Increasing model vertical resolution may not necessarily lead to improved atmospheric predictability

Area of Science:

• Atmospheric science
• Chaos theory
• Numerical modeling

Background:

• Atmospheric predictability is limited by sensitive dependence on initial conditions, a feature of chaos.
• The Lorenz system, a simplified model of thermal convection, first demonstrated this chaotic behavior.
• Previous studies on generalized Lorenz systems suggested predictability might depend non-monotonically on model dimensions.

Purpose of the Study:

• To investigate if the non-monotonic dependence of predictability on model dimensions, observed in generalized Lorenz systems, also occurs in numerical weather prediction (NWP) models.
• To explore the relationship between model vertical resolution and atmospheric predictability.

Main Methods:

• Ensemble experiments were conducted using high-dimensional Lorenz system generalizations.
• Real-case simulations were performed using a numerical weather prediction model.
• Predictability was measured by deviation time, quantifying the time for solutions with slightly different initial conditions to diverge.

Main Results:

• The study demonstrated that atmospheric predictability can exhibit a non-monotonic dependence on model vertical resolution.
• This non-monotonicity was observed in both generalized Lorenz systems and NWP models.
• The spatial distribution of deviation time was identified as a significant factor contributing to this observed non-monotonicity.

Conclusions:

• Chaos, characterized by sensitive dependence on initial conditions, influences atmospheric predictability.
• Model vertical resolution can non-monotonically affect deviation time and, consequently, overall atmospheric predictability.
• Findings suggest that refining vertical resolution in NWP models may not always lead to improved predictability.