组合光学和微波遥感的耕地土壤含盐量反演

doi:10.12118/j.issn.1000-6060.2023.375

摘要/Abstract

摘要：

耕地保护关系到国家粮食安全和经济社会可持续发展，对生态环境保护具有重要作用，快速精准的获取耕地土壤盐分含量及空间分布信息是耕地保护的必然要求。以宁夏平罗县为研究区，利用Landsat 9 OLI和Sentinel-1遥感影像，提取光谱指数和雷达极化组合指数，基于变量投影重要性法与灰度关联法筛选特征变量，然后运用反向传播神经网络、支持向量机和随机森林3种机器学习算法构建模型，并用最佳模型反演耕地土壤含盐量空间分布情况。结果表明：（1）利用变量投影重要性法筛选变量建立的模型验证集决定系数（R²）大于灰度关联法筛选变量建立的模型。（2）利用随机森林算法，组合光谱指数和雷达极化组合指数协同反演模型效果最佳，建模集R²为0.791，均方根误差（RMSE）为1.016，R²较单一数据源模型分别提高0.065和0.085，RMSE分别降低0.147和0.189；验证集R²为0.780，RMSE为1.132，R²较单一数据源模型分别提高0.091和0.237，RMSE分别降低0.175和0.377。（3）平罗县耕地轻度盐渍化和中度盐渍化土壤分布范围广，占比分别为23.77%和33.54%，重度盐渍化达15.37%。研究结果发现，组合多源遥感数据建模能够有效提高土壤含盐量反演精度，可为干旱区耕地土壤含盐量的反演和当地农业可持续发展提供有效的技术参考。

关键词: 光学和微波遥感, 机器学习, 耕地, 土壤含盐量反演

Abstract:

The safeguarding of cultivated land is paramount in ensuring national food security, sustainable economic and social development, and the preservation of the ecological environment. Rapid and accurate acquisition of cultivated soil salinity and spatial distribution information is imperative for the protection of cultivated land. This study focuses on cultivated land in Pingluo County, Ningxia, China, as the research object and discusses the feasibility of combining optical remote sensing and microwave remote sensing to predict the accuracy of soil salt content compared with single remote sensing data. The methodology involved the extraction of the spectral indices from Landsat 9 OLI and radar polarization combination indices from Sentinel-1. Variable projection importance and gray correlation degree were used to screen and combine characteristic variables. Three machine learning algorithms (back propagation neural network, support vector machine, and random forest) were used to construct the soil salt content prediction model. The best model was used to predict the spatial distribution of the soil salt content in cultivated land. The results show the following facts: (1) The model, validated using the variable projection importance method for screening variables, generally exhibited a higher determination coefficient (R²) than the model established using the gray correlation method for characteristic variables. (2) Using the random forest algorithm, the model combining the spectral index and radar polarization combination index demonstrated the best effect. The modeling set exhibited an R² of 0.791 and a root mean square error (RMSE) of 1.016. This represented an increase in R² by 0.065 and 0.085 compared with the single data source model, with corresponding decreases in RMSE by 0.147 and 0.189. The validation set showed an R² of 0.780 and an RMSE of 1.132, indicating a respective increase in R² by 0.091 and 0.237 and a decrease in RMSE by 0.175 and 0.377 compared with the single data source model. (3) The distribution range of mildly salinized and moderately salinized soil of cultivated land in Pingluo County covered wide areas, accounting for 23.77% and 33.54%, respectively, whereas severely salinized soil constituted 15.37%. This underscores the effectiveness of modeling by combining multisource remote sensing data in improving the prediction accuracy of soil salt content. The outcomes offer a valuable technical reference for predicting soil salt content in arid areas and contribute to the sustainable development of local agriculture.

Key words: optical and microwave remote sensing, machine learning, cultivated land, inversion of soil salt content

刘瑞亮, 贾科利, 李小雨, 陈睿华, 王怡婧, 张俊华. 组合光学和微波遥感的耕地土壤含盐量反演[J]. 干旱区地理, 2024, 47(3): 433-444.

LIU Ruiliang, JIA Keli, LI Xiaoyu, CHEN Ruihua, WANG Yijing, ZHANG Junhua. Inversion of soil salt content by combining optical and microwave remote sensing in cultivated land[J]. Arid Land Geography, 2024, 47(3): 433-444.

图/表 13

图1

表1

图2

表2

光谱指数计算公式"

光谱指数	公式
盐分指数（SI）^[18]	$B l u e × R e d$
盐分指数1（SI1）^[19]	$G r e e n × R e d$
盐渍化指数1（S1）^[20]	Blue/Red
盐渍化指数2（S2）^[21]	(Blue-Red)/(Blue+Red)
盐渍化指数5（S5）^[22]	Blue×Red/Green
扩展差值植被指数（EDVI）^[4]	NIR+SWIR1-Red
扩展比值植被指数（ERVI）^[4]	(NIR+SWIR1)/Green
大气阻抗植被指数（ARVI）^[23]	(NIR-2Red+Blue)/(NIR+2Red-Blue)
冠层盐度响应植被指数（CRSI）^[24]	$N I R × R e d - (G r e e n × B l u e) N I R × R e d + (G r e e n × B l u e)$
扩展增强型植被指数（EEVI）^[4]	$2.5 (N I R + S W I R 1 - R e d) (N I R + S W I R 1 + 6 R e d - 7.5 B l u e + 1)$

表2

表3

图3

图4

图5

表4

表5

图6

图7

表6

参考文献 45

[1]	Abdelaziz M E, Abdelsattar M, Abdeldaym E A, et al. Piriformospora indica alters Na⁺/K⁺ homeostasis, antioxidant enzymes and LeNHX1 expression of greenhouse tomato grown under salt stress[J]. Scientia Horticulturae, 2019, 256: 108532, doi: 10.1016/j.scienta.2019.05.059.
[2]	Singh A. Soil salinity: A global threat to sustainable development[J]. Soil Use and Management, 2022, 38(1): 39-67. doi: 10.1111/sum.v38.1
[3]	李星佑, 张飞, 王筝. 土壤盐渍化遥感监测模型构建方法现状与发展趋势[J]. 自然资源遥感, 2022, 34(4): 11-21.
	[Li Xingyou, Zhang Fei, Wang Zheng. Present situation and development trend in building remote sensing monitoring models of soil salinization[J]. Remote Sensing for Natural Resources, 2022, 34(4): 11-21.]
[4]	陈红艳, 赵庚星, 陈敬春, 等. 基于改进植被指数的黄河口区盐渍土盐分遥感反演[J]. 农业工程学报, 2015, 31(5): 107-114.
	[Chen Hongyan, Zhao Gengxing, Chen Jingchun, et al. Remote sensing inversion of saline soil salinity based on modified vegetation index in estuary area of Yellow River[J]. Transactions of the Chinese Society of Agricultural Engineering, 2015, 31(5): 107-114.]
[5]	Wang J Z, Ding J L, Ma X K, et al. Capability of Sentinel-2 MSI data for monitoring and mapping of soil salinity in dry and wet seasons in the Ebinur Lake region, Xinjiang, China[J]. Geoderma, 2019, 353: 172-187. doi: 10.1016/j.geoderma.2019.06.040
[6]	赵孟银. 遥感影像去云方法研究[D]. 天津: 天津科技大学, 2016.
	[Zhao Mengyin. Study of clouds removal methods on remote sensing images[D]. Tianjin: Tianjin University of Science and Technology, 2016.]
[7]	李彪, 王耀强. 土壤盐渍化雷达反演模拟研究[J]. 干旱区资源与环境, 2015, 29(8): 180-184.
	[Li Biao, Wang Yaoqiang. Radar inversion and simulation of salty soil salinization[J]. Journal of Arid Land Resources and Environment, 2015, 29(8): 180-184.]
[8]	马驰. 基于Sentinel 1双极化雷达影像的土壤含盐量反演[J]. 农业工程学报, 2018, 34(2):153-158.
	[Ma Chi. Quantitative retrieval of soil salt content based on Sentinel-1 dual polarization radar image[J]. Transactions of the Chinese Society of Agricultural Engineering, 2018, 34(2): 153-158.]
[9]	张智韬, 杜瑜燕, 劳聪聪, 等. 基于雷达遥感的不同深度土壤含盐量反演模型[J]. 农业机械学报, 2020, 51(10): 243-251.
	[Zhang Zhitao, Du Yuyan, Lao Congcong, et al. Inversion model of soil salt content in different depths based on radar remote sensing[J]. Transactions of the Chinese Society for Agricultural Machinery, 2020, 51(10): 243-251.]
[10]	肖森天, 依力亚斯江·努尔麦麦提, 努尔比耶·穆合塔尔, 等. 基于光学和雷达多源遥感的于田绿洲土壤盐渍化时空分析[J]. 干旱区研究, 2023, 40(1): 59-68.
	[Xiao Sentian, Nurmemet Ilyas, Muhetaer Nuerbiye, et al. Spatial and temporal analysis of soil salinity in Yutian Oasis by combined optical and radar multi-source remote sensing[J]. Arid Zone Research, 2023, 40(1): 59-68.]
[11]	王飞, 丁建丽, 魏阳, 等. 基于Landsat系列数据的盐分指数和植被指数对土壤盐度变异性的响应分析——以新疆天山南北典型绿洲为例[J]. 生态学报, 2017, 37(15): 5007-5022.
	[Wang Fei, Ding Jianli, Wei Yang, et al. Sensitivity analysis of soil salinity and vegetation indices to detect soil salinity variation by using Landsat series images: Applications in different oases in Xinjiang, China[J]. Acta Ecologica Sinica, 2017, 37(15): 5007-5022.]
[12]	王海峰. 基于高光谱遥感的土壤水盐监测模型研究[D]. 咸阳: 西北农林科技大学, 2019.
	[Wang Haifeng. Hyperspectral remote sensing based models for soil moisture and salinity prediction[D]. Xianyang: Northwest A & F University, 2019.]
[13]	刘恩, 王军涛, 常步辉, 等. 小开河引黄灌区土壤盐渍化定量遥感反演[J]. 中国农村水利水电, 2019(12): 20-24.
	[Liu En, Wang Juntao, Chang Buhui, et al. Quantitative remote sensing inversion of soil salinization in Xiaokaihe Yellow River irrigation district[J]. China Rural and Hydropower, 2019(12): 20-24.]
[14]	刘亚秋. 黄河三角洲盐渍土盐分及组分定量光谱估测研究[D]. 泰安: 山东农业大学, 2017.
	[Liu Yaqiu. Quantitative estimation of soil salinity and its composition based on spectra in Yellow River Delta[D]. Tai'an: Shandong Agricultural University, 2017.]
[15]	曹肖奕, 丁建丽, 葛翔宇, 等. 基于不同卫星光谱模拟的土壤电导率估算研究[J]. 干旱区地理, 2020, 43(1): 172-181.
	[Cao Xiaoyi, Ding Jianli, Ge Xiangyu, et al. Estimation of soil conductivity based on spectral simulation of different satellites[J]. Arid Land Geography, 2020, 43(1): 172-181.]
[16]	魏慧敏, 贾科利, 张旭, 等. 基于机器学习和多光谱遥感的银川平原土壤盐分预测[J]. 干旱区地理, 2023, 46(1): 103-114.
	[Wei Huimin, Jia Keli, Zhang Xu, et al. Prediction of soil salinity based on machine learning and multispectral remote sensing in Yinchuan Plain[J]. Arid Land Geography, 2023, 46(1): 103-114.]
[17]	李小雨, 贾科利, 魏慧敏, 等. 基于随机森林算法的土壤含盐量预测[J]. 干旱区研究, 2023, 40(8): 1258-1267.
	[Li Xiaoyu, Jia Keli, Wei Huimin, et al. Prediction of soil salt content based on random forest algorithm[J]. Arid Zone Research, 2023, 40(8): 1258-1267.]
[18]	Khan N M, Rastoskuev V V, Sato Y, et al. Assessment of hydrosaline land degradation by using a simple approach of remote sensing indicators[J]. Agricultural Water Management, 2005, 77(1-3): 96-109. doi: 10.1016/j.agwat.2004.09.038
[19]	樊彦国, 张维康, 刘敬一. 基于植被指数-盐分指数特征空间的黄河三角洲盐渍化遥感监测研究[J]. 山东农业科学, 2016, 48(5): 137-141.
	[Fan Yanguo, Zhang Weikang, Liu Jingyi. Remote sensing monitoring model of soil salinization in the Yellow River Delta zone based on vegetation index-salt index feature space[J]. Shandong Agricultural Sciences, 2016, 48(5): 137-141.]
[20]	Sahana M, Rehman S, Patel P P, et al. Assessing the degree of soil salinity in the Indian Sundarban Biosphere Reserve using measured soil electrical conductivity and remote sensing data-derived salinity indices[J]. Arabian Journal of Geosciences, 2020, 13(24): 1289, doi: 10.1007/s12517-020-06310-w.
[21]	Nguyen K A, Liou Y A, Tran H P, et al. Soil salinity assessment by using near-infrared channel and vegetation soil salinity index derived from Landsat 8 OLI data: A case study in the Tra Vinh Province, Mekong Delta, Vietnam[J]. Progress in Earth and Planetary Science, 2020, 7(1): 1-16. doi: 10.1186/s40645-019-0311-0
[22]	孙亚楠, 李仙岳, 史海滨, 等. 基于多源数据融合的盐分遥感反演与季节差异性研究[J]. 农业机械学报, 2020, 51(6): 169-180.
	[Sun Ya’nan, Li Xianyue, Shi Haibin, et al. Remote sensing inversion of soil salinity and seasonal difference analysis based on multi-source data fusion[J]. Transactions of the Chinese Society for Agricultural Machinery, 2020, 51(6): 169-180.]
[23]	Brunner P, Li H T, Kinzelbach W, et al. Generating soil electrical conductivity maps at regional level by integrating measurements on the ground and remote sensing data[J]. International Journal of Remote Sensing, 2007, 28(15-16): 3341-3361. doi: 10.1080/01431160600928641
[24]	何宝忠, 丁建丽, 刘博华, 等. 渭库绿洲土壤盐渍化时空变化特征[J]. 林业科学, 2019, 55(9): 185-196.
	[He Baozhong, Ding Jianli, Liu Bohua, et al. Spatiotemporal variation of soil salinization in Weigan-Kuqa River delta oasis[J]. Scientia Silvae Sinicae, 2019, 55(9): 185-196.]
[25]	郭晓静. 内蒙古河套灌区冻融期土壤盐分的多极化雷达响应分析[D]. 呼和浩特: 内蒙古农业大学, 2014.
	[Guo Xiaojing. Response analysis of multi-polarization radar on freeze-thaw period soil salt in Inner Mongolia Hetao irrigation district[D]. Hohhot: Inner Mongolia Agricultural University, 2014.]
[26]	于海云, 王志军, 李彪, 等. 内蒙古河套灌区融解期土壤盐分多极化雷达响应分析[J]. 长江科学院院报, 2015, 32(11): 19-24.
	[Yu Haiyun, Wang Zhijun, Li Biao, et al. Response analysis of multi-polarization radar on soil salt during melting period in Hetao irrigation district of Inner Mongolia[J]. Journal of Yangtze River Scientific Research Institute, 2015, 32(11): 19-24.]
[27]	李生勇, 霍轶珍, 王海霞. 内蒙古自治区河套灌区冻结期土壤盐分的多极化雷达响应分析[J]. 江苏农业科学, 2016, 44(3): 348-352.
	[Li Shengyong, Huo Yizhen, Wang Haixia. Response analysis of multi-polarization radaron soil salinity during freezing period in Hetao irrigation district of Inner Mongolia Autonomous Region[J]. Jiangsu Agricultural Sciences, 2016, 44(3): 348-352.]
[28]	Oussama A, Elabadi F, Platikanov S, et al. Detection of olive oil adulteration using FT-IR spectroscopy and PLS with variable importance of projection (VIP) scores[J]. Journal of the American Oil Chemists Society, 2012, 89(10): 1807-1812. doi: 10.1007/s11746-012-2091-1
[29]	Maimaitiyiming M, Ghulam A, Bozzolo A, et al. Early detection of plant physiological responses to different levels of water stress using reflectance spectroscopy[J]. Remote Sensing, 2017, 9(7): 745, doi: 10.3390/rs9070745.
[30]	刘思峰, 蔡华, 杨英杰, 等. 灰色关联分析模型研究进展[J]. 系统工程理论与实践, 2013, 33(8): 2041-2046. doi: 10.12011/1000-6788(2013)8-2041
	[Liu Sifeng, Cai Hua, Yang Yingjie, et al. Advance in grey incidence analysis modelling[J]. Systems Engineering-Theory and Practice, 2013, 33(8): 2041-2046.]
[31]	杨淑娥, 黄礼. 基于BP神经网络的上市公司财务预警模型[J]. 系统工程理论与实践, 2005, 25(1): 12-18. doi: 10.12011/1000-6788(2005)1-12
	[Yang Shu'e, Huang Li. Financial crisis warning model based on BP neural network[J]. Systems Engineering-Theory & Practice, 2005, 25(1): 12-18.]
[32]	赵亚琴. 基于模糊神经网络的火灾识别算法[J]. 计算机仿真, 2015, 32(2): 369-373.
	[Zhao Yaqin. Forest fire recognition algorithm based on fuzzy neural network[J]. Computer Simulation, 2015, 32(2): 369-373.]
[33]	汪海燕, 黎建辉, 杨风雷. 支持向量机理论及算法研究综述[J]. 计算机应用研究, 2014, 31(5): 1281-1286.
	[Wang Haiyan, Li Jianhui, Yang Fenglei. Overview of support vector machine analysis and algorithm[J]. Application Research of Computers, 2014, 31(5): 1281-1286.]
[34]	Biau G, Scornet E. A random forest guided tour[J]. Test, 2016, 25(2): 197-227. doi: 10.1007/s11749-016-0481-7
[35]	杨练兵, 陈春波, 郑宏伟, 等. 基于优化随机森林回归模型的土壤盐渍化反演[J]. 地球信息科学学报, 2021, 23(9): 1662-1674. doi: 10.12082/dqxxkx.2021.200711
	[Yang Lianbing, Chen Chunbo, Zheng Hongwei, et al. Retrieval of soil salinity content based on optimized random forest regression model[J]. Journal of Geo-information Science, 2021, 23(9): 1662-1674.]
[36]	余祥伟, 薛东剑. 星载SAR与OLI影像融合方法研究及质量评价[J]. 西华师范大学学报(自然科学版), 2020, 41(1): 98-103.
	[Yu Xiangwei, Xue Dongjian. Research and quality evaluation of spaceborne SAR and OLI image fusion methods[J]. Journal of China West Normal University (Natural Sciences Edition), 2020, 41(1): 98-103.]
[37]	姜红, 玉素甫江·如素力, 拜合提尼沙·阿不都克日木, 等. 基于支持向量机回归算法的土壤水分光学与微波遥感协同反演[J]. 地理与地理信息科学, 2017, 33(6): 30-36.
	[Jiang Hong, Rusuli Yusufujiang, Abudoukerimu Baihetinisha, et al. Soil moisture retrieval by synergizing optical and microwave remote sensing data based on support vector machine regression algorithm[J]. Geography and Geo-information Science, 2017, 33(6): 30-36.]
[38]	李明亮. 土壤有机质高光谱灰色关联度估测模型研究[D]. 泰安: 山东农业大学, 2017.
	[Li Mingliang. Study on the grey relation degree estimation on model of soil organic matter based on hyperspectral[D]. Tai'an: Shandong Agricultural University, 2017.]
[39]	Viscarra Roseel R A, Webster R. Predicting soil properties from the Australian soil visible-near infrared spectroscopic database[J]. European Journal of Soil Science, 2012, 63(6): 848-860. doi: 10.1111/ejs.2012.63.issue-6
[40]	张智韬, 魏广飞, 姚志华, 等. 基于无人机多光谱遥感的土壤含盐量反演模型研究[J]. 农业机械学报, 2019, 50(12): 151-160.
	[Zhang Zhitao, Wei Guangfei, Yao Zhihua, et al. Soil salt inversion model based on UAV multispectral remote sensing[J]. Transactions of the Chinese Society for Agricultural Machinery, 2019, 50(12): 151-160.]
[41]	Chen B L, Zheng H W, Luo G P, et al. Adaptive estimation of multi-regional soil salinization using extreme gradient boosting with Bayesian TPE optimization[J]. International Journal of Remote Sensing, 2022, 43(3): 778-811. doi: 10.1080/01431161.2021.2009589
[42]	Wang L Y, Hu P, Zheng H W, et al. Integrative modeling of heterogeneous soil salinity using sparse ground samples and remote sensing images[J]. Geoderma, 2023, 430: 116321, doi: 10.1016/j.geoderma.2022.116321.
[43]	Xu H T, Chen C B, Zheng H W, et al. AGA-SVR-based selection of feature subsets and optimization of parameter in regional soil salinization monitoring[J]. International Journal of Remote Sensing, 2020, 41(12): 4470-4495. doi: 10.1080/01431161.2020.1718239
[44]	张旭. 银川平原土壤盐分空间分布及驱动因子研究[D]. 银川: 宁夏大学, 2022.
	[Zhang Xu. Study on spatial distribution and driving factors of soil salinity in Yinchuan Plain[D]. Yinchuan: Ningxia University, 2022.]
[45]	杨劲松, 姚荣江, 王相平, 等. 中国盐渍土研究: 历程, 现状与展望[J]. 土壤学报, 2022, 59(1): 10-27.
	[Yang Jinsong, Yao Rongjiang, Wang Xiangping, et al. Research on salt-affected soils in China: History, status quo and prospect[J]. Acta Pedologica Sinica, 2022, 59(1): 10-27.]

样本等级（含盐量/g·kg^-1）	样本数量/个	含盐量/g·kg^-1			变异系数/%
样本等级（含盐量/g·kg^-1）	样本数量/个	平均值	最大值	最小值	变异系数/%
非盐渍化（<1）	15	0.775	0.999	0.510	16.712
轻度盐渍化（1~2）	27	1.326	1.963	1.008	19.913
中度盐渍化（2~4）	32	2.853	3.795	2.066	18.081
重度盐渍化（4~6）	15	4.966	5.909	4.060	12.385
盐土（>6）	15	8.208	14.231	6.145	25.133
总样本	104	3.251	14.231	0.510	79.154

变量	模型类别	建模集		验证集
变量	模型类别	R²	RMSE	R²	RMSE
光谱指数	VIP-BPNN	0.325	1.828	0.464	1.926
	VIP-SVM	0.552	1.487	0.476	1.840
	VIP-RF	0.726	1.163	0.689	1.307
	GC-BPNN	0.328	1.822	0.433	1.979
	GC-SVM	0.488	1.591	0.461	1.904
	GC-RF	0.649	1.317	0.327	2.158
雷达极化组合指数	VIP-BPNN	0.228	1.954	0.291	2.214
	VIP-SVM	0.482	1.475	0.524	1.814
	VIP-RF	0.706	1.205	0.543	1.509
	GC-BPNN	0.220	1.964	0.346	2.127
	GC-SVM	0.509	1.440	0.356	1.910
	GC-RF	0.518	1.543	0.413	2.015

盐渍化程度	非盐渍化	轻度盐渍化	中度盐渍化	重度盐渍化	盐土
像元数/个	138897	250353	353408	161882	148863
面积/km²	124.95	225.18	317.73	145.60	133.86
占比%	13.19	23.77	33.54	15.37	14.13