Extraction of Rice Phenological Differences under Heavy Metal Stress Using EVI Time-Series from HJ-1A/B Data

Chris H.q. Ding (丁宏强), Shuyuan Liu, Xiangnan Liu, Meiling Liu, Ling Wu, Chao Ding, Zhi HuangSensors 2017 · DOI

TL;DR

This study introduces a novel remote sensing approach to monitor heavy metal stress in rice crops by analyzing changes in their phenological stages. By using enhanced vegetation index (EVI) time-series data from HJ-1A/B satellites and applying advanced signal processing techniques, the researchers successfully identified and quantified phenological differences between rice grown under mild and severe heavy metal contamination. This work offers a promising, non-invasive method for large-scale agricultural monitoring and food security assessment.

What Problem Does This Paper Address?

Heavy metal pollution poses a significant threat to agricultural production and food security, particularly in regions with intensive industrial activity. While traditional methods for assessing heavy metal stress in crops often involve direct sampling and laboratory analysis, these are time-consuming, labor-intensive, and not easily scalable for widespread monitoring. Phenology, the study of cyclical and seasonal natural phenomena, especially in relation to climate, is a sensitive indicator of environmental changes. Heavy metals can interfere with plant physiological processes, potentially altering their growth stages and timing. However, effectively and efficiently monitoring these subtle phenological shifts across large agricultural areas, especially under varying levels of heavy metal stress, remains a challenge. This paper aims to bridge this gap by developing a remote sensing-based methodology to extract and analyze rice phenological differences influenced by heavy metal stress.

Methods

The study employs a multi-step approach to extract and analyze rice phenology:

  • Data Acquisition and Preprocessing: HJ-1A/B satellite CCD images were acquired for four experimental rice fields in Zhuzhou, China, covering the entire growing season. These images, with a 30m spatial resolution, were radiometrically calibrated, atmospherically corrected (using FLAASH), and geometrically corrected.
  • Vegetation Index Calculation: The Enhanced Vegetation Index (EVI) was chosen over NDVI due to its improved sensitivity to vegetation characteristics and reduced influence from soil background and atmospheric effects, especially in high biomass areas. The EVI formula is provided.
  • EVI Time-Series Reconstruction: Due to noise from clouds and shadows, the EVI time-series data were smoothed and reconstructed using the asymmetric Gaussian (AG) model function within the TIMESAT software. This method is known for its fidelity and ability to capture key seasonality parameters.
  • Phenological Period Detection: First derivative (FD) analysis of the smoothed EVI time-series was used to identify three key phenological periods: tillering, heading, and maturation. Peaks and valleys in the FD curves correspond to the maximum rates of vegetation growth and decline, respectively.
  • Phenological Metrics Development: To quantify phenological differences under varying heavy metal stress, two types of metrics were developed:
  • Date-Intervals (L1, L2, L3): These metrics represent the time duration (in days) between consecutive phenological periods (tillering to heading, heading to maturation, and tillering to maturation). This approach aims to normalize for variations in transplanting dates.
  • Time-Integrated EVI (TIEVI1, TIEVI2, TIEVI3): These metrics represent the integral areas under the EVI time-series curves between the defined phenological periods. They are considered indicators of net primary production and ecosystem dynamics.
  • Comparison and Analysis: The developed metrics were compared between areas classified as having mild and severe heavy metal stress to assess their effectiveness in distinguishing stress levels.

Datasets

  • HJ-1A/B Satellite CCD Images: 16 cloud-free images were used, covering the period from June 13 to September 29, 2013. These images provided a spatial resolution of 30m.
  • Soil Samples: Collected from five sample plots in each of the four experimental areas at three key phenological stages. These samples were analyzed for heavy metal concentrations (Cadmium, Lead, Arsenic) to classify the stress levels (mild/severe).

Key Insights and Ideas

  • Phenology as a Sensitive Indicator: The study strongly supports the idea that plant phenology is a highly sensitive indicator of environmental changes, including heavy metal stress. Subtle shifts in growth stages can be detected and quantified.
  • EVI's Advantage: EVI proves to be a more robust vegetation index for this application than NDVI, particularly in capturing the dynamics of high biomass crops like rice and mitigating atmospheric and soil influences.
  • Dual Metric Approach: The combination of date-interval metrics (L1, L2, L3) and time-integrated EVI metrics (TIEVI1, TIEVI2, TIEVI3) provides a comprehensive way to assess phenological differences. While date-intervals capture timing shifts, time-integrated EVI reflects the overall growth vigor within those periods.
  • Stress-Induced Shortening: A key finding is that severe heavy metal stress tends to shorten the duration of phenological periods (e.g., L3, the total growing season) compared to mild stress. This suggests that heavy metals can accelerate or decelerate certain growth phases.
  • TIEVI3 as a Strong Indicator: The TIEVI3 metric (total integrated EVI from tillering to maturation) showed a particularly good ability to differentiate between mild and severe heavy metal stress levels.

Key Results

  • Phenological Periods Identified: The tillering, heading, and maturation periods were successfully extracted from the EVI time-series data for each experimental area.
  • Stress Level Correlation: Rice grown under severe heavy metal stress exhibited smaller values for the phenological metrics (both date-intervals and time-integrated EVI) compared to rice grown under mild stress. For instance, the total growing season length (L3) was shorter under severe stress.
  • Metric Performance:
    • The date-interval metrics (L1, L2, L3) showed differences between mild and severe stress, with L3 (total duration) being a more pronounced indicator.
  • The time-integrated EVI metrics (TIEVI1, TIEVI2, TIEVI3) also reflected stress levels, with TIEVI3 demonstrating a strong ability to distinguish between mild and severe conditions. For example, TIEVI3 values were around 32 for severe stress and higher for mild stress.
  • Feasibility Demonstrated: The study successfully demonstrated the feasibility of using remote sensing technology and phenological analysis to monitor heavy metal stress in rice plants.

Limitations and Future Work

  • Limited Phenological Stages: The study focused on three key phenological periods. Future work could explore other stages, such as transplanting and panicle development, for a more comprehensive analysis.
  • Data Resolution: While HJ-1A/B data provided 30m resolution, higher spatial resolution data might offer more detailed insights, especially for smaller fields.
  • Ground Truth Verification: The study relied on statistical properties and comparative analysis. Further verification through detailed field observation data and ground-based spectrometers is recommended.
  • Metric Generalizability: The identified metrics and thresholds are specific to the studied rice variety, environmental conditions, and heavy metals. Further research is needed to assess the generality of these findings across different regions, crops, and stress types.
  • Qualitative Analysis: The analysis of the metrics was largely qualitative. More detailed quantitative analysis and validation are suggested.
  • Other Environmental Factors: While efforts were made to control for other environmental factors, their potential influence on phenology alongside heavy metals warrants further investigation.

Summary updated

BibTeX
@article{2017extraction,
  author = {Chris H.Q. Ding (丁宏强) and Shuyuan Liu and Xiangnan Liu and Meiling Liu and Ling Wu and Chao Ding and Zhi Huang},
  title = {{Extraction of Rice Phenological Differences under Heavy Metal Stress Using EVI Time-Series from HJ-1A/B Data}},
  journal = {Sensors},
  year = {2017},
  doi = {10.3390/s17061243},
}
Figure