Spatio-temporal Index Based On Stress Features For Identifying Heavy Metal Contamination In Rice With Remote Sensing Imagery

Rice is one of the most important maple foods around the world.However,heavy metals will accumulate in rice during the process of rice growth,entering human bodies through food chain,and eventually have adverse effects on human health.Although tra-ditional field surveys can detect heavy metal stress in rice,it is usually time-consuming,labor-intensive,and do not facilitate mapping the extent of heavy metal contamination for large regions.Meanwhile,remote sensing technology enables heavy metal stress detection in a more accurate,up-to-date,and wider coverage manner.As a result,re-mote sensing techniques for assessing heavy metal stress in rice has aroused widespread concern around scholars both home and abroad.Like other remote sensing techniques which monitor other environmental stresses in rice,namely pest,disease,water,and nutrition stresses,remote sensing heavy metal stress is usually accomplished by modeling physiological-biochemical changes or spec-tral responses of rice.However,heavy metal stress is difficult to distinguish from other stresses,using only rice spectral feature at a single phase or multiple phases.This is true especially for in-year persistent stresses like water and nutrition stresses because their spectral signals and in-year temporal characteristics are extremely similar.Based on the differences of spatial-temporal features between multiple stress types,a spatial-temporal index based on stress features is proposed in this article to remote sensing heavy metal stress in rice.Inter-annual changes based on time series of rice spectral index is introduced in this article to precisely detect heavy metal stress in rice,along with in-year fluctuation characteristics and spatial clustering features were ex-tracted,and a spatial-temporal index for heavy metal stress is constructed.Research focuses and findings include:(1)Time series modelling of rice Enhanced Vegetation Index(EVI)series.In this study,an additive model,which is based on a linear model and a two-order har-monic model in time series analysis,is used to decompose rice EVI series.This model has higher accuracies than those of other mathematic models.(2)Introducing year-to-year change as rice inter-annual features.A year-to-year change is introduced in this research to assess the differences between rice EVI series in adjacent years.Compared to methods which only take in-year persistent signals into consideration,introducing inter-annual changes will help eliminate signals of other in-year persistent stresses,namely nutrition stress,from heavy metal stress signals.(3)Rice heavy metal stress index based on spatial-temporal features.A spatial-temporal index based on in-year,inter-annual and spatial characteristics is pro-posed to efficiently discriminate heavy metal stress from other common types of stress in rice.