Final Report Summary - MDRAF (Effects of Mediterranean desert dust outbreaks on radiation, atmospheric dynamics and forecasting accuracy of a numerical mesoscale model)
The scientific goals of the MDRAF project were: (i) the description of the Mediterranean desert dust outbreaks’ 3D structure based on remote sensing data and (ii) the assessment of intense Mediterranean desert dust outbreaks’ direct impact on the Earth-Atmosphere system’s radiation budget, based on regional model simulations.
According to the main objectives, described briefly above, the research activities of the MDRAF project have been organized in two phases. At the first one, focus is given on the description of the Mediterranean desert dust outbreaks’ vertical structure on an annual and seasonal basis. To this aim, an objective and dynamic algorithm (flowchart of Figure 1 in the attached file) has been developed in order to identify strong and extreme desert dust (DD) episodes, over the broader Mediterranean basin, during the period 1 March 2000 – 28 February 2013. As inputs to the satellite algorithm, aerosol optical properties providing information about aerosols’ load (aerosol optical depth), size (Ångström exponent, Fine Fraction, Effective radius) and nature (Aerosol Index) are utilized. The aforementioned satellite retrievals, available on a daily basis and at 1° x 1° spatial resolution, have been derived from the MODIS-Terra/Aqua, EP-TOMS and OMI-Aura databases. The performance of the satellite algorithm has been evaluated versus optical/microphysical properties retrieved by 109 AERONET stations and surface PM10 concentrations measured at 22 sites. Through this thorough assessment, it is confirmed the appropriateness of the applied methodology ensuring thus an accurate three dimensional view of the intense Mediterranean dust outbreaks. The obtained results from the evaluation analysis are discussed in detail in the midterm report of the MDRAF project.
Based on the satellite algorithm outputs, the DD episodes’ characteristics, namely the frequency of occurrence (in terms of episodes yr-1) and intensity (in terms of AOD550nm) are analyzed. The aforementioned results, representative for the periods Mar. 2000 – Feb. 2013 and 2003 – 2012 when the satellite algorithm operates based on MODIS-Terra and MODIS-Aqua retrievals, respectively, are depicted in Figures 2 (frequency of occurrence) and 3 (intensity). From the long-term averaged geographical distributions, it is apparent that DD episodes’ frequency of occurrence is decreased from south-to-north while for the strong DD episodes it is evident also a west-to-east gradient. Strong DD episodes occur more frequently in the W. Mediterranean (about 10 episodes yr-1) while the maximum frequencies for the extreme ones (3.3 episodes yr-1) are recorded over the central parts of the Mediterranean Sea. The intensities (in terms of AOD550nm) of strong and extreme DD episodes reach up to 1.5 and 4.1 respectively, over the Gulf of Sidra and the Libyan Sea, along the northern African coasts (Figure 3).
The main limitation of the applied satellite algorithm is that operates utilizing columnar satellite retrievals as inputs prohibiting thus the vertical description of dust outbreaks. In order to overcome this issue, the CALIOP-CALIPSO vertical profiles of the Vertical Feature Mask (VFM, aerosol subtype classification scheme) and the total backscatter coefficient at 532nm (β532nm) are used as a complementary tool to the satellite algorithm’s outputs. The obtained results, for the CALIOP VFM and β532nm are presented in Figure 4-i and 4-ii, respectively, over the period Jun. 2006 – Feb. 2013. Based on this synergistic approach of passive and active satellite sensors, it is found that dust particles over the Mediterranean are mainly confined between 0.5 and 6 km, following an ascending mode, up to 40° N, leaving from the sources areas. Among the Mediterranean sub-regions, it is found that dust layers’ base height is recorded at 2 km in the western parts of the basin, being decreased down to 0.5 km over the central and eastern sectors (Fig. 4-i) reflecting the impact of the areal topography. Air masses carrying mineral particles towards the western Mediterranean (mainly in summer) are “convected” towards higher altitudes due to the existence of the Atlas Mountains Range. On the contrary, when dust transport takes place in the central and eastern Mediterranean, air masses carrying dust aerosols travel at lower altitudes because of the absence of significant topographical objects on their route.
The intensity of dust outbreaks, in terms of β532nm, is maximized (up to 0.006 km-1 sr-1) below 2 km and at the southern parts (30° N - 34° N) of the study region (Fig. 4-ii). However, considerably high β532nm values (~ 0.004 km-1 sr-1) are observed between 2 and 4 km in the latitudinal zone extending from 35° N and 42° N. These dust layers are observed in spring when strong low pressure systems (e.g. Saharan depressions) favor the transport of massive loads of mineral particles from the sources areas (Sahara desert and Middle East) towards the Mediterranean. From the 3D visualizations, it is revealed that the Mediterranean desert dust outbreaks are characterized by a multilayered structure with several dust layers of variable geometrical characteristics and intensities.
In order to give a better insight of how the dust outbreaks’ vertical distribution affects the level of agreement between MODIS columnar AOD and surface PM10 concentrations, specific desert dust outbreaks of different geometrical characteristics are investigated. The desert dust outbreaks have been selected when MODIS columnar retrievals, surface PM10 concentrations and CALIOP vertical profiles are available concurrently. According to the defined criteria, 13 desert dust outbreaks which took place in 9 Mediterranean PM10 stations during the period Jun. 2006 – Feb. 2013, are analyzed. Among the selected cases, it is found that when a compact dust layer extends at the lowest tropospheric levels (Fig. 5) then the agreement between satellite columnar AOD and surface PM10 concentrations is satisfactory. On the contrary, when the dust layer is elevated (Fig. 6-i) or is not equally distributed in vertical terms (Fig. 6-ii) then a disagreement is appeared when attempting comparisons between MODIS AOD and surface PM10 concentrations.
The scientific goal of the second phase of the MDRAF project was the assessment of the Mediterranean desert dust outbreaks’ direct impact on the Earth-Atmosphere system’s radiation budget, based on regional short-term (84 hours) simulations. For this purpose, 20 widespread and intense Mediterranean desert dust outbreaks have been selected based on the satellite algorithm outputs. The direct radiative effects (DREs) have been calculated at the top of the atmosphere (TOA), into the atmosphere (ATM), for the downwelling (SURF) and the absorbed radiation at the surface (NETSURF), for the shortwave (SW), longwave (LW) and NET (SW+LW) radiation. The regional model which has been used is the NMMB/BSC-Dust and the simulation domain covers the northern African deserts (sources), the Mediterranean basin (short-range transport) and most of the European continent (long-range transport). In the horizontal plane, the grid point spacing is equal to 0.25° x 0.25° while in vertical, 40 sigma pressure levels up to 50 hPa are used. For each experiment, after a 10-day spin-up period, a forecast cycle (initialized at 00 UTC of the desert dust outbreak day identified by the satellite algorithm) of 84 hours starts and the model outputs are produced every 3 hours. Two configurations of the model, namely RADON (activated dust-radiation interactions) and RADOFF (deactivated dust-radiation interactions) are used.
In Figure 7, are presented the instantaneous NET (SW+LW) DREs, induced by a desert dust outbreak which took place on 2nd August 2012. At TOA, during daytime, negative DREs (cooling effect) up to 250 Wm-2 are found mainly over the eastern parts of the Atlantic Ocean, while over the western parts of the Sahara positive DREs (warming effect) up to 50 Wm-2 are computed. This discrepancy is attributed to the higher/lower albedos across the desert/maritime areas enhancing/reducing thus the atmospheric warming. Over bright surfaces, the atmospheric warming dominates over the surface cooling resulting thus to positive DREs at TOA whereas over sea surfaces this condition is reversed. During daytime, due to the attenuation (through scattering and absorption) of the incoming SW radiation by mineral particles, dust outbreaks induce a strong surface cooling (DREs up to 300 Wm-2). On the contrary, during nighttime, due to the emission of LW radiation by dust particles both downwelling and absorbed radiation at the surface are increased. The warming/cooling effects into the atmosphere, throughout the day, are regulated by the absorption and emission of SW and LW radiation, respectively, by dust aerosols. Due to this fact, the suspended mineral particles warm the atmosphere (DREATM up to 200 Wm-2) during day and cool the atmospheric layers in which they are confined during night (DREATM down to -50 Wm-2). From the obtained results, it is evident that the geographical patterns of DREs are driven by the corresponding ones of the dust plumes.
The DREs have been also calculated at a regional scale, under clear sky conditions, for the broader Mediterranean area and the Sahara desert, for the NET, SW and LW radiation. The obtained results, averaged from the 20 desert dust outbreaks, for TOA, SURF, NETSURF and ATM are displayed in Figure 8. For the NET (SW+LW) radiation, in both sub-regions, it is evident a diurnal cycle of the DREATM, DRESURF and DRENETSURF values. More specifically, the maximum atmospheric warming (up to 30 Wm-2) is observed at noon while the maximum atmospheric cooling (down to -5 Wm-2) is found at night. Dust outbreaks’ radiative impacts on the surface budget are maximized at noon-late noon also; however, the effects are reversed compared to the corresponding ones found into the atmosphere. The downwelling radiation (SURF) at the surface is reduced by up to 60 Wm-2 and is increased by up to 5 Wm-2 during daytime and nighttime, respectively. The diurnal variation of DRENETSURF is similar with the corresponding one for DRESURF but the magnitude is lower. At TOA, the minimum DREs (down to -20 Wm-2) are observed at sunrise and sunset while the maximum ones (up to 5 Wm-2) are found at noon over the Sahara desert. Between the two sub-regions, at noon, it is apparent that slightly positive DRETOA values (planetary warming) are calculated over the Sahara while negative DRETOA values (planetary cooling) are found over the Mediterranean. This difference is mainly attributed to the higher surface reflectiveness across the desert areas in comparison to the darker surfaces of the Mediterranean. During daytime, the atmospheric warming induced by dust layers suspended over bright surfaces (e.g. deserts) is attributed to the absorption of the incoming solar radiation as well as to the absorption of the back reflected radiation from the ground. Under these conditions, the atmospheric warming dominates over the surface cooling leading thus to a warming of the Earth-Atmosphere system (positive DRETOA values). The diurnal variations of the SW DREs are identical with the corresponding ones for the NET radiation; however, their magnitudes are slightly higher. Reverse effects of lower magnitudes are found for the LW radiation indicating that the dust outbreaks’ radiative impacts on the incoming solar radiation are more pronounced than those on the outgoing terrestrial radiation. Due to the emission of LW radiation by dust aerosols, there is an atmospheric cooling (up to -5 Wm-2) and a surface warming (up to 8 Wm-2) and the compensation of these two reverse impacts leads to a planetary warming (up to 3 Wm-2).
Dust aerosols reduce/increase the downwelling SW/LW radiation reaching at the ground affecting thus the energy fluxes (sensible and latent heat) from the surface to atmosphere, which in turn will affect the temperature at 2 meters. As an example, in Figure 9 is presented the impact on temperature at 2 meters attributed to a dust outbreak that affected the western parts of the Sahara and the Mediterranean, on 2nd August 2012. It is apparent a reduction, along the western African coasts, by up to 4 °C during daytime (+12h and +36h) while an increase of similar magnitude is found (western Sahara) during nighttime (+24h and +48h). Both facts, either solely or combined, indicate that due to the interaction of mineral particles with the SW and LW radiation, the diurnal temperature range is reduced. Note, that the impacts on temperature at 2 meters are more pronounced over land than over sea, due to the lower heat capacity (land), resulting thus to rapid temperature anomalies caused by the radiative fluxes’ perturbations. Moreover, the spatial patterns of these anomalies, under clear sky conditions, apart from the dust plumes’ features in the horizontal plane are also determined by the dust layers’ vertical distribution.
The possible feedbacks on dust AOD and dust emission have been also investigated and the obtained results are presented in Figures 10-i and 10-ii, respectively. From the 20 desert dust outbreaks, the mean and the associated standard deviation values of the regional dust AOD550nm (averaged over the whole simulation domain) are computed for the RADON and RADOFF experiments (Fig. 10-i). The positive RADOFF-RADON differences amplify for increasing forecast hours, resulting to a reduction of the regional dust AOD550nm by 6.3% over the forecast cycle (84 hours), when the dust radiative effects are considered into the numerical simulations (RADON). The reduction of dust AOD is attributed to the reduced total emitted amount of dust, being evident at noon-late noon (Fig. 10-ii); however, the removal mechanisms (wet and dry deposition) of aerosols from the atmosphere must be considered also. The reduced outgoing surface sensible heat flux results in a reduction of turbulent momentum transfer into the atmosphere and consequently dust emission. Over the forecast period, the total emitted amount of mineral particles is decreased by about 20% for the RADON with regards to the RADOFF experiment. From the present analysis, it is evident that negative feedbacks on dust AOD and dust emission are revealed when the dust radiative impacts are considered into the numerical simulations (RADON).
At the last stage of the MDRAF project, focus is given on the model’s forecasting ability, for the RADON and RADOFF simulations, in terms of reproducing the downwelling SW/LW radiation at the ground and the temperature fields at 2 m as well as at several altitudes into the atmosphere. The former evaluation is made utilizing as reference data radiation observations, derived from the Baseline Surface Radiation Network (BSRN), measured at 6 stations. As it concerns the temperature fields, both at 2 m and at several altitudes, the corresponding model outputs are evaluated versus reanalysis datasets, meteorological observations at weather stations and vertical profiles derived by radiosondes.
The simulated downwelling SW and LW radiation at the ground, are compared against surface measurements, derived in Sede Boker (south Israel), over the 84 hours forecast period initialized at 00 UTC on 24th February 2007. The aforementioned results are presented in Figure 11-i (SW) and 11-ii (LW) while in Figure 11-iii, the model’s dust AOD is compared against ground AERONET total AOD using Ångström exponent (also retrieved by the ground sun-photometer) as an indicator of coarse/fine particles’ predominance. Between the two configurations of the model, the positive and negative biases for the SW and LW radiation, respectively, are lower for the RADON compared to the RADOFF simulation. Similar conclusions (biases reduction) are drawn from the comparison of the model outputs against other BSRN stations affected by desert dust outbreaks. However, a more careful look reveals deficiencies of the model in terms of reproducing aerosols and clouds fields. More specifically, at noon of the first day, the simulated (RADON) downwelling SW radiation is overestimated by 50 Wm-2, with regards to surface measurements, attributed to the underestimation of aerosols’ load (less radiation is attenuated by the mineral particles), as it is revealed through the model-AERONET AODs comparison (Fig. 11-iii). On the contrary, during the third day, the simulated downwelling SW radiation, for both model configurations, is underestimated by up to 600 Wm-2, compared to the measured radiation at the surface, attributed to the development of low level clouds based on model simulations.
As it has been mentioned above, the forecasting ability of the model, in terms of reproducing the temperature fields, is assessed through the comparison of its outputs against reanalysis datasets, weather observations and radiosondes profiles. For brevity reasons, only the obtained results drawn by the comparison between model and NCEP-FNL temperature vertical profiles are presented here. In Figure 12, are displayed the regional temperature biases, representative for the whole simulation domain, between model (RADON or RADOFF) and FNL, calculated at 17 pressure levels, 24 (i) and 48 (ii) hours after the initialization of the forecast cycle. In addition, the regional dust concentration (in kg m-3), averaged over the whole simulation domain at each pressure level, is also provided. It must be clarified that only grid points where the dust AOD is higher/equal than 0.5 are considered in the analysis. Moreover, the shaded areas correspond to the standard deviation values computed from the 20 desert dust outbreaks, which are analyzed. The warm RADOFF-FNL biases, being more pronounced between 850 hPa and 600 hPa, are decreased by up to 0.3 °C (red line is getting closer to the blue thick line) when the dust-radiation interactions are considered into the numerical simulations (RADON), revealing thus an improvement of the model’s forecasting ability. These temperature “corrections” are found at altitudes, particularly between 850 and 700 hPa, where the dust aerosols’ concentration is maximized. During nighttime, mineral particles emit longwave radiation reducing thus the temperature of atmospheric layers in which they are confined. Due to this process, the model warm biases are decreased, being almost negligible at specific pressure levels, 48 hours after the initialization of the forecast cycle, when the dust-radiation interactions are activated (RADON). Through the assessment of the model’s performance, both for radiation and temperature, it is highlighted the importance of dust radiative impacts’ inclusion into the numerical simulations since they can improve the forecasting ability of a regional model.
According to the main objectives, described briefly above, the research activities of the MDRAF project have been organized in two phases. At the first one, focus is given on the description of the Mediterranean desert dust outbreaks’ vertical structure on an annual and seasonal basis. To this aim, an objective and dynamic algorithm (flowchart of Figure 1 in the attached file) has been developed in order to identify strong and extreme desert dust (DD) episodes, over the broader Mediterranean basin, during the period 1 March 2000 – 28 February 2013. As inputs to the satellite algorithm, aerosol optical properties providing information about aerosols’ load (aerosol optical depth), size (Ångström exponent, Fine Fraction, Effective radius) and nature (Aerosol Index) are utilized. The aforementioned satellite retrievals, available on a daily basis and at 1° x 1° spatial resolution, have been derived from the MODIS-Terra/Aqua, EP-TOMS and OMI-Aura databases. The performance of the satellite algorithm has been evaluated versus optical/microphysical properties retrieved by 109 AERONET stations and surface PM10 concentrations measured at 22 sites. Through this thorough assessment, it is confirmed the appropriateness of the applied methodology ensuring thus an accurate three dimensional view of the intense Mediterranean dust outbreaks. The obtained results from the evaluation analysis are discussed in detail in the midterm report of the MDRAF project.
Based on the satellite algorithm outputs, the DD episodes’ characteristics, namely the frequency of occurrence (in terms of episodes yr-1) and intensity (in terms of AOD550nm) are analyzed. The aforementioned results, representative for the periods Mar. 2000 – Feb. 2013 and 2003 – 2012 when the satellite algorithm operates based on MODIS-Terra and MODIS-Aqua retrievals, respectively, are depicted in Figures 2 (frequency of occurrence) and 3 (intensity). From the long-term averaged geographical distributions, it is apparent that DD episodes’ frequency of occurrence is decreased from south-to-north while for the strong DD episodes it is evident also a west-to-east gradient. Strong DD episodes occur more frequently in the W. Mediterranean (about 10 episodes yr-1) while the maximum frequencies for the extreme ones (3.3 episodes yr-1) are recorded over the central parts of the Mediterranean Sea. The intensities (in terms of AOD550nm) of strong and extreme DD episodes reach up to 1.5 and 4.1 respectively, over the Gulf of Sidra and the Libyan Sea, along the northern African coasts (Figure 3).
The main limitation of the applied satellite algorithm is that operates utilizing columnar satellite retrievals as inputs prohibiting thus the vertical description of dust outbreaks. In order to overcome this issue, the CALIOP-CALIPSO vertical profiles of the Vertical Feature Mask (VFM, aerosol subtype classification scheme) and the total backscatter coefficient at 532nm (β532nm) are used as a complementary tool to the satellite algorithm’s outputs. The obtained results, for the CALIOP VFM and β532nm are presented in Figure 4-i and 4-ii, respectively, over the period Jun. 2006 – Feb. 2013. Based on this synergistic approach of passive and active satellite sensors, it is found that dust particles over the Mediterranean are mainly confined between 0.5 and 6 km, following an ascending mode, up to 40° N, leaving from the sources areas. Among the Mediterranean sub-regions, it is found that dust layers’ base height is recorded at 2 km in the western parts of the basin, being decreased down to 0.5 km over the central and eastern sectors (Fig. 4-i) reflecting the impact of the areal topography. Air masses carrying mineral particles towards the western Mediterranean (mainly in summer) are “convected” towards higher altitudes due to the existence of the Atlas Mountains Range. On the contrary, when dust transport takes place in the central and eastern Mediterranean, air masses carrying dust aerosols travel at lower altitudes because of the absence of significant topographical objects on their route.
The intensity of dust outbreaks, in terms of β532nm, is maximized (up to 0.006 km-1 sr-1) below 2 km and at the southern parts (30° N - 34° N) of the study region (Fig. 4-ii). However, considerably high β532nm values (~ 0.004 km-1 sr-1) are observed between 2 and 4 km in the latitudinal zone extending from 35° N and 42° N. These dust layers are observed in spring when strong low pressure systems (e.g. Saharan depressions) favor the transport of massive loads of mineral particles from the sources areas (Sahara desert and Middle East) towards the Mediterranean. From the 3D visualizations, it is revealed that the Mediterranean desert dust outbreaks are characterized by a multilayered structure with several dust layers of variable geometrical characteristics and intensities.
In order to give a better insight of how the dust outbreaks’ vertical distribution affects the level of agreement between MODIS columnar AOD and surface PM10 concentrations, specific desert dust outbreaks of different geometrical characteristics are investigated. The desert dust outbreaks have been selected when MODIS columnar retrievals, surface PM10 concentrations and CALIOP vertical profiles are available concurrently. According to the defined criteria, 13 desert dust outbreaks which took place in 9 Mediterranean PM10 stations during the period Jun. 2006 – Feb. 2013, are analyzed. Among the selected cases, it is found that when a compact dust layer extends at the lowest tropospheric levels (Fig. 5) then the agreement between satellite columnar AOD and surface PM10 concentrations is satisfactory. On the contrary, when the dust layer is elevated (Fig. 6-i) or is not equally distributed in vertical terms (Fig. 6-ii) then a disagreement is appeared when attempting comparisons between MODIS AOD and surface PM10 concentrations.
The scientific goal of the second phase of the MDRAF project was the assessment of the Mediterranean desert dust outbreaks’ direct impact on the Earth-Atmosphere system’s radiation budget, based on regional short-term (84 hours) simulations. For this purpose, 20 widespread and intense Mediterranean desert dust outbreaks have been selected based on the satellite algorithm outputs. The direct radiative effects (DREs) have been calculated at the top of the atmosphere (TOA), into the atmosphere (ATM), for the downwelling (SURF) and the absorbed radiation at the surface (NETSURF), for the shortwave (SW), longwave (LW) and NET (SW+LW) radiation. The regional model which has been used is the NMMB/BSC-Dust and the simulation domain covers the northern African deserts (sources), the Mediterranean basin (short-range transport) and most of the European continent (long-range transport). In the horizontal plane, the grid point spacing is equal to 0.25° x 0.25° while in vertical, 40 sigma pressure levels up to 50 hPa are used. For each experiment, after a 10-day spin-up period, a forecast cycle (initialized at 00 UTC of the desert dust outbreak day identified by the satellite algorithm) of 84 hours starts and the model outputs are produced every 3 hours. Two configurations of the model, namely RADON (activated dust-radiation interactions) and RADOFF (deactivated dust-radiation interactions) are used.
In Figure 7, are presented the instantaneous NET (SW+LW) DREs, induced by a desert dust outbreak which took place on 2nd August 2012. At TOA, during daytime, negative DREs (cooling effect) up to 250 Wm-2 are found mainly over the eastern parts of the Atlantic Ocean, while over the western parts of the Sahara positive DREs (warming effect) up to 50 Wm-2 are computed. This discrepancy is attributed to the higher/lower albedos across the desert/maritime areas enhancing/reducing thus the atmospheric warming. Over bright surfaces, the atmospheric warming dominates over the surface cooling resulting thus to positive DREs at TOA whereas over sea surfaces this condition is reversed. During daytime, due to the attenuation (through scattering and absorption) of the incoming SW radiation by mineral particles, dust outbreaks induce a strong surface cooling (DREs up to 300 Wm-2). On the contrary, during nighttime, due to the emission of LW radiation by dust particles both downwelling and absorbed radiation at the surface are increased. The warming/cooling effects into the atmosphere, throughout the day, are regulated by the absorption and emission of SW and LW radiation, respectively, by dust aerosols. Due to this fact, the suspended mineral particles warm the atmosphere (DREATM up to 200 Wm-2) during day and cool the atmospheric layers in which they are confined during night (DREATM down to -50 Wm-2). From the obtained results, it is evident that the geographical patterns of DREs are driven by the corresponding ones of the dust plumes.
The DREs have been also calculated at a regional scale, under clear sky conditions, for the broader Mediterranean area and the Sahara desert, for the NET, SW and LW radiation. The obtained results, averaged from the 20 desert dust outbreaks, for TOA, SURF, NETSURF and ATM are displayed in Figure 8. For the NET (SW+LW) radiation, in both sub-regions, it is evident a diurnal cycle of the DREATM, DRESURF and DRENETSURF values. More specifically, the maximum atmospheric warming (up to 30 Wm-2) is observed at noon while the maximum atmospheric cooling (down to -5 Wm-2) is found at night. Dust outbreaks’ radiative impacts on the surface budget are maximized at noon-late noon also; however, the effects are reversed compared to the corresponding ones found into the atmosphere. The downwelling radiation (SURF) at the surface is reduced by up to 60 Wm-2 and is increased by up to 5 Wm-2 during daytime and nighttime, respectively. The diurnal variation of DRENETSURF is similar with the corresponding one for DRESURF but the magnitude is lower. At TOA, the minimum DREs (down to -20 Wm-2) are observed at sunrise and sunset while the maximum ones (up to 5 Wm-2) are found at noon over the Sahara desert. Between the two sub-regions, at noon, it is apparent that slightly positive DRETOA values (planetary warming) are calculated over the Sahara while negative DRETOA values (planetary cooling) are found over the Mediterranean. This difference is mainly attributed to the higher surface reflectiveness across the desert areas in comparison to the darker surfaces of the Mediterranean. During daytime, the atmospheric warming induced by dust layers suspended over bright surfaces (e.g. deserts) is attributed to the absorption of the incoming solar radiation as well as to the absorption of the back reflected radiation from the ground. Under these conditions, the atmospheric warming dominates over the surface cooling leading thus to a warming of the Earth-Atmosphere system (positive DRETOA values). The diurnal variations of the SW DREs are identical with the corresponding ones for the NET radiation; however, their magnitudes are slightly higher. Reverse effects of lower magnitudes are found for the LW radiation indicating that the dust outbreaks’ radiative impacts on the incoming solar radiation are more pronounced than those on the outgoing terrestrial radiation. Due to the emission of LW radiation by dust aerosols, there is an atmospheric cooling (up to -5 Wm-2) and a surface warming (up to 8 Wm-2) and the compensation of these two reverse impacts leads to a planetary warming (up to 3 Wm-2).
Dust aerosols reduce/increase the downwelling SW/LW radiation reaching at the ground affecting thus the energy fluxes (sensible and latent heat) from the surface to atmosphere, which in turn will affect the temperature at 2 meters. As an example, in Figure 9 is presented the impact on temperature at 2 meters attributed to a dust outbreak that affected the western parts of the Sahara and the Mediterranean, on 2nd August 2012. It is apparent a reduction, along the western African coasts, by up to 4 °C during daytime (+12h and +36h) while an increase of similar magnitude is found (western Sahara) during nighttime (+24h and +48h). Both facts, either solely or combined, indicate that due to the interaction of mineral particles with the SW and LW radiation, the diurnal temperature range is reduced. Note, that the impacts on temperature at 2 meters are more pronounced over land than over sea, due to the lower heat capacity (land), resulting thus to rapid temperature anomalies caused by the radiative fluxes’ perturbations. Moreover, the spatial patterns of these anomalies, under clear sky conditions, apart from the dust plumes’ features in the horizontal plane are also determined by the dust layers’ vertical distribution.
The possible feedbacks on dust AOD and dust emission have been also investigated and the obtained results are presented in Figures 10-i and 10-ii, respectively. From the 20 desert dust outbreaks, the mean and the associated standard deviation values of the regional dust AOD550nm (averaged over the whole simulation domain) are computed for the RADON and RADOFF experiments (Fig. 10-i). The positive RADOFF-RADON differences amplify for increasing forecast hours, resulting to a reduction of the regional dust AOD550nm by 6.3% over the forecast cycle (84 hours), when the dust radiative effects are considered into the numerical simulations (RADON). The reduction of dust AOD is attributed to the reduced total emitted amount of dust, being evident at noon-late noon (Fig. 10-ii); however, the removal mechanisms (wet and dry deposition) of aerosols from the atmosphere must be considered also. The reduced outgoing surface sensible heat flux results in a reduction of turbulent momentum transfer into the atmosphere and consequently dust emission. Over the forecast period, the total emitted amount of mineral particles is decreased by about 20% for the RADON with regards to the RADOFF experiment. From the present analysis, it is evident that negative feedbacks on dust AOD and dust emission are revealed when the dust radiative impacts are considered into the numerical simulations (RADON).
At the last stage of the MDRAF project, focus is given on the model’s forecasting ability, for the RADON and RADOFF simulations, in terms of reproducing the downwelling SW/LW radiation at the ground and the temperature fields at 2 m as well as at several altitudes into the atmosphere. The former evaluation is made utilizing as reference data radiation observations, derived from the Baseline Surface Radiation Network (BSRN), measured at 6 stations. As it concerns the temperature fields, both at 2 m and at several altitudes, the corresponding model outputs are evaluated versus reanalysis datasets, meteorological observations at weather stations and vertical profiles derived by radiosondes.
The simulated downwelling SW and LW radiation at the ground, are compared against surface measurements, derived in Sede Boker (south Israel), over the 84 hours forecast period initialized at 00 UTC on 24th February 2007. The aforementioned results are presented in Figure 11-i (SW) and 11-ii (LW) while in Figure 11-iii, the model’s dust AOD is compared against ground AERONET total AOD using Ångström exponent (also retrieved by the ground sun-photometer) as an indicator of coarse/fine particles’ predominance. Between the two configurations of the model, the positive and negative biases for the SW and LW radiation, respectively, are lower for the RADON compared to the RADOFF simulation. Similar conclusions (biases reduction) are drawn from the comparison of the model outputs against other BSRN stations affected by desert dust outbreaks. However, a more careful look reveals deficiencies of the model in terms of reproducing aerosols and clouds fields. More specifically, at noon of the first day, the simulated (RADON) downwelling SW radiation is overestimated by 50 Wm-2, with regards to surface measurements, attributed to the underestimation of aerosols’ load (less radiation is attenuated by the mineral particles), as it is revealed through the model-AERONET AODs comparison (Fig. 11-iii). On the contrary, during the third day, the simulated downwelling SW radiation, for both model configurations, is underestimated by up to 600 Wm-2, compared to the measured radiation at the surface, attributed to the development of low level clouds based on model simulations.
As it has been mentioned above, the forecasting ability of the model, in terms of reproducing the temperature fields, is assessed through the comparison of its outputs against reanalysis datasets, weather observations and radiosondes profiles. For brevity reasons, only the obtained results drawn by the comparison between model and NCEP-FNL temperature vertical profiles are presented here. In Figure 12, are displayed the regional temperature biases, representative for the whole simulation domain, between model (RADON or RADOFF) and FNL, calculated at 17 pressure levels, 24 (i) and 48 (ii) hours after the initialization of the forecast cycle. In addition, the regional dust concentration (in kg m-3), averaged over the whole simulation domain at each pressure level, is also provided. It must be clarified that only grid points where the dust AOD is higher/equal than 0.5 are considered in the analysis. Moreover, the shaded areas correspond to the standard deviation values computed from the 20 desert dust outbreaks, which are analyzed. The warm RADOFF-FNL biases, being more pronounced between 850 hPa and 600 hPa, are decreased by up to 0.3 °C (red line is getting closer to the blue thick line) when the dust-radiation interactions are considered into the numerical simulations (RADON), revealing thus an improvement of the model’s forecasting ability. These temperature “corrections” are found at altitudes, particularly between 850 and 700 hPa, where the dust aerosols’ concentration is maximized. During nighttime, mineral particles emit longwave radiation reducing thus the temperature of atmospheric layers in which they are confined. Due to this process, the model warm biases are decreased, being almost negligible at specific pressure levels, 48 hours after the initialization of the forecast cycle, when the dust-radiation interactions are activated (RADON). Through the assessment of the model’s performance, both for radiation and temperature, it is highlighted the importance of dust radiative impacts’ inclusion into the numerical simulations since they can improve the forecasting ability of a regional model.
