Instantaneous retrieval of aerosol properties from geostationary imagers using i-AERUS-GEO

Instantaneous retrieval of aerosol properties from geostationary imagers using i-AERUS-GEO

Instantaneous retrieval of aerosol properties from geostationary imagers using i-AERUS-GEO
Instantaneous retrieval of aerosol properties from geostationary imagers using i-AERUS-GEO

This study continued previous work to provide instantaneous retrievals of aerosol load from SEVIRI observations.

Last Updated

05 March 2021

Published on

08 November 2018

Satellite remote sensing is a unique tool to study atmospheric aerosols and their impacts at the regional and global scales. In 2010, the Aerosol and surface albEdo Retrieval Using a directional Splitting method-application to GEOstationary data (AERUS-GEO) approach was developed by the Centre National de Recherches Météorologiques (CNRM) of Météo-France to estimate aerosol optical depth (AOD) based on MSG/SEVIRI observations at the daily frequency (Carrer et al., JGR, 2010). In 2014, the method was implemented in the ICARE/AERIS data and services centre for the operational production of the daily-averaged AOD (d-AOD) corresponding to the SEVIRI channel VIS06 at 635 nanometers (see Figure 1). The method was also improved to give an estimation of the Angström exponent based on the SEVIRI channels in the near-infrared (Carrer et al., GRL, 2014). More information is available at the ICARE Data and Services Center website (free registration is required to access the data).

The load and type of aerosols can change rapidly in time and space. Hence, instantaneous variations of aerosol properties are required by the climate and weather communities to take into account the diurnal cycle of aerosols. In particular, the diurnal variation of AOD is of great importance for the study of sudden and rapid aerosol events and for the improvement of chemical transport models outputs through assimilation, for example. With this purpose, the CNRM initiated in 2016 the development of a new version of the AERUS-GEO algorithm to provide estimates of the instantaneous AOD (i-AOD) from SEVIRI data (every 15 minutes). This study was done in the framework of the Federated Activity LSA-SAF_2015-02 with the support of EUMETSAT and ICARE/AERIS.

The work accomplished in this study consisted of the set up of the framework of an instantaneous retrieval methodology, which provided encouraging results even if several limitations were identified. The developed algorithm prototype exploited the potential of geostationary platforms to study the sub-daily variation of the aerosol load thanks to their high number of observations during the day.

Figure 1: For March 30th, 2010, (left) SEVIRI RGB composite at noon and (right) daily AOD at 0.635 μm retrieved by AERUS-GEO
Figure 1: For 30 March 2010, (left) SEVIRI RGB composite at noon and (right) daily AOD at 0.635 μm retrieved by AERUS-GEO

Objectives

The objective of the project was to continue the work initiated in the Federated Activity LSA-SAF_2015-02 to provide instantaneous retrievals of aerosol load from SEVIRI observations.

The main goal was to continue the development of the new version of the AERUS-GEO algorithm and to evaluate its accuracy through the comparison of the retrieved i-AOD values with ground measurements from the AERONET network. For this exercise, 10 years of SEVIRI data were processed in order to carry out a validation over the AERONET stations encompassed within the MSG Earth’s disc.

Another goal of the evaluation was to use other satellite aerosol products, such as the GRASP retrievals obtained from the PARASOL satellite.

The advantages and limitations of the new AERUS-GEO approach had to be detailed at the end of the project, as well as the recommendations for its future adaptation to the MTG-I/FCI mission.


Overview

The project was undertaken during 12 months to prove the feasibility of accurately retrieving the aerosol load from every observation acquired by the MSG/SEVIRI platform. The results obtained by the CNRM during this project confirmed the viability of such a challenge.

The algorithm named i-AERUS-GEO was developed to perform the instantaneous estimation of AOD every 15 minutes. The retrieval is based on the robust inversion of an accurate radiative transfer-based parameterisation of the satellite signal using (i) spatio-temporal constraints and (ii) accurate optical properties to describe the aerosol mixture. SEVIRI radiances are corrected for molecular effects in a first step using ancillary atmospheric data (see Figure 2).

Each retrieved value of i-AOD is provided with a confidence measure, which is constructed at the time of the retrieval based on the robustness of the inversion and the sensitivity of SEVIRI observations to the aerosol signal. The latter parameter is derived using the concept of the critical surface albedo, which corresponds to the value of albedo for which the TOA signal becomes insensitive of the aerosol load.

The newly developed algorithm i-AERUS-GEO is an evolution of the original AERUS-GEO approach working at the daily frequency. The instantaneous and the daily algorithms are now bridged together by the use of the surface BRDF estimated by the original AERUS-GEO approach into the instantaneous aerosol retrieval performed by i-AERUS-GEO.

Figure 2: Flow chart summarising the i-AERUS-GEO algorithm.
Figure 2: Flow chart summarising the i-AERUS-GEO algorithm.

The i-AERUS-GEO method was evaluated thanks to the processing of 10 years of SEVIRI observations corresponding to the location of the AERONET stations encompassed by the MSG disc. Results confirmed the good performances of i-AERUS-GEO in providing accurate i-AOD retrievals (average correlation of 0.79, bias of -0.02, and RMSE of 0.08) when suitable conditions for aerosol remote sensing are satisfied (i.e. frequent clear sky conditions and mild aerosol activity, see Figure 3).

Figure 3: Density scatter plot (left) and map of RMSE (right) comparing two months of i-AERUS-GEO retrievals against measurements of eight AERONET stations
Figure 3: Density scatter plot (left) and map of RMSE (right) comparing two months of i-AERUS-GEO retrievals against measurements of eight AERONET stations

In these situations, the combination of the high frequency of SEVIRI measurements and the strengths of i-AERUS-GEO allowed the retrieval of the diurnal cycle of the aerosol load (see Figure 4).

Figure 4: Time series of i-AOD from i-AERUS-GEO and AERONET
Figure 4: Time series of i-AOD from i-AERUS-GEO and AERONET

Some limitations exist today in the current version of i-AERUS-GEO such as the use of a fixed continental aerosol type or the characterisation of the BRDF of ocean surfaces with a Lambertian model. These limitations may become important under some conditions (e.g. intense aerosol activity due to non-continental aerosols or proximity to water bodies), which may result in less accurate AOD retrievals. When considering all conditions (suitable and less favourable), i-AERUS-GEO still gave promising estimates of i-AOD (average correlation of 0.66, bias of -0.02, and RMSE of 0.12). These scores were obtained from the most robust retrievals, which are selected thanks to the confidence measure provided by the algorithm.

These i-AOD retrievals also compared well with the state-of-the-art GRASP/POLDER satellite product in terms of accuracy (see Figure 5). However, i-AERUS-GEO was able to provide information on the diurnal cycle thanks to the high temporal frequency of the MSG/SEVIRI geostationary platform with respect to the polar orbiting satellite PARASOL.

Figure 5: Maps of average RMSE for the comparison of i-AERUS-GEO (left) and GRASP/POLDER (right) instantaneous AOD retrievals for 2012
Figure 5: Maps of average RMSE for the comparison of i-AERUS-GEO (left) and GRASP/POLDER (right) instantaneous AOD retrievals for 2012

In the future, the known limitations of the current version of i-AERUS-GEO will be tackled in order to improve the accuracy of the retrieved AOD, especially in less favourable situations for aerosol remote sensing.

Furthermore, the performances of i-AERUS-GEO in the spatial domain will be evaluated with the goal of providing robust and homogeneous maps of aerosol load. This future work will be carried out using available full disc observations of SEVIRI.

In addition, the characteristics of the future FCI/MTG-I platform will already be taken into account in the corresponding developments. This will start the preparation for the adaptation of the current i-AERUS-GEO to the next generation of EUMETSAT geostationary satellites. In this regard, the improved spectral resolution and the increased number of bands of FCI will be exploited to perform a more accurate characterisation of the aerosol mixture, which in turn should provide better estimations of the aerosol load and the surface BRDF.

The next version of i-AERUS-GEO will include all these improvements and may be tested on observations of the geostationary platform Himawari/AHI, which shares many similarities with the upcoming MTG-I/FCI .

Other operational missions algorithm studies