Do Subsampling Strategies Reduce the Confounding Effect of Errors in Bispectral Retrievals on Estimates of Aerosol Cloud Interactions?

Journal of Geophysical Research Atmospheres, Journal Year: 2024, Volume and Issue: 129(19)

Published: Sept. 27, 2024

Abstract Bi‐spectral retrievals of droplet effective radius and cloud optical depth are widely utilized to estimate aerosol interactions (ACI) in warm clouds the marine boundary layer. Here, we assess effect retrieval errors due neglect 3D radiative transfer during process on this analysis ACI. We use an ensemble stochastically‐modeled fields simulations study at a solar zenith angle 30°. Simulated biases number concentration ( N d ) for all three MODIS channels vary systematically from +35% −80% as heterogeneity increases. Pixel‐level can be much larger. Commonly subsampling strategies do not reduce systematic variation error. Negative error correlations between produce spuriously negative slopes logarithm liquid water path (−1.0 −0.3). Pixels center (8 km) 2 patches that overcast have relative bias −50%. The frequency these biased pixels varies linearly with clear fraction data form basis simple parameterization (CF). Using parameterization, synthetic experiments indicate estimates first indirect tropical ocean overestimated by up 30%, CF is 50%, neglecting correlation microphysics.

Language: Английский

---

Reply on RC2

Published: Jan. 12, 2023

Abstract. Our global understanding of clouds and aerosols relies on the remote sensing their optical, microphysical, macrophysical properties using, in part, scattered solar radiation. These retrievals assume form plane-parallel, homogeneous layers utilize 1D radiative transfer (RT) models, limiting detail that can be retrieved about 3D variability cloud aerosol fields inducing biases for highly heterogeneous structures such as cumulus smoke plumes. To overcome these limitations, we introduce validate an algorithm retrieving optical or microphysical atmospheric particles using multi-angle, multi-pixel radiances a RT model. The retrieval software, which have made publicly available, is called Atmospheric Tomography with Radiative Transfer (AT3D). It uses iterative, local optimization technique to solve generalized least-squares problem thereby find best-fitting state. iterative fast, approximate Jacobian calculation, extended from Levis et al. (2020) accommodate open well periodic horizontal boundary conditions (BC) improved treatment non-black surfaces. We validated accuracy calculation derivatives respect both volume extinction coefficient parameters controlling across media range depths single scattering it accurate majority over oceanic Relative root-mean-square errors cloud-like increase 2 % 12 Maximum Optical Depths (MOD) medium increases 0.2 100.0 surfaces Lambertian albedos < 0.2. Over 0.7, 20 %. Errors exceed 50 unless plane-parallel providing are very optically thin (~0.1). use theory linear inverse provide insight into physical processes control tomography identify its supported by numerical experiments. show matrix becomes increasing ill-posed size forward peak phase function decreases. This suggests tomographic will become increasingly difficult becoming thicker. Retrievals asymptotically thick likely require other sources information successful. In Part this study, examine how varies target synthetic data. do explore error nature inversion limit affects retrieval. develop method improve limit. also assess surface irradiances compare them transfer.

Language: Английский

---

Reply on RC1

Published: Jan. 12, 2023

Abstract. Our global understanding of clouds and aerosols relies on the remote sensing their optical, microphysical, macrophysical properties using, in part, scattered solar radiation. These retrievals assume form plane-parallel, homogeneous layers utilize 1D radiative transfer (RT) models, limiting detail that can be retrieved about 3D variability cloud aerosol fields inducing biases for highly heterogeneous structures such as cumulus smoke plumes. To overcome these limitations, we introduce validate an algorithm retrieving optical or microphysical atmospheric particles using multi-angle, multi-pixel radiances a RT model. The retrieval software, which have made publicly available, is called Atmospheric Tomography with Radiative Transfer (AT3D). It uses iterative, local optimization technique to solve generalized least-squares problem thereby find best-fitting state. iterative fast, approximate Jacobian calculation, extended from Levis et al. (2020) accommodate open well periodic horizontal boundary conditions (BC) improved treatment non-black surfaces. We validated accuracy calculation derivatives respect both volume extinction coefficient parameters controlling across media range depths single scattering it accurate majority over oceanic Relative root-mean-square errors cloud-like increase 2 % 12 Maximum Optical Depths (MOD) medium increases 0.2 100.0 surfaces Lambertian albedos < 0.2. Over 0.7, 20 %. Errors exceed 50 unless plane-parallel providing are very optically thin (~0.1). use theory linear inverse provide insight into physical processes control tomography identify its supported by numerical experiments. show matrix becomes increasing ill-posed size forward peak phase function decreases. This suggests tomographic will become increasingly difficult becoming thicker. Retrievals asymptotically thick likely require other sources information successful. In Part this study, examine how varies target synthetic data. do explore error nature inversion limit affects retrieval. develop method improve limit. also assess surface irradiances compare them transfer.

Language: Английский

---

Comment on amt-2023-44

Published: May 1, 2023

Abstract. Our global understanding of clouds and aerosols relies on the remote sensing their optical, microphysical, macrophysical properties using, in part, scattered solar radiation. Current retrievals assume form plane-parallel, homogeneous layers utilize 1D radiative transfer (RT) models. These assumptions limit detail that can be retrieved about 3D variability cloud aerosol fields induce biases for highly heterogeneous structures such as cumulus smoke plumes. In Part 1 this two-part study, we validated a tomographic method utilizes multi-angle passive imagery to retrieve distributions species using RT overcome these issues. That validation characterized uncertainty approximate Jacobian used retrieval over wide range atmospheric surface conditions several horizontal boundary conditions. Here 2, test algorithm’s effectiveness synthetic data whether accuracy is limited by use Jacobian. We volume extinction coefficient (σ_3D) at 40 m resolution from multi-angle, mono-spectral 35 derived stochastically-generated ‘cumuliform’ (1 km)³ domains. The are idealized neglect forward modelling instrumental errors with exception radiometric noise; thus reported lower bounds. σ_3D with, average, Relative Root Mean Square Error (RRMSE) < 20 % bias 0.1 Maximum Optical Depth (MOD) 17, RRMSE radiances 0.5 %, indicating very high shallow As MOD increases 80, worsen 60 −35 respectively, reaches 16 incomplete convergence. This expected increasing ill-conditioning inverse problem decreasing mean-free-path predicted theory discussed 1. tested model better conditioned but less accurate due more aggressive delta-M scaling. reduces radiance 9 σ_3D −8 ~80, no improvement σ_3D. illustrates significant sensitivity numerical configuration which, least our circumstances, improves accuracy. All ensemble-averaged results robust inclusion noise during retrieval. However, individual realizations have large deviations up 18 mean which indicates uncertainties optically thick limit. Using tomography also accurately infer optical depths (OD) spanning majority oceanic, (MOD 80) provides OD than 36 respectively. RT, between −30 −23 29 80 here. Prior information or other sources will required improve limit, where shown strong spatial structure varies viewing geometry.

Language: Английский

---

Reply on RC2

Published: June 26, 2023

Abstract. Our global understanding of clouds and aerosols relies on the remote sensing their optical, microphysical, macrophysical properties using, in part, scattered solar radiation. Current retrievals assume form plane-parallel, homogeneous layers utilize 1D radiative transfer (RT) models. These assumptions limit detail that can be retrieved about 3D variability cloud aerosol fields induce biases for highly heterogeneous structures such as cumulus smoke plumes. In Part 1 this two-part study, we validated a tomographic method utilizes multi-angle passive imagery to retrieve distributions species using RT overcome these issues. That validation characterized uncertainty approximate Jacobian used retrieval over wide range atmospheric surface conditions several horizontal boundary conditions. Here 2, test algorithm’s effectiveness synthetic data whether accuracy is limited by use Jacobian. We volume extinction coefficient (σ_3D) at 40 m resolution from multi-angle, mono-spectral 35 derived stochastically-generated ‘cumuliform’ (1 km)³ domains. The are idealized neglect forward modelling instrumental errors with exception radiometric noise; thus reported lower bounds. σ_3D with, average, Relative Root Mean Square Error (RRMSE) < 20 % bias 0.1 Maximum Optical Depth (MOD) 17, RRMSE radiances 0.5 %, indicating very high shallow As MOD increases 80, worsen 60 −35 respectively, reaches 16 incomplete convergence. This expected increasing ill-conditioning inverse problem decreasing mean-free-path predicted theory discussed 1. tested model better conditioned but less accurate due more aggressive delta-M scaling. reduces radiance 9 σ_3D −8 ~80, no improvement σ_3D. illustrates significant sensitivity numerical configuration which, least our circumstances, improves accuracy. All ensemble-averaged results robust inclusion noise during retrieval. However, individual realizations have large deviations up 18 mean which indicates uncertainties optically thick limit. Using tomography also accurately infer optical depths (OD) spanning majority oceanic, (MOD 80) provides OD than 36 respectively. RT, between −30 −23 29 80 here. Prior information or other sources will required improve limit, where shown strong spatial structure varies viewing geometry.

Language: Английский

---