1980-2014
Historical trend window used for model and observational comparisons
Model Development and Evaluation
Historical Precipitation Trends
Large-ensemble analyses separate forced and internal components of precipitation trends, showing that regional observations must be interpreted in a variability-aware framework.
AMIP6 Ensembles
Forced vs Internal Variability
Observed Trend Comparison
AMIP6 Multi-Model
Large ensembles isolate forced responses from internal atmospheric noise
ESSD (2025)
Confronting historical precipitation trends with observations
Related Publication
Liang, Y., et al., 2025: Confronting historical precipitation trends in models with observations: forced signal and atmospheric internal variability. Earth System Science Data.
Scientific Logic
- Question: How much of observed regional precipitation trend is externally forced versus internal variability?
- Method: Large-ensemble AMIP6-style analysis separating forced signal from internal atmospheric variability and comparing with observations.
- Mechanism: Forced thermodynamic moistening interacts with circulation variability, so regional trends can deviate strongly from the ensemble-mean forced pattern.
- Main Findings: Many observed regional trends fall within modeled variability bounds, emphasizing ensemble-context interpretation before model rejection.
Scientific Question
How much of observed regional precipitation change is externally forced versus internally generated, and how consistently do atmospheric models reproduce those signals?
Approach
- Analyze AMIP6 model ensembles to estimate forced precipitation-trend components.
- Compare ensemble means and spread with observational products and pattern-correlation metrics.
- Diagnose zonal-mean and regional wetting/drying structures across hemispheres.
Key Findings
- Internal variability can dominate regional trend realizations even under the same external forcing.
- Observed regional trends often fall within modeled variability envelopes.
- Model evaluation of precipitation trend skill should use ensemble context rather than single-realization comparison.
Figure
Ensemble-Mean Trend Pattern
