1.西北农林科技大学水土保持科学与工程学院(水土保持研究所),水土保持与荒漠化整治全国重点实验室, 陕西 杨凌 712100
2.国网山东省电力公司经济技术研究院,济南 250021
刘璇皓(2000—),男,硕士研究生,主要从事遥感应用研究。E-mail:lxh00113@nwafu.edu.cn
张晓萍(1971—),女,博士,教授,主要从事水土保持及遥感应用研究。E-mail:zhangxp@ms.iswc.ac.cn
收稿:2026-01-16,
修回:2026-03-03,
录用:2026-03-10,
网络首发:2026-05-21,
纸质出版:2026-10-28
移动端阅览
刘璇皓,张晓萍,文艳,等.基于非线性变换与分数阶微分的土壤有机碳光谱估测[J].水土保持学报,2026,40(5):225-237.
Liu Xuanhao, Zhang Xiaoping, Wen Yan, et al. Spectral estimation of soil organic carbon based on nonlinear transformations and fractional differentiation[J]. Journal of Soil and Water Conservation,2026,40(5):225-237.
刘璇皓,张晓萍,文艳,等.基于非线性变换与分数阶微分的土壤有机碳光谱估测[J].水土保持学报,2026,40(5):225-237. DOI: 10.13870/j.cnki.stbcxb.2026.05.003. CSTR: 32310.14.stbcxb.2026.05.003.
Liu Xuanhao, Zhang Xiaoping, Wen Yan, et al. Spectral estimation of soil organic carbon based on nonlinear transformations and fractional differentiation[J]. Journal of Soil and Water Conservation,2026,40(5):225-237. DOI: 10.13870/j.cnki.stbcxb.2026.05.003. CSTR: 32310.14.stbcxb.2026.05.003.
目的
2
以陕西省延安市吴起县和志丹县的213个林地表层(0~20 cm)土壤样本为研究对象,采集可见-近红外(Vis-NIR)光谱及实测土壤有机碳(SOC)质量分数,探究非线性数学变换结合分数阶微分(FOD)处理对光谱特征挖掘及模型性能的影响,为提高黄土高原地表SOC质量分数的光谱估测精度及智能化辨识提供支持。
方法
2
对原始光谱(
R
)采取倒数(1/
R
)、对数(lg
R
)、对数倒数(1/lg
R
)、平方根(
<math id="M1"><mroot><mrow><mi>R</mi></mrow><mrow/></mroot></math>
https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=120708477&type=
https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=120708525&type=
5.16466665
3.55599999
)等4种非线性变换,及0~2阶(步长0.1)的FOD处理,利用竞争性自适应重加权采样(CARS)和连续投影算法(SPA)筛选特征波段,构建并对比偏最小二乘回归(PLSR)、随机森林(RF)和极限梯度提升(XGBoost)3种模型的反演效果。
结果
2
研究区林地土壤SOC质量分数为2.50~23.11 g/kg,平均值为7.70 g/kg。在原始反射光谱(
R
)中,SOC质量分数较高的样本整体反射率水平较低,光谱反射率随SOC升高呈整体降低趋势,光谱特征波段集中在550~800 nm及900 nm附近。相较于整数阶和
>
1.5的高阶微分,采用0.3~0.9阶的FOD数据预处理,能有效去除基线漂移并凸显细部光谱特征。对数与倒数的非线性变换可以显著提升光谱对SOC的辨识程度。相比SPA,CARS在测试集上
R
2
平均提升0.052,RMSE平均降低0.147 g/kg,RPD平均提升0.132。在所有模型中,PLSR的预测性能整体优于RF和XGBoost,基于非线性变换与FOD处理的CARS-PLSR模型测试集
R
2
可达0.843。
结论
2
研究证实,非线性变换结合中低阶分数阶微分及CARS-PLSR建模策略,可有效提升黄土高原土壤有机碳的光谱估测精度。
Objective
2
Using 213 forest soil samples (0-20 cm depth) from Wuqi and Zhidan counties in Yan′an City, Shaanxi Province, visible-near infrared (Vis-NIR) spectra and measured soil organic carbon (SOC) content are collected to investigate the effects of nonlinear mathematical transformations combined with fractional-order differential (FOD) processing on spectral feature extraction and model performance. This study aims to provide support for improving the accuracy of spectral estimation and intelligent identification of surface SOC content in the Loess Plateau.
Methods
2
Four nonlinear transformations-reciprocal (1/
R
), logarithmic (lg
R
), reciprocal logarithmic (1/lg
R
), and square root (
<math id="M2"><mroot><mrow><mi>R</mi></mrow><mrow/></mroot></math>
https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=120708529&type=
https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=120708528&type=
5.16466665
3.55599999
)-were applied to raw spectra (
R
), followed by fractional-order differential (FOD) processing at orders 0-2 (increments of 0.1). The competitive adaptive reweighted sampling (CARS) and successive projection algorithm (SPA) were employed to screen feature bands. Partial least squares regression (PLSR), random forest (RF), and extreme gradient boosting (XGBoost) models were constructed, and their performance was compared.
Results
2
The
soil organic carbon (SOC) content in the forestland of the study area ranged from 2.50 g/kg to 23.11 g/kg, with an average value of 7.70 g/kg. In the raw reflectance spectrum (
R
), samples with higher SOC content exhibited lower overall reflectance levels. Spectral reflectance showed an overall decreasing trend with increasing SOC, with the spectral feature bands concentrated around 550-800 nm and 900 nm. Compared to integer-order and higher-order differentials (
>
1.5), preprocessing with FOD of orders 0.3-0.9 more effectively eliminated baseline drift and highlighted subtle spectral features. Nonlinear transformations, such as logarithmic and reciprocal transformations, significantly enhanced the spectral identification of soil organic carbon. Compared to SPA, CARS achieved an average
R
² improvement of 0.052, an average RMSE reduction of 0.147 g/kg, and an average RPD increase of 0.132 on the test set. Among all models, PLSR consistently outperformed RF and XGBoost in predictive performance. The CARS-PLSR model, incorporating nonlinear transformations and FOD processing, attained an
R
² of 0.843 on the test set.
Conclusion
2
The findings confirm that combining nonlinear transformations with low-to-medium-order fractional differentials and the CARS-PLSR modeling strategy effectively enhances the spectral estimation accuracy for soil organic carbon in the Loess Plateau.
Basile-Doelsch I , Balesdent J , Pellerin S . Reviews and syntheses: The mechanisms underlying carbon storage in soil [J]. Biogeosciences , 2020 , 17 ( 21 ): 5223 - 5242 .
McBratney A , Field D J , Koch A . The dimensions of soil security [J]. Geoderma , 2014 , 213 : 203 - 213 .
Chen Yun , Wang Jinliang , Liu Guangjie , et al . Hyperspectral estimation model of forest soil organic matter in northwest Yunnan Province, China [J]. Forests , 2019 , 10 ( 3 ): e217 .
Li Hongyi , Yang Jiangtao , Hu Bifeng , et al . Improving in-situ spectral estimation of wetland soil organic carbon by integrating multiple optimization strategies [J]. Catena , 2025 , 255 : e109078 .
Odebiri O , Odindi J , Mutanga O . Basic and deep learning models in remote sensing of soil organic carbon estimation: A brief review [J]. International Journal of Applied Earth Observation and Geoinformation , 2021 , 102 : e102389 .
褚小立 , 袁洪福 , 陆婉珍 . 近红外分析中光谱预处理及波长选择方法进展与应用 [J]. 化学进展 , 2004 , 16 ( 4 ): 528 - 542 .
Chu Xiaoli , Yuan Hongfu , Lu Wanzhen . Progress and application of spectral data pretreatment and wavelength selection methods in NIR analytical technique [J]. Progress in Chemistry , 2004 , 16 ( 4 ): 528 - 542 .
周磊 , 何秀芹 , 贾德伟 , 等 . 基于CWT-CARS的黄土高原东部陕州区农田表层有机碳含量高光谱反演 [J]. 遥感技术与应用 , 2025 , 40 ( 1 ): 60 - 68 .
Zhou Lei , He Xiuqin , Jia Dewei , et al . Hyperspectral retrieval of organic carbon content in farmland surface in Shanzhou district of eastern Loess Plateau based on CWT-CARS [J]. Remote Sensing Technology and Application , 2025 , 40 ( 1 ): 60 - 68 .
曾远文 , 范文武 . 不同光谱变换形式对土壤有机质偏最小二乘估算模型精度的影响 [J]. 安徽农学通报 , 2025 , 31 ( 15 ): 89 - 93 .
Zeng Yuanwen , Fan Wenwu . Influence of different spectral transformation forms on the accuracy of partial least squares estimation model of soil organic matter [J]. Anhui Agricultural Science Bulletin , 2025 , 31 ( 15 ): 89 - 93 .
李天乐 , 赵泉华 , 贾淑涵 , 等 . 土壤有机质含量反演方法:土壤高光谱数据微分-瞬时频率变换 [J]. 地球信息科学学报 , 2024 , 26 ( 7 ): 1733 - 1744 .
Li Tianle , Zhao Quanhua , Jia Shuhan , et al . Soil organic matter content inversion methods: Differential-instantaneous frequency transform of soil hyperspectral data [J]. Journal of Geo-Information Science , 2024 , 26 ( 7 ): 1733 - 1744 .
Hong Yongsheng , Guo Long , Chen Songchao , et al . Exploring the potential of airborne hyperspectral image for estimating topsoil organic carbon: Effects of fractional-order derivative and optimal band combination algorithm [J]. Geoderma , 2020 , 365 : e114228 .
郭艳萍 , 王雪梅 , 赵枫 , 等 . 基于最优数学和小波变换的绿洲耕层土壤盐分RF高光谱反演 [J]. 农业工程学报 , 2025 , 41 ( 3 ): 83 - 93 .
Guo Yanping , Wang Xuemei , Zhao Feng , et al . Hyperspectral inversion of the RF model for soil salinity in oasis tillage layer based on optimal mathematics and wavelet transform [J]. Transactions of the Chinese Society of Agricultural Engineering , 2025 , 41 ( 3 ): 83 - 93 .
邓昀 , 王君 , 陈守学 , 等 . 基于PCA-DBO-SVR的林地土壤有机质高光谱反演模型 [J]. 光谱学与光谱分析 , 2025 , 45 ( 2 ): 569 - 583 .
Deng Yun , Wang Jun , Chen Shouxue , et al . Hyperspectral inversion model of forest soil organic matter based on PCA-DBO-SVR [J]. Spectroscopy and Spectral Analysis , 2025 , 45 ( 2 ): 569 - 583 .
Morellos A , Pantazi X E , Moshou D , et al . Machine learning based prediction of soil total nitrogen, organic carbon and moisture content by using VIS-NIR spectroscopy [J]. Biosystems Engineering , 2016 , 152 : 104 - 116 .
郑家兴 , 郭兴燕 , 郭金锋 , 等 . 基于分数阶微分和混合变量选择算法的水稻土有机碳含量反演 [J]. 激光与光电子学进展 , 2026 , 63 ( 2 ): 389 - 402 .
Zheng Jiaxing , Guo Xingyan , Guo Jinfeng , et al . Inversion of organic carbon content in paddy soil based on fractional order differential and hybrid variable selection algorithm [J]. Laser and Optoelectronics Progress , 2026 , 63 ( 2 ): 389 - 402 .
史舟 , 王乾龙 , 彭杰 , 等 . 中国主要土壤高光谱反射特性分类与有机质光谱预测模型 [J]. 中国科学:地球科学 , 2014 , 44 ( 5 ): 978 - 988 .
Shi Zhou , Wang Qianlong , Peng Jie , et al . Classification of hyperspectral reflectance characteristics of main soils in China and spectral prediction model of organic matter [J]. Scientia Sinica (Terrae) , 2014 , 44 ( 5 ): 978 - 988 .
Zhu H F , Xu Z J , Jing Y D , et al . Spatial variation and predictions of soil organic matter and total nitrogen based on VNIR reflectance in a basin of Chinese Loess Plateau [J]. Journal of Soil Science and Plant Nutrition , 2018 , 18 ( 4 ): 1126 - 1141 .
倪玲珊 , 岳荣 , 方怒放 . 基于中红外漫反射光谱法预测黄土碳酸钙含量 [J]. 水土保持研究 , 2020 , 27 ( 2 ): 142 - 148 .
Ni Lingshan , Yue Rong , Fang Nufang . Estimation of calcium carbonate in loess by using mid-infrared diffuse reflectance spectroscopy [J]. Research of Soil and Water Conservation , 2020 , 27 ( 2 ): 142 - 148 .
王可琳 . 黄土高原典型流域土壤有机碳遥感反演及时空变化特征研究 [D]. 陕西 杨凌 : 西北农林科技大学 , 2024 .
Wang Kelin . Remote sensing inversion and spatio-temporal variation of soil organic carbon in a typical basin of the Loess Plateau [D]. Yangling, Shaanxi : Northwest A&F University , 2024 .
鲍士旦 . 土壤农化分析 [M]. 3版 . 北京 : 中国农业出版社 , 2000 .
Bao Shidan . Soil and agricultural chemistry analysis [M]. 3rd edition . Beijing : China Agriculture Press , 2000 .
熊超群 , 荀咪 , 李健 , 等 . 基于数学变换和连续小波变换的苹果园土壤总氮含量高光谱反演 [J]. 湖北农业科学 , 2025 , 64 ( 11 ): 165 - 170 .
Xiong Chaoqun , Xun Mi , Li Jian , et al . Hyperspectral inversion of soil total nitrogen content in an apple orchard based on mathematical transformation and continuous wavelet transform [J]. Hubei Agricultural Sciences , 2025 , 64 ( 11 ): 165 - 170 .
李长春 , 施锦锦 , 马春艳 , 等 . 基于小波变换和分数阶微分的冬小麦叶绿素含量估算 [J]. 农业机械学报 , 2021 , 52 ( 8 ): 172 - 182 .
Li Changchun , Shi Jinjin , Ma Chunyan , et al . Estimation of chlorophyll content in winter wheat based on wavelet transform and fractional differential [J]. Transactions of the Chinese Society for Agricultural Machinery , 2021 , 52 ( 8 ): 172 - 182 .
丁启东 , 王怡婧 , 张俊华 , 等 . 基于高光谱信息的宁夏土壤有机质含量估算 [J]. 生态学杂志 , 2024 , 43 ( 8 ): 2531 - 2538 .
Ding Qidong , Wang Yijing , Zhang Junhua , et al . Estimation of soil organic matter content in Ningxia based on hyperspectral information [J]. Chinese Journal of Ecology , 2024 , 43 ( 8 ): 2531 - 2538 .
赵启东 , 葛翔宇 , 丁建丽 , 等 . 结合分数阶微分技术与机器学习算法的土壤有机碳含量光谱估测 [J]. 激光与光电子学进展 , 2020 , 57 ( 15 ): 245 - 253 .
Zhao Qidong , Ge Xiangyu , Ding Jianli , et al . Combination of fractional order differential and machine learning algorithm for spectral estimation of soil organic carbon content [J]. Laser and Optoelectronics Progress , 2020 , 57 ( 15 ): 245 - 253 .
国佳欣 , 赵小敏 , 郭熙 , 等 . 基于PLSR-BP复合模型的红壤有机质含量反演研究 [J]. 土壤学报 , 2020 , 57 ( 3 ): 636 - 645 .
Guo Jiaxin , Zhao Xiaomin , Guo Xi , et al . Inversion of organic matter content in red soil based on PLSR-BP composite model [J]. Acta Pedologica Sinica , 2020 , 57 ( 3 ): 636 - 645 .
梁永刚 , 杨金燕 . 高光谱土壤成分反演的特征波段提取研究进展 [J]. 传感器与微系统 , 2025 , 44 ( 11 ): 7 - 11 .
Liang Yonggang , Yang Jinyan . Research progress on feature band extraction for hyperspectral inversion of soil composition [J]. Transducer and Microsystem Technologies , 2025 , 44 ( 11 ): 7 - 11 .
刘佳琳 , 王飞 , 韩剑桥 , 等 . 基于分数阶微分的土壤含盐量光谱特征及定量估算研究 [J]. 遥感技术与应用 , 2025 , 40 ( 2 ): 344 - 358 .
Liu Jialin , Wang Fei , Han Jianqiao , et al . Study on spectral characteristics and quantitative estimation of soil salinity based on fractional order derivative [J]. Remote Sensing Technology and Application , 2025 , 40 ( 2 ): 344 - 358 .
Bentéjac C , Csörgő A , Martínez-Muñoz G . A comparative analysis of gradient boosting algorithms [J]. Artificial Intelligence Review , 2021 , 54 ( 3 ): 1937 - 1967 .
许小明 , 田起隆 , 孙景梅 , 等 . 子午岭地区植被演替的土壤碳汇效应 [J]. 水土保持学报 , 2022 , 36 ( 3 ): 159 - 165 .
Xu Xiaoming , Tian Qilong , Sun Jingmei , et al . Effects on soil carbon sink of vegetation succession in the Ziwuling area [J]. Journal of Soil and Water Conservation , 2022 , 36 ( 3 ): 159 - 165 .
何挺 , 王静 , 程烨 , 等 . 土壤氧化铁光谱特征研究 [J]. 地理与地理信息科学 , 2006 , 22 ( 2 ): 30 - 34 .
He Ting , Wang Jing , Cheng Ye , et al . Study on spectral features of soil Fe 2 O 3 [J]. Geography and Geo-Information Science , 2006 , 22 ( 2 ): 30 - 34 .
李武耀 , 买买提·沙吾提 , 买合木提·巴拉提 . 基于分数阶微分的土壤有机质含量高光谱反演研究 [J]. 激光与光电子学进展 , 2023 , 60 ( 7 ): 404 - 411 .
Li Wuyao , Maimait·Sawuti , Maihemuti·Balati . Fractional differential-based hyperspectral inversion of soil organic matter content [J]. Laser and Optoelectronics Progress , 2023 , 60 ( 7 ): 404 - 411 .
Yang Meihua , Xu Dongyun , Chen Songchao , et al . Evaluation of machine learning approaches to predict soil organic matter and pH using vis-NIR spectra [J]. Sensors , 2019 , 19 ( 2 ): e263 .
0
浏览量
0
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621