← Back to Dashboard

Scoring Methodology

The 4-Level Severity Scale

LevelScoreCriteriaExample
SEVERE76–100Irreversible or catastrophic environmental damageMajor oil spill, species extinction event
SIGNIFICANT51–75Serious damage requiring years of recoveryLarge-scale deforestation, coral bleaching event
MODERATE26–50Notable impact with recovery possible in monthsLocalized pollution incident, habitat fragmentation
MINIMAL0–25Minor or localized impact, easily reversibleSmall chemical spill contained quickly

Three Dimensions

Each topic is scored across three dimensions, weighted by their relative importance:

Ecological Impact (40% weight)

Damage to ecosystems, biodiversity, and natural resources. Considers scale, reversibility, and cascading effects.

Health Impact (35% weight)

Direct and indirect effects on human health. Includes air quality, water contamination, disease vectors, and long-term exposure risks.

Economic Impact (25% weight)

Financial consequences including cleanup costs, lost productivity, supply chain disruptions, and long-term economic damage.

Why These Weights

Ecological damage receives the highest weight (40%) because it is often the hardest to reverse — extinct species cannot be restored and destroyed ecosystems take decades to recover. Health impact (35%) reflects the direct human cost of environmental events. Economic impact (25%) captures downstream financial consequences, which are significant but typically more recoverable than ecological or health damage.

How the Overall Score Works

The overall severity score is a weighted average of the three dimensions:

Score = (Eco × 0.40) + (Health × 0.35) + (Econ × 0.25)

Each dimension score ranges from 0 to 100. The resulting overall score also ranges from 0 to 100.

Urgency Levels

The overall score maps to an urgency level displayed as a colored badge:

Score RangeUrgencyBadge Color
80–100BreakingRed
60–79CriticalOrange
30–59ModerateYellow
0–29InformationalGreen

Data Sources

Articles are collected daily from major news sources via the GNews service. Each article is analyzed by a large language model (LLM) that classifies its environmental topic and scores the severity across all three dimensions. Scores are then aggregated at the topic level using the weighted formula above.

Limitations

  • LLM accuracy: Scores are generated by AI models, which can occasionally misinterpret context, severity, or nuance in news articles.
  • News source bias: Coverage varies by region and topic. Events in underreported areas may receive fewer articles and less accurate scoring.
  • Scoring subjectivity: Environmental severity is inherently subjective. Our rubric provides a consistent framework, but reasonable people may disagree on specific scores.
  • Temporal lag: Articles are processed in daily batches. Rapidly evolving situations may not reflect the latest developments.