SBAS Knowledge Base

Ionospheric Model Validation

3 min read

Ionospheric Model Validation

Definition status

This note defines ionospheric model validation as the systematic comparison of empirical or physics-based ionospheric model predictions against independent observational data to quantify model accuracy, identify systematic biases, and assess regional or conditional validity.

Boundary:

  • This is a methodology concept note, not a specific model evaluation.
  • It does not advocate for one model over another.
  • It captures the generic validation framework applied in the Indonesian GNSS-RO study.
  • It does not define operational SBAS ionospheric correction, GIVE, service-volume, or integrity-monitoring requirements. Route those claims through SBAS Corrections and Integrity Separation, SBAS Integrity, and standards/service-provider source notes.

Working definition

Model validation proceeds by:

  1. Computing model predictions for the same spatial, temporal, and geometric conditions as the observations
  2. Calculating error metrics between predicted and observed values
  3. Stratifying errors by relevant conditions (altitude, local time, season, geomagnetic state)
  4. Interpreting whether errors are systematic (bias), random (variance), or conditional (state-dependent)

Standard error metrics

MetricFormulaInterpretation
Biasmean(observed − predicted)Systematic over/under-estimation
RMSE√(mean((obs−pred)²))Total error magnitude
MAEmean(|obs − pred|)Average absolute error
Correlation (r)Pearson correlationLinear relationship strength
R²1 − (SS_res / SS_tot)Explained variance (negative = worse than mean)

Relationship to the vault

Parent domain

Sibling domains

Implementation connection

Methodological considerations

Spatial representativeness

A validation result at one location does not automatically generalize to another. The Indonesian study specifically found:

  • Western Indonesia (Sumatra/Java): bias 98.0 TECU
  • Central Indonesia (Kalimantan): bias 114.8 TECU
  • Eastern Indonesia (Sulawesi/Papua): bias 106.6 TECU

Temporal coverage

Sparse sampling (e.g., 5 or 8 days) captures seasonal snapshots but may miss:

  • Post-sunset equatorial irregularity events
  • Geomagnetic storm responses
  • Solar-cycle variation

The vault explicitly flags the post-sunset gap (18–21 LT) in the Indonesian dataset as an integrity-relevant limitation, not an evidence of low risk.

Model scope

IRI is a monthly-median empirical model. It does not predict day-to-day variability or storm conditions. Comparing IRI to instantaneous RO measurements is methodologically valid for assessing climatological bias, but not for evaluating operational forecast skill.

Current source anchors

Open provenance questions

  • Should model validation studies separate climatological accuracy from operational real-time accuracy?
  • How should validation results be aggregated across spatially and temporally heterogeneous regions?
  • What is the appropriate baseline for “acceptable” model error in an SBAS context?

See also

Graph View

Backlinks