Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Probabilistic Emulation of the Community Radiative Transfer Model Using Machine Learning

Lucas Howard, Aneesh C. Subramanian, Gregory Thompson, Benjamin Johnson, Thomas Auligne

TL;DR

The study investigates extratropical baroclinicity by focusing on the near-surface gradient of potential temperature and its associated growth rate through $B_x$ and $B_y$, with $B_{GRMax}$ linking static stability and vertical shear. It employs ERA-40 and NCEP–NCAR reanalyses to compute $B$ from $\nabla \theta^{2m}$ and $N$ over 1957–2002, constructing a long Northern Hemisphere dataset for climatology and variance. The results show $B_x$ as the dominant component in the North Pacific, with maxima along the Kuroshio Extension storm track and strong winter amplification, while $B_y$ is weaker. The analysis highlights the critical roles of front structure, SST anomaly scale, and land-surface complicating factors in shaping the baroclinic response, underscoring the need to consider spatial scale and surface interactions when interpreting SST–atmosphere links.

Abstract

The continuous improvement in weather forecast skill over the past several decades is largely due to the increasing quantity of available satellite observations and their assimilation into operational forecast systems. Assimilating these observations requires observation operators in the form of radiative transfer models. Significant efforts have been dedicated to enhancing the computational efficiency of these models. Computational cost remains a bottleneck, and a large fraction of available data goes unused for assimilation. To address this, we used machine learning to build an efficient neural network based probabilistic emulator of the Community Radiative Transfer Model (CRTM), applied to the GOES Advanced Baseline Imager. The trained NN emulator predicts brightness temperatures output by CRTM and the corresponding error with respect to CRTM. RMSE of the predicted brightness temperature is 0.3 K averaged across all channels. For clear sky conditions, the RMSE is less than 0.1 K for 9 out of 10 infrared channels. The error predictions are generally reliable across a wide range of conditions. Explainable AI methods demonstrate that the trained emulator reproduces the relevant physics, increasing confidence that the model will perform well when presented with new data.

Probabilistic Emulation of the Community Radiative Transfer Model Using Machine Learning

TL;DR

The study investigates extratropical baroclinicity by focusing on the near-surface gradient of potential temperature and its associated growth rate through and , with linking static stability and vertical shear. It employs ERA-40 and NCEP–NCAR reanalyses to compute from and over 1957–2002, constructing a long Northern Hemisphere dataset for climatology and variance. The results show as the dominant component in the North Pacific, with maxima along the Kuroshio Extension storm track and strong winter amplification, while is weaker. The analysis highlights the critical roles of front structure, SST anomaly scale, and land-surface complicating factors in shaping the baroclinic response, underscoring the need to consider spatial scale and surface interactions when interpreting SST–atmosphere links.

Abstract

The continuous improvement in weather forecast skill over the past several decades is largely due to the increasing quantity of available satellite observations and their assimilation into operational forecast systems. Assimilating these observations requires observation operators in the form of radiative transfer models. Significant efforts have been dedicated to enhancing the computational efficiency of these models. Computational cost remains a bottleneck, and a large fraction of available data goes unused for assimilation. To address this, we used machine learning to build an efficient neural network based probabilistic emulator of the Community Radiative Transfer Model (CRTM), applied to the GOES Advanced Baseline Imager. The trained NN emulator predicts brightness temperatures output by CRTM and the corresponding error with respect to CRTM. RMSE of the predicted brightness temperature is 0.3 K averaged across all channels. For clear sky conditions, the RMSE is less than 0.1 K for 9 out of 10 infrared channels. The error predictions are generally reliable across a wide range of conditions. Explainable AI methods demonstrate that the trained emulator reproduces the relevant physics, increasing confidence that the model will perform well when presented with new data.

Paper Structure

This paper contains 9 sections, 3 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. The sea surface temperature
  3. A subtle point
  4. Changes in the temperature
  5. Complicating factor
  6. Data and calculation procedures
  7. Climatology and variance
  8. Sample appendix section head
  9. Sample appendix section head

Figures (3)

  • Figure 1: Climatology of $Bx (10^{-6} s^{-1}$, color) and $U^{200}$(m s$^-1$, contours) for (a) February and (b) August; $\overline{V'\theta'}^{850}$ (K m s$^-1$, color) and $\overline{V'V'}^{200}$ (m$^2$ s$^{-1}$, contours) for (c) February and (d) August; MR$^{z850}$ (10$^{-3}$ m$^2$ s$-2$, color) and $U^{1000}$ (m s$^{-1}$, contours) for (e) February and (f) August; and SST (K, color) and $F_h$ [10$^5$ J m$^{-2}$ (6 h)$^{-1}$] for (g) February and (h) August. Red rectangles indicate the domain of EOF calculations.
  • Figure 2: As in Fig. 10, but for (a),(b) September EOF1; (c),(d) September EOF2; (e),(f) October EOF1; (g),(h) October EOF2; and (i),(j) December EOF2.
  • Figure 3: Here is an appendix, single column figure caption.