A machine learning approach for evaluating Southern Ocean cloud radiative biases in a global atmosphere model DOI Creative Commons
Sonya L. Fiddes, Marc Mallet, Alain Protat

et al.

Geoscientific model development, Journal Year: 2024, Volume and Issue: 17(7), P. 2641 - 2662

Published: April 11, 2024

Abstract. The evaluation and quantification of Southern Ocean cloud–radiation interactions simulated by climate models are essential in understanding the sources magnitude radiative bias that persists for this region. To date, most methods focus on specific synoptic or cloud-type conditions do not consider entirety Ocean's cloud regimes at once. Furthermore, it is difficult to directly quantify complex non-linear role different properties have modulating effect. In study, we present a new method model evaluation, using machine learning can once identify complexities within system individual contributions. this, use an XGBoost (eXtreme Gradient Boosting) predict nudged version Australian Community Climate Earth System Simulator – Atmosphere-only model, property biases as predictive features. We find explain up 55 % from these alone. then apply SHAP (SHapley Additive exPlanations) feature importance analysis each plays predicting bias. liquid water path largest contributor over Ocean, though important regional dependencies exist. test usefulness evaluating perturbations clearly responses, including compensating errors.

Language: Английский

New particle formation in coastal New Zealand with a focus on open-ocean air masses DOI Creative Commons
Maija Peltola, Clémence Rose,

Jonathan V. Trueblood

et al.

Atmospheric chemistry and physics, Journal Year: 2022, Volume and Issue: 22(9), P. 6231 - 6254

Published: May 13, 2022

Abstract. Even though oceans cover the majority of Earth, most aerosol measurements are from continental sites. We measured particle number size distribution at Baring Head, in coastal New Zealand, over a total period 10 months to study properties and new formation, with special focus on formation open-ocean air masses. Particle concentrations were higher land-influenced compared clean marine all classes, sub-10 nm cloud condensation nuclei sizes. When classifying distributions traditional methods designed for sites, was observed station throughout year an average event frequency 23 %. While these days had some land influence, we also growth starting nucleation mode 16 % data air, least part this connected boundary layer. Sub-10 particles accounted 29 concentration larger than 1 during spring. This shows that is frequent enough influence concentration. more intense rates what found air. primary emissions increased as function time spent first 1–2 d. After this, seems start getting suppressed by pre-existing population, but accumulation keeps increasing, likely due emissions. Further work showed NPF events favoured sunny conditions low relative humidity wind speeds. In temperatures, humidity, speeds could happen even night. Our future will mechanisms responsible Head different chemical precursor species. sheds light both masses coming Southern Ocean local Zealand.

Language: Английский

Citations

10
How Accurately Can Warm Rain Realistically Be Retrieved with Satellite Sensors? Part I: DSD Uncertainties DOI

Richard Schulte,

Christian D. Kummerow,

Christian Klepp

et al.

Journal of Applied Meteorology and Climatology, Journal Year: 2022, Volume and Issue: 61(9), P. 1087 - 1105

Published: May 13, 2022

Abstract A significant part of the uncertainty in satellite-based precipitation products stems from differing assumptions about drop size distributions (DSDs). Satellite radar-based retrieval algorithms rely on DSD that may be overly simplistic, whereas radiometers further struggle to distinguish cloud water rain. We utilize Ocean Rainfall and Ice-phase Precipitation Measurement Network (OceanRAIN), version 1.0, dataset examine impact variability ability satellite measurements accurately estimate rates warm rainfall. use binned disdrometer counts a simple model atmosphere simulate observations for three architectures. Two are similar existing instrument combinations GPM Core Observatory CloudSat , third is theoretical triple-frequency radar–radiometer architecture. Using an optimal estimation framework, we find assumed shape can have large retrieved rain rate. three-parameter normalized gamma sufficient describing retrieving DSDs observed OceanRAIN dataset. Assuming simpler single-moment models lead biases rate, order 100%. Differing could thus plausibly explain portion disagreement estimates.

Language: Английский

Citations

8
Retrieval of the sea spray aerosol mode from submicron particle size distributions and supermicron scattering during LASIC DOI Creative Commons
Jeramy L. Dedrick, Georges Saliba, A. Williams

et al.

Atmospheric measurement techniques, Journal Year: 2022, Volume and Issue: 15(14), P. 4171 - 4194

Published: July 20, 2022

Abstract. Improved quantification of sea spray aerosol concentration and size is important for determining effects on clouds the climate, though attempts to accurately capture distribution mode remain limited by availability supermicron distributions. In this work, we introduce a new approach retrieving lognormal fit parameters combining submicron distributions with scattering measurements using Mie inversion. Submicron were measured an ultra-high-sensitivity spectrometer (UHSAS), was taken as difference between <10 µm <1 three-wavelength integrating nephelometer (NEPH). This UHSAS-NEPH method applied during background marine periods Department Energy Atmospheric Radiation Measurement Layered Atlantic Smoke Interactions Clouds (LASIC) campaign Ascension Island (November 2016–May 2017), when contribution expected represent large fraction mass total scattering. Lognormal modal retrieved from comparisons lookup table theory-simulated coefficients low-error solutions that minimized 0.4–1 residual in UHSAS distribution. We evaluated set clean North included chemical measurements, showing are needed constrain number but reasonably characterized For LASIC, properties approximately 88 % times variability particle low (<± 5 Mm−1 <400 cm−3, respectively), mean diameter ranging 0.6 1.9 (1.47 ± 0.17 µm), width 1.1 3.97 (2.4±0.3), 0.18 23.0 µg m−3 (8.37. 4.1 m−3). The at three wavelengths found marginally largest sizes absence additional defining solutions. Comparing retrievals those fitting algorithm only showed correlations available mass-based tracers (coarse scattering, wind speed, chloride) not considered. work demonstrates added value reasonable concentrations, providing best-available observationally constrained estimate available.

Language: Английский

Citations

8
New particle formation leads to enhanced cloud condensation nuclei concentrations on the Antarctic Peninsula DOI Creative Commons
Jiyeon Park,

Hyojin Kang,

Yeontae Gim

et al.

Atmospheric chemistry and physics, Journal Year: 2023, Volume and Issue: 23(21), P. 13625 - 13646

Published: Nov. 2, 2023

Abstract. Few studies have investigated the impact of new particle formation (NPF) on cloud condensation nuclei (CCN) in remote Antarctica, and none has elucidated relationship between NPF CCN production. To address that knowledge gap, we continuously measured number size distribution 2.5–300 nm particles concentrations at King Sejong Station Antarctic Peninsula from 1 January to 31 December 2018. Ninety-seven events were detected throughout year. Clear annual seasonal patterns observed: high concentration frequency nucleation-mode summer (December–February: 53 cases) undetected winter (June–August: no cases). We estimated spatial scale by multiplying time during which a distinct nucleation mode can be observed sampling site locally wind speed. The median around was found approximately 155 km, indicating large events. Air back-trajectory analysis revealed 80 cases associated with air masses originating over ocean, followed sea-ice (12 cases), multiple (3 land (2 regions. present discuss three major categories: (1) marine NPF, (2) (3) NPF. Satellite estimates for sea-surface dimethylsulfoniopropionate (DMSP; precursor gaseous dimethyl sulfide) data showed production oceanic biogenic precursors could key component events, whereas halogen compounds released ice-covered areas contribute Terrestrial sources (wildlife colonies, vegetation, meltwater ponds) Antarctica affect aerosol masses. Out 97 83 characterized simultaneous increase 2 %–270 % (median 44 %) following 36 h 8 h) after Overall, significant source different physical characteristics related Peninsula, subsequently grew nuclei.

Language: Английский

Citations

3
A machine learning approach for evaluating Southern Ocean cloud radiative biases in a global atmosphere model DOI Creative Commons
Sonya L. Fiddes, Marc Mallet, Alain Protat

et al.

Geoscientific model development, Journal Year: 2024, Volume and Issue: 17(7), P. 2641 - 2662

Published: April 11, 2024

Abstract. The evaluation and quantification of Southern Ocean cloud–radiation interactions simulated by climate models are essential in understanding the sources magnitude radiative bias that persists for this region. To date, most methods focus on specific synoptic or cloud-type conditions do not consider entirety Ocean's cloud regimes at once. Furthermore, it is difficult to directly quantify complex non-linear role different properties have modulating effect. In study, we present a new method model evaluation, using machine learning can once identify complexities within system individual contributions. this, use an XGBoost (eXtreme Gradient Boosting) predict nudged version Australian Community Climate Earth System Simulator – Atmosphere-only model, property biases as predictive features. We find explain up 55 % from these alone. then apply SHAP (SHapley Additive exPlanations) feature importance analysis each plays predicting bias. liquid water path largest contributor over Ocean, though important regional dependencies exist. test usefulness evaluating perturbations clearly responses, including compensating errors.

Language: Английский

Citations

0