A New Satellite-Based Retrieval of Low-Cloud Liquid-Water Path Using Machine Learning and Meteosat SEVIRI Data

Miae Kim,Roland Stirnberg,Hendrik Andersen,Julia Fuchs,Jan Cermak

doi:10.3390/rs12213475

Miae Kim, Roland Stirnberg + Show 3 more

Open Access

https://doi.org/10.3390/rs12213475

Copy DOI

Journal: Remote sensing	Publication Date: Oct 22, 2020
Citations: 6	License type: CC BY 4.0

Affiliation: Karlsruhe Institute of Technology

Abstract

Clouds are one of the major uncertainties of the climate system. The study of cloud processes requires information on cloud physical properties, in particular liquid water path (LWP). This parameter is commonly retrieved from satellite data using look-up table approaches. However, existing LWP retrievals come with uncertainties related to assumptions inherent in physical retrievals. Here, we present a new retrieval technique for cloud LWP based on a statistical machine learning model. The approach utilizes spectral information from geostationary satellite channels of Meteosat Spinning-Enhanced Visible and Infrared Imager (SEVIRI), as well as satellite viewing geometry. As ground truth, data from CloudNet stations were used to train the model. We found that LWP predicted by the machine-learning model agrees substantially better with CloudNet observations than a current physics-based product, the Climate Monitoring Satellite Application Facility (CM SAF) CLoud property dAtAset using SEVIRI, edition 2 (CLAAS-2), highlighting the potential of such approaches for future retrieval developments.

Highlights

The Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC) stated that clouds remain one of the largest uncertainties of the climate system [1]
We found that liquid water path (LWP) predicted by the machine-learning model agrees substantially better with CloudNet observations than a current physics-based product, the Climate Monitoring Satellite Application Facility (CM SAF) CLoud property dAtAset using Spinning-Enhanced Visible and Infrared Imager (SEVIRI), edition 2 (CLAAS-2), highlighting the potential of such approaches for future retrieval developments
The fitted line of the scatter plot for the homogeneous situation has a slope of 0.47 and an intercept of 38.5, as the higher values tend to be underestimated. These values differ from the CLAAS-2 LWP scatter plot, and from those reported by Greuell and Roebeling [2]

Summary

Introduction

The Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC) stated that clouds remain one of the largest uncertainties of the climate system [1]. Passive microwave satellite data have been used to estimate LWP as passive microwave sensors have the ability to penetrate clouds, allowing to directly measure liquid cloud condensate amount [5]. Kostsov et al [9] mentioned the inhomogeneity of cloud fields as an error source for the differences between satellite and ground-based observations This inhomogeneity involves the interaction between types of the underlying surface and the atmospheric conditions affecting LWP values as well, which further links to a seasonal dependence of accuracy. Both LWP retrievals are evaluated and compared with the CloudNet in-situ measurements. The decisions of the machine learning model are analyzed in detail to examine the influence of input variables on the LWP retrieval, and biases of model-derived LWP are discussed

Meteosat-9 SEVIRI Data

CloudNet Data

CLAAS-2 Data

Gradient Boosting Regression Trees

Statistics for Model Performance

Bias Analysis for LWP Retrieval

Summary and Conclusions