Atmospheric aerosols have been known to be a major source of uncertainties in

Carbon dioxide is the most important greenhouse gas in our atmosphere. It accounts for 76 % of the total anthropogenic greenhouse gas emissions in 2010, according to the latest Assessment Report (2014) of the IPCC (Intergovernmental Panel on Climate Change). In an international effort to mitigate climate change, 195 countries signed the Paris Agreement (United Nations Framework Convention on Climate Change, 2015) which aims to limit global temperature rise to less than 2

The CO2M mission is designed as a constellation of up to three satellites with imaging capabilities, providing a global coverage with a revisit time of 5 d. Each satellite carries a primary sounder, that is, a nadir-looking spectrometer that will deliver measurements of column-averaged dry-air mole fraction of carbon dioxide

Scattering by aerosols and cirrus has long been identified as one of the main sources of uncertainties in retrieving

In this paper, we explore the potential of having auxiliary aerosol-dedicated measurements alongside the CO2M spectrometer measurements to help achieve the required

In the next section, we present the generic instrument description of the spectrometer and the two MAP instrument concepts used in this study. Section

For the CO2M mission, a three-band spectrometer is envisaged to be the main instrument that provides measurements necessary for the

Setup of the CO2M spectrometer.

In our study, we consider two MAP instrument concepts, i.e. MAP-mod and MAP-band. Here, the MAP-mod instrument is inherited from the SPEXone instrument

To enable

In this retrieval, the concept of inverse modelling applies, in which state vector

The forward model computes the Stokes vector, which describes the radiance and polarization state of light, at a certain wavelength and at a certain viewing angle for a specific atmospheric and geophysical scene. Degree of linear polarization (DLP) is then derived from the first three components of the vector

Stokes parameters are computed from the optical properties via the radiative transfer model based on the work of

The model atmosphere consists of 15 predefined height layers. Atmospheric vertical profiles of temperature,

The refractive index of each aerosol mode is defined by a linear combination of two aerosol types, such that the complex refractive index of a mode as a function of wavelength becomes

To account for the reflection and polarization properties of the surface, the retrieval algorithm employs semi-empirical bidirectional reflectance distribution function (BRDF) and bidirectional polarization distribution function (BPDF) models. The BRDF is characterized by the surface total reflectances

Although our main focus in this retrieval is

State variables in the joint retrieval.

Aerosol properties in the ensemble. The numbers in square brackets specify the interval from which a random value is drawn, whereas a single number indicates a fixed value.

The goal of the retrieval is to find

Due to the non-linearity of the forward model, the minimization problem is solved in an iterative manner. At every iteration step,

The linearization in Eq. (

The solution is found using an iterative Gauss–Newton method and expressed in terms of the departure

At each iteration step, we compute a fast and simplified forward model using a combination of five possible

Linear error analysis allows

In order to estimate

We compute for each scenario two Jacobian matrices. One of the Jacobians is associated to retrieval step 1 with a given MAP setup (

Errors on the retrieved aerosol properties from step 1 comprise the smoothing errors and the MAP-measurement-noise-induced error (retrieval noise). The smoothing error is formulated as

The matrix

Finally,

To perform

As in the joint retrieval, we retrieve the total column of

With only spectrometer data available, we resort to a simplified approximation of aerosols in the retrieval. In the forward model, aerosols are described by a simple model where the size distribution is parameterized by a monomodal power-law function. The power-law distribution is prescribed in

The reflection at the Earth surface is assumed to be Lambertian. Surface reflectance is included in the state vector via the albedo and its wavelength dependence in each window, which are modelled as a first-order polynomial. The prior for the albedo is the Lambertian-equivalent albedo corresponding to the maximum radiance measured in the retrieval window in question. The slope of the polynomial (wavelength dependence of the albedo) is given a prior of 0.0. Additionally, for each spectral window, a spectral shift parameter is retrieved with a prior of 0.0. In total, the state vector consists of 15 variables, i.e. the total columns of three trace gases (

We construct an ensemble of 500 synthetic scenes, characterized by different combinations of trace gas and aerosol content, surface albedo, and solar zenith angle (SZA). Every scene is generated by randomly varying those atmospheric and geophysical properties. The random value is drawn from a uniform distribution within a specific interval. Vertical profiles of pressure, temperature, water vapour, and trace gases are adopted from the AFGL atmospheric profiles

Given the spectral windows of the CO2M spectrometer, the radiance spectra include absorption features due to

Aerosols in every scene are constructed to consist of the fine and the coarse mode. The size distribution of each mode is quantified by a lognormal distribution (Eq.

Solar zenith angle is allowed to take any value between 10 and 70

The simulated spectra for the spectrometer-only retrieval (Sects.

Finally, we add random realizations of the instrument noise to the synthetic measurements. It is this noisy spectra that are given as input data for the retrievals. For the spectrometer, the noise follows the formulation in Sect.

Here, we present the results of RemoTeC iterative retrievals (Sect.

Out of 500 retrievals, 343 converge and meet the

Residual

The trends of

To minimize outlier effects on the statistics, we choose to evaluate the bias and the spread of the

As mentioned above, the accuracy and precision requirements of the CO2M mission are 0.5 and 0.7

Before we can assess the contribution of a MAP instrument in improving the

We evaluate the performance of MAP instrument setups with respect to three aspects, i.e. radiance and polarization measurement uncertainties, number of viewing angles, and the wavelength range. For this purpose, the linear error analysis is applied to a generic set of study scenarios involving a variety of aerosol and surface properties. Below, we define the study cases, followed by the requirement analysis for the MAP-mod concept, which results in the MAP-mod baseline setup. Afterwards, we present the baseline setup for the MAP-band concept that we determine through a separate error analysis similar to that for the MAP-mod.

We introduce three aerosol cases that form the basis of the scenarios used to derive the requirements. They are referred to as “case 1”, “case 2”, and “case 3”. In all cases, the aerosols are modelled according to the bimodal lognormal size distribution, Gaussian height distribution, and the linear superposition of complex refractive index (Eqs.

Aerosol properties adopted in the study cases.

We consider two types of land surface, i.e. soil and vegetation. These are the basic surface types used to create the 500 synthetic scenes (Sect.

To examine the sensitivity of

Figure

Figure

If we change the number of viewing zenith angles (VZAs), we effectively add or remove measurements, and this would certainly influence the aerosol and hence the

List of the viewing angles in the evaluated MAP-mod setups.

The resulting

An odd number of viewing angles is preferred over an even number to allow for symmetry and to include a nadir view. Strictly speaking, a minimum of seven viewing angles are required to have

Looking at heritage missions with different spectral coverage, we assess the effect of varying spectral range on the retrieved

List of the four options for the MAP-mod spectral range.

Figure

Following the assessment above, we adopt the default setup as the MAP-mod baseline setup. For clarity, we summarize this setup in Table

MAP-mod baseline setup.

For the MAP-band instrument, we consider six spectral bands from 410 to 865

Figure

MAP-band baseline setup.

The linear error analysis provides reliable

We apply the same

Residual

Same as Fig.

Figure

The key outcome of this exercise is the stark contrast in the

Compared to the statistics of the spectrometer-only retrievals, the joint retrieval results obviously represent a major improvement in the accuracy and precision of the retrieved

Deployment of a MAP instrument would additionally offer better insight into the surface reflection properties, which are important factors in simulating the radiation at the top of atmosphere, especially for retrievals over land. In Sect.

It should be noted that besides the aerosol-induced errors studied here, there are also other error sources that affect the final performance, most notably due to imperfect spectroscopy

In the context of ESA's CO2M mission, we investigated the need for an aerosol-dedicated instrument (multi-angle polarimeter or MAP) in support of the CO2M spectrometer to achieve the required

In the ensemble of synthetic scenes, aerosol size distribution is described by a bimodal lognormal function, where each mode follows a Gaussian height distribution. The trace gas total column, aerosol and surface properties, and the solar zenith angle are randomly varied within certain limits to generate 500 atmospheric and geophysical scenes.

For the standard retrieval exercise using only spectrometer data, we employed the RemoTeC algorithm that has been widely used for greenhouse gas retrievals from space. In RemoTeC, a simple aerosol model is used; i.e. aerosol size distribution is retrieved following a monomodal power-law parametrization. Out of 500 retrievals, 69 % meet our

Boxplot of

Prior to performing the joint retrieval, we conducted a requirement analysis to construct a baseline setup for each of the two alternative MAP concepts being considered for the CO2M mission, i.e. MAP-mod and MAP-band. The MAP-mod concept is based on a spectral modulation technique where polarization information is encoded in the modulation pattern of the radiance spectrum, while the MAP-band instrument acquires radiance and polarization measurements at specific discrete spectral bands. The MAP-mod instrument is inherited from SPEXone and the MAP-band wavelength channels are inherited from 3MI polarized VNIR bands. The optimal baseline setups for the MAP-mod and for the MAP-band instrument designs are found through a linear error analysis that is formulated to mimic a joint retrieval. In particular, we investigated three aspects of a MAP instruments, i.e. the measurement uncertainties, number of viewing angles, and wavelength range. For the MAP-mod concept, the baseline setup includes five viewing angles (

To implement the joint retrieval, we further developed an existing aerosol retrieval algorithm to include features related to the spectrometer measurements and to the derivation of trace gas total columns. With this tool, and using the combined spectrometer and MAP-mod (MAP-band) measurements, 70 % (78 %) of the 500 retrievals reach convergence according to our

The results of the joint retrieval (for either of the two MAP concepts) represent a significant improvement in the retrieved

The lognormal distribution is used in this paper to describe the size distribution

In all of our retrieval exercises, aerosol properties are fitted alongside

Aerosol optical thickness from spectrometer-only

We begin by comparing aerosol optical depth estimates from the three retrieval approaches, i.e. spectrometer-only, MAP-only, and the joint MAP-(mod)-spectrometer setups. Because spectrometer-only retrieval relies on a simpler aerosol parameterization, the total

To help us gain better insight on the overall retrieval system, Figs.

Aerosol properties retrieved using MAP-only measurements compared to the truth. In each panel, the letter “c” or “f” in parentheses behind the aerosol variable name indicates the aerosol mode, i.e. coarse or fine. The RMSE for each parameter is given in the bottom right corner.

Same as Fig.

The synthetic test dataset is available from the authors upon request.

JL and OH designed the research experiment. SPR developed the joint retrieval tool and performed the retrieval exercises. The requirement study and interpretation of the results were carried out by SPR, OH, and JL. JadB provided assistance with the linear error analysis and technical details of the spectrometer-only retrieval method. GF contributed to the simulation of test data and aspects related to the MAP retrieval. YM provided input to the experiment in the context of MAP deployment for the CO2M mission. SPR wrote the manuscript with feedback from all co-authors.

The authors declare that they have no conflict of interest.

The authors thank Robert Roland Nelson, Annmarie Eldering, the anonymous reviewer, and the editor for their comments and suggestions that helped improve the paper. Stephanie P. Rusli is grateful to Haili Hu for the valuable discussions about RemoTeC implementation.

This paper was edited by Christof Janssen and reviewed by Robert Roland Nelson, Annmarie Eldering, and one anonymous referee.