Pub2.doc.md

Curvature-Aware, Grid-Dependent Stability Function Corrections for Arctic and Stable Boundary Layers

Authors: David England*, Richard T. McNider, Arastoo Pour-Biazar, Bright Student
*Corresponding author: (add email)

Abstract

Operational and climate models exhibit grid-dependent biases in stable boundary layer (SBL) turbulence, contributing to spread in Arctic Amplification projections. We present a curvature-aware analytic correction to Monin–Obukhov Similarity Theory (MOST) stability functions that embeds vertical grid spacing into the functional form, guided by the neutral curvature coefficient (\Delta=\alpha_h\beta_h-2\alpha_m\beta_m) of the gradient Richardson number (Ri_g=\zeta\phi_h/\phi_m^2). The method lengthens stability-function tails for coarse grids while preserving near-neutral curvature. We formalize diagnostic criteria (curvature ratios, inflection height, omission error for variable (L(z))) and propose a graduate research program to (a) extend the correction to heterogeneous Arctic regimes, (b) validate against LES and tower data, and (c) integrate a Hasse–Stirling (HS) series accelerator for φ and inversion ζ(Ri). Initial tests show reduced grid-spacing sensitivity in idealized Arctic SBL profiles. The resulting framework improves physical consistency, supports adaptive vertical refinement triggers, and offers a reproducible route to parameter transparency in climate and numerical weather prediction (NWP) models.

1. Introduction

Arctic Amplification remains underconstrained partly due to vertical resolution limits in representing shallow, strongly stable layers. Classical MOST stability functions assume implicit grid adequacy; coarse discretization inflates bulk Ri, triggering excessive mixing suppression or ad hoc tail extensions. Existing remedies (long-tail empirical forms, critical Ri tuning) lack curvature-based diagnostics tying parameter choices to physical stratification evolution. This work centers on analytic curvature of (Ri_g) to design resolution-aware corrections, minimizing divergence between coarse and fine-grid solutions without arbitrary diffusivity floors.

2. Background

• MOST power-law stable branch: (\phi_{m,h}=(1-\beta_{m,h}\zeta)^{-\alpha_{m,h}}) with (\zeta=z/L).
• Gradient Richardson number: (Ri_g=\zeta\phi_h/\phi_m^2).
• Neutral curvature: ( \partial_\zeta^2 Ri_g|_{0}=2\Delta), controls initial concavity (momentum vs heat dominance).
• Observed grid dependence: discretized Ri increases with layer thickness, altering turbulence cutoffs.
• Previous longer-tailed functions (Beljaars–Holtslag, Louis) add mixing but may distort curvature near neutral and stable transitions.

3. Problem Statement

Vertical coarsening (Δz ≳ 20–100 m) in GCM/NWP SBLs yields:

1. Misrepresentation of early (Ri_g) curvature.
2. Artificially delayed or exaggerated suppression of turbulent exchange.
3. Model spread in Arctic temperature trends and entrainment warming response.

We require a resolution-aware modification: maintain neutral curvature + physically plausible growth for larger ζ while adjusting tail behavior analytically using measured Δz.

4. Methodology Overview

4.1 Curvature Formula

[ \partial_\zeta^2 Ri_g = F\left[2V_{\log}+\zeta(V_{\log}^2-W_{\log})\right],\quad F=\frac{\phi_h}{\phi_m^2},\quad V_{\log}=v_h-2v_m ] with (v_{m,h}=\phi_{m,h}'/\phi_{m,h}), (W_{\log}=V_{\log}').

4.2 Grid-Dependent Correction

Base short-tail form approximated by exponential surrogate ( f_s(\zeta)=e^{-\gamma\zeta} ). Introduce multiplicative correction ( f_c(\zeta, \Delta z; D)=\exp{-D,\zeta,(\Delta z/\Delta z_r)}) blending short/long tails. Composite: [ \phi_m^{}=f_s f_c,\quad \phi_h^{}=Pr_t(\zeta)\phi_m^{*} \text{ (optionally dynamic Prandtl)}. ] Parameter (D) function of curvature measures and local Ri growth slope; calibrated to recover reference fine-grid solution.

4.3 Variable L Mapping Criterion

Use omission metric: [ E_{\text{omit}}=\left|\frac{\zeta''z,\partial\zeta Ri_g}{(\zeta'z)^2,\partial\zeta^2 Ri_g}\right|<\epsilon ] to decide constant-L vs full chain rule.

4.4 HS Series Accelerator (Optional)

Expand (\log(1-\beta\zeta)) using HS-generated coefficients with tail bound (O(\rho^{N+1}/[(N+1)(1-\rho)])); precompute to accelerate φ, F, V_log evaluations and ζ(Ri) inversion (series seed + single Newton step).

5. Data & Experiments

• Idealized Arctic SBL test (GABLS1-type).
• LES ensemble references (temperature, wind, flux profiles).
• Tower observations (wind, temperature, flux) for neutral + stable nights.
• Vertical grids: fine (2 m spacing); coarse sets (10, 30, 60, 100 m).
• Metrics: Ri_g bias, curvature difference, neutral curvature match, inflection detection, flux errors.

6. Diagnostics

1. Neutral curvature error: (|\partial_\zeta^2Ri_g^{*}-2\Delta|/|2\Delta|).
2. Curvature amplification ratio: (A(\zeta)=|\partial_\zeta^2 Ri_g^{*}/\partial_\zeta^2 Ri_g^{ref}|).
3. Inflection presence/shift: (|\zeta_{\text{inf}}^{*}-\zeta_{\text{inf}}^{ref}|).
4. Vertical mapping error (E_{\text{omit}}).
5. Grid convergence: profile norm differences (L2, L∞) between coarse/fine after correction.
6. Flux consistency: surface momentum & heat flux relative error.

7. Research Objectives (Bright Student Focus)

O1. Derive functional form (D(\text{curv},\ Ri, \Delta z)) minimizing coarse-grid curvature bias.
O2. Quantify robustness across φ families (power-law, quadratic, regularized).
O3. Validate diagonal neutral curvature preservation in blended regimes (stable → very stable).
O4. Integrate HS coefficient tables; benchmark speed vs analytic direct evaluation.
O5. Construct adaptive refinement trigger using (|W_{\log}/V_{\log}^2|) and (E_{\text{omit}}).
O6. Deliver open-source module + reproducible notebooks (Arctic case studies).

8. Work Packages

WP1: Literature synthesis & neutral curvature extraction (Weeks 1–2).
WP2: Baseline φ family fitting to LES/tower (Weeks 3–5).
WP3: Grid-dependent correction calibration (optimization over D forms) (Weeks 6–9).
WP4: HS-assisted series implementation + error bound verification (Weeks 10–12).
WP5: Comprehensive validation (multi-season Arctic, Antarctic) (Weeks 13–16).
WP6: Adaptive trigger evaluation; vertical refinement tests (Weeks 17–19).
WP7: Manuscript preparation & code release (Weeks 20–24).

9. Methods Detail

• Optimization: minimize weighted sum (J = w_1\text{CurvBias}+w_2\text{RiBias}+w_3\text{FluxErr}) over D functional candidates (linear, logistic, curvature-driven).
• Surrogates: Q‑SBL quadratic forms for pole-proximate ζ intervals; smooth blend at ζ_b.
• Uncertainty: propagate φ derivative uncertainties through curvature linearization (\delta R''g \approx |F|\big[2\delta V{\log} + \zeta(2|V_{\log}|\delta V_{\log}+\delta W_{\log})\big]).
• Computational efficiency: precompute tables of φ, F, V_log at log-spaced ζ nodes; use interpolation (cubic Hermite) for runtime queries.

10. Expected Results

• Demonstrated reduction in grid-dependent Ri_g curvature error (>40% improvement).
• Consistent neutral curvature across grid sets.
• HS series reduces evaluation cost (wall-clock) by 25–35% at target accuracy.
• Adaptive trigger decreases unnecessary refinement events while capturing true curvature spikes.

11. Risks & Mitigations

Risk	Impact	Mitigation
Overfitting D to single site	Poor generalization	Multi-site cross-validation
Numerical instability near β pole	Divergent curvature	Surrogate / cap + masking
HS series divergence at high ζ	Accuracy loss	Adaptive N selection via tail bound
Data sparsity in extreme stability	Insufficient calibration	Include reanalysis profiles + synthetic LES cases

12. Path to Publication

Targets: Journal of Applied Meteorology and Climate; Secondary: Boundary-Layer Meteorology (methodological focus). Deliverables: main paper (method + validation), dataset/ code archive (Zenodo DOI), short methods note on HS acceleration.

13. Broader Impacts

• Improves SBL representation → tighter Arctic amplification spread.
• Provides transparent curvature metrics for model tuning.
• Supports energy demand forecasting (stable nocturnal inversions).
• Transferable to planetary boundary layers (Mars, Titan) via L scaling.

14. Proposed Timeline (Months)

1–2: WP1 | 3–5: WP2 | 6–9: WP3 | 10–12: WP4 | 13–16: WP5 | 17–19: WP6 | 20–24: WP7.

15. Data Management

Versioned Git repository (profiles, diagnostics). Input data (LES/tower) stored with metadata (site, time window, stability classification). Outputs: CSV + NetCDF curvature spectra; Jupyter notebooks for reproducibility.

16. References

1. Monin, A. S., & Obukhov, A. M. (1954). DOI: 10.1007/BF02247208
2. Businger, J. A. et al. (1971). DOI: 10.1175/1520-0469(1971)028<0669:FLITAT>2.0.CO;2
3. Dyer, A. J. (1974). DOI: 10.1002/qj.49710042515
4. Beljaars, A. C. M., & Holtslag, A. A. M. (1991). DOI: 10.1007/BF00119965
5. Högström, U. (1988). DOI: 10.1007/BF00120624
6. Cheng, Y., & Brutsaert, W. (2005). DOI: 10.1175/JAM2302.1
7. Grachev, A. A., & Fairall, C. W. (1997). DOI: 10.1175/1520-0450(1997)036<0403:POTMBL>2.0.CO;2
8. Cuxart, J. et al. (2006). DOI: 10.1007/s10546-006-9061-9
9. McNider, R. T. et al. (2012). DOI: 10.1175/JAMC-D-11-0167.1
10. Bintanja, R. et al. (2012). DOI: 10.1038/ngeo1342
11. Van de Wiel, B. J. H. et al. (2017). DOI: 10.1175/JAMC-D-16-0240.1
12. Tastula, E.-M. et al. (2015). DOI: 10.1007/s10546-015-0056-9
13. Louis, J.-F. (1979). DOI: 10.1002/qj.49710544304
14. Pleim, J. (2007). DOI: 10.1175/JAM2440.1
15. Screen, J. A. et al. (2012). DOI: 10.1029/2011JD016545
16. Van de Wiel, B. J. H. et al. (2002). DOI: 10.1175/1520-0469(2002)059<2665:ISBATN>2.0.CO;2

17. Student Outcomes

• Thesis integrating analytic correction + HS acceleration.
• Open dataset (curvature diagnostics across grid sets).
• Two manuscripts: (1) Method; (2) Arctic multi-season application.

18. Conclusion

Embedding curvature metrics into stability-function design yields a principled, resolution-aware correction, reducing grid dependence while preserving physically meaningful Ri_g evolution. The proposed research path provides a tractable, high-impact Masters/PhD project with direct relevance to climate model uncertainty reduction.