Abstract: Abstract: Rice is one of the most important food crops around the world. In recent years, the accumulation of heavy metals in rice crops has posed a great threat to rice growth and quality, even to human bodies through food chains. Traditional field surveys cannot map the extent of heavy metal contamination for large regions, although they can detect heavy metal stress in rice. Remote sensing can real-time monitor the heavy metal stress in crops with wider coverage. However, most previous studies focused mainly on the physiological and biochemical characteristics of rice, or spectral response similar to other remote sensing monitoring of environmental stress, such as diseases and pests, or water and nutrition stress. Moreover, the growth of rice depends usually on various types of environmental stresses with complex conditions in a large-scale eco-environment system, such as heavy metal stresses, disease, drought, or flood. Specifically, there are different stress types in various spaces, where multiple stresses covered one or several growth stages in the same space. It is also difficult to distinguish heavy metal stress from others in the farmland, due to the similarity of canopy spectral variation induced by multiple stresses. Particularly, heavy metal stress in rice is characterized by the stability in space and time, leading to a difference from other transient stress. Sentinel-2 satellite can be used to monitor crops under multiple stresses. Sentinel-2 data includes three bands in the "red edge", providing more important information on the vegetation state. Taking Zhuzhou, Hunan Province of China as a study area, Sentinel-2 satellite images were collected for the field data from 2017 to 2019. First, the leaf area index (LAI) time series of the year was simulated using the World Food Studies (WOFOST) model. Second, an ensemble empirical mode decomposition (EEMD) was used to disassemble a series of LAI in rice growth stages, and thereby to obtain different intrinsic mode functions (IMF). The time series that may contain heavy metal stress can be achieved through analysis and composition. Third, the time series of healthy rice was considered to be a reference sequence after EEMD. Dynamic time warping (DTW) was used to compare with two time series, further to determine the level of heavy metal stress in rice. Finally, two neighboring time series were compared to measure the stability of heavy metal stress. Results demonstrated that: 1) The combined EEMD and DTW can be used to distinguish the specific levels of heavy metal stress. A high value of normalized stress index referred to that the rice was subjected to a high level of heavy metal stress. There was a relatively high proportion of planting areas with a low level of heavy metal stress in the study area. 2) Heavy metal stress in rice was characterized by spatio-temporal stability. There was a 0.17 percentage point difference in the region proportion with severe stress in 2017 and 2018 and a difference of 0.91 percentage point in 2018 and 2019. Rice under high heavy metal stress was distributed in the western, northeast, and southeast study areas. 3) There was generally a small value in the normalized inter-annual variation index on the rice between adjacent years, indicating an excellent indicator for the stability of rice in space and time.