Owing to the advantages of fast and non-destructive measurement, visible and near-infrared reflectance (VNIR) spectra have been widely used in the study of heavy metal pollution. However, few studies have focused on the estimation of soil heavy metal concentration by the enhanced joint architecture of spectral indices and metal elements enriched in clay minerals. In this work, a new composite index, namely Three-band Spectral and Metal Element Index (TSMEI), is proposed to retrieve arsenic (As) in soil by utilizing the multi-view spectral information. Based on obtained data, including spectra and the concentration of iron (Fe), potassium (K), aluminum (Al), magnesium (Mg) and arsenic (As) of the soil around the open-pit coal mine area, the three-band spectral index (TBSI) for As, K, Fe, Mg and Al were calculated from four types of spectral data, that is, raw reflectance (R), the first-order derivative of the spectrum (FD), spectral continuum removal (CR) and spectral reciprocal logarithmic (RL). Then, the metal element index (MEI) for As concentration were constructed via estimated content of K, Fe, Mg and Al based on their TBSIs. Finally, the optimized TBSIs and MEIs were used to construct TSMEI, and it was combined with random forest to invert the As concentration. The following conclusions are drawn: TBSIs is significantly better than that dual-band spectral index and single-band spectrum for estimating As content, and the correlations between the TBSIs based on the FD and the As concentration perform best (r ≥ 0.7684). In addition, two/three element MEIs show higher correlation coefficients with As concentration compared to individual metal element. Furthermore, the proposed TSMEI allow high-precision estimation of As content, which acquired highest correlation coefficient and lowest RMSE (r = 0.9732, RMSE = 0.0703). The results confirm that the TSMEI is significantly effective in estimating soil As content.

由于可见光和近红外反射（VNIR）光谱具有快速、无损测量的优点，在重金属污染的研究中得到了广泛的应用。然而，很少有研究关注通过光谱指数和富含粘土矿物的金属元素的增强联合结构来估计土壤重金属浓度。在这项工作中，提出了一种新的复合指数，即三波段光谱和金属元素指数（TSMEI），利用多视图光谱信息检索土壤中的砷（As）。根据获得的数据，包括光谱和露天煤矿区周围土壤的铁（Fe）、钾（K）、铝（Al）、镁（Mg）和砷（As）的浓度，三个- As、K、Fe、Mg 和 Al 的波段光谱指数 (TBSI) 由四种光谱数据计算得出，即原始反射率 (R)、光谱的一阶导数 (FD)、光谱连续谱去除 (CR) 和光谱倒数对数 (RL)。然后，通过基于 TBSI 估计 K、Fe、Mg 和 Al 的含量，构建 As 浓度的金属元素指数 (MEI)。最后，利用优化后的TBSIs和MEIs构建TSMEI，并结合随机森林对As浓度进行反演。得出以下结论：TBSIs在估计As含量方面明显优于双波段光谱指数和单波段光谱，并且基于FD和As浓度的TBSIs之间的相关性表现最好（As 浓度的金属元素指数 (MEI) 是通过基于 TBSI 估计 K、Fe、Mg 和 Al 的含量构建的。最后，利用优化后的TBSIs和MEIs构建TSMEI，并结合随机森林对As浓度进行反演。得出以下结论：TBSIs在估计As含量方面明显优于双波段光谱指数和单波段光谱，并且基于FD和As浓度的TBSIs之间的相关性表现最好（As 浓度的金属元素指数 (MEI) 是通过基于 TBSI 估计 K、Fe、Mg 和 Al 的含量构建的。最后，利用优化后的TBSIs和MEIs构建TSMEI，并结合随机森林对As浓度进行反演。得出以下结论：TBSIs在估计As含量方面明显优于双波段光谱指数和单波段光谱，并且基于FD和As浓度的TBSIs之间的相关性表现最好（r  ≥ 0.7684)。此外，与单个金属元素相比，二/三元素 MEI 与 As 浓度显示出更高的相关系数。此外，所提出的 TSMEI 允许对 As 含量进行高精度估计，获得最高的相关系数和最低的 RMSE（r  = 0.9732，RMSE = 0.0703）。结果证实 TSMEI 在估算土壤砷含量方面非常有效。