ESA Climate Change Initiative Aerosol cci - Phase 2 2014 ...cci.esa.int/sites/default/files/Aerosol_cci2_FR_v1.2.pdf · Aerosol_cci2_FR_v1.2.pdf . Aerosol_cci2 Final Report REF :

Aerosol_cci2

Final Report

REF : Aerosol FR ISSUE : 1.2 DATE : 09.01.2019 PAGE : I

ESA Climate Change Initiative

Aerosol_cci - Phase 2

2014 - 2018

Thomas Popp, Gerrit de Leeuw, Miriam Kosmale, Rosa Astoreca, Christine Bingen, Adam Bourassa, Christoph Brühl, Virginie Capelle, Lieven Clarisse,

Martin De Graaf, Enza Di Tomaso, Jacques Descloitres, Oleg Dubovik, Jürgen Fischer, Yves Govaerts,

Don Grainger, Jan Griesfeller, Andreas Heckel, Stefan Kinne, Klaus Klingmüller, Lars Klüser, Pekka Kolmonen, Arve Kylling, Luca Lelli,

Pavel Litvinov, Marta Lufarelli, Linlu Mei, Julian Meyer-Arnek, David Neubauer, Peter North, Caroline Poulson, Adam Povey,

Simon Proud, Charles Robert, Jennifer Schallock, Michael Schulz, Larisa Sogacheva, Kerstin Stebel, Johanna Tamminen,

Gareth Thomas, Gijsbert Tilstra, Sophie Vandenbussche, Pepijn Veefkind, Marco Vountas, Yong Xue

ESA Technical Officer: Simon Pinnock

Final Report

Version 1.2

Aerosol_cci2_FR_v1.2.pdf

Aerosol_cci2

Final Report

REF : Aerosol FR ISSUE : 1.2 DATE : 09.01.2019 PAGE : II

partner acronym role city country

German Aerospace Center DLR Main contract Oberpfaffenhofen D

Finnish Meteorological Institut FMI Sub contract Helsinki FI

Barcelona Supercomputing Center BSC Sub contract Barcelona ES

Belgium Institute for Space Aeronomie BIRA Sub contract Uccle B

Centre National de la Recherche Scientifique ICARE Sub contract Lille F

Eidgenössische Technische Hochschule Zürich ETHZ Sub contract Zürich CH

Freie Universität Berlin FUB Sub contract Berlin D

Laboratoire de Meteorologie Dynamique LMD Sub contract Paris F

Lille Observatoire Atmospherique LOA Sub contract Lille F

London Metropolitan university LonMet Sub contract London UK

Max Planck Institut MPI Sub contract Hamburg / Mainz D

Meteorological Institute of Norway MetNo Sub contract Oslo NO

Norwegian Institut for Air Research NILU Sub contract Oslo NO

Oxford university UOx Sub contract Oxford UK

Rayference RF Sub contract Brusseles B

Royal Meteorological Institute KNMI Sub contract De Bilt NL

Rutherford Appleton Laboratory RAL Sub contract Harwell UK

Swansea university SU Sub contract Swansea UK

Universität Bremen UB Sub contract Bremen D

Universite Libre de Bruxelles ULB Sub contract Brusseles B

University of Derby UD Sub contract Derby UK

Technical University Delft TUD Sub-sub contract Delft NL

University of Saskatchewan US Sub-sub contract Saskatoon CAN

Aerosol_cci2

Final Report

REF : Aerosol FR ISSUE : 1.2 DATE : 09.01.2019 PAGE : III

TABLE OF CONTENTS

TABLE OF CONTENTS ....................................................................................................... III

EXECUTIVE SUMMARY .................................................................................................... IV

1 INTRODUCTION .......................................................................................................... 1

2 AEROSOL_CCI PHASE 2 PRODUCTS ..................................................................... 3

3 VALIDATION AND UNCERTAINTIES .................................................................... 5

4 USER CASE STUDIES ................................................................................................ 14

5 ASSESSING CHALLENGING NEW PROPERTIES .............................................. 22

6 OVERALL ACHIEVEMENTS ................................................................................... 25

7 CONCLUSIONS AND OUTLOOK ............................................................................ 27

8 PUBLICATIONS .......................................................................................................... 29

Aerosol_cci2

Final Report

REF : Aerosol FR ISSUE : 1.2 DATE : 09.01.2019 PAGE : IV

EXECUTIVE SUMMARY

When this project “Aerosol_cci - Phase 2” within the ESA Climate Change Initiative (CCI) started in spring 2014 it was based on its Phase 1 (2010 – 2014), where already a substantial part of the European satellite aerosol retrieval community had collaborated. During Phase 1 major improvements had been achieved for several total column aerosol retrieval algorithms for European sensors providing mainly Aerosol Optical Depth (AOD); a round-robin inter-comparison exercise for eight different algorithms had been conducted and a concept for iterative algorithm development and assessment had been developed and demonstrated with limited length global datasets (4 months covering all seasons and subsequently extending to one 12-month full annual cycle). At its end, one first long-term AOD record (1995 – 2012) from ATSR-2 and AATSR (together also referred to as “ATSR”) had been processed and evaluated. Finally, a concept for pixel-level uncertainties had been developed and demonstrated. Additionally, a global demonstration dataset covering stratospheric extinction profiles was produced.

This report summarizes the achievements and conclusions of the Phase 2 project. In its baseline part this project focused on processing and evaluating consistent full mission data records of several European sensors with several different algorithms. The sensor suite included dual view radiometers (ATSR-2, AATSR; first demonstration for SLSTR) as well as a polarimeter (POLDER), but also as a new element a thermal infrared spectrometer (IASI); a star occultation instrument (GOMOS) was used to infer stratospheric extinction profiles. This set of different sensors provides complementary information content which allows to retrieve different parameters of the global aerosol distribution: Total aerosol optical depth and its fine mode part (AOD, FMAOD from ATSR instruments), dust AOD (from IASI), consistently retrieved AOD, FMAOD together with absorption information (single scattering albedo SSA or Absorbing AOD AAOD, from POLDER). Furthermore, an absorbing aerosol index (AAI) from a series of UV-VIS spectrometers (TOMS, GOME, GOME-2, SCIAMACHY, OMI) was prepared with support of this project and the stratospheric extinction profiles (GOMOS) were processed. With this set of algorithms, climate data records covering 10 years (GOMOS, IASI; POLDER for selected regions and one year over Africa), 17 years (ATSR instruments), and even 30 years (AAI) were processed and evaluated against ground-based reference networks (AERONET, NDAAC lidars), other satellite retrievals (NASA MODIS, MISR, SeaWIFS) and modelling datasets (AeroCOM, MACC, EMAC). In several cases more than one algorithm was used for the retrieval from one sensor since no single algorithm proved best under all conditions. One further important task of the project was to conduct altogether eight user case studies for different applications, in which the strengths and limitations of the different datasets could be assessed.

Several optional tasks were added in which additional algorithms were added, including nadir-only AOD retrievals to increase spatial coverage (MERIS), geostationary radiometer to assess the diurnal cycle (SEVIRI onboard MSG). Other additional work packages compared approaches to invert difficult additional aerosol properties (absorption, layer height). At the end of the baseline project (12/2017) a one-year bridging option was added, which enabled to continue on smaller scale algorithm development and assessment with limited data periods (4-12 months) for ESA sensors with existing data record and in-orbit successor instruments, namely AATSR (SLSTR) and MERIS (OLCI).

At its end this project has led to a visible integration of the European satellite aerosol retrieval community with close links to various user communities, it established several European aerosol data records in the international community, it achieved acceptance of its new concept for pixel-level uncertainties and its iterative algorithm evolution cycle. By initiating the International Satellite Aerosol Science Network AEROSAT (which has just held its 6th annual workshop in collaboration with the modelling community workshop of AeroCOM) this project has contributed to strengthening the international collaboration on the use of satellite data records for aerosol science.

Aerosol_cci2

Final Report

REF : Aerosol FR ISSUE : 1.2 DATE : 09.01.2019 PAGE : 1

1 INTRODUCTION

The predecessor project “Aerosol_cci - Phase 1” (2010 – 2014) within the ESA Climate Change Initiative (CCI) had already involved a substantial part of the European satellite aerosol retrieval community. In its course major improvements had been achieved for several total column aerosol retrieval algorithms for European sensors providing mainly Aerosol Optical Depth (AOD); a round-robin inter-comparison exercise for eight different algorithms had been conducted. At its end, one first long-term ATSR AOD record (1995 – 2012) had been processed and evaluated. Finally, a concept for pixel-level uncertainties had been developed and demonstrated. Additionally, a global demonstration dataset covering stratospheric extinction profiles was produced. With regard to cross-consistency with other satellite Essential Climate Variables an assessment of potential gaps and overlaps with cloud products inferred from the same sensor had been made. A concept for iterative algorithm development and assessment within a science-driven agile de-centralised system had been developed and demonstrated with limited length global datasets (4 months covering all seasons and subsequently extending to one 12-month full annual cycle) and a few selected core users had been included to better understand their needs.

Based on these earlier experiences this Phase 2 project (shortly “Aerosol_cci2”) continued and extended the best practices of the Phase 1 with an extended Climate Research Group (with applications related to aerosol model evaluation, stratospheric research, aerosol-cloud interaction, data assimilation through Climate Model User Group, radiative forcing), maintaining the set of validation tools and concepts by independent experts (adding the assessment of data record stability and aerosol type variables beyond AOD), using Round robin exercises where appropriate (new IASI products, options of absorption and layer height products), extended consistency analysis with other ECV projects, harmonized developments of uncertainties (with the support of metrologists), and continuing the iterative algorithm development and processing approach (with foci on cloud masks and aerosol properties).

The foremost focus of the Phase 2 project was on long time series production of full mission time series for all datasets; adding new datasets with complementary information (IASI dust only AOD) and testing the use of the sensor with highest information content (POLDER) as spatially continuous quasi-reference between ground stations. A set of dedicated user case studies was included to test different applications. A sustainable but distributed system embedded the datasets of Aerosol_cci2 into the common framework of the CCI program with its system engineering working group, common standards (CF metadata convention, netCDF file format) and tools (e.g. filename and variable conventions, versioning) and model community standards to facilitate dataset usage (Obs4MIPS). Additionally, preparations for future sensors were started (Copernicus Sentinel family) and difficult retrieval aspects were tested (e.g. joint aerosol cloud retrieval, layer height, absorption retrieval).

Figure 2 shows the main elements of the project workflow in its annual cycle. Based on the results of the user requirements analysis, algorithms were improved and thus far immature or new algorithms were qualified through round robin exercises as part of algorithm development (with independent validation experts already involved for the assessment). Using mature, as determined from the round-robin results, algorithms in an evolving processing system of de-central processing chains full-mission ECV time series were produced and made available in the project archive. Subsequently these products were comprehensively validated and inter-compared to external reference data with different complementary tools. The loop closed with user assessment of the resulting products in dedicated case studies and exposure to the wider user community in annual user workshops, international conferences and peer-reviewed publications. In the center of the project cycle the

Aerosol_cci2

Final Report


annual work plan defined priorities based on dialogue within the project science team (all EO, validation and user partners) which was embedded into the international AEROSAT network to ensure a wide support from the aerosol retrieval community. Depending on maturity and status at its respective start, the entry into this loop occurred at different points (e.g. for AATSR already one full time series was available for in depth analysis of the overlapping period, whereas for IASI a round robin exercise needed first to be conducted to qualify and inter-compare aerosol algorithms for a first time). Specific algorithm work and validation was conducted on aerosol type, uncertainties, analysis of consistency with other ECV products, and joint aerosol-cloud retrieval. Documentation from all steps (requirements, algorithms, validation, system, user assessment) was made publicly available in the project web portal. Specific activities (lighter shade of grey) were only conducted in one of the three project years: use case studies (year 2) and Sentinel preparation (year 3).

Figure 2: Overview of Aerosol_cci2 approach (from the Aerosol_cci2 proposal)

Aerosol_cci2

Final Report


2 AEROSOL_CCI PHASE 2 PRODUCTS

In conformance with the thematic requirements of the ESA tender for Phase 2 for the aerosol ECVs the project aimed at producing a set of validated aerosol ECV products for the full time series from a set of European satellite instruments with substantially different characteristics complementary to each other. Each product includes a pixel-wise error characterization. The aerosol ECV product specifications were updated based on interactions with the user communities and user feedback, resulting in yearly updates of user requirements.

Table 2 provides the mature climate data records produced and qualified in this project.

Table 2: Mature Climate Research Data Package

algorithm version sensor(s) responsible provider

Main aerosol parameters

Resolution coverage

period(s)

ADV / ASV

2.31 (also 2.30_plume)

AATSR FMI AOD, FMAOD 10km, 1° global

2002-2012

2.31 ( also 2.30_plume)

ATSR-2 FMI AOD, FMAOD 10km, 1° global

1995-2003

ORAC 4.01 AATSR UOxford / RAL AOD, FMAOD 10km, 1° global

2002-2012

4.01 ATSR-2 UOxford / RAL AOD, FMAOD 10km, 1° global

1995-2003

SU 4.3 AATSR USwansea AOD, FMAOD 10km, 1° global

2002-2012

4.3 ATSR-2 USwansea AOD, FMAOD 10km, 1° global

1995-2003

ensemble 2.6 AATSR DLR AOD 10km, 1° global

2002-2012

2.6 ATSR-2 DLR AOD 10km, 1° global

1995-2001

AERGOM 3.00 GOMOS BIRA stratospheric extinction profiles + AOD

10x2,5° 1km vertical global

2002-2012

IMARS 5.2 IASI DLR Dust AOD 12km, 1° global

2007-2015

MAPIR 3.51 IASI BIRA Dust AOD 12km, 1° “Greater Sahara”

2007-2015

ULB 7 IASI ULB Dust AOD 12km, 1° global

2007-2015

LMD 1.3 IASI LMD Dust AOD 12km, 1° 60S – 60N

2007-2015

GRASP 0.07.00 POLDER LOA AOD, SSA regions of 10x10 deg

Entire Africa

2005-2013 12/2007-11/2008

Table 3 contains a list of algorithms (including tests of less mature algorithms which were then not included into the mature climate research data package) with sensors they were applied to.

Aerosol_cci2

Final Report


Table 3: Algorithms and sensors targeted in Aerosol_cci2

sensor algorithm

OR

AC

AD

V/A

SV

Swan

sea

SYN

AER

GR

ASP

AA

I

Stra

tosp

he

ric

ext

inct

ion

MA

PIR

IMA

RS

ULB

LMD

XB

AER

SEA

WIF

S4M

ERIS

CIS

AR

Dual view radiometers (1995 – 2012, 2016)

AATSR x x x x

ATSR-2 x x x

SLSTR x x x

Multi-angle polarimeter (2005 - 2013)

POLDER x

UV-VIS spectrometers (1978 – 2015)

TOMS x

SCIAMACHY x x

GOME x

GOME-2 x

OMI x

TROPOMI x

Star occultation spectrometer (2002 - 2012)

GOMOS x

Thermal infrared spectrometer (2007 - ongoing)

IASI x x x x

Single-view radiometer (2002 - 2012)

MERIS x x

Geostationary radiometer (2006 - ongoing)

SEVIRI x

ESA Aerosol_cci products are publically and freely available at http://www.icare.univ-lille1.fr/archive/?dir=CCI-Aerosols and at ftp.icare.univ-lille1.fr. Unrestricted access is provided without user registration – a user shall use the common CCI user account: account name: cci; password: cci

The following license is valid for using the Aerosol_cci products:

The products provided on this server are openly and freely available. No warranty is given by their providers. Users are obliged to acknowledge the ESA Climate Change Initiative and in particular its Aerosol_cci project together with the individual algorithm developer.

We encourage interaction with the algorithm developers on proper use of the products and would like to receive a copy of all reports and publications using the datasets. An offer of co-authorship should be considered, if the CCI datasets constitute a major component of a scientific publication.

Aerosol_cci2

Final Report


3 VALIDATION AND UNCERTAINTIES

In each development cycle the datasets were evaluated by independent experts in comparison to ground-based reference measurements at globally distributed stations (AERONET total column, NDACC lidar profile measurements) resulting in statistical validation quantities for level2 (satellite projection pixels) and level3 (gridded) dataset versions. Furthermore, the datasets were inter-compared with other satellite datasets to understand common features and to assess their evolution and maturity.

The following table 4 summarizes validation results for the latest mature datasets at satellite pixel resolution level2) for the Aerosol_cci2 products (which were specified in the Product Specification Document (PSD)). Datasets are global and cover the entire periods for all sensors (POLDER only selected regions). The detailed validation is described in the Product Validation and Inter-comparison Report (PVIR)).

Table 4: Summary of Aerosol_cci2 validation results for the mature datasets as in table 2 (as per end of 2017 in the Product Validation and Inter-comparison Report)

algorithm spatial resolution coverage

rmse/bias land rmse/bias ocean

ADV (AOD level2)

AATSR 10km, 1° global 0.11 / - 0.03 0.09 / + 0.02 ATSR-2 10km, 1° global 0.14 / + 0.02 0.29 / + 0.11

ORAC (AOD level2)

AATSR 10km, 1° global 0.14 / + 0.01 0.16 / + 0.11 ATSR-2 10km, 1° global 0.11 / + 0.03 0.15 / + 0.05

SU (AOD level2)

AATSR 10km, 1° global 0.14 / + 0.02 0.06 / + 0.01 ATSR-2 10km, 1° global 0.16 / + 0.06 0.09 / + 0.08

ensemble (AOD level2)

AATSR 10km, 1° global 0.12 / - 0.01 0.07 / + 0.01 ATSR-2 10km, 1° global 0.14 / + 0.02 0.18 / + 0.08

IMARS Dust AOD

IASI 12km, 1° global dust belt 0.26 / - 0.02

LMD Dust AOD


MAPIR Dust AOD

IASI 12km, 1° global dust belt 0.30 / + 0.17

ULB Dust AOD


GRASP AOD SSA ANG

POLDER 6km, 1° regions land 1) 0.00 … 0.12 / + 0.08 … 0.23 0.05 … 0.08 / + 0.00 … 0.04 0.27 … 0.37 / - 0.05 … + 0.04

AAI OMI 0,25° global Direct validation is not possible (no geo-physical quantity); evaluation through model inter-comparison

AERGOM (extinction level3)

GOMOS 60 longitude X 5° latitude 1 km vertical global

< 1 · 10-4 km-1 (above 15 km altitude; few validation matches only)

1) POLDER GRASP validation includes very difficult bright / desert stations / separate validation for 4 regions (Banizoumbou, Kanpur, Mongu, Beijing)

In the following figures 3 - 8 example validation results for the different Aerosol_cci2 products (from the Product Validation and Inter-comparison Report and its underlying analysis) are shown

Aerosol_cci2

Final Report


graphically also illustrating the different types of assessments. Figure 3 shows scatter plots of level2 retrievals compared to AERONET and maps of the three AATSR retrievals (AOD and FMAOD). Figure 4 shows results of a scoring approach (which quantifies bias and the capabilities of an algorithm to reproduce regional and seasonal patterns via assessing spatio-temporal correlations) for the same 3 AATSR retrievals and the uncertainty-weighted ensemble of them.

Figure3: Total AOD error statistics (left ) for different AATSR retrievals and retrieved total annual AOD (center) and fine-mode AOD (right) versions for the year 2008.

Figure 4: Regional year 2008 combination scores of monthly AOD for the 4 AATSR datasets (from left: ensemble, Swansea, ADV, ORAC) when applying an aerosol climatology for AOD (AERONET / MAN + modeling) as reference. Green colors indicate a higher retrieval skill and red colors a poorer skill.

Aerosol_cci2

Final Report


The stability of the ATSR-2 / AATSR datasets is assessed in figure 5 (total AOD) and 6 (fine mode AOD) with time series of relative bias to a selected subset of 43 AERONET ground stations which provide sufficient coverage over the entire period (note that the mean global AOD is on the order of 0.15 and thus 10% bias means 0.015 in AOD absolute values, for fine mode AOD the global mean value is about half of the total AOD).

Figure 5: Time series of yearly normalized mean bias (NMB) for ATSR AOD retrievals at up to 43 AERONET sites using daily AOD matches for ATSR-2 (dashed) and AATSR (solid).

Figure 6: Time series of yearly normalized mean bias (NMB) for ATSR fine mode AOD retrievals at up to 43 AERONET sites using daily AOD matches for ATSR-2 (dashed) and AATSR (solid).

Aerosol_cci2

Final Report


Figure 7 shows maps and level3 / gridded validation statistics for the 4 IASI datasets during the IASI round robin exercise for the dust belt.

Figure 7: Annual average dust AOD maps (left) and scatter statistics (right) from monthly comparisons to AERONET site statistics for the four IASI retrievals: ULB 7.0 (row 1), LMD 2.1 (row 2), DLR 5.2 (row 3) and MAPIR 3.0 (row 4).

Aerosol_cci2

Final Report


Figure 8 shows the validation of POLDER/GRASP which has as its outstanding figure the capability to retrieve consistently several aerosol properties, namely total abundance (as AOD), size / shape (fine mode AOD and non-spherical AOD) and absorption (absorbing AOD or single scattering albedo).

Figure 8: Validation with scatter plots against AERONET of POLDER/GRASP monthly means 2008 over four selected regions for different aerosol properties (from left: total AOD, fine mode AOD, absorbing AOD) Validation results for the last sensor line for which algorithm development was conducted, namely the nadir-only MERIS radiometer are shown in figure 9 (here comparisons of monthly mean AOD to a reference climatology dataset based on a multi-model median for its spatial patterns and on AERONET measurements for quantitative values).

Figure 9: Monthly AOD differences in 2008 for different MERIS retrievals with respect to MACv2_08 climatology: from left to right: GRASP (outside Aerosol_cci2); XBAER, SeaWIFS4MERIS; and for reference SeaWIFS. Red colors show overestimates and blue colors suggest underestimates.

Aerosol_cci2

Final Report


As final validation example, figure 10 shows the validation of extinction profiles in the stratosphere against four ground-based NDACC lidar observation stations (note that only few stations provide such stratospheric extinction profiles). The most important issue encountered during this validation was the choice of the lidar ratio, an indispensable parameter to convert the lidar backscatter into an extinction coefficient.

Garmisch Partenkirchen Mauna Loa

Dumont d’Urville – Background aerosols Dumont d’Urville – PSCs

Figure 10: Comparison between GOMOS level 2 (blue) and lidar profiles (black) for Garmisch Partenkirchen (upper left panels), Mauna Loa observatory (upper right panels) and Dumont D’Urville identified as background aerosol (lower left panels), and as observations affected by PSC influence (lower right panels). Shown are the median profiles for the entire period (plus standard deviations), the median and inter-quartile range of the relative differences between GOMOS and the lidar data [in %]. The numbers of co-located datasets, which were found in the respective altitude bins, are given on the right y-axis.

Aerosol_cci2

Final Report


Aerosol product validation summary

In summary, the independent validation of new dataset versions was a continuous source of feedback to define next steps of necessary algorithm development. The status at the end of the project validation and inter-comparison showed the significant improvements of almost all algorithms through the iterative development cycle. The best dual view (ATSR) product is becoming similar in quality (but not in coverage) to most established NASA datasets, but detailed evaluation shows remaining limitations (e.g. for high AOD or over bright surfaces). Other products (e.g. MERIS with its comparatively low information content) could not yet reach the envisaged level of quality (e.g. set by a similar NASA SeaWIFS product). IASI aerosol datasets as rather new product could be established but still show higher uncertainties (in particular related to the needed transfer from thermal to visible spectral range).

For ATSR (AOD, fine mode AOD) there is a general tendency to overestimate lower AOD and to underestimate higher AOD (except for SU over desert). Retrievals of AOD are more accurate than retrievals for fine mode AOD. SU offers overall the best ATSR retrieval (despite too much dust and fine-mode AOD), but overestimates strongly near dust sources and in fine mode AOD. ADV underestimates AOD over land but overestimates AOD over oceans, while ORAC has improved much over land (at the expense of an increased ocean bias). The uncertainty-weighted ensemble has not the best skill over oceans, but good skill over land; its extra coverage is small. Pixel-level uncertainty estimates are in parts too low. Most important it was shown that AATSR data are consistent over time, less so with the earlier sensor ATSR-2. Future improvements can be made by assessing co-retrieved parametes such as fine mode fraction or surface reflectance as shown for some regions in the bridging option at the end of the project.

IASI is capable of providing a complementary product which covers dust AOD alone. Two of the retrievals (ULB, LMD) show good agreement with AERONET coarse mode, while the two other retrievals need further improvements (MAPIR has large overestimates, IMARS has low coverage). The MERIS (AOD) products are still not yet mature despite of strong improvements between the last versions (XBAER underestimates over coastal oceans, SeaWIFS4MERIS has large overestimates over continents in particular in dust source regions). The SeaWIFS potential did not yet materialize (however underestimates are found for the SeaWIFS4MERIS algorithm similar to the SeaWIFS original dataset due to the minimal reflectance approach failing in constantly polluted areas). An external MERIS dataset retrieved with the GRASP algorithm shows the smallest noise of all MERIS datasets.

The GRASP / POLDER dataset contains the largest set of variables (AOD, SSA, size-distribution, surface reflectance) and its AOD attribution to fine and coarse mode is much better than in other retrievals. Its level2 AOD local comparisons do well at high AOD cases over continents, while level3 AOD regional comparisons however reveal unique large AOD biases, where more evaluations are needed before an AOD reference status can be considered. The GOMOS (stratospheric extinction) dataset provides a meaningful complementary product which despite its sparse coverage (at 5 day, 5x60 degree latitude / longitude bins) shows good agreement to limited reference data within error bounds.

The best Aerosol_cci2 data quality scores are offered by AATSR / SU and POLDER / GRASP. Comparing their skill scores to those of MISR and MODIS collection 6, it appears that they are competitive to the better NASA retrievals for AOD. At lower AOD, Aerosol_cci2 scores are on average even better (due to AOD overestimates over oceans by MISR and MODIS). However, coverage (ATSR) and lack of processing (GRASP) remain major handicaps in direct competition. Another important

Aerosol_cci2

Final Report


aspect is that AATSR (with SLSTR), POLDER (with 3MI) and MODIS (with VIIRS) have successor sensors.

Validation of pixel level uncertainties

An estimate of the uncertainty on a measurement is necessary to make appropriate use of the information conveyed by a measurement. Awareness of the importance of pixel-level uncertainty estimates is beginning to pervade the aerosol remote sensing community through efforts such as the yearly AeroSAT workshops. By comparing the performance of the algorithms evaluated within the Aerosol_cci2 project to each other and validation data, the quality of their uncertainty estimates could be evaluated and potential areas for improvement were illuminated in the iterative development cycles. This work led to better homogenisation of uncertainty calculations among the different algorithms for the ATSR sensors. The IASI algorithms are in an earlier stage of development, but are already including the estimation of uncertainty – its validation is even more complex (need to use AERONET coarse mode AOD (at 550 nm) as proxy for IASI (at 10 micron) retrievals which include as challenging step the conversion from 10 micron to 550 nm). It should be noted that aspects such as uncertainties induced by errors in the cloud masks are not yet included into the pixel-level uncertainty estimates. Figure 11 shows the statistical comparisons of histograms of predicted uncertainties (in blue) and of the best guess for the true error (difference to AERONET relying on the significantly lower uncertainty of AERONET AOD measurements), where overall a good match can be seen for most products. Figure 12 shows the temporal evolution of the fraction of pixels where the ratio of predicted uncertainty and true error is better than 1 (this fraction should be 63% for a Gaussian distribution) – here a good stability over time can be seen, but also a constant deviation of fractions for some of the products which indicates general over/under-estimation of uncertainties.

Aerosol_cci2

Final Report


Figure 11: Uncertainty histograms for ATSR-2 land / coast (column 1 / 2), AATSR land / coast (column 3 / 4),: SU 4.3 (row 1), FI 2.3 (row 2) and OX 4.01 (row3). The predicted retrieval uncertainty statistics are in blue and the observed error statistics from observations are in red.

Figure 12: Percentage of pixels per year where the ratio of predicted uncertainty and true error ( Δ) lies within [−1, +1] for the ADV v230_plume, ORAC v401, SU v43 and Ensemble v26 algorithm for ATSR-2 and AATSR over land (upper plot) and coastal (lower plot) sites. The black line shows the optimum percentage value. The dotted lines, without symbols, are showing years where less than 100 co-locations were found, so these comparisons are not significant.

Aerosol_cci2

Final Report


4 USER CASE STUDIES

Work in the Aerosol_cci2 project was based on the user requirements of the Global Climate Observing System (GCOS) as summarized in table 5. Regarding variables, Aerosol_cci2 provides several AOD products (ATSR, POLDER, MERIS) and stratospheric extinction profiles (GOMOS). Further variables were demonstrated / tested: absorption information / SSA (in particular from POLDER), layer height (in particular from IASI). In addition to the GCOS required variables, aerosol compositional information is highly requested from the AeroCOM modelling community – here Aerosol_cci2 provided Fine Mode AOD (from ATSR and POLDER) and Dust AOD (from IASI). While the horizontal and vertical resolution of the Aerosol_cci2 products can meet the GCOS requirements, the temporal resolution below 24 hours cannot be fulfilled with polar orbiting satellites; here the demonstration of hourly AOD / FMAOD from SEVIRI geostationary observations was made in Aerosol_cci2. Requirements for accuracy and stability are highly challenging and cannot be met entirely with any satellite aerosol product; however through the iterative development cycles accuracy was improved and came closer to the GCOS requirements and a first AOD record stability assessment was made which also showed reasonable agreement with the requirement.

Table 5: GCOS requirements for aerosol properties

variable resolution accuracy stability per decade horizontal

(km) vertical

(km) temporal

AOD (column) 5-10 N / A 4 h max (0.03, 10%) 0.02 SSA (column) 5-10 N / A 4 h 0.03 0.01 layer height 5-10 N / A 4 h 1 km 0.5 km extinction (profile)

200-500 1 (at ~10km) 2 (at ~30km)

1 week 10% 20%

Table 6 shows another type of requirements in this case from the Copernicus Atmosphere Monitoring Service (CAMS) for data assimilation of total AOD products. Here timeliness and coverage are key requirements together with a comprehensive quantification of pixel level uncertainties.

Table 6: Data assimilation (CAMS) requirements for aerosol properties (level2, near-real time)

threshold

requirement target

requirement

coverage and sampling

geographic coverage global global

temporal sampling 500 observed

locations per hour 1000 observed locations per hour

temporal extent 1991-present 1982-present

resolution

horizontal resolution 20km 5km vertical resolution N/A N/A

error / uncertainty

precision 0.05 0.02 accuracy 0.05 0.02 stability N/A N/A

error characteristics global statistics sample statistics

Aerosol_cci2

Final Report


Starting from the requirements discussed above, the Aerosol_cci2 team interacted during several workshops and conferences with a number of potential user communities to further enhance the understanding of requirements and to inform them on the achieved capabilities of aerosol datasets. A summary of relevant user communities is provided in table 7.

Table 7: Overview of Aerosol_cci2 user communities / applications

user segment application major user need user examples process studies understand processes,

improve parameterization in models

high resolution Obs4MIPs, CMIP

aerosol-cloud interaction understand processes, reduce associated IPCC uncertainty

high resolution associated products of aerosol and cloud variables (most important: cloud droplet number concentration)

ACPC

model development initialize, evaluate models, provide additional information on emissions

several year datasets high information content (e.g. fine mode AOD, dust AOD, stratosphere)

AeroCOM EMAC SPARC

radiative forcing monitor trends and variability vertical profiles absorption

GCOS, EMAC

aerosol monitoring monitor trends and variability stable long-term time series GCOS reanalysis consistent time series, data

assimilation bias-free data pixel-level uncertainty

CAMS

forecasting warning, data assimilation NRT data ICAP, CAMS, SDS-WAS

Eight different user case studies as listed in table 8 were chosen to investigate the potential value of new Aerosol_cci2 data products for the climate community in data applications. More specifically, in these user case studies aerosol schemes in global models were evaluated, data were applied in assimilations to demonstrate their impact of improved (aerosol) forecasts, spatially and temporally associated aerosol and cloud property retrievals were combined to establish observational constraints to evaluate and to identify deficiencies in (aerosol-cloud) processes in global modelling and finally aerosol retrieval data were analysed to better quantify their sources, and were applied in radiative transfer modelling to determine associated impacts on climate.

Table 8: Overview of Aerosol_cci2 user case studies

user case study using institute lead author Temporal trends in AOD ATSR MetNo Michael Schulz Aerosol direct radiative forcing ATSR MPI-Met Stefan Kinne Aerosol-cloud interactions ATSR, MODIS ETHZ David Neubauer Long-term data record on UV aerosol index AAI KNMI Pepijn Veefkind Assimilations of IASI dust AOD IASI BSC Enza Di Tomaso Evaluation and improvement of the CCM EMAC GOMOS

IASI, ATSR MPI-C Christoph Brühl

Temporal trends in (natural) coarse-mode AOD ATSR MetNo Jan Griesfeller Aerosol-Cloud relations in satellite data ATSR MPI-Met Stefan Kinne

The following figures 13 – 20 show example results (as described in detail in the Climate Assessment Report (CAR) and associated publications) of these user case studies including long-term regional records of the absorbing aerosol index (figure 13), assimilation of IASI dust AOD into an atmospheric dust model (figure 14), model evaluation with combined ATSR and IASI datasets in an atmospheric model to improve dust parameterization (figure 15), use of GOMOS observations to fill gaps in volcano emission inventories based on MIPAS and for additional information on aerosol type via the spectral properties (figure 16), consistent multi-sensor records of AOD, fine and coarse mode AOD

Aerosol_cci2

Final Report


from ATSR algorithms (figure 17), aerosol-cloud interaction susceptibilities from modelling and various satellite datasets (figure 18), assessments of the relations between fine mode aerosol (as a proxy for cloud condensation nuclei) and cloud droplet number concentrations (figure 19), and using those relations in radiative forcing calculations of the Twomey effect (figure 20).

Figure 13: Temporal long-term trends of the satellite data-based UV-AI index data for southern central Africa (top panel, labelled ‘West Africa’) and for West Africa (lower panel, labelled ‘Sahel’). The green dashed line is a fit to the seasonal minima to demonstrate the stability of the time-series. The blue dashed line is a fit to the seasonal maxima, indicating the trend in aerosol source strength (assuming similar absorption and altitude) over time. Different colors indicate datasets from different satellite sensors: TOMS (black), GOME-1 (red), SCIAMACHY (brown), OMI (green), GOME-2A/B (blue).

Figure 14: NMMB-MONARCH v1.0 model results for March – June 2015 obtained by assimilating Level 3 IASI dust AOD (550 nm) data of the ULB v8 (upper left), LMD v2.1 (upper right), MAPIR v3.5 (lower left) and IMARS v5.2 (lower right).

Aerosol_cci2

Final Report


Figure 15: Annual mean 10 μm dust AOD for 2011 simulated by EMAC with revised dust emission scheme (upper panel) and the ULB.v8 retrieval for the IASI sensor (bottom panel).

Figure 16: comparison of stratospheric aerosol extinctions (in log10 scale) based on GOMOS retrievals (top) and EMAC model simulations (bottom) at 550nm wavelength

Aerosol_cci2

Final Report


Figure 17: Time-series and trends in ATSR AOD (upper left), fine (upper right) and coarse (lower right) mode AOD from three ATSR retrievals for 1997 – 2011 and AATSR and MODIS total AOD for 2003-2011 over South Asia.

Figure 18: Susceptibilities of cloud liquid water path (cloud top pressure > 500 hPa, cloud top temperature > 273.15 K) to aerosol number concentration (approximated by the aerosol index AI = AOD*Ångström). Values for ln(LWP)/ln(AI) are compared for different environmental regimes based on model results (ECHAM6-HAM2, green, ECHAM6-HAM2 with AI-dry, violet), on satellite retrievals by AATSR-CAPA (red) and MODIS-CERES (blue). The left panel displays non-raining scenes and the right panel displays raining scenes.

Aerosol_cci2

Final Report


Figure 19: fine mode AOD vs CDNC relationships using monthly 1x1 lat/lon retrieval averages of two different CDNC retrievals with MODIS sensor data (upper row) and the same relationship only with data for the boreal summer of MODIS and ATSR sensor data (lower row). The probability frequency is indicated by the color and by fine mode AOD bins data are summarized by average (*), median (-) and uncertainty (boxes establishing 25th and 75th percentiles and whisker ends 10th and 90th percentiles). Functions of best fit logarithmic curves are presented at the top right corner of each panel.

Figure 20: Annual maps for present-day indirect (Twomey) effects by anthropogenic aerosol on solar (left column) and IR (right column) radiative net-fluxes at the top of the atmosphere (top) and at the surface (bottom). Impacts with (standard) anthropogenic AOD based on AeroCom 2 simulations (left block, AC2) are compared to impacts with anthropogenic AOD based on AeroCom 1 emissions (right block, AC1). Blue colours indicate a ‘cooling’ and red colours a ‘warming’. Values below the labels are global averages.

Aerosol_cci2

Final Report


Main conclusions from the user case studies

Temporal Trends in AOD

A weak global positive trend of fine mode AOD but not for total AOD was found. Also a global shift to smaller aerosol sizes is suggested as a weakly negative trend to coarse mode AOD is deduced while fine mode AOD slightly increases. Furthermore, regional shifts in fine mode AOD from Europe and Northern America to Southern and Eastern Asia are detected. These shifts are consistent with known regional changes in anthropogenic emissions, but the shifts are smaller than simulated by global modelling. Positive total AOD trends of AATSR (2003-2011) in the Middle East are consistent with retrievals of coarse mode AOD trends by MODIS (and ground-based AERONET data), but apparently the AATSR size attribution of AOD is incorrect as no significant coarse mode AOD trend is derived. Remaining inconsistencies between ATSR-2 and AATSR sensor regional averages during the overlap period need to be adjusted before any analysis, which is difficult. One obvious strength of the ATSR data record is the capability to document regional and seasonal AOD anomalies (e.g. pollution and wildfire as major contributors to total and fine mode AOD); regional anomalies indicate that this variability has decreased, partly because wildfire seasons (especially over South America) are less intense than in the 1990's. Longer time-series spanning several decades are needed (e.g. SLSTR data to extend the AATSR data record) to obtain consistent trends for aerosol properties; more capable sensors (higher number of independent observables) help reduce this required period.

Aerosol-Cloud Interactions

For studying aerosol-cloud interactions different satellite datasets (e.g. from MODIS, ATSR) agree on a positive logarithmic relationship between aerosol number (approximated by fine mode AOD) and cloud droplet number. This relationship translates into an indirect TOA forcing of -0.8 W/m2, much smaller than in most global models. IASI retrievals for both large size mineral dust and (ice-) clouds – even as function of altitude are a potential testbed to extract observational relationships between elevated (dust) aerosol and ice-clouds. The liquid water path relationship to aerosol number, which is affected by aerosol water uptake in humid environments around clouds, could be improved for AATSR and MODIS datasets with a new Cloud-Aerosol Pairing Algorithm (CAPA) also reducing effects by cloud contamination, aggregation or 3D effects. Altogether this led to a weaker liquid water path relationship to aerosol number (similar to model simulations where the aerosol water uptake was removed when computing the liquid water path relationship to aerosol number).

Aerosol in the EMAC Model

For model development (EMAC), simulated dust AOD distributions generally agree with the satellite data in the visible (ATSR and MODIS) and the infrared (IASI ULB) for both, the tropospheric and the stratospheric setup. A sophisticated modeling of dust and organic aerosol as well as a detailed volcano data set can reproduce the (aerosol) extinction in the lower stratosphere observed by GOMOS at three different wavelengths while also simulated total AOD in the mid-visible is very sensitive to aerosol water and composition of sea salt. In the modal model the bulk fraction has to be increased compared to ions to reduce artifacts. The satellite data help finding the best parameters.

Assimilation of Dust AOD

Assimilating ULB IASI retrievals tends to produce an analysis that underestimates dust AOD near source regions, with the exception of MAPIR retrievals which produce an overestimation of dust AOD everywhere; further studies are needed to optimally characterize observation uncertainty for IASI data assimilation.

Aerosol_cci2

Final Report


Evaluation of the Long Term Absorbing Aerosol Index

The AAI long term data record composed from different sensor data is relatively stable, but it provides (only) qualitative information on elevated aerosol absorption. Applications over central Africa and Southern America regions with seasonal wildfire activity reproduce the seasonality of elevated aerosol absorption very well. The use of the AAI data records by the climate modeling community is not straightforward and requires a use of an observation operator (AAI simulator).

Direct Aerosol Radiative Forcing

Finally, in radiative forcing the ATSR AOD climate data record (1996-2011) reveals that direct aerosol radiative effects have a strong regional and seasonal character.

Aerosol_cci2

Final Report


5 ASSESSING CHALLENGING NEW PROPERTIES

Capabilities to retrieve several additional aerosol properties were tested with limited datasets and round robin exercises. This included layer height inferred from IASI measurements (for mineral dust) and from UV-VIS spectrometers (total aerosol). One of the complexities was the limited availability of matching reference datasets for validation (in particular for the UV-VIS test cases). Figure 21 shows one analysis of test cases for IASI layer height with 4 algorithms (LISA complementary dataset provided by this institute outside the project consortium). In the upper row scatter plots of the CALIOP mean height (obtained by extinction-weighting) versus dust layer height from the various algorithms is shown. The color indicates the density of points. Also given are the Pearson’s correlation coefficient and root mean square error (RMSE). The center row provides the difference between the passive algorithm (IASI) and CALIOP extinction-weighted heights as function of the CALIOP column extinction. The color indicates the density of points. The bottom row displays frequency distributions of the difference between the height from the various algorithms and the CALIOP extinction-weighted height. Included is also a normal distribution fitted to the difference. The mean and standard deviation (σ) of the normal distribution together with the number of data points are given in each plot. Data are shown for the BIRA-IASB (first column), DLR (second column), LISA (third column) and LMD (fourth column) algorithms. Obviously there exist significant biases and in one case (DLR) a concentration near two distinct values, while the rmse of the algorithms lies near the GCOS required value of 1 km.

Figure 21: Analysis of trajectory-propagated comparisons of IASI layer height test data with CALIPSO averaged extinction layer height values (from report on layer height round robin).

Aerosol_cci2

Final Report


Another challenging aerosol property tested in a round robin exercise with several case studies concerned aerosol absorption. Figure 22 shows an analysis of three different algorithms over Southern Africa which reveal significant differences between the three products assessed but reasonable agreement with AERONET zonal means (not collocated, but regionally averaged). It can be concluded that there is some information on absorption in the data and by looking at the quantity “absorbing AOD” the high uncertainty of cases with low AOD (where the retrievals have the lowest information content) matter less.

Figure 22: Evaluation of absorption AOD from 3 algorithms PARASOL/GRASP, AATSR/Swansea and AATSR/ORAC) in Southern Africa: (left) maps of absorbing AOD, (centre column) zonal means of absorbing AOD and (right column) zonal means of total AOD together with AERONET ground-based values (from Aerosol_cci2 report on absorption round robin)

As an additional test, hourly AOD was retrieved from geostationary SEVIRI measurements for 2008 over Europe and North Africa. Using hourly observations from this radiometer allows to make up for its small spectral information content by taking benefit from the varying solar zenith angle and thus from the changing observation geometry throughout the day, so that meaningful AOD retrieval and FMAOD retrieval could be demonstrated. In response to a user requirement for studying rapidly evolving aerosol-cloud interactions, a combined demonstration product was prepared (as shown in figure 23), which includes FMAOD from the Aerosol_cci2 test datasets and Cloud Droplet Number Concentrations (courtesy of the Climate SAF) both inferred from SEVIRI.

Aerosol_cci2

Final Report


Figure 23: Combined demonstration product of SEVIRI Fine mode AOD (red colour bar) and cloud droplet number density (courtesy of CM SAF) for 1st July 2008 at 6 AM.

Aerosol_cci2

Final Report


6 OVERALL ACHIEVEMENTS

The main achievements of the Aerosol_cci Phase 2 project were: - substantial improvement of Aerosol ECV dataset quality - consistent long-term records over 1-3 decades - comprehensive evaluation of the datasets by independent experts including assessment of

their stability and validation of the pixel-level uncertainties - complementary parameters (Dust AOD, Fine Mode AOD, Total AOD, Absorbing AOD,

stratospheric extinction) with different coverage from different sensors AATSR, IASI, POLDER, MERIS, GOMOS

- extension of a volcanic eruption inventory, with a better quantitative estimate of sulfur emissions and an improved modelled stratospheric aerosol radiative forcing

- demonstration of the information content for layer height (IASI), and diurnal cycles (SEVIRI) - establishment of a concept for pixel-level uncertainties - establishment of AEROSAT as international forum - integration and visibility of the European aerosol retrieval community - evaluation of the usefulness of the datasets in eight user case studies - transfer of the routine tasks to the Copernicus Climate Change Service (C3S) - qualitative understanding of reasons for differences between datasets processed with

different algorithms (cloud masking, quality filtering, trade-off between accuracy and coverage)

- derivation of robust trends from different algorithms (with remaining biases between them)

Figure 24: (Upper plot) Time series of total AOD derived with 3 algorithms and an uncertainty-weighted ensemble of them from the two subsequent sensors ATSR-2 and AATSR covering the period 1995 – 2012 with overlapping period of the two sensors in 2002 /2003. (Lower plot) Time series of one of the above datasets (Swansea) and the ensemble for the common period with MODIS (2002 -2012). (extracted using AeroCOM tools)

Aerosol_cci2

Final Report


As one example figure 24 shows the large consistency of monthly global climate data records of AOD from 4 Aerosol_cci2 datasets, each including two sensors, and also with a NASA MODIS data record. There are remaining constant biases of the order of 0.02 and the strength of the seasonal cycles also shows some differences, but the long-term trends appear to be strongly correlated for those independently retrieved data records. This project yielded several lessons learned, which are briefly summarized here:

- Bringing together a critical mass of experts exposing / sharing openly their problems and solutions was highly beneficial to all algorithms and led to significant improvements

- The active involvement of users in this process was also very valuable

- The transfer of the prototype system “near science” to the operational Copernicus Service has been managed successfully; with strongly overlapping consortia transfer of further future algorithm developments is facilitated (but only as long as these are funded from various sources)

- The three AATSR algorithms are performing almost equal (with somewhat different trade-off between coverage and quality) and not a single one of them is performing best everywhere; approaches for a best combination of the three datasets into an uncertainty-weighted ensemble show similar quality to the best algorithm with slightly increased coverage

- Pixel-level uncertainties can be systematically produced as meaningful information including their validation; cross-ECV collaboration allowed to develop and demonstrate a consistent framework for uncertainties propagation

- Aerosol type information is emerging with fine mode AOD capabilities of radiometers, complementary dust AOD from thermal instruments and absorption capabilities of polarimeters; all this information is of high relevance for several applications (e.g. aerosol-cloud interaction, radiative forcing)

- Having several AOD algorithms and different instruments is very valuable as it can provide a better picture of the aerosol distribution and the uncertainties than a single algorithm and cover complementary variables needed to fully describe the atmospheric aerosol distribution

- International collaboration facilitates progress as challenge by other communities drives improvement and critical feedback on uncertainties helped refine the concepts

Aerosol_cci2

Final Report


7 CONCLUSIONS AND OUTLOOK

This project worked with several different sensor lines which cover complementary aerosol information:

• Dual view radiometers ATSR-2 (1995 – 2003), AATSR (2002 – 2012), SLSTR (2016 - ): Total AOD, Fine Mode AOD (or Ångström exponent)

• Multi-spectral, multi-angle polarimeter POLDER (2005 – 2013) and 3MI (2021 - ): Total AOD, Fine / Coarse Mode AOD (or Ångström exponent), Single Scattering Albedo (or Absorbing AOD), stereo aerosol layer height

• Thermal infrared spectrometers IASI (2007 - ): Dust AOD, and effective / mean geometric dust layer height

• Star occultation spectrometer GOMOS (2002 – 2012), continuation option ALTIUS(2022 - ): stratospheric aerosol extinction profiles and multi-spectral stratospheric AOD (size, PSCs)

• Nadir multi-spectral radiometers MERIS (2002 – 2012), OLCI (2016 - ), [AVHRR (1980 - , on a series of platforms)]: Total AOD

• Geostationary radiometer (SEVIRI, 2006 -): hourly AOD • UV absorbing spectrometers (GOME, SCIAMACHY, GOME-2, OMI, TROPOMI): aerosol

absorbing index (AAI) The project did not include limb sounding radiometers (e.g. SAGE, OSIRIS or, active sensors (lidars such as CALIOP) and focused on European instruments while using other similar established sensors, in particular from NASA (multi-spectral nadir instruments: MODIS, SeaWIFs; multi-angle instrument MISR) for comparison. One major future development line is the extension of its records by using the new generation sensors of the Copernicus Sentinel family (and other future missions such as METOP-SG 3M/I for POLDER or ALTIUS for GOMOS) which provide continuation of some of the sensor families treated by Aerosol_cci2 (SLSTR for AATSR, OLCI for MERIS, TROPOMI for UVVIS spectrometers) while the operational meteorological IASI and geostationary sensor lines are continuing. At the end of this project following gaps are identified, where further / continuing work is needed:

• Multi-mission consistency of underlying FCDRs • Validation reference data gaps (Southern hemisphere, open ocean, near clouds, coastal

waters) • Inversion reference validation data gaps (aerosol type for low AOD, e.g. SSA from AERONET) • In situ reference data usage gaps (humidity + profile closure) • Vertical profile reference data gaps (sparse balloon data, lidar time series available but badly

known lidar ratio dependance) • Satellite data record length (10 years are still short for significant trend analysis) • MERIS / OLCI: no thermal channels - weak cloud masking capability • AATSR / SLSTR: forward / backward view inconsistency, SLSTR remaining level1 calibration

issues • POLDER / GRASP: (so far) limited geographical coverage of high accuracy version • IASI: assessment of uncertainties and stability • GOMOS, POLDER: no sensor until 2022 (ALTIUS, 3MI)

There are further general development needs for improving speciated information (at least fine / coarse mode, but as demonstrated for IASI also mineralogical composition, and stratospheric aerosol composition), for further improved accuracy (combined sensors or multi-pixel algorithms, use of

Aerosol_cci2

Final Report


model auxiliary data), for a direct link of validation results to the GCOS accuracy requirement for AOD (here we defined the GCOS fraction), for consistent integration of information from different sensors, and for comprehensive integration of several datasets from one sensor / ensembles, and for improving the stratospheric aerosol size information.

Regarding uncertainties further validation needs to be done (IASI, GOMOS, POLDER, MERIS) and approaches need to be developed for estimating uncertainties due to tropospheric and stratospheric cloud contamination, for validation near clouds, on further assessment of systematic uncertainties / biases, and to develop uncertainty propagation to level3 / gridded datasets (taking into account sampling effects and correlation structures).

The modelling community has several more challenging requirements (which are different for different applications) including extending the satellite-based records into the past (back to early 1980s), consolidating overall consistency of trends (understand biases). There is a continuously stated wish from the model community for “a satellite median” or at least guidance on best use for different applications; here ensemble approaches with best use of uncertainty information and precise definition of aerosol type / components play an important role.

The dual view sensor line (where Aerosol_cci and Aerosol_cci2 have invested most of their work since 2010), a bridging of the gap between AATSR and SLSTR (2012 – 2016 without data) and full consistency between different forward (AATSR) / rearward (SLSTR) views which lead to different angular sampling need to be worked out. Further required work includes correction or patching for SLSTR level 1 problems, improved quality of higher AOD, improved quality over bright desert, improve quality of fine mode AOD, correcting overestimates near dust sources, improved consistent integration of the ensemble (especially over ocean), further improved uncertainty estimates, improved understanding on regional level of biases between the four datasets, and use of the new spectral channels of the SLSTR instrument.

The polarimeter sensor line needs assessment of its quality of full global coverage, improved AOD regional biases (compared to best model estimates), consolidated uncertainties, further accelerated processing for global application with highest accuracy, assess of record stability.

Regarding IASI for the first time a comprehensive comparison of 4 aerosol algorithms was made. Two of the algorithms of the thermal sensor line have to work on correcting significant issues which showed up in the Aerosol_cci2 evaluation. Further work is needed to improve the conversion factor from thermal 10 µm retrieved AOD to mid-visible AOD at 0.55 µm and to quantify its uncertainty, to fully evaluate uncertainty estimates and to develop a more comprehensive validation of layer height.

The star occultation sensor line needs future comprehensive validation including routine approaches with limited available reference data, and further development and evaluation of algorithms on aerosol size and type. Further, GOMOS remains an indispensible testbench for the preparation of ALTIUS that will use stellar occultation as well, but also poses new challenges such as the use of a hyperspectral imager, and the absence of fast photometers helping for the scintillation removal.

For the nadir sensor line (with the weakest information content) improved cloud masking (better use of blue band and / or O2-A bands), improved surface reflectance / bi-directionality treatment, improved uncertainty characterisation and establishing consistency of a combined record from MERIS – OLCI is required.

Note that some of these gaps have been addressed within the context of or for C3S (e.g. IASI stability assessment, PODLER geographical coverage and its evaluation, IASI/MAPIR has updated its radiative transfer, IASI/IMARS has corrected a preprocessing bug to increase its coverage) or AEROSAT (assessment of multi-dataset consistency) during 2018.

Aerosol_cci2

Final Report


8 PUBLICATIONS

1) Bai, J., G. de Leeuw, R. van der A, I. De Smedt, N. Theys, M. Van Roozendael, L. Sogacheva and W. Chai (2018). Variations and photochemical transformations of atmospheric constituents in North China. Atmospheric Environment 189 (2018) 213–226; https://doi.org/10.1016/j.atmosenv.2018.07.004.

2) Bingen Christine, Charles E. Robert, Kerstin Stebel, Christoph Brühl, Jennifer Schallock, Filip Vanhellemont, Nina Mateshvili, Michael Höpfner, Thomas Trickl, John E. Barnes, Julien Jumelet, Jean-Paul Vernier, Thomas Popp, Gerrit de Leeuw, and Simon Pinnock, 2017, Stratospheric aerosol data records for the Climate Change Initiative: development, validation and application to Chemistry-Climate Modelling, Remote Sensing for Environment, http://dx.doi.org/10.1016/j.rse.2017.06.002

3) Brühl Christoph, Jennifer Schallock, Klaus Klingmüller, Charles Robert, Christine Bingen, Lieven Clarisse, Andreas Heckel, Peter North, and Landon Rieger: Stratospheric aerosol radiative forcing simulated by the chemistry climate model EMAC using Aerosol CCI satellite data, Atmos. Chem. Phys., 18, 12845–12857, 2018 https://doi.org/10.5194/acp-18-12845-2018.

4) Che, Y., Xue, Y., Mei, L., Guang, J., She, L., Guo, J., Hu, Y., Xu, H., He, X., Di, A., and Fan, C., 2016, Technical note: Intercomparison of three AATSR Level 2 (L2) AOD products over China, Atmos. Chem. Phys., 16, 9655-9674, doi:10.5194/acp-16-9655-2016

5) Chedin A., Capelle V., and Scott N.A. Detection of IASI dust AOD trends over Sahara: how many years of data required? Atmospheric Research, 212, 120–129, doi: 10.1016/j.atmosres.2018.05.004 (2018)

6) de Leeuw, G., Sogacheva, L., Rodriguez, E., Kourtidis, K., Georgoulias, A. K., Alexandri, G., Amiridis, V., Proestakis, E., Marinou, E., Xue, Y., and van der A, R.: Two decades of satellite observations of AOD over mainland China using ATSR-2, AATSR and MODIS/Terra: data set evaluation and large-scale patterns, Atmos. Chem. Phys., 18, 1573-1592, https://doi.org/10.5194/acp-18-1573-2018, 2018.

7) de Leeuw, G., T. Holzer-Popp, S. Bevan, W. Davies, J. Descloitres, R.G. Grainger, J. Griesfeller, A. Heckel, S. Kinne, L. Klüser, P. Kolmonen, P. Litvinov, D. Martynenko, P.J.R. North, B. Ovigneur, N. Pascal, C. Poulsen, D. Ramon, M. Schulz, R.Siddans, L. Sogacheva, D. Tanré, G.E. Thomas, T.H. Virtanen, W. von Hoyningen Huene, M.Vountas, S. Pinnock, 2015, Evaluation of seven European aerosol optical depth retrieval algorithms for climate analysis, Remote Sensing of Environment 162 (2015) 295–315. DOI: 10.1016/j.rse.2013.04.023

8) Kauppi Anu, Pekka Kolmonen, Marko Laine, Johanna Tamminen, 2017, Aerosol type retrieval and uncertainty quantification from OMI data, Atmospheric Measurement Techniques discussions, doi: 10.5194/amt-2017-47

9) Klüser Lars, Thomas Popp, 2017, Large-scale analysis of relationships between mineral dust, ice cloud properties and precipitation from satellite observations using a Bayesian approach: Theoretical basis and first results for the tropical Atlantic Ocean, Advances in Meteorology, Vol. 2017, Article ID 5278120, 18 pages, doi:10.1155/2017/5278120

10) Klüser, L., C. Di Biagio, P.D. Kleiber, P. Formenti, V.H. Grassian, 2016, Optical properties of non-spherical desert dust particles in the terrestrial infrared – An asymptotic

Aerosol_cci2

Final Report


approximation approach, Journal of Quantitative Spectroscopy & Radiative Transfer, 2016, 178, pp. 209-223, http://dx.doi.org/10.1016/j.jqsrt.2015.11.020

11) Kolmonen, P., L. Sogacheva, T. H. Virtanen, G. de Leeuw and M. Kulmala, 2016, The ADV/ASV AATSR aerosol retrieval algorithm: current status and presentation of a full-mission AOD data set, International Journal of Digital Earth, DOI 10.1080/17538947.2015.1111450

12) Kwinten Maes, Sophie Vandenbussche, Lars Klüser, Nicolas Kumps, Martine de Mazière, 2016, Vertical Profiling of Volcanic Ash from the 2011 Puyehue Cordón Caulle Eruption Using IASI, Remote Sensing 2016, 8(2), 103; doi: 10.3390/rs8020103

13) Lauer Axel, Veronika Eyring, Mattia Righi, Michael Buchwitz, Pierre Defourny, Martin Evaldsson, Pierre Friedlingstein, Richard de Jeu, Gerrit de Leeuw, Alexander Loew, Christopher J. Merchant, Benjamin Müller, Thomas Popp, Maximilian Reuter, Stein Sandven, Daniel Senftleben, Martin Stengel, Michel Van Roozendael, Sabrina Wenzel, Ulrika Willén, 2017, Benchmarking CMIP5 models with a subset of ESA CCI Phase 2 data using the ESMValTool, Remote Sensing of Environment, 191, http://dx.doi.org/10.1016/j.rse.2017.01.007, 2017

14) Mei, L. L., Rozanov, V., Vountas, M., Burrows, J., Levy, R., Lotz, W., Retrieval of aerosol optical properties using MERIS observations: algorithm and some first results, Remote Sensing of Environment, 2017, http://dx.doi.org/10.1016/j.rse.2016.11.015, 197, 125-140, 2017

15) Mei, L. L., Rozanov, V., Vountas, M., Burrows, J., Levy, R., Lotz, W., A Cloud masking algorithm for the XBAER aerosol retrieval using MERIS data , Remote Sensing of Environment,2017 http://dx.doi.org/10.1016/j.rse.2016.11.016,197, 141-160, 2017

16) Mei, L., Strandgren, J., Vountas, M., Burrows, J. P., Lyapustin, A., and Wang, Y.: Study of satellite retrieved aerosol optical depth spatial resolution effect on particulate matter concentration prediction, Interonational Journal of Remote sensing (accepted), 2018

17) Mei, L.L., Rozanov, V., Vountas, M., Burrows, J. P., Richter, A., XBAER-derived aerosol optical thickness from OLCI/Sentinel-3 observation, Atmos. Chem. Phys., 18, 2511-2523, doi:10.5194/acp-18-2511-2018, 2018.

18) Mei, L.L., Rozanov, V., Vountas, M., Burrows, J.P., The retrieval of ice cloud parameters from multi-spectral satellite observations of reflectance using a modified XBAER algorithm, Remote Sensing of Environmen, 215(15),128-144, 2018

19) Merchant Christopher J., Frank Paul, Thomas Popp, Michael Ablain, Sophie Bontemps, Pierre Defourny, Rainer Hollmann, Thomas Lavergne, Alexandra Laeng, Gerrit de Leeuw, Jonathan Mittaz, Caroline Poulsen, Adam C. Povey, Max Reuter, Shubha Sathyendranath, Stein Sandven, Viktoria F. Sofieva and Wolfgang Wagner, 2017, Uncertainty information in climate data records from Earth observation, Earth System Science Data, 9, 511–527, https://doi.org/10.5194/essd-9-511-2017

20) Merchant, C. J., G. de Leeuw and W. Wagner, 2015, Selecting algorithms for Earth observation of climate within the European Space Agency Climate Change Initiative: Introduction to a special issue, Remote Sensing of Environment 162 (2015) 239-241

21) Mielonen, T., Hienola, A., Kühn, T., Merikanto, J., Lipponen, A., Bergman, T., Korhonen, H., Kolmonen, P., Sogacheva, L., Ghent, D., Pitkänen, M.R.A., Arola, A., de Leeuw, G., and Kokkola, H. (2018). Summertime aerosol radiative effects and their dependence on

Aerosol_cci2

Final Report


temperature over the southeastern USA. Atmosphere 2018, 9, 180; doi:10.3390/atmos9050180.

22) Popp T., G. de Leeuw, C. Bingen, C. Brühl, V. Capelle, A. Chedin, L. Clarisse, O. Dubovik, R. Grainger, J. Griesfeller, A. Heckel, S. Kinne, L. Klüser, M. Kosmale, P. Kolmonen, L. Lelli, P. Litvinov, L. Mei, P. North, S. Pinnock, A. Povey, C. Robert, M. Schulz, L. Sogacheva, K. Stebel, D. Stein Zweers, G. Thomas, L.G. Tilstra, S. Vandenbussche, P. Veefkind, M. Vountas, and Y. Xue, 2016, Development, production and evaluation of aerosol Climate Data Records from European satellite observations (Aerosol_cci), Remote Sensing, 8, 421; doi:10.3390/rs8050421, 2016

23) Povey A. C., R. G. Grainger, 2015, Known and unknown unknowns: uncertainty estimation in satellite remote sensing, Atmos. Meas. Tech, 8, 4699–4718, doi:10.5194/amt-8-4699-2015

24) Robert C. E., C. Bingen, F. Vanhellemont, N. Mateshvili, E. Dekemper, C. Tétard, D. Fussen, A. Bourassa, and C. Zehner, 2016, AerGOM, an improved algorithm for stratospheric aerosol retrieval from GOMOS observations, Part 2: Intercomparisons, Atmospheric Measurement Techniques, 2016, 9, 4701-4718

25) Rodríguez A., E., Kolmonen, P., Virtanen, T. H., Sogacheva, L., Sundström, A.-M., and de Leeuw, G, 2015, Indirect estimation of absorption properties for fine aerosol particles using AATSR observations: a case study of wildfires in Russia in 2010, Atmos. Meas. Tech., 8, 3075-3085, doi:10.5194/amt-8-3075-2015

26) Rosenfeld, D., M. O. Andreae, A. Asmi, M. Chin, G. de Leeuw, D. Donovan, R. Kahn, S. Kinne, N. Kivekäs, M. Kulmala, W. Lau, S. Schmidt, T. Suni, T. Wagner, M. Wild, J. Quaas, 2014, Global observations of aerosol-cloud-precipitation-climate interactions, Rev. Geophys., 52, 750–808, doi:10.1002/2013RG000441

27) Sergey M. Khaykin , Sophie Godin-Beekmann , Philippe Keckhut , Alain Hauchecorne , Julien Jumelet , Jean-Paul Vernier , Adam Bourassa , Doug A. Degenstein, Landon A. Rieger , Christine Bingen , Filip Vanhellemont , Charles Robert , Matthew DeLand , and Pawan K. Bhartia, 2017, Variability and evolution of the midlatitude stratospheric aerosol budget from 22 years of ground-based lidar, and satellite observations, Atmos. Chem. Phys., 17, 1829–1845, 2017 doi:10.5194/acp-17-1829-2017

28) She Lu, Linlu Mei, Yong Xue, Yahui Che and Jie Guang, 2017, SAHARA: A Simplified AtmospHeric Correction AlgoRithm for Chinese gAofen Data: 1. Aerosol Algorithm, Remote Sens. 2017, 9(3), 253; doi:10.3390/rs9030253

29) Sogacheva, L., Kolmonen, P., Virtanen, T. H., Rodriguez, E., Sundström, A.-M. and de Leeuw, G, 2015,Determination of land surface reflectance using the AATSR dual-view capability,Atmos. Meas. Tech., 8, 891-906, doi:10.5194/amt-8-891-2015

30) Sundström, A.-M., Arola, A., Kolmonen, P., Xue, Y., de Leeuw, G., and Kulmala, M, 2015, On the use of a satellite remote-sensing-based approach for determining aerosol direct radiative effect over land: a case study over China, Atmos. Chem. Phys., 15, 505-518, doi:10.5194/acp-15-505-2015

31) Sundström, A.-M., Nikandrova, A., Atlaskina, K., Nieminen, T., Vakkari, V., Laakso, L., Beukes, J. P., Arola, A., van Zyl, P. G., Josipovic, M., Venter, A. D., Jaars, K., Pienaar, J. J., Piketh, S., Wiedensohler, A., Chiloane, E. K., de Leeuw, G., and Kulmala, M, 2015,

Aerosol_cci2

Final Report


Characterization of satellite-based proxies for estimating nucleation mode particles over South Africa, Atmos. Chem. Phys., 15, 4983-4996, doi:10.5194/acp-15-4983-2015

32) Vanhellemont F., N. Mateshvili, L. Blanot, C. E. Robert, C. Bingen, V., Sofieva, F. Dalaudier, C. Tétard, D. Fussen, E. Dekemper, E. Kyrölä, M. Laine, J. Tamminen, and C. Zehner, 2016, AerGOM, an improved algorithm for stratospheric aerosol extinction retrieval from GOMOS observations. Part 1: Algorithm description, Atmospheric Measurement Techniques, 2016, 9, 4687-4700

33) Xie Yanqing, Yong Xue, Yahui Che, Jie Guang, Linlu Mei, Dave Voorhis, Cheng Fan, Lu She, and Hui Xu, 2017, Ensemble of ESA/AATSR AOD products based on the likelihood estimate method with uncertainties, Transaction on Geoscience and Remote Sensing, 56(2) pp. 997-1007, 10.1109/TGRS.2017.2757910

34) Xu, H. , J. Guang, Y. Xue, G. de Leeuw, Y. H. Che, J. Guo and X. He, 2014, A consistent Aerosol Optical Depth (AOD) Dataset over China by Integration of Various AOD Products, Atm Env. 114, 48-56

35) Xue Yong, Xingwei He, Gerrit de Leeuw, Linlu Mei, Yahui Che, Wayne Rippin, Jie Guang, Yincui Hu, 2017, Long-time series ,aerosol optical depth retrieval from AVHRR data over land in North China and Central Europe, Remote Sensing of Environment, https://doi.org/10.1016/j.rse.2017.06.036

36) Zieger, P., Aalto, P. P., Aaltonen, V., Äijälä, M., Backman, J., Hong, J., Komppula, M., Krejci, R., Laborde, M., Lampilahti, J., de Leeuw, G., Pfüller, A., Rosati, B., Tesche, M., Tunved, P., Väänänen, R., and Petäjä, T., 2015, Low hygroscopic scattering enhancement of boreal aerosol and the implications for a columnar optical closure study, Atmos. Chem. Phys., 15, 7247-7267, doi:10.5194/acp-15-7247-2015

Aerosol_cci2

Final Report


Main project documents (available from ESA at http://cci.esa.int/aerosol) User Requirements Document, v3., 12.03.2017 Product Specification Document, v3.2, 12.03.2017 ATBD package, v4.2, 23.08.2017 Product User Guide, v2.3, 08.01.2018 Product Validation and Inter-comparison Report, v3.41, 17.12.2017 Bridging option Product Validation and Inter-comparison Report, v4.2, 09.01.2019 Climate Assessment Report, v2.7, 08.01.2018 Comprehensive Error Characterization Report, v3.2, 18.08.2017 Algorithm Development Report, v1.5, 13.12.2017 Bridging Option Algorithm Development Report, v2.0, 19.11.2018 Data Access Requirements Document, v3.3, 02.05.2017

Aerosol_cci2

Final Report


End of the document

Documents

ESA Climate Change Initiative Aerosol cci - Phase 2 2014 ...cci.esa.int/sites/default/files/Aerosol_cci2_FR_v1.2.pdf · Aerosol_cci2_FR_v1.2.pdf . Aerosol_cci2 Final Report REF :