Research on Chromium Ion Content Inversion of GF-5 Satellite Images Based on Grid Search Optimization CatBoost Model

doi:10.16258/j.cnki.1674-5906.2024.09.013

Ecology and Environment ›› 2024, Vol. 33 ›› Issue (9): 1460-1470.DOI: 10.16258/j.cnki.1674-5906.2024.09.013

• Research Article [Environmental Science] • Previous Articles Next Articles

Research on Chromium Ion Content Inversion of GF-5 Satellite Images Based on Grid Search Optimization CatBoost Model

LIU Dongyi¹(), QU Yonghua¹^,²^,^*(), FENG Yaowei¹, QU Ran³

1. Beijing Normal University/State Key Laboratory of Remote Sensing Science, Beijing 100875, P. R. China
2. Beijing Engineering Research Center for Global Land Remote Sensing Products/Institute of Remote Sensing Science and Engineering, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, P. R. China
3. Satellite Application Center for Ecology and Environment, Beijing 100094, P. R. China

Received:2024-05-13 Online:2024-09-18 Published:2024-10-18
Contact: QU Yonghua

基于网格搜索优化CatBoost模型的GF-5卫星影像铬离子含量反演研究

刘东宜¹(), 屈永华¹^,²^,^*(), 冯耀伟¹, 屈冉³

1.北京师范大学/遥感科学国家重点实验室，北京 100875
2.北京市陆表遥感数据产品工程技术研究中心/北京师范大学地理科学学部遥感科学与工程研究院，北京 100875
3.生态环境部卫星环境应用中心，北京 100094

通讯作者: 屈永华
作者简介:刘东宜（2001年生），男，硕士研究生，研究方向为灾害遥感。E-mail: liudy@mail.bnu.edu.cn
基金资助:
国家自然科学基金重大项目课题(42192581)

Abstract

Abstract:

Hyperspectral remote sensing technology has attracted attention because of its advantages of low cost, high efficiency, wide coverage, macroscopicity, and dynamic monitoring capability, which make it possible to carry out large-scale soil heavy metal inversion. Hyperspectral remote sensing technology was used to effectively monitor the contamination status of hexavalent chromium ions in the soil of a tailing pond in Xinhe Township, Antu County. Hyperspectral remote sensing images from the Gaofen-5 (GF-5) satellite and measured soil samples were used as data sources. Based on the Pearson's correlation coefficient, band addition, and ratio index, features related to changes in the heavy metal content of the soil were extracted to construct an inversion model. First, the saturated, noise and moisture absorption bands in the pre-processed remote sensing images were excluded, and the addition and ratio indices were constructed algebraically using the reflectance of each band to calculate the Pearson correlation coefficients with the measured chromium ion content data. Then, based on the Pearson's correlation coefficient of the top 50 most relevant feature variables, the CatBoost regression optimized by grid search was used to build an inversion model of soil hexavalent chromium ion content, and the SHAP method was used to evaluate the importance of the features and explore the important bands affecting the inversion of soil heavy metals by GF-5. The results showed that under the same spectral transformation conditions, the grid search optimized CatBoost model was the most effective with a goodness of fit of 0.92 for the training set and 0.88 for the validation set compared with partial least squares regression (PLSR), random forest regression (RFR), support vector machine regression (SVMR), convolutional neural network (CNN), and multivariate stepwise linear regression models. The results were obtained with a goodness of fit of 0.92 and a goodness of fit of 0.88 for the training set and the validation set respectively. The search optimized CatBoost regression model was used to invert the soil hexavalent chromium ion content of the chromium slag landfill site in Antu, Jilin Province, and the results showed that hexavalent chromium ion contamination around the tailings pond mining area in the region was serious, which is consistent with the actual situation of mining and tailings dumping in the area. This study provides an important technical and scientific basis for monitoring soil heavy metal pollution and environmental management.

Key words: hyperspectral inversion, hyperspectral remote sensing, GF-5, hexavalent chromium ions, soil heavy metals, CatBoost regression model

摘要：

高光谱遥感技术以其低成本、高效率、广泛覆盖范围、宏观性强和动态监测能力等优势而受到关注。利用高光谱遥感技术，对安图县新合乡尾矿库土壤中六价铬离子污染状况进行有效监测；以高分五号（GF-5）卫星的高光谱遥感影像及实测土壤样本为数据源，基于皮尔逊相关系数、波段加法和比值指数，提取与土壤重金属含量变化相关的特征，构建反演模型。首先，剔除预处理后的遥感影像中饱和、噪声和水汽吸收波段，同时利用各波段反射率进行代数运算构建加法和比值指数，计算其与实测铬离子含量数据的皮尔逊相关系数；然后，基于皮尔逊相关系数得到的前50最高相关特征变量，采用网格搜索优化的CatBoost回归，建立土壤六价铬离子含量反演模型，并使用SHAP方法评估特征重要性，探究影响GF-5反演土壤重金属重要波段。结果显示，在相同光谱变换条件下，与偏最小二乘回归（PLSR）、随机森林回归（RFR）、支持向量机回归（SVMR）、卷积神经网络（CNN）和多元逐步线性回归模型相比，网格搜索优化的CatBoost模型效果最好，训练集拟合优度为0.92，验证集为0.88。利用网格搜索优化CatBoost回归模型对吉林省安图铬渣填埋场进行了土壤六价铬离子含量反演，结果显示该区域尾矿库开采区域周边六价铬离子污染严重，这与该地矿山开采和尾矿堆放实际情况基本一致。该研究为土壤重金属污染监测和环境管理提供了重要的技术手段和科学依据。

关键词: 高光谱反演, 高光谱遥感, GF-5, 六价铬离子, 土壤重金属, CatBoost回归模型

CLC Number:

TP7
X833

LIU Dongyi, QU Yonghua, FENG Yaowei, QU Ran. Research on Chromium Ion Content Inversion of GF-5 Satellite Images Based on Grid Search Optimization CatBoost Model[J]. Ecology and Environment, 2024, 33(9): 1460-1470.

刘东宜, 屈永华, 冯耀伟, 屈冉. 基于网格搜索优化CatBoost模型的GF-5卫星影像铬离子含量反演研究[J]. 生态环境学报, 2024, 33(9): 1460-1470.

Figures/Tables 10

Figure 1 Geographical location and field measured soil sample distribution map of the study area

Figure 2 Distribution of saturated pixels in the study area and Raw DN value image of GF-5 sample points

Figure 3 Flow Chart for inversion of hexavalent chromium ion concentration

Figure 4 Pearson correlation coefficient obtained from single band calculation

Figure 5 Pearson correlation coefficient obtained from band index calculation

Figure 6 Scatter plot of soil hexavalent chromium ion concentration inversion

Table 1 Comparison Table of Fitting Effect Parameters

回归模型	训练集		验证集
回归模型	R²	σ_RMSE	R²	σ_RMSE
随机森林回归	0.74	155.05	0.8	35.94
支持向量机回归	0.92	83.77	−0.89	109.73
偏最小二乘回归	0.32	250.73	−0.79	106.71
卷积神经网络	0.07	193.62	0.5	56.58
多元逐步回归	0.24	265.26	−4.65	189.54
CatBoost 回归	0.92	87.73	0.88	27.94

Figure 7 Ranking of band importance with method of CatBoost

Figure 8 Feature importance evaluation of SHAP

Figure 9 CatBoost inversion results of remote sensing images in the study area

References 34

[1]	CAMPILLO-CORA C, RODRÍGUEZ-GONZÁLEZ L, ARIAS-ESTÉVEZ M, et al., 2021. Influence of physicochemical properties and parent material on chromium fractionation in soils[J]. Processes, 9(6): 1073.
[2]	CHEN L H, LAI J, TAN K, et al., 2022. Development of a soil heavy metal estimation method based on a spectral index: Combining fractional-order derivative pretreatment and the absorption mechanism[J]. Science of The Total Environment, 813: 151882.
[3]	CHOE E, VAN DER MEER F, VAN RUITENBEEK F, et al., 2008. Mapping of heavy metal pollution in stream sediments using combined geochemistry, field spectroscopy, and hyperspectral remote sensing: A case study of the Rodalquilar mining area, SE Spain[J]. Remote Sensing of Environment, 112(7): 3222-3233.
[4]	DING Y, CHEN Z Q, LU W F, et al., 2021. A CatBoost approach with wavelet decomposition to improve satellite-derived high-resolution PM_2.5 estimates in Beijing-Tianjin-Hebei[J]. Atmospheric Environment, 249: 118212.
[5]	GE X Y, DING J L, TENG D X, et al., 2022. Exploring the capability of Gaofen-5 hyperspectral data for assessing soil salinity risks[J]. International Journal of Applied Earth Observation and Geoinformation, 112: 102969.
[6]	HONG Y S, CHEN S C, LIU Y L, et al., 2019. Combination of fractional order derivative and memory-based learning algorithm to improve the estimation accuracy of soil organic matter by visible and near-infrared spectroscopy[J]. CATENA, 174: 104-116.
[7]	LEE S H, VO T P, THAI H T, et al., 2021. Strength prediction of concrete-filled steel tubular columns using categorical gradient boosting algorithm[J]. Engineering Structures, 238: 112109.
[8]	LI Z Y, MA Z W, VAN DER KUIJP T J, et al., 2014. A review of soil heavy metal pollution from mines in China: Pollution and health risk assessment[J]. Science of The Total Environment, 468-469: 843-853.
[9]	LIU P, LIU Z H, HU Y M, et al., 2019. Integrating a hybrid back propagation neural network and particle swarm optimization for estimating soil heavy metal contents using hyperspectral data[J]. Sustainability, 11(2): 419.
[10]	MENG X T, BAO Y L, YE Q, et al., 2021. Soil organic matter prediction model with satellite hyperspectral image based on optimized denoising method[J]. Remote Sensing, 13(12): 2273.
[11]	PARK J, LEE W H, KIM K T, et al., 2022. Interpretation of ensemble learning to predict water quality using explainable artificial intelligence[J]. Science of The Total Environment, 832: 155070.
[12]	PROKHORENKOVA L, GUSEV G, VOROBEV A, et al., 2018. CatBoost: Unbiased boosting with categorical features[EB/OL]. arXiv, [2018-10-24]. https://arxiv.org/abs/1810.11363.
[13]	ZHANG B, GUO B, ZOU B, et al., 2022. Retrieving soil heavy metals concentrations based on GaoFen-5 hyperspectral satellite image at an opencast coal mine, Inner Mongolia, China[J]. Environmental Pollution, 300: 118981.
[14]	ZHANG M, YUAN L Y, 2023. High-precision estimation of hourly PM_2.5 concentration based on a grid scale of satellite-derived products[J]. Atmospheric Pollution Research, 14(4): 101724.
[15]	ZHANG S W, SHEN Q, NIE C J, et al., 2019. Hyperspectral inversion of heavy metal content in reclaimed soil from a mining wasteland based on different spectral transformation and modeling methods[J]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 211: 393-400.
[16]	ZHAO S H, WANG Q, LI Y, et al., 2017. An overview of satellite remote sensing technology used in China’s environmental protection[J]. Earth Science Informatics, 10(2): 137-148.
[17]	柏晗, 杨耘, 崔琴芳, 等, 2022. 基于GA-XGBoost模型的GF-5卫星影像土壤重金属含量反演研究[J]. 激光与光电子学进展, 59(12): 515-524.
	BAI H, YANG Y, CUI Q F, et al., 2022. Retrieval of heavy metal content in soil using GF-5 satellite images based on GA-XGBoost model[J]. Laser & Optoelectronics Progress, 59(12): 515-524.
[18]	陈点点, 陈芸芝, 冯险峰, 等, 2022. 基于超参数优化CatBoost算法的河流悬浮物浓度遥感反演[J]. 地球信息科学学报, 24(4): 780-791. DOI
	CHEN D D, CHEN Z Y, FENG X F, et al., 2022. Retrieving suspended matter concentration in rivers based on hyperparameter optimized catboost algorithm[J]. Journal of Geo-Information Science, 24(4): 780-791.
[19]	陈宇波, 薛云, 邹滨, 等, 2020. 有色金属矿区土壤铬污染遥感反演研究[J]. 中南大学学报(自然科学版), 51(10): 2876-2884.
	CHEN Y B, XUE Y, ZOU B, et al., 2020. Research on remote sensing retrieval of soil chromium pollution in nonferrous metal mining area[J]. Journal of Central South University (Science and Technology), 51(10): 2876-2884.
[20]	龚绍琦, 王鑫, 沈润平, 等, 2010. 滨海盐土重金属含量高光谱遥感研究[J]. 遥感技术与应用, 25(2): 169-177. DOI
	GONG S Q, WANG X, SHEN R P, et al., 2010. Study on heavy metal element content in the coastal saline soil by hyperspectral remote sensing[J]. Remote Sensing Technology and Application, 25(2): 169-177.
[21]	孔卓, 杨海涛, 郑逢杰, 等, 2022. 高光谱遥感图像大气校正研究进展[J]. 自然资源遥感, 34(4): 1-10.
	KONG Z, YANG H T, ZHENG F J, et al., 2022. Research advances in atmospheric correction of hyperspectral remote sensing images[J]. Remote Sensing for Natural Resources, 34(4): 1-10.
[22]	刘银年, 2021. 高光谱成像遥感载荷技术的现状与发展[J]. 遥感学报, 25(1): 439-459.
	LIU Y N, 2021. Development of hyperspectral imaging remote sensing technology[J]. Journal of Remote Sensing, 25(1): 439-459.
[23]	苏红军, 2022. 高光谱遥感影像降维: 进展、挑战与展望[J]. 遥感学报, 26(8): 1504-1529.
	SU H J, 2022. Downscaling hyperspectral remote sensing images: progress, challenges and prospects[J]. Journal of Remote Sensing, 26(8): 1504-1529.
[24]	孙丽丽, 方宏彬, 朱星星, 等, 2021. 基于网格搜索优化的XGBoost模型的股票预测[J]. 阜阳师范大学学报(自然科学版), 38(2): 97-101.
	SUN L L, FANG H B, ZHU X X, et al., 2021. Stock prediction using XGBoost model based on grid search optimization[J]. Journal of Fuyang Normal University (Natural Science), 38(2): 97-101.
[25]	唐洪钊, 肖晨超, 梁树能, 等, 2020. 资源一号02D卫星在轨辐射定标精度验证与分析[J]. 航天器工程, 29(6): 142-147.
	TANG H Z, XIAO C C, LIANG S N, et al., 2020. On-orbit radiometric calibration and validation of ZY-1-02D satellite[J]. Spacecraft Engineering, 29(6): 142-147.
[26]	田义超, 郑丹琳, 张强, 等, 2024. 基于国产资源一号02D卫星和机器学习算法的钦州湾滨海土壤盐分反演[J]. 中国环境科学, 44(1): 371-385.
	TIAN Y C, ZHENG D L, ZHANG Q, et al., 2024. Inversion of coastal soil salinity in Qinzhouwan based on domestic ZY1-02D satellite and machine learning algorithm[J]. China Environmental Science, 44(1): 371-385.
[27]	王雪梅, 玉米提·买明, 毛东雷, 等, 2021. 干旱区绿洲耕层土壤重金属铬含量的高光谱估测[J]. 生态环境学报, 30(10): 2076-2084. DOI
	WANG X M, YUMITI M, MAO D L, et al., 2021. Hyperspectral estimation of heavy metal chromium content in arable soil of arid area oasis[J]. Journal of Ecology and Environment, 30(10): 2076-2084.
[28]	闫峰, 2010. 影响土壤中Cr(Ⅵ)吸持与Cr(Ⅲ)氧化的主要土壤理化性质分析[D]. 咸阳: 西北农林科技大学.
	YAN F, 2010. Analysis of the main soil physicochemical properties affecting Cr(VI) adsorption and Cr(III) oxidation in soil[D]. Xianyang: Northwest A & F University.
[29]	颜祥照, 姚艳敏, 张霄羽, 等, 2021. 星载高分五号高光谱耕地主要土壤类型土壤有机质含量估测——以黑龙江省建三江农垦区为例[J]. 中国土壤与肥料 (5): 10-20.
	YAN X Z, YAO Y M, ZHANG X Y, et al., 2021. Estimation of soil organic matter content in major soil types of star-borne Gaofen-5 hyperspectral cultivated land: A case study of Jiansanjiang reclamation area in Heilongjiang province[J]. Soil and Fertilizer Sciences in China (5): 10-20.
[30]	杨灵玉, 高小红, 张威, 等, 2016. 基于Hyperion影像植被光谱的土壤重金属含量空间分布反演——以青海省玉树县为例[J]. 应用生态学报, 27(6): 1775-1784. DOI
	YANG L Y, GAO X H, ZHANG W, et al., 2016. Inversion of spatial distribution of soil heavy metal content based on hyperion image vegetation spectra: A case study of Yushu County in Qinghai Province[J]. Chinese Journal of Applled Ecology, 27(6): 1775-1784.
[31]	袁自然, 魏立飞, 张杨熙, 等, 2020. 优化CARS结合PSO-SVM算法农田土壤重金属砷含量高光谱反演分析[J]. 光谱学与光谱分析, 40(2): 567-573.
	YUAN Z R, WEI L F, ZHANG Y X, et al., 2020. Optimization of CARS combined with PSO-SVM algorithm for hyperspectral inversion analysis of heavy metal arsenic in farmland soil[J]. Spectroscopy and Spectral Analysis, 40(2): 567-573.
[32]	张明月, 张奇栎, 王璐, 等, 2019. 东北黑土区土壤铬含量高光谱反演研究[J]. 遥感技术与应用, 34(2): 313-322.
	ZHANG M Y, ZHANG Q L, WANG L, et al., 2019. Research on chromium retrieval of black soil with hyperspectral imagery in northeast of China[J]. Remote Sensing Technology and Application, 34(2): 313-322.
[33]	赵瑞, 崔希民, 刘超, 2020. GF-5高光谱遥感影像的土壤有机质含量反演估算研究[J]. 中国环境科学, 40(8): 3539-3545.
	ZHAO R, CUI X M, LIU C, et al., 2020. Inverse estimation of soil organic matter content from GF-5 hyperspectral remote sensing images[J]. China Environmental Science, 40(8): 3539-3545.
[34]	郑家桐, 王鹏, 石航源, 等, 2023. 基于CatBoost模型和SHAP解释方法的土壤重金属影响因素与程度定量分析[J]. 环境科学学报, 43(4): 448-456.
	ZHENG J T, WANG P, SHI H Y, et al., 2023. Quantitative analysis of the influence factors and extent of soil heavy metals based on CatBoost model and SHAP interpretation method[J]. Acta Scientiae Circumstantiae, 43(4): 448-456.

Research on Chromium Ion Content Inversion of GF-5 Satellite Images Based on Grid Search Optimization CatBoost Model

基于网格搜索优化CatBoost模型的GF-5卫星影像铬离子含量反演研究

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 34

Related Articles 6

Recommended Articles

Metrics

Comments

[1]	WU Wenwei, SHEN Cheng, SHA Chenyan, LIN Kuangfei, WU Jian, XIE Yuqing, ZHOU Xuan. Soil Heavy Metal Enrichment Characteristics, Risk Assessment, and Source Analysis in Redevelopment Areas during Urban Industrial Plots [J]. Ecology and Environment, 2024, 33(5): 791-801.
[2]	CHEN Minyi, ZHU Hanghai, SHE Weiduo, YIN Guangcai, HUANG Zuzhao, YANG Qiaoling. Health Risk Assessment and Source Apportionment of Soil Heavy Metals at A Legacy Shipyard Site in Pearl River Delta [J]. Ecology and Environment, 2023, 32(4): 794-804.
[3]	XIAO Jieyun, ZHOU Wei, SHI Peiqi. Hyperspectral Inversion of Soil Heavy Metals [J]. Ecology and Environment, 2023, 32(1): 175-182.
[4]	LIU Di, SU Chao, ZHANG Hong, QIN Guanyu. Pollution Characteristics and Risk Assessment of Heavy Metal Pollution in A Typical Coal-based Industrial Cluster Zone [J]. Ecology and Environment, 2022, 31(2): 391-399.
[5]	TA Weiyuan, KANG Zhen, MENG Zhaojun, JIN Shenghua, YANG Xing, GUO Longfei, ZHAO Dongxu, ZHANG Xin. Research of Pollution Characteristics of Heavy Metals in Soil of Typical Closed Zinc Smelting Enterprises in Qinling Mountains [J]. Ecology and Environment, 2021, 30(7): 1513-1521.
[6]	CHA Lijuan, ZHOU Dandan, FENG Hongjuan, ZHAO Shuyuan, FENG Kaiping. Research on the Bioaccumulation Characteristics of Two Kinds of Wild Edible Fungi to Soil Heavy Metals [J]. Ecology and Environment, 2021, 30(10): 2093-2099.