Atmosphere

The Advanced Infra-Red Water Vapour Estimator, or AIRWAVE, is a novel algorithm developed by CNR and ESA for the accurate and precise estimate of ocean/cloud-free Total Column Water Vapour (TCWV). The algorithm is based on the exploitation of Thermal Infra-Red channels of the ATSR-like instruments, such as SLSTR on board Sentinel 3, in combination with advanced radiative transfer models and a sea surface spectral emissivity database. In particular, the simultaneous use of forward and nadir measurements minimise the impact of changes in sea surface temperature and in atmospheric radiation on the quality of retrieved TCWV for all illumination conditions (day and night).

The AIRWAVE-PP prototype processor has been designed and integrated in the ESA GRID environment (GPOD) for testing purposes, and for the bulk processing of the ATSR missions (1991-2012). It is written in Python 2.7, and based on standard Python libraries, such as Numpy and Matplotlib. The BEAT-CODA software is used for the decoding of ESA Level 1 data format.

The tool’s logical operation flow can be summarised as follows: 1) ATSR Level 1 product decoding and pre-processing; 2) sea and cloud-free data screening using L1 product flags; 3) brightness temperature to radiance conversion; 4) TCWV retrieval at 1x1 km² and 0.25°x0.25° spatial resolution; 5) exporting of TCWV results into NetCDF-CF (Climate Forecast) file products. The NetCDF-CF convention has been adopted in order to facilitate the readability of dataset by means of freely available software. For each ATSR orbit file, AIRWAVE-PP produces two outputs: a) NetCDF-CF file containing the TCWV data; b) a PNG file (image) showing the TCWV vs. latitude plot (quick-look), the latter used for a fast check on the quality of the retrieval.

The AIRWAVE-PP implementation will be discussed in detail, along with the preliminary results of the ATSR vs. SSM/I TCWV inter-comparison exercise.

Within the ESA Climate Change Initiative (CCI) project Aerosol_cci (Phase 1: 2010 –2014; Phase 2: 2014-2017) intensive work has been conducted to improve algorithms for the retrieval  of aerosol information from European sensors ATSR-2 (ERS-2), AATSR (3 algorithms), MERIS (3 algorithms), synergetic AATSR/SCIAMACHY, GOMOS (all on ENVISAT), PARASOL and OMI (EOS-Aura) (both part of NASA’s A-Train). Whereas OMI and GOMOS were used to derive absorbing aerosol index and stratospheric extinction profiles, respectively, Aerosol Optical Depth (AOD) and Ångström coefficient were retrieved from the other sensors. The cooperation between the project partners, including both the retrieval teams and independent validation teams, has resulted in a strong improvement of most algorithms. In particular the AATSR retrieved AOD is qualitatively similar to that from MODIS, usually taken as the standard, MISR and SeaWiFS. This conclusion has been reached based on the evaluation of results from several different ways of validation of the L2 and L3 products, using AERONET sun photometer data as the common ground-truth. Both ‘traditional’ statistical techniques and a ‘scoring’ technique based on spatial and temporal correlations were applied. Quantitatively, the limited AATSR swath width of 512km results in a smaller amount of data. Nevertheless, the assimilation of AATSR-retrieved AOD, together with MODIS data, contributes to improving ECMWF / MACC climate model results. In addition to the multi-spectral AOD, and thus the Ångström Exponent, also a per-pixel uncertainty is provided and validated. By the end of Aerosol_cci Phase 1 the ATSR algorithms have been applied to both ATSR-2 and AATSR resulting in an AOD time series of 17 years dating back to 1995.

In phase 2 this work is continued with a focus on the further improvement of both the ATSR algorithms and those for the other instruments, mentioned above, which in phase 1 were considered less mature. The first efforts are on the further characterization of the uncertainties and on better understanding of the cloud screening in the various algorithms. Other efforts will focus on surface treatment and possible improvement of aerosol models used in the retrieval. A yearly re-processing of the full 17-year global ATSR-2/AATSR data set is planned to evaluate the effect of different changes and to monitor further improvement.

The successful launch of Sentinel-3, planned for the autumn of 2015, will mean an important step to further extend the 17-years ATSR time series with SLSTR data products. The SLSTR instrument is very similar to AATSR with some important differences such as the forward view instead of backward and extra wavebands which can be beneficial for aerosol retrieval. In addition to a wider swath in both viewing directions, also the launch of the second SLSTR instrument will largely increase data coverage with respect to AATSR. This will render the data set much more attractive for modeling applications such as data assimilation.

In addition to SLSTR, also OLCI on Sentinel-3, similar to MERIS, has great promise for aerosol retrieval. One aerosol retrieval algorithm for MERIS (ALAMO), developed as part of Aerosol_cci for application over ocean only, has been judged mature enough to contribute to the climate data record. For application over land, currently no MERIS algorithm is mature enough but efforts are planned to develop it. This will provide potential to provide a MERIS/OLCI time series starting with ENVISAT and continued with Sentinel-3 data.

An important aspect of Aerosol_cci Phase 2 are use cases where representatives of users communities, climate, stratospheric aerosol and aerosol-cloud interaction, will evaluate the use of Aerosol_cci products in their own work as regards the usefulness and added value. This will be done in close cooperation with the data providers to further improve the products and meet users’ needs, both as regards data quality and presentation. The latter also requires data availability and easy accessibility through good data management which is another important aspect in Aerosol_cci. Clearly, newly developed Sentinel-3 aerosol products will be offered to these users for scientific application.

In the S34Science Conference an overview will be presented of the latest developments in the Aerosol_cci project, including time series over different areas and the preparation for the transition from ENVISAT to Sentinel-3.

A spatially high resolved total column water vapour (TCWV) dataset over land surfaces with uncertainty estimates for each pixel has been derived for the period January 2003 to March 2012 in frame of the ESA DUE GlobVapour project [Lindstrot et al., 2014]. It is based on near-infrared measurements of Medium Resolution Imaging Spectrometer (MERIS) onboard ESA’s Environmental Satellite (ENVISAT). The dataset was already validated against radiosonde, ground-based microwave and GPS derived TCWV [Lindstrot et al., 2012]. Now, it is statistically analysed in order to identify trends within the previously mentioned time frame. Therefore, monthly composites on a rectangular longitude-latitude
grid with a spatial resolution of 0.5° x 0.5° have been used in a multiple linear regression, which also
takes a possible lagged El-Nino – Southern Oscillation (ENSO) influence into account. Subsequent hypothesis testing delivered spatially high resolved significant positive and negative regional trend patterns over land surfaces, such as an increase in water vapour for the east coast of Australia and a decrease for parts of Mexico. The purpose of the presentation will be the description of the methodology and the illustration of MERIS TCWV trends.

References:

Lindstrot, R., Preusker, R., Diedrich, H., Doppler, L., Bennartz, R. and Fischer, J., 1D-Var retrieval
of daytime total columnar water vapour from MERIS measurements, Atmos. Meas. Tech., 5, 631-
646, doi:10.5194/amt-5-631-2012, 2012.

Lindstrot, R., Stengel, M., Schröder, M., Fischer, J., Preusker, R., Schneider, N., Steenbergen, T.,
and Bojkov, B.R., A global climatology of total columnar water vapour from SSM/I and MERIS,
Earth Syst. Sci. Data, 6, 221-233, doi:10.5194/essd-6-221-2014, 2014.

The AATSR Dual View (ADV) aerosol retrieval algorithm used at FMI is based on the stereo-viewing capability of AATSR. In the k-ratio approach the radiances measured in nadir andforward views are used to eliminate the surface reflectance contribution from the top of atmosphere (TOA) reflectance, allowing us to retrieve the aerosol contribution without any prior knowledge of the surface properties. We are preparing to use the dual-view algorithm for aerosol retrieval with the SLSTR nadir and backward view data. In this paper we consider the effects the changing viewing geometry may have on the aerosol retrieval.

The k-ratio is the ratio of the forward and nadir view surface reflectance. In the ADV algorithm two assumptions are made on the k-ratio: (1) the k-ratio is assumed to be independent of wavelength, and (2) the aerosol contribution to TOA reflectance is negligible at 1.6 µm (in first approximation). The k-ratio is then obtained from the TOA reflectance at the AATSR wavelength of 1.61 µm, and the same ratio is used for 555 nm and 659 nm. For 865 nm assumption (1) does not hold, and retrieval is not attempted at this wavelength. It is also possible to iterate the k-ratio (at 1.61 µm) to better account for the aerosol contribution during the retrieval.

The oblique viewing angle changes form a 55ºforward view in AATSR to a 55ºbackward view in SLSTR. This affects the TOA reflectance in two ways: (i) the aerosol signal may change due to changing scattering angles; (ii) the surface reflectance (and thus the k-ratio) will change. In principle the algorithm can handle all reasonable viewing geometries (for solar zenith angles up to 75º). However, there is some evidence of anomalous latitude dependence in the existing ADV aerosol retrievals. We suspect that this anomaly is related to the k-ratio and scattering angle, which are latitude dependent. This anomaly may have unexpected effects on the retrievals with SLSTR, which will have different latitude dependence for the k-ratio and scattering angles.

To study the possible effects of the changing viewing geometry prior to the launch of Sentinel-3, we use MISR data. MISR has nine viewing angles, one in nadir, and four facing forward and backward directions. Even though MISR lacks the 1.6 µm NIR channel, two of its wavebands (558 nm and 672 nm) are close to the AATSR channels used in ADV, and can be used to study the effect of the viewing direction, at a 60.0º viewing angle. A customized two-wavelength version of ADV is run with near-simultaneous data from AATSR and MISR. First, the results of retrieval using the MISR nadir and 60.0º forward view cameras are compared to AATSR results. Next, a comparison is made with MISR retrievals using the 60.0º backward view camera.

The results show that the 1.6 µm channel is essential for the ADV algorithm, and the two-wavelength version cannot be used for reliable retrieval of aerosol properties. However, the preliminary studies show that there are significant differences in performance of ADV between different MISR viewing directions. Further work on the subject is ongoing.

In order to fill the gap in the total column water vapour (TCWV) time-line between the Medium Resolution Imaging Spectrometer (MERIS) on ENVISAT and its follow-on Ocean and Land Color imager (OLCI) on Sentinel-3, a new TCWV retrieval for the Moderate-resolution Imaging Spectroradiometer (MODIS) on Terra and Aqua has been established. The procedure uses radiance measurements in 5 channels in the near-infrared (NIR) around 900nm and has been adapted from the latest MERIS retrieval developed in the ESA-DUE-Globvapour project. It provides TCWV and realistic error estimates on a pixel-by-pixel basis with a spatial resolution of 1x1km² over cloud-free land surfaces and is ready to be upgraded for ocean surfaces. Because of its universal architecture, the adaption for the OLCI channel setup is feasible.

The presentaion will show the general funtionality of the MODIS retrieval and will focus on validation results against several ground-based measurements (ARM-Microwave Radiometer, GNSS-water vapour stations, GUAN Radiosondes, and AERONET sun photometers) and a comparison to TCWV from MERIS. Additionally, the reduction of uncertainties of the future OLCI retrieval due to the new 940 nm channel, especially over dark ocean surfaces, will be demonstrated.

Knowledge of the distribution and variability of the chemical composition of atmospheric aerosol is an important element when attempting to assess the aerosol direct and indirect radiative effects. As for all the aerosol physical and chemical properties, in situ observations are often too sparse and not continuous, thus a view from space is an attractive way to complement the information gained from the ground. Several works proposed methods to infer information on the aerosol composition from satellite using observations from the ultra-violet to the infrared. The results are usually presented in terms of a group of possible “aerosol types” or “aerosol models” (e.g. dust-like, continental, maritime, soot-containing, etc.) that may be identified by the procedure. The selection of one such aerosol model is also often implied in the retrieval algorithm of related quantities such as the aerosol optical depth, and may induce significant uncertainty on the results.

Here we explore the possibility of identifying a different set of aerosol classes, more closely comparable to in situ measurements and model simulations. The classes are inorganic ions (such as sulfate, ammonium nitrate), organic matter, black/brown carbon, soil dust and sea salt. To this end we produced a simulated set of paired aerosol scenes and reflectances at top-of -atmosphere, in order to train a neural network to recognize the aerosol classes from the reflectances. The aerosol fields are taken sampling randomly the 1-year simulation of a global chemistry-transport model (GEOS-Chem), thus representing many possible aerosol mixtures occurring over the globe. Only scenes over the oceans are selected in this first experiment to enhance contrast with the surface. The optical properties are calculated from simulated aerosol fields under different alternative assumptions on physical-chemical properties of the components (e.g. refractive indices, density, hygroscopic properties) and mixing state (external or internal mixtures). The reflectances are then simulated using a radiative transfer model (libRadtran) at 1 nm spectral resolution in the range 350-1000 nm, assuming observation geometry typical of a polar orbiting satellite (such as MODIS, Hyperion, Sentinel-3, etc.). We discuss the potential for hyper- and multi-spectral satellite observations, including those that can be provided by Sentinel-3,  to constraint the relative abundance of the selected aerosol classes. We also discuss potential limitations to the identification procedure dependent on the observation characteristics.

Surface reflectance has a central role in the analysis of land surface for a broad variety of Earth System studies. The appropriate selection of aerosol type and loading is one key requirement for a reliable surface reflectance estimation from satellite observations, a process often called atmospheric correction. The aerosol type is defined  by  its physical and chemical properties, while the aerosol loading is usually described by the optical thickness at 550 nm. The aim of this work is to evaluate the radiative impact of the aerosol type on the surface reflectance obtained from CHRIS (Compact High Resolution Imaging Spectrometer) hyperspectral data over land. The physically-based algorithm CHRIS@CRI (CHRIS Atmospherically Corrected Reflectance Imagery) was developed specifically for CHRIS data by using the vector version of 6S (6SV) radiative transfer model.  The aerosol impact on the atmospheric correction was investigated comparing the reflectance alternatively using different aerosol types, all else being the same; we used the AERONET inversion products, which provides estimated of the column-average aerosol microphysical properties, as the reference, three 6SV aerosol standard types and one calculated from the output of a comprehensive global chemistry-transport model (GEOS-Chem).

The general quality of the CHRIS@CRI algorithm was benchmarked against results obtained with the standard BEAM software on several images over Brussels (Belgium) and Cabauw (The Netherlands); we found differences in the spectral surface reflectance of less than 10%, which leans confidence in the use of the algorithm. We then intercompared the surface reflectance obtained with aerosol input alternatively from AERONET or from one “aerosol model” as mentioned above. While the choice of the observation-based AERONET input seem to be the most reliable, we found that, in the absence of AERONET data on the scene, the best performing model is the detailed one, because it better reproduces the characteristics, specific in space and time, of the aerosol and the radiative field on the particular scene. Our work thus suggests that the use of output from chemistry-transport models, which is now becoming an operationally available quality product, may be fruitfully exploited in satellite retrieval algorithms.

Simultaneous retrieval of aerosol and marine parameters using inverse techniques based on a coupled atmosphere-ocean radiative transfer model (CRTM) and optimal estimation can yield considerably improved retrieval accuracy based on radiances measured by MERIS, MODIS, and future instruments like OLCI compared with traditional methods. As an example, we discuss simultaneous retrieval in a Norwegian coastal environment from MERIS data using a one-step nonlinear optimal estimation method instead of the traditional two-step look up table approach. To increase retrieval speed without loss of accuracy we replace the forward CRTM by a radial basis function neural network. Five parameters are obtained from the retrieval: aerosol optical depth, aerosol bi-modal fraction, chlorophyll concentration, absorption by colored dissolved organic matter, and backscattering coefficient. The water leaving radiance is provided as a by-product. We demonstrate the accuracy of this simultaneous retrieval approach through a comparison with match ups from a Norwegian coastal area.

The scenario of data availability for the upcoming years will confront the scientific and industrial communities with a strong increase of satellite missions and related data. This is in-particular the case for the Atmospheric sciences communities, with the upcoming Copernicus Sentinels and scientific satellites and their integration with ground based observations and models, and poses the basis for a new data exploitation paradigm opening new research and commercial opportunities.

As a preparation activity supporting scientific data exploitation for Earth Explorer and Sentinel atmospheric mission, ESA is funding the technology study and prototype implementation of the “Technology and Atmospheric Mission Platform” (TAMP)

The TAMP test-bed environment will allow handling visualization and analysis purposes of the “data triangle” consisting of EO satellite products (available within the system, generated inside the system, and user provided), model data provided by chemical weather forecast models (e.g. hourly 3D distribution of PM10, O3, NO2 pollutants at European and regional scales), reference / validation data sets (as standard validation datasets e.g. AERONET, EARLINET, NDAAC or Pandonia).

Services offered by the system will comprise data access, viewing, processing and analysis services. They will be developed along use cases defined with users from different scientific and operational fields and implemented according to their requirements to ensure acceptance of TAMP by the atmospheric community. These services and tools are provided following the “virtual workspace” concept, meaning that source data are becoming available on the virtual platform. All resources (processing, data, collaboration tools) are provided as “remote services”, accessible through a standard web browser, to avoid the download of big data volumes and for allowing utilization of provided infrastructure for computation.

The present work aims at presenting the TAMP concept to the Sentinel-3 scientific community, introducing the platform and stimulating interest and discussions on the need of a thematic platform for atmospheric sciences to better exploit the upcoming missions data.

Natural aerosols are a key component of the Earth’s environmental system. Dust aerosols interact directly with both shortwave and longwave radiation, influencing the flux and transport of heat across the atmosphere-ocean boundary. While quantifying the impact of dust aerosols on the Earth's climate remains a challenge, their effect is potentially significant for the circulation and heat budget of regional seas over different time scales. We investigate the impact of summertime dust aerosols on the Red Sea using a regional climate model. A monthly climatology of direct radiative forcing derived from 5 years of SEVIRI retrievals is used as a proxy for atmospheric dust. Aerosol optical depth estimates show enhanced aerosol loading and the development of a north to south gradient in the summer relative to the winter months over the region. Shortwave cooling typically dominates the net radiative effect over the basin and is particularly pronounced in the summer, exceeding 130 Wm^-2 at the surface. We use these dust aerosol effects to perturb the radiative forcing in a regional ocean model and find that the Red Sea is highly sensitive to dust induced reductions in radiative fluxes. A summertime sea surface temperature (SST) decrease of -1.1^oC in the northern Red Sea is associated with a net radiation reduction of -40Wm^-2 while a similar summertime SST decrease of -2^oC in the southern Red Sea is associated with a net radiation reduction of -90Wm^-2. The summer mixed layer temperature decreases by -0.4^oC (north) and -1.7^oC (south), respectively. Larger cooling occurs below the mixed layer, at 75m, in autumn, -1.2^oC (north) and – 1.9^oC (south). Surface cooling relaxes more rapidly than subsurface cooling, especially in the south, due to feedbacks from surface longwave, latent and sensible heat fluxes compensating for the cooler surface. Preferential subsurface cooling, especially in autumn, leads to a stronger gradient in stratification, reduced mixed layer depth and increased southward heat transport in the south. This is accompanied by an intensification of anticyclonic eddies and of the southern coastal currents. Overall, the average annual total heat flux over the basin changes from a loss of heat (-17.3Wm^-2) to a small gain of heat (+4.2Wm^-2) in response to summertime dust perturbations. Our study supports the notion that dust aerosols play an important role in regulating the surface and subsurface temperature of the Red Sea, and provides the first quantitative estimate of this effect.