Monitoring of soil heavy metals based on hyperspectral remote sensing: A review

Hyperspectral remote sensing (HRS) has emerged as a promising technique for monitoring the spatiotemporal distribution of soil heavy metal (SHM) contamination, owing to its extensive coverage and high-efficiency. Nevertheless, the monitoring technique for SHMs based on HRS still faces challenges in engineering applications, including the unclear theoretical basis, unstable inversion methods, and uncertain application scenarios for remote sensing platforms. In this paper, we conducted a comprehensive review of 529 critical publications to identify research gaps and uncertainties related to this monitoring technique. Firstly, the theoretical basis of SHM monitoring is elucidated, highlighting the crucial role played by the association characteristics between soil active ingredients (SAIs) and SHMs as the core theoretical support for SHM inversion. Subsequently, we systematically review the current state of research on key steps in the monitoring technique, encompassing spectral pre-processing, spectral characteristic identification, and inversion modeling. This is because the application results of these steps often lack uniformity or even exhibit contradictions. Notably, the results show that representative characteristic bands, despite being identified within the wavelength ranges of 400–1000, 1200–1450, and 1700–2400 nm, exhibit significant variations across different SHMs and application scenarios. Machine learning methods (R2: 0.54–0.94) employed in SHM inversion modeling are not significantly superior to empirical statistical models (R2: 0.52–0.91). Moreover, the historical development survey of remote sensing platforms indicates the increasing utilization of airborne and spaceborne HRS platforms, particularly HyMAP and GF-5, for SHM monitoring. However, achieving a balance between monitoring accuracy and breadth remains a fundamental challenge for each HRS platform. We recommend emphasizing research directions of this monitoring technique around the guidance framework of “mechanism-data-model-application”, thereby effectively aiding in the prevention and control of SHM contamination in the future. [Display omitted] •Correlation between SAIs and SHMs is core theoretical basis for SHM inversion.•Key characteristic bands of SHMs in diverse scenarios were summarized.•SHM inversion accuracy is comparable for machine learning and statistical models.•High-accuracy, large-scale SHM monitoring needs space-air-ground spectral fusion.•Future research focuses on