Identifying rice stress on a regional scale from multi-temporal satellite images using a Bayesian method

Area of Science:

• Agricultural remote sensing
• Environmental monitoring
• Plant stress detection

Background:

• Crop stress, including heavy metal contamination, threatens food security and human health.
• Distinguishing heavy metal stress from other stressors is challenging due to complex crop responses.
• Satellite-derived vegetation indices offer potential for large-scale crop stress assessment.

Purpose of the Study:

• To infer the probability of heavy metal (cadmium) stress in rice on a regional scale.
• To integrate satellite-derived vegetation indices with spatio-temporal stressor characteristics using a Bayesian method.
• To differentiate heavy metal stress from other environmental stressors in crops.

Main Methods:

• Utilized Sentinel-2 satellite imagery to calculate the normalized difference red-edge index (NDRE) for rice growth stages.
• Derived coefficients of spatio-temporal variation (CSTV) from NDRE to identify stress patterns.
• Applied a Bayesian method integrating NDRE and CSTV to infer cadmium (Cd) pollution in rice.

Main Results:

• Normalized difference red-edge index (NDRE) proved to be a sensitive indicator of rice stress levels.
• The CSTV, with a threshold of 2.7, effectively detected regional Cd-induced stress and abrupt stress events.
• Achieved a high map accuracy of 81.57% for identifying Cd-induced stress in rice.

Conclusions:

• Satellite-derived vegetation indices, like NDRE, are valuable tools for capturing crop stress.
• The Bayesian method, incorporating spatio-temporal stressor characteristics, effectively distinguishes specific stressors.
• This approach provides a robust framework for regional crop stress monitoring and management.