery in the Amazon rainforest[J].Earth System Dynamics,2024,15(3):763-778.
DONGL H, LIF R, JIAW W,et al.Compatible biomass models for main tree species with measurement error in Heilongjiang Province of Northeast China[J].Chinese Journal of Applied Ecology,2011,22(10):2653-2661.
[37]
LYUY L, HANN, DUH Q.Estimation of bamboo forest aboveground carbon using the RGLM model based on object-based multiscale segmentation of SPOT-6 imagery[J].Remote Sensing,2023,15(10):2566.
[38]
LIH M, ZHANGG L, ZHONGQ C,et al.Prediction of urban forest aboveground carbon using machine learning based on Landsat 8 and Sentinel-2:A case study of Shanghai,China[J].Remote Sensing,2023,15(1):284.
[39]
WANGT L, WANGG Y, INNESJ L,et al.ClimateAP:An application for dynamic local downscaling of historical and future climate data in Asia Pacific[J].Frontiers of Agricultural Science and Engineering,2017,4(4):448-458.
DUY C, LIM Z, FANW Y,et al.Estimation of forest stand age based on GWR model and forest fire remote sensing data[J].Scientia Silvae Sinicae,2019,55(6):184-194.
WEIG R, LIM Z, QUANY,et al.Geographically weighted random forest approach to predict forest carbon storage by remote sensing in Heilongjiang[J].Journal of Central South University of Forestry & Technology,2024,44(7):64-76.
[44]
CHENGK, SUY J, GUANH C,et al.Mapping China’s planted forests using high resolution imagery and massive amounts of crowdsourced samples[J].ISPRS Journal of Photogrammetry and Remote Sensing,2023,196:356-371.
[45]
FRIEDLM, SULLA-MENASHED.MODIS/terra+aqua land cover type yearly L3 global 500m SIN grid V061[M].NASA Land Processes Distributed Active Archive Center,2022.
LIH K.Research advance of forest carbon sink assessment methods and carbon sequestration potential estimation under carbon neutral vision[J].Geological Survey of China,2021,8(4):79-86.
[50]
CHENT Q, GUESTRINC.XGBoost:A scalable tree boosting system[J].Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.2016:785-94.
CHENZ Y, LIUJ B, LIC,et al.Ultra short-term power load forecasting based on combined LSTM-XGBoost model[J].Power System Technology,2020,44(2):614-620.
[53]
HEN P, WEND, ZHUJ X,et al.Vegetation carbon sequestration in Chinese forests from 2010 to 2050[J].Global Change Biology,2016,23(4):1575-1584.
[54]
KINGA W, POSTW M, WULLSCHLEGERS D.The potential response of terrestrial carbon storage to changes in climate and atmospheric CO2 [J].Climatic Change,1997,35(2):199-227.
[55]
ADAMSB, WHITEA, LENTONT M.An analysis of some diverse approaches to modelling terrestrial net primary productivity[J].Ecological Modelling,2004,177(3/4):353-391.
[56]
NORBYR J, DELUCIAE H, GIELENB,et al.Forest response to elevated CO2 is conserved across a broad range of productivity[J].Proceedings of the National Academy of Sciences,2005,102(50):18052-18056.
[57]
LEAKEYA D B, AINSWORTHE A, BERNACCHIC J,et al.Elevated CO2 effects on plant carbon,nitrogen,and water relations:Six important lessons from FACE[J].Journal of Experimental Botany,2009,60(10):2859-2876.
[58]
AINSWORTHE A, LONGS P.What have we learned from 15 years of free-air CO2 enrichment (FACE)? A meta-analytic review of the responses of photosynthesis,canopy properties and plant production to rising CO2 [J].New Phytologist,2005,165(2):351-371.
[59]
BIANS J, WANGB, LIM,et al.Relationship between CO2 fertilization effects,and stand age,stand type,and site conditions[J].Remote Sensing,2023,15(17):4197.
[60]
VERHULSTM, HEREMANSS, BLASCHKOM B,et al.Temporal transferability of tree species classification in temperate forests with Sentinel-2 time series[J].Remote Sensing,2024,16(14):2653.
[61]
GRAYP C, CHAMORROD F, RIDGEJ T,et al.Temporally generalizable land cover classification:A recurrent convolutional neural network unveils major coastal change through time[J].Remote Sensing,2021,13(19):3953.
HANZ H, LIG C, LIUD D,et al.Carbon storage and carbon sink capacity of major arbor forest types in Heilongjiang Province of Northeastern China[J].Journal of Beijing Forestry University,2024,46(11):10-23.
[64]
ZHUZ, PIAOS, MYNENIR B,et al.Greening of the Earth and its drivers[J].Nature Climate Change,2016,6(8):791-795.
[65]
PEÑUELASJ, CIAISP, CANADELLJ G,et al.Shifting from a fertilization-dominated to a warming-dominated period[J].Nature Ecology & Evolution,2017,1(10):1438-1445.
[66]
YAOY T, PIAOS L, WANGT.Future biomass carbon sequestration capacity of Chinese forests[J].Science Bulletin,2018,63(17):1108-1117.
[67]
CHENGF S, TIANJ X, HEJ Y,et al.China’s future forest carbon sequestration potential under different management scenarios[J].Trees,Forests and People,2024,17:100621.
[68]
YUZ, LIUS R, LIH K,et al.Maximizing carbon sequestration potential in Chinese forests through optimal management[J].Nature Communications,2024,15(1):3154.