data-centers/database-tables.md

# Database Tables Documentation

## Database Configuration

**Database Name**: `data_centers`  
**Type**: PostgreSQL with PostGIS extension  
**Connection**: Environment variables from `~/.zsh_secrets`
- `PGWEB_HOST`: Database host
- `PGWEB_PORT`: Database port (typically 5432)
- `PGWEB_USER`: Database user
- `PGWEB_PASSWORD`: Database password
- `PGWEB_DATABASE`: Database name (`data_centers`)

## Table Organization

Tables are organized into four categories:
1. **Core Data Center Tables** - Master inventories and source data
2. **Enrichment Tables** - Data centers joined with contextual data
3. **Base Layer Tables** - Geographic and demographic reference layers
4. **Infrastructure Tables** - Energy and connectivity infrastructure

---

## Core Data Center Tables

### `master_data_centers`
**Rows**: 1,833  
**Purpose**: Canonical data center inventory - deduplicated merge of curated + OSM sources

**Key Columns**:
- `id` (INTEGER) - Unique identifier
- `name` (TEXT) - Facility name
- `address` (TEXT) - Street address
- `city` (TEXT) - City
- `state` (TEXT) - State code
- `latitude` (DOUBLE PRECISION) - Latitude
- `longitude` (DOUBLE PRECISION) - Longitude
- `geom` (GEOMETRY) - PostGIS point geometry (EPSG:4326)
- `operator` (TEXT) - Operator/owner
- `power_mw` (DOUBLE PRECISION) - Power capacity in megawatts (sparse: 5.9% populated)
- `source` (TEXT) - Data source (`curated`, `osm`, or `both`)
- `osm_id` (TEXT) - OpenStreetMap ID if applicable
- `geocode_method` (TEXT) - Geocoding provenance

**Notes**: 
- 108 of 1,833 facilities have power ratings
- 45 facilities use city-precision fallback coordinates
- Operator strings have fragmentation issues ("Meta" vs. "Meta, Inc.")

### `us_dc_sample_geocoded`
**Rows**: 1,489  
**Purpose**: Original curated sample with geocoding provenance (superseded by `master_data_centers`)

**Key Columns**:
- `name`, `address`, `city`, `state`, `zip`
- `latitude`, `longitude`, `geom`
- `operator`, `power_mw`
- `census_lat`, `census_lon` - Census TIGER geocode results
- `nominatim_lat`, `nominatim_lon` - Nominatim fallback results
- `geocode_source` - Which geocoder was used

### `osm_data_centers`
**Rows**: 1,549  
**Purpose**: Raw OpenStreetMap-derived facilities

**Key Columns**:
- `osm_id` (TEXT) - OSM element ID
- `osm_type` (TEXT) - `node`, `way`, or `relation`
- `name` (TEXT) - OSM name tag
- `latitude`, `longitude`, `geom`
- `tags` (JSONB) - All OSM tags as JSON
- `operator` (TEXT) - Extracted from OSM tags
- `city`, `state`, `country`

**Notes**: Fetched via Overpass API with query for `telecom=data_center` or `building=data_center`

### `master_data_center_spatial_clusters`
**Rows**: 1,831  
**Purpose**: DBSCAN cluster assignments for master data centers

**Key Columns**:
- All columns from `master_data_centers`
- `cluster_id` (INTEGER) - Cluster assignment (-1 = noise/singleton)
- `cluster_size` (INTEGER) - Number of facilities in cluster
- `cluster_label` (TEXT) - Human-readable cluster name

**Notes**: DBSCAN parameters: eps=15 km, min_samples=2

---

## Enrichment Tables

### `data_center_census_tracts_2024`
**Rows**: 1,815  
**Purpose**: Per-facility demographics from containing Census tract

**Key Columns**:
- All columns from `master_data_centers`
- `geoid` (TEXT) - 11-digit Census tract GEOID
- `state_fips`, `county_fips`, `tract`
- **Population**: `total_population`, `population_density_sq_mi`
- **Age**: `median_age`, `under_18_pct`, `over_65_pct`
- **Race/Ethnicity**: `white_nh_pct`, `black_nh_pct`, `asian_nh_pct`, `hispanic_pct`
- **Economics**: `median_household_income`, `per_capita_income`, `poverty_rate`
- **Education**: `bachelors_or_higher_pct`, `high_school_or_higher_pct`
- **Housing**: `median_home_value`, `median_rent`, `homeownership_rate`
- **Broadband**: `broadband_pct` - Households with broadband subscription

**Source**: ACS 2024 5-year estimates

**Notes**: 
- 18 of 1,833 facilities failed tract join (geocoding issues)
- Data from `_dc_census_tract_acs_2024` base table

### `data_center_watershed_huc8`
**Rows**: 1,833  
**Purpose**: Per-facility watershed assignment

**Key Columns**:
- All columns from `master_data_centers`
- `huc8` (TEXT) - 8-digit Hydrologic Unit Code
- `watershed_name` (TEXT) - Watershed name
- `watershed_area_sq_km` (DOUBLE PRECISION)
- `states` (TEXT) - States intersecting watershed

**Source**: USGS Watershed Boundary Dataset

**Notes**: 257 unique HUC8 watersheds contain at least one data center

### `data_center_nri_exposure`
**Rows**: 1,833  
**Purpose**: Per-facility FEMA National Risk Index hazard exposure scores

**Key Columns**:
- All columns from `master_data_centers`
- `nri_id` (TEXT) - Census tract GEOID (matches `geoid` from demographics)
- `risk_score` (DOUBLE PRECISION) - Overall NRI risk score
- `social_vulnerability` (DOUBLE PRECISION) - Social vulnerability index
- **Hazard-specific risk scores** (18 hazards):
  - `avalanche_risk`, `coastal_flooding_risk`, `cold_wave_risk`
  - `drought_risk`, `earthquake_risk`, `hail_risk`
  - `heat_wave_risk`, `hurricane_risk`, `ice_storm_risk`
  - `landslide_risk`, `lightning_risk`, `riverine_flooding_risk`
  - `strong_wind_risk`, `tornado_risk`, `tsunami_risk`
  - `volcanic_activity_risk`, `wildfire_risk`, `winter_weather_risk`

**Source**: FEMA National Risk Index (December 2025 release)

### `data_center_rdh_precinct_vote_matches`
**Rows**: Varies  
**Purpose**: Per-facility precinct-level election results

**Key Columns**:
- Data center identifiers
- `precinct_name`, `precinct_id`
- `election_year`, `office`
- `candidate`, `party`, `votes`
- `vote_share_pct`

**Source**: Redistricting Data Hub precinct shapefiles

**Notes**: Spatial join to voting precincts (point-in-polygon)

---

## Base Layer Tables

### `_dc_census_tract_acs_2024`
**Rows**: 85,382  
**Purpose**: ACS 2024 demographics for all Census tracts in states with data centers

**Key Columns**:
- `geoid` (TEXT) - 11-digit tract GEOID (PRIMARY KEY)
- `name` (TEXT) - Tract name
- `state_fips`, `county_fips`, `tract`
- **Full ACS 5-year estimates** (85+ columns):
  - Population by age, sex, race/ethnicity
  - Households, families, housing units
  - Income, poverty, education, employment
  - Housing values, rents, costs
  - Broadband, computer access
  - Commuting, vehicles

**Source**: Census ACS 2024 5-year estimates API

**Notes**: Universe limited to 46 states with data centers (excludes DC-free states)

### `_dc_census_tract_boundaries_2024`
**Rows**: 85,058  
**Purpose**: TIGER 2024 tract polygons for data center states

**Key Columns**:
- `geoid` (TEXT) - 11-digit tract GEOID
- `name` (TEXT) - Tract name
- `state_fips`, `county_fips`, `tract_code`
- `geom` (GEOMETRY) - Polygon geometry (EPSG:4326)
- `area_land_sq_m` (DOUBLE PRECISION) - Land area in square meters
- `area_water_sq_m` (DOUBLE PRECISION) - Water area in square meters

**Source**: Census TIGER/Line 2024

### `ruca_codes_2020_tract`
**Rows**: 85,528  
**Purpose**: USDA Rural-Urban Commuting Area codes for metro/rural classification

**Key Columns**:
- `geoid` (TEXT) - 11-digit tract GEOID (matches Census tracts)
- `ruca_code` (TEXT) - Primary RUCA code (1-10)
- `ruca_category` (TEXT) - Simplified category:
  - `Metropolitan` (codes 1-3)
  - `Micropolitan` (codes 4-6)
  - `Small town` (codes 7-9)
  - `Rural` (code 10)
- `ruca_description` (TEXT) - Full RUCA code description
- `population_2020` (INTEGER)

**Source**: USDA Economic Research Service RUCA 2020

**Notes**: 
- Based on 2020 Census tracts and 2010-2020 commuting patterns
- 7 data centers failed RUCA join (Puerto Rico / non-US)

### `watershed_huc8`
**Rows**: 2,139  
**Purpose**: USGS HUC8 subbasin polygons for water-stress analysis

**Key Columns**:
- `huc8` (TEXT) - 8-digit Hydrologic Unit Code (PRIMARY KEY)
- `name` (TEXT) - Watershed name
- `geom` (GEOMETRY) - Polygon geometry (EPSG:4326)
- `area_sq_km` (DOUBLE PRECISION)
- `states` (TEXT) - Comma-separated state codes
- `dc_count` (INTEGER) - Number of data centers in watershed

**Source**: USGS Watershed Boundary Dataset

**Notes**: 
- 257 of 2,139 watersheds contain at least one data center
- Top 15 watersheds contain 50% of all US data centers

### `nri_census_tracts`
**Rows**: ~84,000  
**Purpose**: Full FEMA National Risk Index by Census tract

**Key Columns**:
- `nri_id` (TEXT) - Census tract GEOID
- `state_name`, `county_name`, `tract_name`
- **460+ columns** including:
  - Overall risk scores and ratings
  - Expected annual loss (dollars and building value %)
  - Social vulnerability components (15 factors)
  - Community resilience score
  - Individual hazard risk scores (18 hazards)
  - Exposure, annualized frequency, historic loss ratios per hazard

**Source**: FEMA National Risk Index v2.1 (December 2025)

**Notes**: 
- Massive table with comprehensive natural hazard risk data
- Join to data centers via `geoid` field
- See [FEMA NRI Technical Documentation](https://hazards.fema.gov/nri/)

---

## Infrastructure Tables

### Energy Infrastructure

#### `energy_eia_operating_generator_capacity_flat`
**Rows**: 4.7 million  
**Purpose**: EIA generator inventory with lat/lon/MW (monthly 2008-2026)

**Key Columns**:
- `plant_id` (INTEGER) - EIA plant ID
- `generator_id` (TEXT) - Generator unit ID
- `plant_name` (TEXT)
- `latitude`, `longitude`, `geom`
- `state`, `county`
- `utility_name`, `operator_name`
- `nameplate_capacity_mw` (DOUBLE PRECISION)
- `technology` (TEXT) - Generation technology
- `energy_source_1`, `energy_source_2` - Primary fuel codes
- `operating_month`, `operating_year` - When unit became operational
- `status` (TEXT) - Operating, standby, retired, etc.
- `report_month`, `report_year` - Data snapshot date

**Source**: EIA Form 860 via API

**Notes**:
- "Flat" means denormalized for fast spatial queries
- Each generator-month is a row (4.7M rows from monthly snapshots)
- Use for proximity analysis (e.g., "all generators within 50 km of data center")

#### `energy_eia_facility_fuel_flat`
**Rows**: Varies  
**Purpose**: Monthly generation by plant/fuel

**Key Columns**:
- `plant_id`, `plant_name`
- `report_month`, `report_year`
- `energy_source` (TEXT) - Fuel code
- `net_generation_mwh` (DOUBLE PRECISION)
- `fuel_consumed_mmbtu` (DOUBLE PRECISION)

**Source**: EIA Form 923 via API

#### `energy_eia_seds_flat`
**Rows**: 2.57 million  
**Purpose**: Annual state energy consumption/production (1960-2024)

**Key Columns**:
- `state_code` (TEXT)
- `year` (INTEGER)
- `msn` (TEXT) - Mnemonic series names (e.g., `TETCB` = total energy consumption)
- `value` (DOUBLE PRECISION) - Energy in trillion BTU
- `unit` (TEXT)
- `description` (TEXT) - Human-readable MSN description

**Source**: EIA State Energy Data System (SEDS)

**Notes**: 
- Annual aggregates by state
- Use for state-level energy context analysis

---

### Connectivity Infrastructure

#### `internet_cables`
**Rows**: 693  
**Purpose**: Submarine cable routes

**Key Columns**:
- `cable_id` (TEXT) - Unique cable identifier
- `cable_name` (TEXT) - Official cable name
- `geom` (GEOMETRY) - LineString geometry (EPSG:4326)
- `rfs_year` (INTEGER) - Ready For Service year
- `length_km` (DOUBLE PRECISION)
- `owners` (TEXT[]) - Array of owner names
- `landing_points` (TEXT[]) - Array of landing point names

**Source**: TeleGeography-style cable database

**Notes**: 
- 693 unique submarine cables
- Geometry is approximate route (not exact seabed path)

#### `internet_cable_landing_points`
**Rows**: 3,361  
**Purpose**: Cable landing points (where cables come ashore)

**Key Columns**:
- `landing_point_id` (TEXT) - Unique identifier
- `name` (TEXT) - Landing point name
- `city`, `country`
- `latitude`, `longitude`, `geom`
- `cables` (TEXT[]) - Array of cable names landing at this point
- `cable_count` (INTEGER)

**Source**: TeleGeography-style cable database

**Notes**: 
- Used for proximity analysis (how close are data centers to cable landings?)
- **Key finding**: Data centers are NOT systematically closer to cables than ordinary US cities

#### `internet_city_dominance`
**Rows**: 4,552  
**Purpose**: City-level IPs/capacity (internet hub strength proxy)

**Key Columns**:
- `city` (TEXT)
- `country` (TEXT)
- `latitude`, `longitude`, `geom`
- `ip_addresses` (INTEGER) - Number of routable IP addresses
- `capacity_rank` (INTEGER) - Relative capacity ranking

**Source**: Internet topology datasets

**Notes**: Proxy for "internet hub" strength (not directly used in main analyses)

---

### Broadband

#### `fcc_bdc_location_provider_aggregates`
**Rows**: Varies  
**Purpose**: FCC BDC provider availability aggregated by county/tract

**Key Columns**:
- `geoid` (TEXT) - County or tract GEOID
- `geography_level` (TEXT) - `county` or `tract`
- `provider_count` (INTEGER)
- `technology_counts` (JSONB) - Count by technology type
- `max_download_mbps`, `max_upload_mbps`

**Source**: FCC Broadband Data Collection (BDC)

#### `fcc_bdc_broadband_connection_table`
**Rows**: Varies  
**Purpose**: Per-data-center broadband provider availability

**Key Columns**:
- Data center identifiers
- `provider_id`, `provider_name`
- `technology` (TEXT)
- `max_advertised_download_speed`, `max_advertised_upload_speed`
- `low_latency` (BOOLEAN)

**Source**: FCC BDC, joined to data center locations

**Notes**: Built by `build_fcc_bdc_broadband_connection_table.py`

---

### Other Tables

#### `opposition_cases_geocoded`
**Rows**: 18  
**Purpose**: Geocoded community-opposition cases against data center builds

**Key Columns**:
- `case_id` (TEXT) - Unique identifier
- `developer` (TEXT) - Proposed developer/operator
- `investment_billions` (DOUBLE PRECISION) - Investment amount in billions
- `outcome` (TEXT) - Case outcome (approved, rejected, pending)
- `governance_response` (TEXT) - Government response
- `latitude`, `longitude`, `geom`

**Source**: Compiled from news archives

**Notes**: Loaded but currently unused - see research-ideas.md for proposed analyses

#### `census_tract_huc8_link`
**Rows**: 806  
**Purpose**: Tract↔HUC8 spatial overlap table

**Key Columns**:
- `geoid` (TEXT) - Census tract GEOID
- `huc8` (TEXT) - HUC8 watershed code
- `overlap_pct` (DOUBLE PRECISION) - Percentage of tract overlapping watershed

**Notes**: Useful for downstream tract-level water-stress joins; limited to tracts containing data centers

#### `im3_state_projected_moderate_50`
**Rows**: 328  
**Purpose**: PNNL IM3 projected data center siting (moderate growth, gravity weight 0.50)

**Key Columns**:
- `facility_id` (TEXT)
- `state` (TEXT)
- `cost_millions` (DOUBLE PRECISION)
- `it_mw` (DOUBLE PRECISION) - IT load in megawatts
- `cooling_water_demand_gal_per_day` (DOUBLE PRECISION)
- `latitude`, `longitude`, `geom`

**Source**: PNNL Integrated Multisector Multiscale Modeling (IM3)

**Notes**: Loaded but unused - potential for forward-projection analysis

#### `im3_projected_state_demand_summary`
**Rows**: 31  
**Purpose**: State-level rollup of IM3 projected facility counts, IT MW, and cooling demand

**Key Columns**:
- `state` (TEXT)
- `facility_count` (INTEGER)
- `total_it_mw` (DOUBLE PRECISION)
- `total_cooling_demand_mgd` (DOUBLE PRECISION) - Million gallons per day

**Source**: IM3 model outputs

#### `utility_rate_tracker_2025_2028`
**Rows**: 374  
**Purpose**: Utility rate-increase tracker by provider × state × service type

**Key Columns**:
- `provider` (TEXT) - Utility provider name
- `state` (TEXT)
- `service_type` (TEXT)
- `effective_date` (DATE)
- `monthly_increase_dollars` (DOUBLE PRECISION)
- `percent_increase` (DOUBLE PRECISION)

**Source**: Utility rate tracker database

**Notes**: Loaded but unused in demographic/energy analysis

#### `energy_atlas_layers_catalog`
**Rows**: ~5  
**Purpose**: Metadata catalog of EIA layers ingested

**Key Columns**:
- `table_name` (TEXT)
- `source_url` (TEXT)
- `import_timestamp` (TIMESTAMP)
- `row_count` (INTEGER)

**Notes**: Created by `ingest_eia_energy_layers.py`

---

## Commonly Used Joins

### Data Center to Demographics
```sql
SELECT 
    dc.*,
    ct.median_household_income,
    ct.bachelors_or_higher_pct,
    ct.broadband_pct
FROM master_data_centers dc
JOIN data_center_census_tracts_2024 ct 
    ON dc.id = ct.id;
```

### Data Center to Watershed
```sql
SELECT 
    dc.*,
    w.huc8,
    w.watershed_name
FROM master_data_centers dc
JOIN data_center_watershed_huc8 dw ON dc.id = dw.id
JOIN watershed_huc8 w ON dw.huc8 = w.huc8;
```

### Data Center to Energy Infrastructure (50 km radius)
```sql
SELECT 
    dc.id,
    dc.name,
    SUM(eg.nameplate_capacity_mw) AS total_capacity_50km
FROM master_data_centers dc
JOIN energy_eia_operating_generator_capacity_flat eg
    ON ST_DWithin(
        dc.geom::geography,
        eg.geom::geography,
        50000  -- 50 km in meters
    )
WHERE eg.status = 'OP'  -- Operating only
GROUP BY dc.id, dc.name;
```

### Data Center to FEMA Hazard Risk
```sql
SELECT 
    dc.*,
    nri.risk_score,
    nri.wildfire_risk,
    nri.drought_risk,
    nri.heat_wave_risk
FROM master_data_centers dc
JOIN data_center_census_tracts_2024 ct ON dc.id = ct.id
JOIN nri_census_tracts nri ON ct.geoid = nri.nri_id;
```

---

## Table Naming Conventions

- **`master_*`** - Canonical, deduplicated tables (use these for analysis)
- **`data_center_*`** - Data center-specific enrichment tables
- **`_dc_*`** - Base layers scoped to data center states (underscore prefix = private/internal)
- **`energy_eia_*`** - EIA energy data
- **`internet_*`** - Connectivity infrastructure
- **`fcc_bdc_*`** - FCC Broadband Data Collection

---

## Indexes and Performance

All tables have spatial indexes on `geom` columns for fast spatial joins:
```sql
CREATE INDEX idx_tablename_geom ON tablename USING GIST(geom);
```

Key `geoid` columns are indexed for fast demographic joins:
```sql
CREATE INDEX idx_tablename_geoid ON tablename(geoid);
```

---

## Maintenance Notes

### Updating Data Centers
1. Run `load_postgis_osm_data_centers.py` to refresh OSM data
2. Run `build_master_data_centers.py` to rebuild master table
3. Run enrichment scripts to update joins

### Updating Demographics
1. Update `_dc_census_tract_acs_2024` from Census API
2. Run `create_data_center_census_tract_table.py --replace-final`

### Updating Energy Data
```bash
python3 ingest_eia_energy_layers.py --category power --update
```

---

## Schema Export

To export the full schema:
```bash
pg_dump -h $PGWEB_HOST -U $PGWEB_USER -d data_centers --schema-only > schema.sql
```

To list all tables:
```sql
SELECT schemaname, tablename, pg_size_pretty(pg_total_relation_size(schemaname||'.'||tablename))
FROM pg_tables
WHERE schemaname = 'public'
ORDER BY pg_total_relation_size(schemaname||'.'||tablename) DESC;
```

---

## Contact

For database access or questions, contact the repository owner.