An interactive dashboard for visualizing citation metrics, research impact, and model capability levels across NASA JPL Earth science models and observation missions.
- Overview
- Supported Models & Missions
- Features
- Technologies
- Setup and Installation
- Scripts
- Data Sources
- Citation Data Verification
- Deployment
- Contributing
- Governance
- Citation
- License
The JEME Dashboard is a React-based web application that provides interactive visualizations of how NASA JPL Earth science models are cited and used in the scientific literature. It covers two enterprise views:
- JEME (JPL Earth Modeling Enterprise) — 8 science models
- JEOE (JPL Earth Observation Enterprise) — NASA Earth observation missions
| Model | Full Name |
|---|---|
| RAPID | River Application for Parallel computation of Discharge |
| CARDAMOM | CARbon DAta MOdel framework |
| CMS-Flux | Carbon Monitoring System Flux |
| ECCO | Estimating the Circulation and Climate of the Ocean |
| ISSM | Ice-sheet and Sea-level System Model |
| MOMO-CHEM | Chemistry-Climate Model |
| LES | Large Eddy Simulation |
| EDMF | Eddy Diffusivity Mass Flux |
| Mission | Full Name |
|---|---|
| GRACE | Gravity Recovery and Climate Experiment |
| SWOT | Surface Water and Ocean Topography |
| TROPESS | TRopospheric Ozone and its Precursors from Earth System Sounding |
- Multi-Model Dashboard — Overview of all models with cross-model network analysis
- Citation Trends — Annual and cumulative citations over time per model
- Research Domain Analysis — Distribution of citations across scientific fields
- Engagement Levels — Depth of model usage in research (L1–L4)
- Geographic Impact — Map-based visualization of global research impact
- Satellite Missions — Track which missions/instruments are referenced in citations
- Journal Distribution — See which journals cite each model most frequently
- Network Analysis — Cross-model connectivity via bridge papers, shared authors, domain overlap
- Model Capability Levels (MCL) — Maturity assessment for each science model
- Uncertainty Quantification — 3-phase UQ pipeline with stochastic variance and skeptic agent review
- Private Sector Engagement — Alumni network and industry adoption explorer
- CSV Export — Export filtered citation data
- React 18 (Create React App)
- Recharts for data visualization
- D3 / TopoJSON for geographic maps
- Tailwind CSS v2 for styling
- Lucide React for icons
- Node.js v14+
- npm
git clone https://github.jpl.nasa.gov/kyun/jeme.git
cd jeme
npm install
npm startOpen http://localhost:3000/science-model-dashboard to view the dashboard.
| Command | Description |
|---|---|
npm start |
Start development server (binds to 0.0.0.0:3000) |
npm run build |
Build for production |
npm test |
Run tests |
npm run deploy |
Build and deploy to GitHub Pages (JPL) |
node scripts/clean_citation_data.js --all --dry-run |
Preview citation data cleanup across all models |
node scripts/clean_citation_data.js --model ECCO |
Clean a specific model's data |
python3 scripts/verify_peer_review.py --all --dry-run |
Preview non-peer-reviewed paper removal |
python3 scripts/verify_peer_review.py --model TROPESS |
Remove non-peer-reviewed papers for a specific model |
python3 scripts/compute_uncertainty.py --model RAPID |
Run Phase 1 uncertainty scoring |
python3 scripts/phase2_llm_confidence.py --model RAPID --sample 10 |
Run Phase 2 LLM confidence sampling |
python3 scripts/phase3_skeptic_agent.py --model RAPID |
Run Phase 3 skeptic agent review |
python3 scripts/fetch_ecco_venues.py |
Fetch missing journal/venue info for ECCO citations |
Citation data is stored as JSON in public/data/{MODEL_NAME}_analyzed.json. Two formats coexist:
- Crossref format: Used by RAPID, CMS-Flux, ISSM, CARDAMOM — fields:
title[],author[],DOI,container-title[] - Simplified format: Used by ECCO, LES, EDMF, MOMO-CHEM — fields:
title,authors,doi,venue,research_domain,engagement_level,missions_instruments[]
Always use extractPublicationData() from src/utils/dataUtils.js to normalize entries before processing.
Citation data undergoes a multi-agent verification pipeline that cross-validates entries using multiple external APIs and heuristic analysis.
| Agent | Source | Role |
|---|---|---|
| Semantic Scholar | Paper metadata, citation graphs | Primary data source; provides citing_team_paper links |
| Crossref API | DOI resolution, journal metadata | Venue enrichment; validates DOI existence and journal attribution |
| Keyword Classifier | Title + abstract text | Topical relevance scoring via domain-specific keyword matching |
| Team Paper Categorizer | Team paper titles | Classifies team papers into relevance tiers (Core / Infrastructure / Related / Unrelated) |
| Deduplication | DOI + title matching | Removes duplicate entries across data sources |
All entries are peer-reviewed publications. Non-peer-reviewed entries (preprints, theses, conference abstracts, technical reports) are excluded.
| Model | Papers in Dashboard | Confidence |
|---|---|---|
| RAPID | 336 | High (96% keyword match) |
| CARDAMOM | 514 | High (95% keyword match) |
| CMS-Flux | 646 | High (95% keyword match) |
| ECCO | 16,320 | Medium-High |
| ISSM | 3,491 | High (96% keyword match; metadata repaired) |
| MOMO-CHEM | 983 | High (95% keyword match) |
| LES | 168 | High (93% keyword match) |
| EDMF | 810 | High (91% keyword match) |
| GRACE | 3,084 | High |
| SWOT | 371 | High |
| TROPESS | 771 | High |
- Overall false removal rate: ~8–12% of removed entries may have been legitimate citations with indirect relevance.
- Overall false retention rate: ~3–5% of retained entries may be false positives (primarily among entries lacking abstracts).
- Keyword-based relevance: Cannot detect papers that use a model's output without naming it. Estimated miss rate: 5–10%.
- Single-pass classification: Each entry is classified once without human-in-the-loop review of borderline cases.
npm run deployDeploys to GitHub Pages at https://github.jpl.nasa.gov/pages/kyun/jeme/.
Please read CODE_OF_CONDUCT.md before contributing. To report bugs or request features, use the issue tracker.
Pull requests are welcome — please use the PR template and ensure npm run build passes before submitting.
See GOVERNANCE.md for project decision-making, maintainership, and release policies.
If you use this dashboard in your research, please cite it using the metadata in CITATION.cff.
This project is licensed under the MIT License — see the LICENSE file for details.