EeLISA: Combating Global Warming Through the Rapid Analysis of Eelgrass Wasting Disease

Brendan H. Rappazzo, Morgan E. Eisenlord, Olivia J. Graham, Lillian R. Aoki, Phoebe D. Dawkins, Drew Harvell, Carla P. GomesAAAI 2021 · OpenReview

Abstract

Global warming is the greatest threat facing our planet, and is causing environmental disturbance at an unprecedented scale. We are strongly positioned to leverage the advancements of Artificial Intelligence (AI) and Machine Learning (ML) which provide humanity, for the first time in history, an analysis and decision making tool at massive scale. Strong evidence supports that global warming is contributing to marine ecosystem decline, including eelgrass habitat. Eelgrass is affected by an opportunistic marine pathogen and infections are likely exacerbated by rising ocean temperatures. The necessary disease analysis required to inform conservation priorities is incredibly laborious, and acts as a significant bottleneck for research. To this end, we developed EeLISA (Eelgrass Lesion Image Segmentation Application). EeLISA enables ecologist experts to train a segmentation module to perform this crucial analysis at human level accuracy, while minimizing their labeling time and integrating into their existing workflow. EeLISA has been deployed for over 16 months, and has facilitated the preparation of four manuscripts including a critical eelgrass study ranging from Southern California to Alaska. These studies, utilizing EeLISA, have led to scientific insight and discovery in marine disease ecology.

Introduction

Eelgrass (Zostera marina) is an ecological powerhouse, sequestering carbon up to 35 times faster than tropical rainforests. However, rising ocean temperatures are fueling the spread of Eelgrass Wasting Disease (EWD). To understand and combat this, ecologists need to measure two key metrics: prevalence (is the plant sick?) and severity (what percentage of the tissue is lesioned?).

The primary bottleneck in this research is the analysis phase. Traditionally, an expert must manually trace healthy and diseased tissue on digital scans—a process that takes roughly 30 minutes per scan. With a single survey generating hundreds of scans, the "human bandwidth" required for large-scale, multi-site studies simply did not exist, effectively stalling critical climate research.

The EeLISA Framework

To address this bottleneck, we developed EeLISA (Eelgrass Lesion Image Segmentation Application). EeLISA is not just a model, but a comprehensive tool consisting of three main components:

  1. Segmentation CNN Module: The core AI that performs pixel-wise classification using architectures like UNet and DeepLab V3.
  2. Labeling/Web Application: A critical "human-in-the-loop" interface that allows experts to review and edit AI-generated masks.
  3. Analysis Script: A tool that automatically calculates ecological metrics from the final masks and exports them for research use.
The EeLISA segmentation pipeline: an input scan is processed by a CNN architecture to produce a pixel-wise mask identifying healthy (green) and diseased (red) tissue.

While the CNN architectures are established, the integration into the ecological workflow is the true innovation. By creating a feedback loop where expert corrections improve future model iterations, we minimized labeling fatigue and ensured high-quality data.

Datasets and Training

The study utilized a newly curated dataset of 1,036 eelgrass scans collected from sites ranging from Southern California to Alaska. This dataset is unique because it captures the natural variation of eelgrass across a vast latitudinal gradient, making the model robust to different environmental appearances.

Representative digital scans of eelgrass blades. Manually analyzing these for lesions is a labor-intensive process that EeLISA automates.

To train the models on this dataset, we employed extensive data augmentation and a combined loss function (Dice Loss + Cross Entropy) to handle the class imbalance between background, healthy tissue, and small lesions. Interestingly, while DeepLab V3 performed best on general segmentation benchmarks, UNet proved superior for specific ecological metrics like Severity, highlighting the importance of domain-specific evaluation.

Key Results and Real-World Impact

EeLISA has demonstrated transformative results for marine ecology research:

  • Massive Speedup: EeLISA analyzed a year's worth of data (694 scans) in roughly 5 minutes. Manually, this would have taken an expert approximately 347 hours, representing a 5,000x increase in efficiency.
  • Human-Level Accuracy: The system achieved a prevalence accuracy of over 96%, matching the consistency of expert researchers.
  • Democratization of Expertise: The AI allows non-experts or smaller research teams to conduct high-level disease analysis that previously required decades of specialized experience.

At the time of publication, EeLISA had been deployed for 16 months, facilitating four scientific manuscripts and a massive NSF-funded study across 32 sites on the US West Coast.

Conclusion

EeLISA represents a shift from reactive analysis to a scalable, AI-driven approach to marine conservation. Future work involves expanding the system to other seagrass species and potentially integrating it with aerial or satellite imagery to monitor eelgrass meadows at a global scale. The ultimate goal is to leverage these tools for predictive modeling of marine disease outbreaks in a warming world.

Summary updated

Abstract

Global warming is the greatest threat facing our planet, and is causing environmental disturbance at an unprecedented scale. We are strongly positioned to leverage the advancements of Artificial Intelligence (AI) and Machine Learning (ML) which provide humanity, for the first time in history, an analysis and decision making tool at massive scale. Strong evidence supports that global warming is contributing to marine ecosystem decline, including eelgrass habitat. Eelgrass is affected by an opportunistic marine pathogen and infections are likely exacerbated by rising ocean temperatures. The necessary disease analysis required to inform conservation priorities is incredibly laborious, and acts as a significant bottleneck for research. To this end, we developed EeLISA (Eelgrass Lesion Image Segmentation Application). EeLISA enables ecologist experts to train a segmentation module to perform this crucial analysis at human level accuracy, while minimizing their labeling time and integrating into their existing workflow. EeLISA has been deployed for over 16 months, and has facilitated the preparation of four manuscripts including a critical eelgrass study ranging from Southern California to Alaska. These studies, utilizing EeLISA, have led to scientific insight and discovery in marine disease ecology.
BibTeX
@inproceedings{rappazzo2021eelisa,
  author = {Brendan H. Rappazzo and Morgan E. Eisenlord and Olivia J. Graham and Lillian R. Aoki and Phoebe D. Dawkins and Drew Harvell and Carla P. Gomes},
  title = {{EeLISA: Combating Global Warming Through the Rapid Analysis of Eelgrass Wasting Disease}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2021},
  url = {https://openreview.net/forum?id=9XHtVDm7o9},
}
Figure