基于多时相遥感物候变量的平缓区土壤有机碳数字土壤制图

doi:10.13287/j.1001-9332.202607.040

摘要/Abstract

摘要： 在地形平缓区域,由于地势起伏较小,传统地形因子对土壤有机碳(SOC)空间变异的解释能力有限,制约了数字土壤制图的精度。本研究以河南省封丘县(黄河冲积平原区)为研究区,引入气候、多时相遥感影像提取物候变量等环境变量,包括不同作物生育阶段归一化植被指数(NDVI)、增强植被指数、土壤调整植被指数及归一化水体指数(NDWI)的峰值、均值等生长趋势参数,采用Boruta算法对环境协变量进行筛选,基于2023年采集的136个样点数据,构建随机森林、随机森林回归-克里格(RFRK)、极端梯度提升等SOC数字模型,通过决定系数(R²)、平均误差、均方根误差和一致性相关系数(CCC)对比分析不同模型及变量组合下的模拟性能以确定最优数字模型,并对2013和2023年SOC进行模拟。结果表明:与水分条件相关的环境因子(如欧几里得距离)、基于NDWI提取的多时相物候变量及NDVI在SOC数字土壤制图中占主导地位。RFRK模型在整体精度(R²为0.45)、一致性(CCC为64.4%)及空间连续性方面为本研究最优模型,且引入多时相物候变量可有效提升平缓区的SOC模拟结果(R²提升0.08,CCC提升14.0%)。RFRK模型模拟表明,2013—2023年间,研究区SOC整体呈增长趋势,其中西北部区域增幅最大,最大增幅为80.6%,但不同区域变化幅度存在差异。综上,在平缓区引入多时相物候变量有助于提高SOC数字土壤制图精度,为平缓区SOC精细化制图提供了方法参考。

关键词: 土壤有机碳, 数字土壤制图, 平缓区, 物候变量

Abstract: The explanatory power of traditional topographic factors on the spatial variability of soil organic carbon (SOC) is limited due to the small topographic relief in flat areas, which restricts the accuracy in digital soil mapping. Based on data collected from 136 sampling points in Fengqiu County in Henan Province (located in the alluvial plain of the Yellow River) in 2023, we constructed random forest, random forest regression-kriging (RFRK), extreme gradient boosting, and other SOC digital models. Within those models, we introduced environmental variables such as climate and multi-temporal remote sensing image-derived phenological variables, including growth trend parameters such as the peak and mean values of normalized difference vegetation index (NDVI), enhanced vegetation index, soil-adjusted vegetation index, and normalized difference water index (NDWI) during different crop growth stages. We used the Boruta algorithm to screen environmental covariates. The simulation performance of different models and variable combinations was compared and analyzed using the coefficient of determination (R²), mean error, root mean square error, and consistency correlation coefficient (CCC) to determine the optimal digital model, and SOC was simulated for the years 2013 and 2023. The results showed that environmental factors related to water conditions (such as Euclidean distance), multi-temporal phenological variables derived from NDWI, and NDVI played dominant roles in SOC digital soil mapping. The RFRK model was the optimal model in terms of overall accuracy (R² of 0.45), consistency (CCC of 64.4%), and spatial continuity. The introduction of multi-temporal phenological variables could effectively improve SOC simulation results in flat areas (R² increased by 0.08, CCC increased by 14.0%). RFRK model simulation results showed that from 2013 to 2023, SOC exhibited an overall increasing trend. There were differences in the magnitude of change across different regions, with the largest increase in the northwest region, reaching 80.6%. In summary, introducing multi-temporal phenological variables in flat areas could improve the accuracy of SOC digital soil mapping and provide a methodological reference for refined mapping of SOC in flat areas.

Key words: soil organic carbon, digital soil mapping, flat area, phenological variable

黄炜杰, 程金凯, 冯永康, 王徳彩. 基于多时相遥感物候变量的平缓区土壤有机碳数字土壤制图[J]. 应用生态学报, 2026, 37(7): 2222-2232.

HUANG Weijie, CHENG Jinkai, FENG Yongkang, WANG Decai. Digital soil mapping of soil organic carbon in flat areas based on multi-temporal remote sensing phenological variables[J]. Chinese Journal of Applied Ecology, 2026, 37(7): 2222-2232.

参考文献

[1] Siewert MB. High-resolution digital mapping of soil organic carbon in permafrost terrain using machine lear-ning: A case study in a sub-Arctic peatland environment. Biogeosciences, 2018, 15: 1663-1682
[2] Nussbaum M, Spiess K, Baltensweiler A, et al. Evaluation of digital soil mapping approaches with large sets of environmental covariates. Soil, 2018, 4: 1-22
[3] Minasny B, Finke P, Stockmann U, et al. Resolving the integral connection between pedogenesis and landscape evolution. Earth-Science Reviews, 2015, 150: 102-120
[4] Richardson AD, Hollinger DY, Dail DB, et al. Influence of spring phenology on seasonal and annual carbon balance in two contrasting New England forests. Tree Physiology, 2009, 29: 321-331
[5] Qiu SY, Wang ZY, Xu JL, et al. Influence of vegetation dynamics on soil organic carbon and its fractions in a coastal wetland. Ecosystem Health and Sustainability, 2023, 9: 16
[6] Zhang L, Cai YY, Huang HL, et al. A CNN-LSTM model for soil organic carbon content prediction with long time series of MODIS-based phenological variables. Remote Sensing, 2022, 14: 4441
[7] Liu XY, Wang J, Song XD. Improving the spatial prediction of soil organic carbon content using phenological factors: A case study in the middle and upper reaches of Heihe River basin, China. Remote Sensing, 2023, 15: 1847
[8] Xiao X, He QJ, Ma S, et al. Environmental variables improve the accuracy of remote sensing estimation of soil organic carbon content. Scientific Reports, 2024, 14: 18964
[9] Forkuor G, Hounkpatin OKL, Welp G, et al. Time-series remotely sensed data improves digital soil mapping. Remote Sensing of Environment, 2020, 251: 112091
[10] Yan K, Wang DC, Feng YK, et al. Digital mapping of soil organic carbon in a plain area based on time-series features. Ecological Indicators, 2025, 171: 113215
[11] Keskin H, Grunwald S. Regression Kriging as a workhorse in the digital soil mapper's toolbox. Geoderma, 2018, 326: 22-41
[12] 王晓峰, 章玥, 周潮伟, 等. 基于可解释机器学习的秦巴山区森林土壤有机碳动态及成因分析. 生态学报, 2026, 46(3): 1193-1207
[13] 彭守璋. 中国1 km分辨率逐月平均气温数据集(1901—2024). 国家青藏高原科学数据中心.(2025-07-02) [2026-01-26]. https://doi.org/10.11888/Meteoro.tpdc.270961
[14] 彭守璋. 中国1 km分辨率逐月降水量数据集(1901—2024). 国家青藏高原科学数据中心. (2025-07-02) [2026-01-26]. https://doi.org/10.5281/zenodo.3114194
[15] Zandi Baghche-Maryam M, Sheklabadi M, Ayoubi S. Covariate selection approaches in spatial prediction of soil quality indices using machine learning models at the watershed scale, west of Iran. Soil and Tillage Research, 2025, 252: 106571
[16] Galvão RKH, Araujo MCU, José GE, et al. A method for calibration and validation subset partitioning. Talanta, 2005, 67: 736-740
[17] Isaaks EH, Srivastava RM. An Introduction to Applied Geostatistics. New York: Oxford University Press, 1989: 1-561
[18] Goovaerts P. Geostatistics for Natural Resources Evaluation. New York: Oxford University Press, 1997: 1-483
[19] Breiman L. Random forests. Machine Learning, 2001, 45: 5-32
[20] 李顿, 王雪梅, 李坤玉, 等. 基于合成影像和多变量的博斯腾湖流域土壤有机碳含量估测. 环境科学, 2025, 46(7): 4428-4440
[21] Hengl T, Nussbaum M, Wright MN, et al. Random forest as a generic framework for predictive modeling of spatial and spatio-temporal variables. PeerJ, 2018, 6: e5518
[22] Chen TQ, Guestrin C. XGBoost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, 2016: 785-794
[23] Amankulova K, Farmonov N, Omonov K, et al. Integrating the Sentinel-1, Sentinel-2 and topographic data into soybean yield modelling using machine learning. Advances in Space Research, 2024, 73: 4052-4066
[24] Han GZ, Zhang GL, Gong ZT, et al. Pedotransfer functions for estimating soil bulk density in China. Soil Science, 2012, 177: 158-164
[25] Lininger KB, Wohl E, Rose JR. Geomorphic controls on floodplain soil organic carbon in the Yukon flats, interior Alaska, from reach to river basin scales. Water Resources Research, 2018, 54: 1934-1951
[26] Wu ZH, Chen YY, Zhu YL, et al. Mapping soil organic carbon in floodplain farmland: Implications of effective range of environmental variables. Land, 2023, 12: 1198
[27] Wang LP, Liu HJ, Wang X, et al. Identifying optimal variables to predict soil organic carbon in sandy, saline, and black soil regions: Remote sensing, terrain, or climate factors? Remote Sensing, 2025, 17: 237
[28] Yang L, He XL, Shen FX, et al. Improving prediction of soil organic carbon content in croplands using phenological parameters extracted from NDVI time series data. Soil and Tillage Research, 2020, 196: 104465
[29] He XL, Yang L, Li AQ, et al. Soil organic carbon prediction using phenological parameters and remote sen-sing variables generated from Sentinel-2 images. Catena, 2021, 205: 105442
[30] He YH, Yang YY, Xu DG, et al. A prediction model of soil organic carbon into river and its driving mechanism in red soil region. Scientific Reports, 2025, 15: 4889
[31] Ho VH, Morita H, Bachofer F, et al. Random forest regression Kriging modeling for soil organic carbon density estimation using multi-source environmental data in central Vietnamese forests. Modeling Earth Systems and Environment, 2024, 10: 7137-7158
[32] Farooq I, Bangroo SA, Bashir O, et al. Comparison of random forest and Kriging models for soil organic carbon mapping in the Himalayan Region of Kashmir. Land, 2022, 11: 2180
[33] Silatsa FBT, Yemefack M, Tabi FO, et al. Assessing countrywide soil organic carbon stock using hybrid machine learning modelling and legacy soil data in Cameroon. Geoderma, 2020, 367: 114260
[34] 姚煜航, 欧铭, 李敏, 等. 基于机器学习的淮河流域水文干旱预测模型的构建及其应用. 灌溉排水学报, 2025, 44(12): 137-147
[35] Beillouin D, Corbeels M, Demenois J, et al. A global meta-analysis of soil organic carbon in the Anthropocene. Nature Communications, 2023, 14: 3700
[36] 李硕, 李有兵, 王淑娟, 等. 关中平原作物秸秆不同还田方式对土壤有机碳和碳库管理指数的影响. 应用生态学报, 2015, 26(4): 1215-1222
[37] 闫桂菀, 董文斌, 李忠义, 等. 绿肥覆盖对果园土壤团聚体及有机碳组分的影响. 应用生态学报, 2024, 35(12): 3427-3434
[38] 河南省人民政府. 河南省高标准农田建设规划(2021—2023年).(2022-03-03)[2026-01-26]. https://www.henan.gov.cn/2022/03-03/2404839.html
[39] 农业农村部. 全国秸秆综合利用实施方案(2016—2020年). (2016-12-25)[2026-01-26]. https://www.moa.gov.cn/