Spectral Separability Analysis and Band Selection for Heavy Metal
Inversion in Water Using Correlation Coefficient Heatmaps
LIANG Ye-heng1, LAO Xiao-min2, DENG Ru-ru1, 3*, NI Hua-hong1, ZHAO Tong-tong1, HUANG An-feng4, GUO Zhi-peng5, LI Yu-hua6, ZHANG Rui-wu1
1. School of Geography and Planning，Sun Yat-sen University，Guangzhou 510006，China
2. Institute of Urban and Sustainable Development，City University of Macau，Macau 999078，China
3. Southern Marine Science and Engineering Guangdong Laboratory（Zhuhai），Guangdong Engineering Research Center of Water Environment Remote Sensing Monitoring，Zhuhai 519082，China
4. Chancheng Branch of Foshan Natural Resources Bureau，Foshan 528000，China
5. Guangzhou Urban Planning and Design Survey Research Institute Limited Company，Guangzhou 510030，China
6. Tianjin Institute of Geological Environment Monitoring，Tianjin 300191，China
Abstract：The concept of “green economy”, which balances rapid urban economic development with environmental protection, has become a societal consensus. Integrating emerging remote sensing technologies (e. g., satellites and drones) with traditional methods can establish a multidimensional monitoring system, enabling a more efficient balance between economic and environmental goals. A critical challenge for expanding remote sensing applications lies in overcoming technical bottlenecks related to parameters that cannot be extracted via remote sensing. Monitoring heavy metals in water, a valuable yet unresolved area of environmental remote sensing, encounters the issue of spectral separability in multi-parameter inversion through satellite remote sensing. To address this, the reflectance spectra of four compounds (copper sulfate, potassium ferricyanide, potassium ferrocyanide, and ferric chloride) were measured using a spectrometer within the wavelength range of 350~1 050 nm. Key findings include: Copper sulfate exhibits a reflectance peak at 441 nm; Ferric chloride shows a “wave-like” gradual increase until 906 nm, followed by a decline; Potassium ferricyanide peaks at 767 nm; Potassium ferrocyanide peaks at 826 nm, with the latter having a trough at 988 nm. The reflectance spectra of these four compounds intersect at two places: copper sulfate and potassium ferrocyanide intersect at a wavelength of 492 nm, while copper sulfate, ferric chloride, and potassium ferricyanide intersect in the wavelength range of 576~584 nm (centered at 580 nm). Consequently, the Ratio Copper Index (RCI) was proposed to differentiate copper sulfate from the three iron compounds. Two remote sensing models were employed to analyze the mathematical principles behind spectral separability. Utilizing a linear spectral unmixing model, pairwise correlation coefficients among ten heavy metal compounds (copper sulfate, potassium ferricyanide, potassium ferrocyanide, ferric chloride, cadmium oxide, lead tetroxide, lead chromate, cadmium sulfide, lead sulfate, and lead sulfide) were calculated across varying spectral resolutions and visualized through heatmaps. The results showed that there are many valuable phenomena, laws, and references about the separability between these ten heavy metal compounds. Ultimately, a dynamic band selection strategy for remote sensing inversion was proposed, which adjusts the number of compounds and bands in accordance with correlation heatmap results. The research results provide a deep discussion of the mathematical nature of spectral separability between compounds and the solution of remote sensing models, further deepening the algorithmic theoretical basis for solving remote sensing models. This research promotes the implementation of heavy metal concentration inversion in water at the satellite image level in the future.
Key words：Remote sensing of environment;Remote sensing of heavy metals in water;Spectral separability;Correlation coefficient heatmap
梁业恒，劳小敏，邓孺孺，倪桦泓，赵彤彤，黄岸峰，郭志鹏，李玉华，张瑞午. 基于相关系数热力图的水中重金属光谱可分性讨论及遥感反演波段选择策略[J]. 光谱学与光谱分析, 2025, 45(11): 3057-3065.
LIANG Ye-heng, LAO Xiao-min, DENG Ru-ru, NI Hua-hong, ZHAO Tong-tong, HUANG An-feng, GUO Zhi-peng, LI Yu-hua, ZHANG Rui-wu. Spectral Separability Analysis and Band Selection for Heavy Metal
Inversion in Water Using Correlation Coefficient Heatmaps. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(11): 3057-3065.
[1] Liu X P，Huang Y H，Xu X C，et al. Nature Sustainability，2020，3：564.
[2] He D，Shi Q，Xue J Q，et al. Remote Sensing of Environment，2023，299：113884.
[3] ZHANG Zhi-xin，DENG Ru-ru，LI Hao，et al（张志新，邓孺孺，李 灏，等）. Remote Sensing for Land and Resources（国土资源遥感），2011，（3）：88.
[4] HE Ying-qing，QIN Yan，FU Qing-hua，et al（何颖清，秦 雁，扶卿华，等）. Tropical Geography（热带地理），2016，36（5）：860.
[5] Yang J Y，Deng R R，Ma Y W，et al. Sensors，2025，25（6）：1728.
[6] Yang J Y，Deng R R，Ma Y W，et al. Water，2025，17（6）：780.
[7] Qiao H Y，Lee Z P，Wang D S，et al. Remote Sensing of Environment，2025，323：114709.
[8] Guo Y，Liang Y H，Deng R R，et al. Heliyon，2022，8（12）：e12033.
[9] LIU Zhi-fei，YANG Xue-mei，WANG Jing-rui，et al（刘志飞，杨雪梅，王景瑞，等）. Arid Zone Research（干旱区研究），2025，42（1）：141.
[10] MENG Chang，HONG Mei，LI Fei（孟 畅，红 梅，李 斐）. Ecology and Environmental Sciences（生态环境学报），2025，34（6）：950.
[11] Wang L，Zhou Y，Zhou Z H，et al. IEEE Transactions on Geoscience and Remote Sensing，2025，63：4504714.
[12] DENG Ru-ru，LIANG Ye-heng，GAO Yi-kang，et al（邓孺孺，梁业恒，高奕康，等）. National Remote Sensing Bulletin（遥感学报），2016，20（1）：35.
[13] LIANG Ye-heng，OUYANG Yu-chun，XU Min-duan，et al（梁业恒，欧阳宇纯，许敏端，等）. Spectroscopy and Spectral Analysis（光谱学与光谱分析），2025，45（8）：2149.
[14] LIANG Ye-heng，DENG Ru-ru，GAO Yi-kang，et al（梁业恒，邓孺孺，高奕康，等）. National Remote Sensing Bulletin（遥感学报），2016，20（1）：27.
[15] DENG Ru-ru，HE Zhi-jian，CHEN Xiao-xiang，et al（邓孺孺，何执兼，陈晓翔，等）. Acta Scientiarum Naturalium Universitatis Sunyatseni（中山大学学报·自然科学版），2002，41（3）：99.
[16] Deng R R，Liu Q H，Ke R P, et al. Acta Oceanologica Sinica，2004，23（1）：119.
