Search

We studied the potential of using machine learning to downscale global-scale climate model output towards ground station data. The aim was to simultaneously analyze both city-level air quality and regional- and global-scale radiative forcing values for anthropogenic aerosols. As the city-level air pollution values are typically underestimated in global-scale models, we used a machine learning approach to downscale fine particulate (PM2.5) concentrations towards measured values. We first simulated the global climate with the aerosol–climate model ECHAM-HAMMOZ and corrected the PM2.5 values for the Indian megacity New Delhi. The downscaling procedure clearly improved the seasonal variation in the model data. The seasonal trends were much better captured in the corrected PM2.5 than in original ECHAM-HAMMOZ PM2.5 when compared to the reference PM2.5 from the ground stations. However, short-term variations showed less extreme values with the downscaling approach. We applied the downscaling model also to simulations where the aerosol emissions were following two different future scenarios: one following the current legislation and one assuming currently maximum feasible emission reductions. The corrected PM2.5 concentrations for the year 2030 showed that mitigating anthropogenic aerosols improves local air quality in New Delhi, with organic carbon reductions contributing most to these improvements. In addition, aerosol emission mitigation also resulted in negative radiative forcing values over most of India. This was mainly due to reductions in absorbing black carbon emissions. For the two future emission scenarios modeled, the radiative forcing due to aerosol–radiation interactions over India was -0.09±0.26 and -0.53±0.31 W m−2, respectively, while the effective radiative forcing values were -2.1±4.6 and 0.06±3.39 W m−2, respectively. Although accompanied by relatively large uncertainties, the obtained results indicate that aerosol mitigation could bring a double benefit in India: better air quality and decreased warming of the local climate. Our results demonstrate that downscaling and bias correction allow more versatile utilization of global-scale climate models. With the help of downscaling, global climate models can be used in applications where one aims to analyze both global and regional effects of policies related to mitigating anthropogenic emissions.

Here we present for the first time a proof of concept for an emulation-based method that uses a large-eddy simulations (LES) to present sub-grid cloud processes in a general circulation model (GCM). We focus on two key variables affecting the properties of shallow marine clouds: updraft velocity and precipitation formation. The LES is able to describe these processes with high resolution accounting for the realistic variability in cloud properties. We show that the selected emulation method is 5 able to represent the LES outcome with relatively good accuracy and that the updraft velocity and precipitation emulators can be coupled with the GCM practically without increasing the computational costs. We also show that the emulators influence the climate simulated by the GCM, but do not consistently improve or worsen the agreement with observations on cloud related properties. Although especially the updraft velocity at cloud base is better captured. A more quantitative evaluation of the emulator impacts against observations would, however, have required model re-tuning, which is a significant task and thus could 10 not be included in this proof-of-concept study. All in all, the approach introduced here is a promising candidate for representing detailed cloud and aerosol related sub-grid processes in GCMs. Further development work together with increasing computing capacity can be expected to improve the accuracy and the applicability of the approach in climate simulations.

A quantitative comparison study for Raman lidar and ceilometer observations, and for model simulations of mass concentration estimates of smoke particles is presented. Layers of biomass burning aerosol particles were observed in the lower troposphere, at 2 to 5 km height on 4 to 6 June 2019, over Kuopio, Finland. These long-range-transported smoke particles originated from a Canadian wildfire event. The most pronounced smoke plume detected on 5 June was intensively investigated. Optical properties were retrieved from the multi-wavelength Raman polarization lidar PollyXT. Particle linear depolarization ratios (PDRs) of this plume were measured to be 0.08±0.02 at 355 nm and 0.05±0.01 at 532 nm, suggesting the presence of partly coated soot particles or particles that have mixed with a small amount of dust or other non-spherical aerosol type. The layer-mean PDR at 355 nm (532 nm) decreased during the day from ∼ 0.11 (0.06) in the morning to ∼ 0.05 (0.04) in the evening; this decrease with time could be linked to the particle aging and related changes in the smoke particle shape properties. Lidar ratios were derived as 47±5 sr at 355 nm and 71±5 sr at 532 nm. A complete ceilometer data processing for a Vaisala CL51 ceilometer is presented from a sensor-provided attenuated backscatter coefficient to particle mass concentration (including the water vapor correction for high latitude for the first time). Aerosol backscatter coefficients (BSCs) were measured at four wavelengths (355, 532, 1064 nm from PollyXT and 910 nm from CL51). Two methods, based on a combined lidar and sun-photometer approach, are applied for mass concentration estimations from both PollyXT and the ceilometer CL51 observations. In the first method, no. 1, we used converted BSCs at 532 nm (from measured BSCs) by corresponding measured backscatter-related Ångström exponents, whereas in the second method, no. 2, we used measured BSCs at each wavelength independently. A difference of ∼ 12 % or ∼ 36 % was found between PollyXT and CL51 estimated mass concentrations using method no. 1 or no. 2, showing the potential of mass concentration estimates from a ceilometer. Ceilometer estimations have an uncertainty of ∼ 50 % in the mass retrieval, but the potential of the data lies in the great spatial coverage of these instruments. The mass retrievals were compared with the Modern-Era Retrospective analysis for Research and Applications, version 2 (MERRA-2) meteorological and aerosol reanalysis. The inclusion of dust (as indicated by MERRA-2 data) in the retrieved mass concentration is negligible considering the uncertainties, which also shows that ceilometer observations for mass retrievals can be used even without exact knowledge of the composition of the smoke-dominated aerosol plume in the troposphere.

Satellite estimates of surface UV irradiance have been available since 1978 from the TOMS UV spectrometer and have continued with significantly improved ground resolution using the Ozone Monitoring Instrument (OMI 2004–current) and Sentinel 5 Precursor (S5P 2017–current). The surface UV retrieval algorithm remains essentially the same: it first estimates the clear-sky UV irradiance based on measured ozone and then accounts for the attenuation by clouds and aerosols, applying two consecutive correction factors. When estimating the total aerosol effect in surface UV irradiance, there are two major classes of aerosols to be considered: (1) aerosols that only scatter UV radiation and (2) aerosols that both scatter and absorb UV radiation. The former effect is implicitly included in the measured effective Lambertian-equivalent scene reflectivity (LER), so the scattering aerosol influence is estimated through cloud correction factor. Aerosols that absorb UV radiation attenuate the surface UV radiation more strongly than non-absorbing aerosols of the same extinction optical depth. Moreover, since these aerosols also attenuate the outgoing satellite-measured radiance, the cloud correction factor that treats these aerosols as purely scattering underestimates their aerosol optical depth (AOD), causing underestimation of LER and overestimation of surface UV irradiance. Therefore, for correction of aerosol absorption, additional information is needed, such as a model-based monthly climatology of aerosol absorption optical depth (AAOD). A correction for absorbing aerosols was proposed almost a decade ago and later implemented in the operational OMI and TROPOMI UV algorithms. In this study, however, we show that there is still room for improvement to better account for the solar zenith angle (SZA) dependence and nonlinearity in the absorbing aerosol attenuation, and as a result we propose an improved correction scheme. There are two main differences between the new proposed correction and the one that is currently operational in OMI and TROPOMI UV algorithms. First, the currently operational correction for absorbing aerosols is a function of AAOD only, while the new correction additionally takes the solar zenith angle dependence into account. Second, the second-order polynomial of the new correction takes the nonlinearity in the correction as a function of AAOD better into account, if compared to the currently operational one, and thus better describes the effect by absorbing aerosols over a larger range of AAOD. To illustrate the potential impact of the new correction in the global UV estimates, we applied the current and new proposed correction for global fields of AAOD from the aerosol climatology currently used in OMI UV algorithm, showing a typical differences of ±5 %. This new correction is easy to implement operationally using information of solar zenith angle and existing AAOD climatology.

Satellite-based aerosol retrievals provide a timely view of atmospheric aerosol properties, having a crucial role in the subsequent estimation of air quality indicators, atmospherically corrected satellite data products, and climate applications. However, current aerosol data products based on satellite data often have relatively large biases compared to accurate ground-based measurements and distinct uncertainty levels associated with them. These biases and uncertainties are often caused by oversimplified assumptions and approximations used in the retrieval algorithms due to unknown surface reflectance or fixed aerosol models. Moreover, the retrieval algorithms do not usually take advantage of all the possible observational data collected by the satellite instruments and may, for example, leave some spectral bands unused. The improvement and the re-processing of the past and current operational satellite data retrieval algorithms would become tedious and computationally expensive. To overcome this burden, we have developed a model-enforced post-process correction approach to correct the existing operational satellite aerosol data products. Our approach combines the existing satellite aerosol retrievals and a post-processing step carried out with a machine-learning-based correction model for the approximation error in the retrieval. The developed approach allows for the utilization of auxiliary data sources, such as meteorological information, or additional observations such as spectral bands unused by the original retrieval algorithm. The post-process correction model can learn to correct for the biases and uncertainties in the original retrieval algorithms. As the correction is carried out as a post-processing step, it allows for computationally efficient re-processing of existing satellite aerosol datasets without fully re-processing the much larger original radiance data. We demonstrate with over-land aerosol optical depth (AOD) and Ångström exponent (AE) data from the Moderate Imaging Spectroradiometer (MODIS) of the Aqua satellite that our approach can significantly improve the accuracy of the satellite aerosol data products and reduce the associated uncertainties. For instance, in our evaluation, the number of AOD samples within the MODIS Dark Target expected error envelope increased from 63 % to 85 % when the post-process correction was applied. In addition to method description and accuracy results, we also give recommendations for validating machine-learning-based satellite data products.

In this study, we present our findings for correcting global model-derived surface fine particulate mass (PM2.5) concentrations with a machine learning approach. We simulated the PM2.5 concentrations with an aerosol-climate model ECHAM-HAMMOZ, and trained a machine learning model to downscale the PM2.5 concentrations modelled for an Indian mega city, New Delhi. This way, we are able to utilize a global atmospheric model for analyzing aerosol emission mitigation effects on both Earth's energy budget and local air quality.Similarly as with many other global-scale models, ECHAM-HAMMOZ underestimates surface PM2.5 at several urban locations. One apparent explanation for this is the coarse grid resolution of global climate models, which results in averaged aerosol concentrations over a much larger area than what urban cities typically cover. Therefore, due to averaging over a large grid box, the very high peak concentrations from urban areas can be evened out. Furthermore, the input fields describing aerosol emissions might lack information of some local emission sources, which can as well affect the simulated surface air pollution levels.We used the random forest (RF) regression algorithm in order to downscale ECHAM-HAMMOZ-derived surface PM2.5 concentrations towards measured PM2.5 values from the New Delhi capital region in India. In addition, we applied the trained RF model to additional simulations where we had future anthropogenic aerosol emissions according to a business as usual scenario and a mitigation scenario. This allowed us to evaluate the effects of aerosol emission reductions on both global radiative balance, and on local air quality in New Delhi.The obtained results indicate that surface PM2.5 concentrations from RF prediction correlate with the measured PM2.5 concentrations much better than the original ECHAM-HAMMOZ particulate concentrations for New Delhi region. However, with the current setup and input variables, the PM2.5 concentrations produced by the RF model seems to be lacking some of the short-term variations and very low and high values.All in all, the downscaling method used in this project shows very promising potential, but requires further adjustment with the selection of input variables and the RF hyperparameters.

Satellite-based aerosol retrievals provide global spatially distributed estimates of atmospheric aerosol parameters that are commonly needed in applications such as estimation of atmospherically corrected satellite data products, climate modelling and air quality monitoring. However, a common feature of the conventional satellite aerosol retrievals is that they have reasonably low spatial resolution and poor accuracy caused by uncertainty in auxiliary model parameters, such as fixed aerosol model parameters, and the approximate forward radiative transfer models utilized to keep the computational complexity feasible. As a result, the improvement and reprocessing of the operational satellite data retrieval algorithms would become a tedious and computationally excessive problem. To overcome these problems, we have developed a machine-learning-based post-process correction approach to correct the existing operational satellite aerosol data products. Our approach combines the existing satellite retrieval data and a post-processing step where a machine learning algorithm is utilized to predict the approximation error in the conventional retrieval. With approximation error, we refer to the discrepancy between the true aerosol parameters and the ones retrieved using the satellite data. Our hypothesis is that the prediction of the approximation error with a finite training dataset is a less complex and easier task than the direct, fully learned machine-learning-based prediction in which the aerosol parameters are directly predicted given the satellite observations and measurement geometry. Our approach does not require reprocessing of the satellite retrieval products; it requires only a computationally fast machine-learning-based post-processing step of the existing retrieval product. Our approach is based on neural networks trained based on collocated satellite data and accurate ground-based Aerosol Robotic Network (AERONET) aerosol data. Based on our post-processing approach, we propose a post-process-corrected high-resolution Sentinel-3 Synergy aerosol product, which gives a spectral estimate of the aerosol optical depth at five different wavelengths with a high spatial resolution equivalent to the native resolution of the Sentinel-3 Level-1 data (300 m at nadir). With aerosol data from Sentinel-3A and 3B satellites, we demonstrate that our approach produces high-resolution aerosol data with clearly better accuracy than the operational Sentinel-3 Level-2 Synergy aerosol product, and it also results in slightly better accuracy than the conventional fully learned machine learning approach. We also demonstrate better generalization capabilities of the post-process correction approach over the fully learned approach.

A novel method has been developed to estimate aerosol optical depth (AOD) from sunshine duration (SD) measurements under cloud-free conditions. It is a physically-based method serving for the reconstruction of the historical evolution of AOD during the last century. In addition to sunshine duration data, it requires daily water vapor and ozone products as inputs taken from the ECMWF twentieth century reanalysis ERA-20C, available at global scale over the period 1900–2010. Surface synoptic cloud observations are used to identify cloud-free days. For sixteen sites over Europe, the accuracy of the estimated daily AOD, and its seasonal variability, is similar or better than those from two earlier methods when compared to AErosol RObotic NETwork measurements. In addition, it also improves the detection of the signal from massive aerosol events such as important volcanic eruptions (e.g., Arenal and Fernandina Island in 1968, Chichón in 1982 and Pinatubo in 1992). Finally, the reconstructed AOD time series are in good agreement with the dimming/brightening phenomenon and also provides preliminary evidence of the early-brightening phenomenon.

During the COVID-19 lockdown, the dramatic reduction of anthropogenic emissions provided a unique opportunity to investigate the effects of reduced anthropogenic activity and primary emissions on atmospheric chemical processes and the consequent formation of secondary pollutants. Here, we utilize comprehensive observations to examine the response of atmospheric new particle formation (NPF) to the changes in the atmospheric chemical cocktail. We find that the main clustering process was unaffected by the drastically reduced traffic emissions, and the formation rate of 1.5 nm particles remained unaltered. However, particle survival probability was enhanced due to an increased particle growth rate (GR) during the lockdown period, explaining the enhanced NPF activity in earlier studies. For GR at 1.5–3 nm, sulfuric acid (SA) was the main contributor at high temperatures, whilst there were unaccounted contributing vapors at low temperatures. For GR at 3–7 and 7–15 nm, oxygenated organic molecules (OOMs) played a major role. Surprisingly, OOM composition and volatility were insensitive to the large change of atmospheric NOx concentration; instead the associated high particle growth rates and high OOM concentration during the lockdown period were mostly caused by the enhanced atmospheric oxidative capacity. Overall, our findings suggest a limited role of traffic emissions in NPF.

We present here an aerosol model selection based statistical method in Bayesian framework for retrieving atmospheric aerosol optical depth (AOD) and pixel-level uncertainty. Especially, we focus on to provide more realistic uncertainty estimate by taking into account a model selection problem when searching for the solution by fitting look-up table (LUT) models to a satellite measured top-of-atmosphere reflectance. By means of Bayesian model averaging over the best-fitting aerosol models we take into account an aerosol model selection uncertainty and get also a shared inference about AOD. Moreover, we acknowledge model discrepancy, i.e. forward model error, arising from approximations and assumptions done in forward model simulations. We have estimated the model discrepancy empirically by a statistical approach utilizing residuals of model fits. We use the measurements from the TROPOspheric Monitoring Instrument (TROPOMI) onboard the Sentinel-5 Precursor in ultraviolet and visible bands, and in one wavelength band 675 nm in near-infrared, in order to study the functioning of the retrieval in a broad wavelength range. We exploit a fundamental classification of the aerosol models as weakly absorbing, biomass burning and desert dust aerosols. For experimental purpose we have included some dust type of aerosols having non-spherical particle shapes. For this study we have created the aerosol model LUTs with radiative transfer simulations using the libRadtran software package. It is reasonably straightforward to experiment with different aerosol types and evaluate the most probable aerosol type by the model selection method. We demonstrate the method in wildfire and dust events in a pixel level. In addition, we have evaluated in detail the results against ground-based remote sensing data from the AErosol RObotic NETwork (AERONET). Based on the case studies the method has ability to identify the appropriate aerosol types, but in some wildfire cases the AOD is overestimated compared to the AERONET result. The resulting uncertainty when accounting for the model selection problem and the imperfect forward modelling is higher compared to uncertainty when only measurement error is included in an observation model, as can be expected.

The number of cloud droplets formed at the cloud base depends both on the properties of aerosol particles and the updraft velocity of an air parcel at the cloud base. As the spatial scale of updrafts is too small to be resolved in global atmospheric models, the updraft velocity is commonly parameterised based on the available turbulent kinetic energy. Here we present alternative methods through parameterising updraft velocity based on high-resolution large eddy simulation (LES) runs in the case of marine stratocumulus clouds. First we use our simulations to assess the accuracy of a simple linear parametrisation where the updraft velocity depends only on cloud top radiative cooling. In addition, we present two different machine learning methods (Gaussian process emulation and random forest) that account for different boundary layer conditions and cloud properties. We conclude that both machine learning parameterisations reproduce the LES-based updraft velocities at about the same accuracy, while the simple approach employing radiative cooling only produce on average lower coefficient of determination and higher root mean square error values. Finally, we apply these machine learning methods to find the key parameters affecting cloud base updraft velocities.

In recent decades, the warming in the Arctic has been much faster than in the rest of the world, a phenomenon known as Arctic amplification (AA). Numerous studies report that Arctic is warming either twice, more than twice, or even three times as fast as the globe on average. However, the lack of consensus of AA definition precludes its precise quantification. Here we show, by using several observational datasets which cover the Arctic region and adopting a simple definition of AA, that during the last 40 years the Arctic has been warming almost four times faster than the globe as a whole, which is a higher ratio than generally reported in literature. Furthermore, we compared the observed AA ratio to the ratio simulated by state-of-the-art climate models, and show that the models largely underestimate the present AA, a finding that is not very sensitive to the exact definition of AA. The underestimation of AA by climate models most likely results from their inability to realistically simulate feedback mechanisms between sea ice melt and atmospheric temperatures. Our results imply that the underestimated AA leads to biased projections of climate change both in the Arctic and mid-latitudes.

Layers of biomass burning aerosol particles were observed in the lower troposphere, at 2 to 5 km height on 4 to 6 June 2019, over Kuopio, Finland. These long-range-transported smoke particles originated from a Canadian wildfire event. The most pronounced smoke plume detected on 5 June was intensively investigated. Optical properties were retrieved from the multi-wavelength Raman polarization lidar PollyXT. Particle linear depolarization ratios of this plume were measured to be 0.08 ± 0.02 at 355 nm and 0.05 ± 0.01 at 532 nm which were slightly higher than the values given in the literature. Non-spherical shaped aged smoke particles and/or the mixing with a small amount of fine dust particles could cause the observed increase in the particle linear depolarization ratios. Lidar ratios were derived as 47 ± 5 sr at 355 nm and 71 ± 5 sr at 532 nm. A complete ceilometer data processing for a Vaisala CL51 is presented, including the water vapor correction for high latitude for the first time, from sensor provided attenuated backscatter coefficient to particle mass concentration. A combined lidar and sun-photometer approach (based on AERONET products) is applied for mass concentration estimations. Mass concentrations were estimated from both PollyXT and the ceilometer CL51 observations, which were of the order of ~ 30 µg m−3 in the morning and decreased to ~ 20 µg m−3 in the night. A difference of ~ 30% was found between PollyXT and CL51 estimated mass concentrations. The mass retrievals were discussed and compared with the Modern-Era Retrospective analysis for Research and Applications, version 2 (MERRA-2) meteorological and aerosol reanalysis. The inclusion of dust in the retrieved mass concentration slightly improved the correspondence between the observations and the MERRA-2 simulations.

Biogenic secondary organic aerosol (BSOA) constitutes a major fraction of aerosol over boreal forests. As the emissions of BSOA precursors are temperature dependent, changes in temperature have potentially important implications on regional aerosol radiative forcing. Here, we have used long-term aerosol composition and temperature data measurements from a boreal forest site together with remote sensing observations of aerosol and cloud properties to investigate the effect of increasing temperature on organic aerosol mass loadings, and further on aerosol direct and indirect radiative effects. The analysis was based on 7 years of measurements done at Hyytiälä, Southern Finland, and they cover the summer months (July-August) between 2012-2018. We limited the analyses to these summer months to isolate the temperature dependence of the organic mass loadings from the seasonal effects arising from the vegetation growth cycle. Our analysis showed that organic aerosol loadings and cloud condensation nuclei concentrations increased in concert with surface temperature. Furthermore, we found that cloud reflectivity increased when the organic aerosol loadings increased. This research presents the first direct observational evidence on the effect of BSOA on cloud properties and their climatic significance.

In this presentation we consider uncertainty in Look-up table (LUT) based technique for retrieving aerosol optical depth (AOD) and aerosol type using TROPOMI/S5P measurements.The LUTs are multi-dimensional tables containing aerosol microphysical properties and they have been constructed using libRadtran simulations. Especially we study difficulty in aerosol microphysical model selection that reflects the retrieval uncertainty. As a source of uncertainty we have also acknowledged so called model discrepancy originating from imperfect forward modeling. The methodology considered is based on Bayesian inference where the retrieved AOD estimate is given as maximum a posterior (MAP) value and uncertainties are described as posterior density functions. We have also combined statistically the most appropriate aerosol microphysical models by Bayesian model averaging when the selection of single best-fitting model is not clear.The motivation is to consider difficulty in aerosol model selection and obtain realistic uncertainty estimates.We have applied this methodology to OMI/Aura measurements in our earlier studies. Here we present results when used higher resolution measurements from TROPOMI/S5P and studied the methodology covering various aerosol conditions including wild fire and dust events.

Satellite estimates of surface UV irradiance have been available since 1978 from TOMS UV spectrometer and continued with significantly improved ground resolution using Ozone Monitoring Instrument (OMI 2004-current) and Sentinel 5 Precursor (S5P 2017-current). The surface UV retrieval algorithm remains essentially the same: it first estimates the clear-sky UV irradiance based on measured ozone and then accounts for the attenuation by clouds and aerosols applying two consecutive correction factors. When estimating the total aerosol effect in surface UV irradiance, there are two major classes of aerosols to be considered: 1) aerosols that only scatter UV radiation and 2) aerosols that both scatter and absorb UV radiation. The former effect is implicitly included in the measured effective Lambertian Equivalent scene reflectivity (LER), so the scattering aerosol influence is estimated through cloud correction factor. Aerosols that absorb UV radiation attenuate the surface UV radiation more strongly than non-absorbing aerosols of the same extinction optical depth (AOD). Moreover, since these aerosols also attenuate the outgoing satellite-measured radiance, the cloud correction factor that treats these aerosols as purely scattering underestimates their AOD causing underestimation of LER and overestimation of surface UV irradiance. Therefore, for correction of aerosol absorption additional information is needed, such as the UV absorbing Aerosol Index (UVAI) or a model-based monthly climatology of aerosol absorption optical depth (AAOD). A correction for absorbing aerosols was proposed almost a decade ago and later implemented in the operational OMI and TROPOMI UV algorithms. In this study, however, we show that there is still room for an improvement to better account for the solar zenith angle dependence and non-linearity in the absorbing aerosol attenuation and as a result we propose an improved correction scheme. There are two main differences between the new proposed correction and the one that is currently operational in OMI and TROPOMI UV-algorithms. First, the currently operational correction for absorbing aerosols is a function of AAOD only, while the new correction takes additionally the solar zenith angle dependence into account. Second, the 2nd order polynomial of the new correction takes better into account the non-linearity in the correction as a function of AAOD, if compared to the currently operational one, and thus better describes the effect by absorbing aerosols over larger range of AAOD. To illustrate the potential impact of the new correction in the global UV estimates, we applied the current and new proposed correction for global fields of AAOD from the aerosol climatology currently used in OMI UV algorithm, showing a typical differences of ±5 %. This new correction is easy to implement operationally using information of solar zenith angle and existing AAOD climatology.

Remote sensing of the atmospheric composition from current and future satellites, such as the Sentinel missions of the Copernicus programme, yields an unprecedented amount of data to monitor air quality, ozone, UV radiation and other climate variables. Hence, full exploitation of the growing wealth of information delivered by spaceborne observing systems requires addressing the technological challenges for developing new strategies and tools that are capable to deal with these huge data volumes. The H2020 AURORA (Advanced Ultraviolet Radiation and Ozone Retrieval for Applications) project investigated a novel approach for synergistic use of ozone profile measurements acquired at different frequencies (ultraviolet, visible, thermal infrared) by sensors onboard Geostationary Equatorial Orbit (GEO) and Low Earth Orbit (LEO) satellites in the framework of the Copernicus Sentinel-4 and Sentinel-5 missions. This paper outlines the main features of the technological infrastructure, designed and developed to support the AURORA data processing chain as a distributed data processing and describes in detail the key components of the infrastructure and the software prototype. The latter demonstrates the technical feasibility of the automatic execution of the full processing chain with simulated data. The Data Processing Chain (DPC) presented in this work thus replicates a processing system that, starting from the operational satellite retrievals, carries out their fusion and results in the assimilation of the fused products. These consist in ozone vertical profiles from which further modules of the chain deliver tropospheric ozone and UV radiation at the Earth&rsquo, s surface. The conclusions highlight the relevance of this novel approach to the synergistic use of operational satellite data and underline that the infrastructure uses general-purpose technologies and is open for applications in different contexts.

Satellite-based aerosol retrievals provide a timely global view of atmospheric aerosol properties for air quality, atmospheric characterization, and correction of satellite data products and climate applications. Current aerosol data products based on satellite data, however, often have relatively large biases relative to accurate ground-based measurements and distinct levels of uncertainty associated with them. These biases and uncertainties are often caused by oversimplified assumptions and approximations used in the retrieval algorithms due to unknown surface reflectance or fixed aerosol models. Moreover, the retrieval algorithms do not usually take advantage of all the possible observational data collected by the satellite instruments and may, for example, leave some spectral bands unused. The improvement and the re-processing of the past and current operational satellite data retrieval algorithms would become a tedious and computationally expensive task. To overcome this burden, we have developed a model enforced post-process correction approach that can be used to correct the existing and operational satellite aerosol data products. Our approach combines the existing satellite aerosol retrievals and a post-processing step carried out with a machine learning based correction model for the approximation error in the retrieval. The developed approach allows for the utilization of auxiliary data sources, such as meteorological information, or additional observations such as spectral bands unused by the original retrieval algorithm. The post-process correction model can learn to correct for the biases and uncertainties in the original retrieval algorithms. As the correction is carried out as a post-processing step, it allows for computationally efficient re-processing of existing satellite aerosol datasets with no need to fully reprocess the much larger original radiance data. We demonstrate with over land aerosol optical depth (AOD) and Angstrom exponent (AE) data from the Moderate Imaging Spectroradiometer (MODIS) of Aqua satellite that our approach can significantly improve the accuracy of the satellite aerosol data products and reduce the associated uncertainties. We also give recommendations for the validation of satellite data products that are constructed using machine learning based models.

A better knowledge of the present–day aerosol forcing requires an accurate estimation of the historical evolution of aerosol optical depth (AOD), which is also crucial to better understand the role played by atmospheric aerosols in the dimming/brightening phenomena that have occurred since the mid-20th century. A physically-based approach using daily sunshine duration and cloud cover measurements is applied over Europe for retrieving AOD (Wandji Nyamsi et al., 2019). Both European Climate Assessment & Dataset (ECA&D) and national meteorological offices/institutes provide suitable measurements, from ~ 1000 ground-based stations, to carry out our study.The retrieved long-term AOD shows reasonable seasonal and annual variabilities including signals induced by major volcanic eruptions. The trends of atmospheric aerosols and associated increase and decrease of AOD over the periods 1960–1984 and 1985–2010, respectively, are in good agreement with the dimming/brightening periods reported before. In addition, a more dominant decrease in AOD including high variability from the early-1900s to the 1950s is observed, which agrees with some earlier studies reporting “early brightening” for this period. The high inter-annual AOD variability during that period may be partly due to the transition from coal to gas in some European countries and also due to the possible influence of the Word Wars I & II.ReferencesWandji Nyamsi, W.; Lipponen, A.; Sanchez–Lorenzo, A.; Wild, M. and Arola, A. (2019), A hybrid method for reconstructing the historical evolution of aerosol optical depth from sunshine duration measurements, submitted.

AURORA (Advanced Ultraviolet Radiation and Ozone Retrieval for Applications) is a three-year project supported by the European Union in the frame of its H2020 Call (EO-2-2015) for &ldquo, Stimulating wider research use of Copernicus Sentinel Data&rdquo, The project addresses key scientific issues relevant for synergistic exploitation of data acquired in different spectral ranges by different instruments on board the atmospheric Sentinels. A novel approach, based on the assimilation of geosynchronous equatorial orbit (GEO) and low Earth orbit (LEO) fused products by application of an innovative algorithm to Sentinel-4 (S-4) and Sentinel-5 (S-5) synthetic data, is adopted to assess the quality of the unique ozone vertical profile obtained in a context simulating the operational environment. The first priority is then attributed to the lower atmosphere with calculation of tropospheric columns and ultraviolet (UV) surface radiation from the resulting ozone vertical distribution. Here we provide details on the surface UV algorithm of AURORA. Both UV index (UVI) and UV-A irradiance are provided from synthetic satellite measurements, which in turn are based on atmospheric scenarios from MERRA-2 (Modern-Era Retrospective analysis for Research and Applications, Version 2) re-analysis. The UV algorithm is implemented in a software tool integrated in the technological infrastructure developed in the context of AURORA for the management of the synthetic data and for supporting the data processing. This was closely linked to the application-oriented activities of the project, aimed to improve the performance and functionality of a downstream application for personal UV dosimetry based on satellite data. The use of synthetic measurements from MERRA-2 gives us also a &ldquo, ground truth&rdquo, against which to evaluate the performance of our UV model with varying inputs. In this study we both describe the UV algorithm itself and assess the influence that changes in ozone profiles, due to the fusion and assimilation, can cause in surface UV levels.

We have developed a Bayesian aerosol retrieval (BAR) algorithm for the retrieval of aerosol optical depth (AOD) over land from the Moderate Resolution Imaging Spectroradiometer (MODIS). In the BAR algorithm, we simultaneously retrieve all dark land pixels in a granule, utilize spatial correlation models for the unknown aerosol parameters, use a statistical prior model for the surface reflectance, and take into account the uncertainties due to fixed aerosol models. The retrieved parameters are total AOD at 0.55 µm, fine-mode fraction (FMF), and surface reflectances at four different wavelengths (0.47, 0.55, 0.64, and 2.1 µm). The accuracy of the new algorithm is evaluated by comparing the AOD retrievals to Aerosol Robotic Network (AERONET) AOD. The results show that the BAR significantly improves the accuracy of AOD retrievals over the operational Dark Target (DT) algorithm. A reduction of about 29 % in the AOD root mean square error and decrease of about 80 % in the median bias of AOD were found globally when the BAR was used instead of the DT algorithm. Furthermore, the fraction of AOD retrievals inside the ±(0.05+15%) expected error envelope increased from 55 to 76 %. In addition to retrieving the values of AOD, FMF, and surface reflectance, the BAR also gives pixel-level posterior uncertainty estimates for the retrieved parameters. The BAR algorithm always results in physical, non-negative AOD values, and the average computation time for a single granule was less than a minute on a modern personal computer.

In simulations of complex physical phenomena, model reductions are often required to decrease the computation time of the simulation model to a feasible level. Model reduction is often obtained by using a reduced model, which may be based on a reduced numerical approximation and simplifications of the underlying accurate model. The use of a reduced model, however, induces errors to the simulation results. In this paper, we describe and evaluate a novel approach for the correction of the approximation errors in reduced simulation models. The key idea is to model the approximation error between the accurate and reduced simulation model as an additive noise term to the reduced model and construct a low-cost predictor model for the approximation error based on statistical learning. In this paper, the approximation error approach is evaluated with the following problems: correction of spatial and temporal discretization errors in a time-varying heat equation--based evolution model, correction of spatial discre...

Fitting a line to two measured variables is considered one of the simplest statistical procedures researchers can carry out. However, this simplicity is deceptive as the line-fitting procedure is actually quite a complex problem. Atmospheric measurement data never come without some measurement error. Too often, these errors are neglected when researchers make inferences from their data. To demonstrate the problem, we simulated datasets with different numbers of data points and different amounts of error, mimicking the dependence of the atmospheric new particle formation rate (J1.7) on the sulfuric acid concentration (H2SO4). Both variables have substantial measurement error and, thus, are good test variables for our study. We show that ordinary least squares (OLS) regression results in strongly biased slope values compared with six error-in-variables (EIV) regression methods (Deming regression, principal component analysis, orthogonal regression, Bayesian EIV and two different bivariate regression methods) that are known to take errors in the variables into account.

Recent years have seen the increasing inclusion of per-retrieval prognostic (predictive) uncertainty estimates within satellite aerosol optical depth (AOD) data sets, providing users with quantitative tools to assist in the optimal use of these data. Prognostic estimates contrast with diagnostic (i.e. relative to some external truth) ones, which are typically obtained using sensitivity and/or validation analyses. Up to now, however, the quality of these uncertainty estimates has not been routinely assessed. This study presents a review of existing prognostic and diagnostic approaches for quantifying uncertainty in satellite AOD retrievals, and it presents a general framework to evaluate them based on the expected statistical properties of ensembles of estimated uncertainties and actual retrieval errors. It is hoped that this framework will be adopted as a complement to existing AOD validation exercises; it is not restricted to AOD and can in principle be applied to other quantities for which a reference validation data set is available. This framework is then applied to assess the uncertainties provided by several satellite data sets (seven over land, five over water), which draw on methods from the empirical to sensitivity analyses to formal error propagation, at 12 Aerosol Robotic Network (AERONET) sites. The AERONET sites are divided into those for which it is expected that the techniques will perform well and those for which some complexity about the site may provide a more severe test. Overall, all techniques show some skill in that larger estimated uncertainties are generally associated with larger observed errors, although they are sometimes poorly calibrated (i.e. too small or too large in magnitude). No technique uniformly performs best. For powerful formal uncertainty propagation approaches such as optimal estimation, the results illustrate some of the difficulties in appropriate population of the covariance matrices required by the technique. When the data sets are confronted by a situation strongly counter to the retrieval forward model (e.g. potentially mixed land–water surfaces or aerosol optical properties outside the family of assumptions), some algorithms fail to provide a retrieval, while others do but with a quantitatively unreliable uncertainty estimate. The discussion suggests paths forward for the refinement of these techniques.

The composition of organic aerosol under different ambient conditions as well as their phase state have been a subject of intense study in the recent years. One way to study the particle properties is to measure the particle size shrinkage in a diluted environment at isothermal conditions. From these measurements it is possible to separate the fraction of low volatility compounds from high volatility compounds. In this work, we analyze and evaluate a method for obtaining particle composition and viscosity from measurements using process models coupled with input optimization algorithms. Two optimization methods, Monte Carlo Genetic Algorithm and Bayesian inference, are used together with process models describing the dynamics of particle evaporation. The process model optimization scheme in inferring particle composition in a volatility-basis-set sense and composition dependent particle viscosity is tested with artificially generated data sets and real experimental data. Optimizing model input so that the output matches these data yields a good match for the estimated quantities. Both optimization methods give equally good results when they are used to estimate particle composition to artificial test data. The time scale of the experiments and the initial particle size are found to be important in defining the range of values that can be identified for the properties from the optimization.

The magnitude of aerosol radiative effects remains the single largest uncertainty in current estimates of anthropogenic radiative forcing. One of the key quantities needed for accurate estimates of...

Fitting a line on a scatterplot of two measured variables is considered as one of the simplest statistical procedures researchers can do. However, this simplicity is deceptive as the line fitting procedure is actually quite a complex problem. Atmospheric measurement data never comes without some measurement error. Too often, these errors are neglected when researchers are making inferences from their data. To demonstrate the problem, we simulated datasets with different amounts of data and error, mimicking the dependence of atmospheric new particle formation rate (J1.7) on sulphuric acid concentration (H2SO4). Both variables have substantial measurement error and thus they are good test variables for our study. We show that ordinary least squares (OLS) regression results in strongly biased slope values compared with six error-in-variables (EIV) regression methods (Deming, Principal component analysis, orthogonal, Bayesian EIV, and two different bivariate regression methods) known to take into account errors in the variables.

With the launch of the Sentinel-5 Precursor (S-5P, lifted-off on 13 October 2017), Sentinel-4 (S-4) and Sentinel-5 (S-5)(from 2021 and 2023 onwards, respectively) operational missions of the ESA/EU Copernicus program, a massive amount of atmospheric composition data with unprecedented quality will become available from geostationary (GEO) and low Earth orbit (LEO) observations. Enhanced observational capabilities are expected to foster deeper insight than ever before on key issues relevant for air quality, stratospheric ozone, solar radiation, and climate. A major potential strength of the Sentinel observations lies in the exploitation of complementary information that originates from simultaneous and independent satellite measurements of the same air mass. The core purpose of the AURORA (Advanced Ultraviolet Radiation and Ozone Retrieval for Applications) project is to investigate this exploitation from a novel approach for merging data acquired in different spectral regions from on board the GEO and LEO platforms. A data processing chain is implemented and tested on synthetic observations. A new data algorithm combines the ultraviolet, visible and thermal infrared ozone products into S-4 and S-5(P) fused profiles. These fused products are then ingested into state-of-the-art data assimilation systems to obtain a unique ozone profile in analyses and forecasts mode. A comparative evaluation and validation of fused products assimilation versus the assimilation of the operational products will seek to demonstrate the improvements achieved by the proposed approach. This contribution provides a first general overview of the project, and discusses both the challenges of developing a technological infrastructure for implementing the AURORA concept, and the potential for applications of AURORA derived products, such as tropospheric ozone and UV surface radiation, in sectors such as air quality monitoring and health.