A New Vegetation Index Based on Multitemporal Sentinel-2 Images for Discriminating Heavy Metal Stress Levels in Rice

Abstract

Heavy metal stress in crops is a worldwide problem that requires accurate and timely monitoring. This study aimed to improve the accuracy of monitoring heavy metal stress levels in rice by using multiple Sentinel-2 images. The selected study areas are in Zhuzhou City, Hunan Province, China. Six Sentinel-2 images were acquired in 2017, and heavy metal concentrations in soil were measured. A novel vegetation index called heavy metal stress sensitive index (HMSSI) was proposed. HMSSI is the ratio between two red-edge spectral indices, namely the red-edge chlorophyll index (CIred-edge) and the plant senescence reflectance index (PSRI). To demonstrate the capability of HMSSI, the performances of CIred-edge and PSRI in discriminating heavy metal stress levels were compared with that of HMSSI at different growth stages. Random forest (RF) was used to establish a multitemporal monitoring model to detect heavy metal stress levels in rice based on HMSSI at different growth stages. Results show that HMSSI is more sensitive to heavy metal stress than CIred-edge and PSRI at different growth stages. The performance of a multitemporal monitoring model combining the whole growth stage images was better than any other single growth stage in distinguishing heavy metal stress levels. Therefore, HMSSI can be regarded as an indicator for monitoring heavy metal stress levels with a multitemporal monitoring model.

Unsupervised Monitoring Vegetation after the Closure of an Ore Processing Site with Multi-Temporal Optical Remote Sensing

“…There are many indices exploiting the red edge. This database includes: Indices specified for global vegetation monitoring such as NDVI [ 77 ] or MTCI (MERIS Terrestrial Chlorophyll Index) [ 74 ]; Indices defined to characterize the impact of contaminants (heavy metals, hydrocarbons...) on vegetation such as EMEN2 (Red and green ratio) [ 66 ] or HSSMI (Heavy Metal Stress Sensitive Index) [ 27 ]; Indices defined to estimate biochemical parameters (chlorophyll, anthocyanin…) such as ARVI [ 62 ] or SIPI [ 79 ]. NDVI, widely used in many applications [ 24 ], is chosen as the reference index and the other indices are compared to this reference index.…”

“…Some of these indices, such as Photochemical Reflectance Index (PRI) or Ratio Vegetation Index (RVI), have been applied on hyperspectral or multispectral reflectance data to detect vegetation stress related to heavy metals [ 9 , 26 ]. Recent indices have also been proposed for HM-stress detection in rice crops such as Heavy Metal Stress Sensitive Index (HMSSI), which is the ratio between two existing red-edge indices CI red-edge (named CIREDEDGE) and PSRI ((Plant Senescence Reflectance Index) or Heavy metal Cd stress-sensitive Spectral Index (HCSI) [ 27 , 28 ]. It is difficult to select one or more indices from the large number existing in the literature, as the index performance depends on the context defined by the species, the pollutant, the pollution level and the environment.…”

“…The Sentinel-2 (S-2) (A and B) multispectral instruments, with a spatial resolution of 10 and 20 m depending on the spectral domains and a high temporal coverage (five days revisit time), are one of the best ways in which to detect inter-annual and intra-annual vegetation changes as well as stressed vegetation [ 7 , 33 , 42 ]. The presence of the red-edge bands provides the opportunity for HM stress monitoring [ 27 ]. The exploitation of multi-temporal data should make it possible to reduce false alarms that may be linked to other sources of stress, such as weather-related stress (flood, drought…) or pests and diseases, also known as abrupt stressors [ 28 ].…”

Spatio-temporal Index Based on Time Series of Leaf Area Index for Identifying Heavy Metal Stress in Rice under Complex Stressors

“…Researchers have attempted to measure heavy metal stress levels by using physiological and spectral features, because heavy metal contaminants have direct or indirect influences on physiological parameters such as leaf area, dry weight, photosynthetic efficiency, and transpiration rate; these influences, in turn affect several spectral values in remote sensing images. Indices based on hyperspectral [6] or multispectral [7] images reflecting stress levels in rice, canopy-air temperature difference [8], crop growth models, like the WOrld FOod STudies (WOFOST) model [9], and the components of time series decomposition [10] have been proposed to discriminate heavy metal stress in rice on the basis of remote sensing images.…”

Use of aerial image in the estimation of volume and biomass of Eucalyptus sp. forest stand

“…Contemporary forest management in forest stands or native forests requires spatially continuous and multitemporal information retrieval with comparable scenarios, providing a basis for the successful implementation of sustainable and continuous long-term tree management (Pasher andKing, 2010, Huang et al, 2018). In the last decades, remote sensing has played a crucial role in forest monitoring and management, disaster management and agricultural applications (White et al, 2016;Mendes et al, 2018), providing an alternative that is low cost, environmentally friendly and fast to monitoring the vegetation (Zhang et al, 2018;Vrieling et al, 2018). Remote sensing includes techniques that use satellite images, allows the observation of the area of interest as a whole.…”