[Paper Review] How different are deterministic physics suites when coupled to fixed model dynamics and why?

Edward Groot, Hannah Christensen+6 more|arXiv (Cornell University)|Jan 6, 2026

Meteorological Phenomena and Simulations0 citations

TL;DR

The study evaluates how four deterministic physics suites behave when forced by a common fixed large-scale dynamics, finding high precipitation agreement across suites but notable under-dispersion relative to a convection-permitting benchmark, with a slight non-linear dynamical feedback evident.

ABSTRACT

It is often difficult to attribute uncertainty and errors in atmospheric models to designated model components. This is because sub-grid parameterised processes interact strongly with the large-scale transport represented by the explicit model dynamics. We carry out experiments with prescribed large-scale dynamics and different sub-grid physics suites. This dataset has been constructed for the Model Uncertainty Model Intercomparison Project (MUMIP), in which each suite forecasts sub-grid tendencies at a 22km grid. The common dynamics is derived from a convection-permitting benchmark: an ICON DYAMOND experiment (2.5km grid). We compare four different physics suites for atmospheric models in an Indian Ocean experiment. We analyse their joint PDFs of precipitation and associated physics tendencies for a full month. Precipitation is selected because it is a dominant uncertainty in the models that redistributes large amounts of heat. We find that all physics suites produce very similar precipitation amounts, with very high correlations between models, which exceed 0.95 at the native grid. However, the convection-permitting benchmark is more dissimilar from each of the physics suites, with correlations of $\approx$0.80. Similarly, we show that the vertically averaged physics tendencies in the free-troposphere are highly similar between the four physics suites, yet different if reconstructed for the benchmark. The water vapour sink is very closely linked with precipitation in the four physics suites. This suggests that the coarse-grid models are overconfident. We hypothese is that variation in unresolved convective structures can lead to variation in the dynamics, following a given amount of latent heating at fine grids, but not in our physics suites. The abstract length limit of ArXiv requires you to proceed in the PDF.

Motivation & Objective

• Assess similarity of deterministic physics suites under fixed large-scale dynamics.
• Quantify how deterministic physics tendencies relate to convective organisation and aggregation.
• Explore potential links between deterministic physics and non-linear dynamical feedbacks seen with convection-permitting models.

Proposed method

• Use Model Uncertainty Model Intercomparison Project (MUMIP) SCM forcing files derived from a 2.5 km DYAMOND ICON simulation.
• Force SCMs with coarse-grained advection tendencies from a shared dynamics to compute physics responses.
• Compute joint PDFs of precipitation and vertical tendencies for 3–6 hour lead times across 44,000 grid boxes (22 km spacing) over 31 days.
• Reconstruct pseudo-ICON physics tendencies from SCM dynamics for comparison with a convection-permitting benchmark.
• Fit an exponential relationship y = a + b x^p between ICON precipitation and SCM precipitation to test non-linearity (p != 1).
• Evaluate correlations of humidity and temperature tendencies in the free troposphere and mixed layer between suites and with the benchmark.

Figure 1: Under conditions of fixed column precipitation rate (as proxy for latent heating rate), the mass divergence rate depends on convective organisation and aggregation in large-eddy and convection-permitting ICON simulations, but not in coarser ICON simulations of the same case study ( Groot e

Experimental results

Research questions

• RQ1 Do different deterministic physics suites produce similar precipitation and physics-tendency responses under fixed dynamics?
• RQ2 Is there under-dispersion in SCMs relative to a convection-permitting benchmark, and how does this relate to sub-grid variability?
• RQ3 Is there evidence of non-linear feedback between convective organisation/dynamics and precipitation when using coarse-grid physics?
• RQ4 How do humidity and temperature tendencies in the free troposphere and mixed layer correlate across suites and with the benchmark?
• RQ5 What role do gravity-wave interactions and convective organisation play in the observed relationships between physics and dynamics?

Key findings

• Precipitation shows very high pairwise correlations across physics suites (0.96–0.98), while correlations with the convection-permitting benchmark are lower (~0.80).
• Reconstructed benchmark precipitation correlates with SCMs at ~0.93–0.94 when coarse-grained to 1.0 degree, but SCM–benchmark correlations remain notably lower than SCM–SCM correlations.
• Humidity tendencies in the free troposphere correlate strongly between suites (≈0.97), yet drop to ≈0.80–0.82 when compared to the benchmark tendencies.
• Humidity tendencies exhibit substantial under-dispersion in SCMs relative to the benchmark, indicating unrepresented sub-grid variability.
• A slightly non-linear (p ≈ 1.05–1.06) relationship best fits ICON precipitation from SCMs for most suites, suggesting a weak but robust non-linear dynamics feedback; GFS shows a weaker, less robust signal.
• Temperature tendencies in the free troposphere are highly correlated between suites (≈0.93–0.97) and less under-dispersed than humidity, implying humidity is the dominant unrepresented sub-grid uncertainty among SCM tendencies.

Figure 2: Joint PDFs of precipitation accumulation over all available columns, left top: for ARPEGE and IFS at lead times of 3hr to 6hr; right top: IFS (same lead time) and ICON over all available columns; bottom: same as right top, after regridding to 1.0 degrees. Blue lines represent the 1:1-relat