gas-risiko.deData & API Access
DE
For analysts & developers

European gas storage data, ready for your workflow

gas-risiko.de publishes structured, machine-readable JSON data covering daily storage levels, 14-day ARIMAX forecasts with full quantile bands, 365-day seasonal outlooks, cross-border flow z-scores, and refill stress scenarios — updated every morning from primary sources.

Request access →Methodology →
Live data status

All datasets show export freshness versus the JSON build time. Production runs twice daily on Europe/Berlin time (≈06:00 morning refresh and ≈23:45 after AGSI+ evening publication), sourcing AGSI+, ENTSO-G, Open-Meteo, SMARD, and TTF. A GitHub Actions backup may run around 00:30 UTC. Amber in the table can mean an upstream source is still flagged stale even when the export is recent.

Available datasets

Four structured feeds

📊
14-day forecast
/api/data?country=DE
Update
2× daily ~06:00 & ~23:45 Berlin
Max lag
< 26 h
fill_level_pctstorage_pct_rankrequired_injection_rate_ppdayP10/P25/P50/P75/P90p_alarmsource_statuscoefficientsdiagnostics
📅
Annual outlook (365d)
/api/annual?country=DE
Update
2× daily ~06:00 & ~23:45 Berlin
Max lag
< 36 h
baseline p10/p50/p90canonical scenarioswinter_minimum (OLS)p_breach_15pctood_boundsess_paths
🔀
Flow corridors
/api/network?country=DE
Update
2× daily ~06:00 & ~23:45 Berlin
Max lag
< 26 h
flow_gwhcapacity_gwhutilization_pcttrailing_30d_meanzscore_30dstress_flag
⚠️
Refill stress scenarios
/api/refill-stress?country=DE
Update
2× daily ~06:00 & ~23:45 Berlin
Max lag
< 48 h
sensitivity_gridcritical_boundaryoct31_fill_pcttarget_shortfall_twhdays_below_15seasonal_rates
Countries covered: DE · NL · FR · IT · AT · CZ · PL — pass ?country=XX to all endpoints.
Sample response

Inspect a real payload

All responses are versioned JSON with a contract_version field checked in CI. Quantiles are always monotone; NaN values are serialised as null.

GET /api/data?country=DE (excerpt)
{
  "contract_version": "3.1",
  "generated_at": "2026-04-26T04:07:33.912Z",
  "latest": {
    "date": "2026-04-24",
    "fill_level_pct": 37.42,
    "gas_in_storage_twh": 182.1,
    "storage_pct_rank": 31.4,
    "required_injection_rate_ppday": 0.284,
    "d_current": 0.91
  },
  "predictions": [
    { "date": "2026-04-25", "P10": 36.8, "P50": 37.6, "P90": 38.4 },
    { "date": "2026-04-26", "P10": 36.2, "P50": 37.1, "P90": 38.0 }
  ],
  "source_status": {
    "agsi": { "stale": false, "last_date": "2026-04-24", "fetched_at": "2026-04-26T03:55Z" },
    "ttf":  { "stale": false, "last_date": "2026-04-25", "fetched_at": "2026-04-26T04:01Z" }
  }
}
Use cases

Who uses this data

📈
Gas & power trading desks
Monitor storage percentile rank vs history, track required injection pace vs current day-over-day change, and catch corridor z-score anomalies before prices move.
🛡️
Energy risk management
Use P10/P50/P90 annual trajectories and winter minimum OLS predictions to stress-test portfolios and validate VAR models against physical storage constraints.
🏦
Structured products & derivatives
The p_breach_15pct_any_winter probability and corridor stress flags provide independent physical reference points for exotic option pricing and winter-delivery contract structuring.
🔬
Academic & policy research
Structured historical fill series with source metadata, model diagnostics (Ljung-Box, VIF, HAC standard errors), and full methodology documentation published with each run.
⚡
Utilities & grid operators
Daily injection pace gap (pp/day needed to meet 90% Oct 31 target) and corridor utilisation ratios integrate directly into dispatch planning and balancing tools.
🤖
Quantitative model input
JSON payloads are versioned and schema-validated in CI. Supply to LLMs, time-series models, or scenario engines without manual data cleaning.
Scientific standards

Rigorous, non-alarmist, transparent

📐
ARIMAX with HAC errors
Country-specific ARIMAX models with Newey-West HAC standard errors. All diagnostics (Ljung-Box, Jarque-Bera, VIF, Durbin-Watson) exposed in API response.
🔄
Conformal calibration
Forecast bands are calibrated using rolling conformal prediction (κ multiplier) to achieve target 80% coverage on held-out windows. Coverage status exposed per country.
📊
Seasonal block bootstrap
365-day annual outlook uses recency-weighted seasonal path ensemble (not point forecast extrapolation). ESS tracks how many historical years dominate.
🚦
OOD detection
Scenario parameters exceeding ±3σ of training data are flagged with ood_flag and ood_variables — so users always know when they are extrapolating.

Integrate gas-risiko.de data into your stack

Write to discuss data licensing, update SLAs, custom country coverage, white-label deployment, or integration support.

marc.harms@mh-analytics.eu →

Data is updated daily from public primary sources (AGSI+, ENTSO-G, Open-Meteo, SMARD, TTF). This site does not constitute financial advice. Forecasts carry explicit uncertainty — always check diagnostics and interval coverage before use in production systems.