An Earth and environmental workflow concept defining repeatable steps used to quantify and report observations. It governs data collection, processing, quality checks, and uncertainty treatment required for defensible results. It does not ensure correctness without documented procedures, verification, and appropriate handling of missing or biased data. It supports transparency and improvement by making results auditable and comparable across time. The concept is generally stable, though automation and data standards evolve over time.

Principle

Validation requires independence from calibration data, pre-specified performance metrics (e.g., ROC, precision, recall, false alarm rate), and transparent reporting of spatial/temporal limits so that users can judge fitness for purpose rather than accepting undocumented or anecdotal claims.

Demonstration

A susceptibility model trained on events up to year N is validated by testing its prediction against landslides that occurred in years N+1 to N+3, calculating hit rate, false alarm ratio, and area under the ROC curve, and reporting where performance degrades by terrain class or storm type.

Misapplication

Using the same dataset for both calibration and validation, or relying only on qualitative field impressions without systematic metrics, gives misleading indicators of performance and overstates confidence in operational decision support.

Consequence

Proper validation defines the operational limits and confidence in landslide products, informs necessary refinements, and provides evidence for stakeholders to adopt, adapt, or reject tools and thresholds for hazard management.

Reversal

The opposite action is calibration: rather than testing independently, calibration alters model or sensor settings to improve fit; validation without prior calibration or calibration without subsequent validation both distort the intended quality assurance cycle.

Boundary

Applies to spatial, temporal, and methodological assessment of products and sensors; excludes informal acceptance without metrics, and is constrained by availability of independent, representative event records and by changes in drivers over time.

Semantic Tension

There is tension between statistical metrics of predictive skill (which may favor some objectives) and practitioner needs for actionable warnings; validation may show low statistical performance yet still produce useful operational cues in combination with expert interpretation.

Synthesis

Landslide validation is the rigorous, independent testing of landslide-related outputs against reserved or later observations using explicit metrics and metadata to establish performance, limits, and confidence for responsible application.