Periodic Reporting for period 1 - DUST-GLASS (Improving global dust prediction and monitoring through data assimilation of satellite-based dust aerosol optical depth)
Reporting period: 2017-05-01 to 2019-04-30
Across the arid and semi-arid regions of the planet, massive amounts of desert dust aerosols (or mineral particles) are emitted in the atmosphere under the impact of winds. Among aerosol species, dust is the most abundant component, in terms of mass, contributing more than half to the global aerosol amount. Dust aerosols play a key role in several aspects of the Earth system, thus explaining the huge scientific efforts on the description of their spatiotemporal features and the investigation of the induced impacts on weather and climate, on marine and terrestrial ecosystems, on humans’ health and on several anthropogenic activities (transportation, agriculture, solar energy production). Given the scientific importance of dust in the Earth system as well as the numerous socioeconomic impacts, it is clearly reflected the imperative need for optimizing dust monitoring and forecasting. For this purpose, contemporary satellite observations, ground-based platforms and atmospheric-dust models have been employed. Considering the advantages and drawbacks of the above-mentioned approaches, the best practice is the synergistic implementation of all the available “tools” for dust research.
Aligned to these daunting and demanding challenges in the relevant scientific field, the research activities of the DUST-GLASS project (Improving global dust prediction and monitoring through data assimilation of satellite-based dust aerosol optical depth) have been designed appropriately in order to fulfil and advance the methodologies for dust monitoring and forecasting. To realize, satellite observations obtained by passive and active remote sensors, reanalysis datasets and a sophisticated chemical weather prediction system, with focus on mineral dust, have been utilized. DUST-GLASS is composed by three key research objectives (ROs), spanning in a wide range of activities, including the development of a multi-year global dust optical depth (DOD) dataset, its integration in a data assimilation (DA) system suitable for global/regional dust simulations and the assessment of DA impact on dust forecasts.
Aligned to these daunting and demanding challenges in the relevant scientific field, the research activities of the DUST-GLASS project (Improving global dust prediction and monitoring through data assimilation of satellite-based dust aerosol optical depth) have been designed appropriately in order to fulfil and advance the methodologies for dust monitoring and forecasting. To realize, satellite observations obtained by passive and active remote sensors, reanalysis datasets and a sophisticated chemical weather prediction system, with focus on mineral dust, have been utilized. DUST-GLASS is composed by three key research objectives (ROs), spanning in a wide range of activities, including the development of a multi-year global dust optical depth (DOD) dataset, its integration in a data assimilation (DA) system suitable for global/regional dust simulations and the assessment of DA impact on dust forecasts.
In the framework of DUST-GLASS, a decadal (2007-2016) global DOD product has been developed via the synergy of MODIS-Aqua columnar aerosol optical depth (AOD), MERRA-2 reanalysis aerosol outputs and CALIOP-CALIPSO vertically-resolved aerosol retrievals. The core concept of the applied methodology is to derive DOD at 550nm on MODIS Level 2 (L2) retrievals in conjunction with the MERRA-2 DOD-to-AOD ratio. The reliability and usefulness of the latter parameter has been justified via a thorough assessment analysis versus CALIOP-CALIPSO dust portion.
Based on MODIS DOD, the regime of dust loads is described through the reproduction of annual (Figure 1) and seasonal (Figure 2) climatological global maps as well as by analyzing the interannual and intra-annual variation of DOD at planetary, hemispherical and regional level. Moreover, the derived DOD has been compared against CALIOP and MERRA-2 DODs showing a satisfactory level of agreement, despite the different techniques applied for the quantification of mineral particles’ amount. Likewise, the MODIS-DOD is highly consistent with AERONET observations, which have been treated appropriately in order to minimize, as much as possible, the “contamination” of other aerosol types (Figure 3).
At the second phase of DUST-GLASS, the developed DOD dataset is utilized in the DA scheme of the NMMB-MONARCH model, operating at the Barcelona Supercomputing Center (BSC). The data assimilation scheme is the Local Ensemble Transform Kalman filter (LETKF), an ensemble-based technique using flow-dependent model error amplitudes and structures which evolve during forecast. A “prognostic” uncertainty model is constructed for the estimation of DOD error, which is expressed by the MODIS-AERONET absolute DOD departures, multiplied with the geometric air mass factor, as a function of MODIS DOD.
The implementation of the ensemble forecast for NMMB-MONARCH is relied on 12 members, generated by applying multi-parameters, multi-physics and multi-meteorological initial and boundary conditions perturbations. High resolution numerical simulations (0.1° x 0.1°) have been performed for two experiments consisting of an ensemble first-guess (FG), and an ensemble analysis (AN), where a first-guess is an analysis-initialized forecast. The domain of interest covers the North Africa, the Middle East and Europe and the simulation period spans from 1st Jan 2012 to 31st Jan 2012, with a spin-up run (without assimilation) for one month (December 2011). In Figure 4, are illustrated the geographical distributions of DOD based on the FG (upper panel) and AN (middle panel) runs while in the bottom panel are depicted the increments (AN-FG). For the assessment of the potential positive impact of DA on short-term (24 hours) dust forecasts, the FG and AN DODs have been compared against coarse AODs obtained by AERONET stations. Overall, the representation of dust fields is improved when the MODIS DOD observations are assimilated in the NMMB-MONARCH model (Figure 5).
Besides the research activities of DUST-GLASS, the developed DOD product has been utilized in the FINDING (“ForecastINg Dust Impact on solar eNergy in EGypt”) project while it will be exploited in the DustClim project (https://bit.ly/2Kj7zBE) which focus on the production of a dust regional model reanalysis for North Africa, Middle East and Europe and the development of dust-related services tailored to specific socio-economic sectors. During the project period, the obtained findings have been presented (7 announcements) at international conferences/workshops while the preparation of two scientific manuscripts (i.e. global dust climatology, updated results for DA by performing numerical simulations for a full year cycle) is in progress. Finally, in order to enhance the dissemination and exploitation actions, the global DOD dataset will be included in the dust observations inventory developed in the framework of the inDust (“International Network to Encourage the Use of Monitoring and Forecasting Dust Products”) COST Action (https://cost-indust.eu/).
Based on MODIS DOD, the regime of dust loads is described through the reproduction of annual (Figure 1) and seasonal (Figure 2) climatological global maps as well as by analyzing the interannual and intra-annual variation of DOD at planetary, hemispherical and regional level. Moreover, the derived DOD has been compared against CALIOP and MERRA-2 DODs showing a satisfactory level of agreement, despite the different techniques applied for the quantification of mineral particles’ amount. Likewise, the MODIS-DOD is highly consistent with AERONET observations, which have been treated appropriately in order to minimize, as much as possible, the “contamination” of other aerosol types (Figure 3).
At the second phase of DUST-GLASS, the developed DOD dataset is utilized in the DA scheme of the NMMB-MONARCH model, operating at the Barcelona Supercomputing Center (BSC). The data assimilation scheme is the Local Ensemble Transform Kalman filter (LETKF), an ensemble-based technique using flow-dependent model error amplitudes and structures which evolve during forecast. A “prognostic” uncertainty model is constructed for the estimation of DOD error, which is expressed by the MODIS-AERONET absolute DOD departures, multiplied with the geometric air mass factor, as a function of MODIS DOD.
The implementation of the ensemble forecast for NMMB-MONARCH is relied on 12 members, generated by applying multi-parameters, multi-physics and multi-meteorological initial and boundary conditions perturbations. High resolution numerical simulations (0.1° x 0.1°) have been performed for two experiments consisting of an ensemble first-guess (FG), and an ensemble analysis (AN), where a first-guess is an analysis-initialized forecast. The domain of interest covers the North Africa, the Middle East and Europe and the simulation period spans from 1st Jan 2012 to 31st Jan 2012, with a spin-up run (without assimilation) for one month (December 2011). In Figure 4, are illustrated the geographical distributions of DOD based on the FG (upper panel) and AN (middle panel) runs while in the bottom panel are depicted the increments (AN-FG). For the assessment of the potential positive impact of DA on short-term (24 hours) dust forecasts, the FG and AN DODs have been compared against coarse AODs obtained by AERONET stations. Overall, the representation of dust fields is improved when the MODIS DOD observations are assimilated in the NMMB-MONARCH model (Figure 5).
Besides the research activities of DUST-GLASS, the developed DOD product has been utilized in the FINDING (“ForecastINg Dust Impact on solar eNergy in EGypt”) project while it will be exploited in the DustClim project (https://bit.ly/2Kj7zBE) which focus on the production of a dust regional model reanalysis for North Africa, Middle East and Europe and the development of dust-related services tailored to specific socio-economic sectors. During the project period, the obtained findings have been presented (7 announcements) at international conferences/workshops while the preparation of two scientific manuscripts (i.e. global dust climatology, updated results for DA by performing numerical simulations for a full year cycle) is in progress. Finally, in order to enhance the dissemination and exploitation actions, the global DOD dataset will be included in the dust observations inventory developed in the framework of the inDust (“International Network to Encourage the Use of Monitoring and Forecasting Dust Products”) COST Action (https://cost-indust.eu/).
The contribution of data assimilation on aerosol research has been proven critical and valuable. This is clearly reflected by the remarkable improvement of the performance of aerosol models used for monitoring and forecasting as well as for the assessment of aerosol-induced impacts. Nevertheless, the vast majority of these efforts is relied on the assimilation of total aerosol observations and not solely on the dust component. Therefore, one novel aspect of DUST-GLASS is the development of a dust optical depth product suitable for data assimilation applications, aiming at advancing dust modelling and prediction. Currently, only a limited number of DOD datasets, derived by spaceborne retrievals or reanalysis datasets, exists. Even though these are available at global scale and for extended time periods, DOD is provided either at coarse spatial resolution or only above desert areas. It must be highlighted that the developed DOD product, in the framework of DUST-GLASS, complements existing observational capabilities through the provision of columnar DOD on a daily basis over a decade, at fine spatial resolution and at global scale, including both continental and maritime areas.