Amor a Colores | Read more | Download

Sensitivity studies of aerosol data assimilation and direct radiative feedbacks in modeling dust aerosols

Sensitivity studies of aerosol data assimilation and direct radiative feedbacks in modeling dust aerosols

Atmospheric Environment 64 (2013) 208e218 Contents lists available at SciVerse ScienceDirect Atmospheric Environment journal homepage: www.elsevier...

3MB Sizes 0 Downloads 11 Views

Atmospheric Environment 64 (2013) 208e218

Contents lists available at SciVerse ScienceDirect

Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv

Sensitivity studies of aerosol data assimilation and direct radiative feedbacks in modeling dust aerosols Hong Wang, Tao Niu* Chinese Academy of Meteorological Sciences (CAMS), Beijing 100081, China

h i g h l i g h t s < 3DV improves modeling dust concentration and 3DV_RAD further corrects the modeling. < 3DV_RAD is more efficiency in enhancing model ability in upper air meteorological fields than that in low and surface. < Dust radiative feedback is negative at daytime and positive at night time.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 18 May 2012 Received in revised form 26 September 2012 Accepted 27 September 2012

In order to study dust aerosol assimilation and radiative forcing in modeling East Asian dust aerosols and their impacts on the regional atmosphere, a three-dimensional variational data assimilation (3DV) and an aerosol radiative feedback scheme (RAD) are online implemented into a mesoscale numerical weather prediction system GRAPES/CUACE_Dust. Four modeling experiments are conducted: one control running (CTL) excluding 3DV and RAD as well as three sensitive running experiments respectively with 3DV, RAD, the integrated 3DV and RAD (3DV_RAD). The results indicate the 3DV-running shows a distinct improvement in the daily averaged dust concentrations, while the 3DV_RAD performs the better modeling during strong dust storms. The comparisons of the model bias for air temperature, pressure and wind speed from the CTL, RAD, and 3DV_RAD experiments present that the dust direct radiation leads to decreases in the lower tropospheric temperature and increases in the upper tropospheric temperature, which results in enhancing air pressure in the lower troposphere and declining air pressure in the upper troposphere. The 3DV_RAD modeling in the middle and upper troposphere is more reasonable than the RAD modeling, which suggests the importance of integration of aerosol assimilation and radiation forcing in modeling aerosols and meteorological fields. Ó 2012 Elsevier Ltd. All rights reserved.

Keywords: Dust radiative feedbacks Aerosol assimilation system Mesoscale dust forecast system

1. Introduction The development of dust emission and transport models aiming on Asian dust episodes began at the 1990s [Kotamarthi and Carmichael, 1993; Liu and Zhou, 1997; Xiao et al., 1997]. Dust concentration and aerosol optical depth (AOD) over middle and east Asia are intensively modeled [Wang et al., 2000; Shao, 2001, 2004; Gong et al., 2003; Uno et al., 2001, 2004; Gong and Zhang, 2008; Zhou et al., 2008; Wang et al., 2009.]. However, large uncertainties still exist in modeling dust aerosols, and the modeled concentration levels could differ as high as 2e4 times and even up to more than two orders of magnitude [Uno et al., 2006]. Among

* Corresponding author. E-mail addresses: [email protected] cams.cma.gov.cn (T. Niu).

(H.

Wang),

[email protected]

1352-2310/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.atmosenv.2012.09.066

the attempts reducing the modeling uncertainties and bias, the online calculation of dust radiative impacts and the introduction of data assimilation system are two eye-catching directions in the dust model development. They are also being regarded as the new steps to improve dust weather forecast [Barker et al., 2004; Perez et al., 2006; Niu et al., 2008; Wang et al., 2010]. Dust aerosol has profound impacts on regional radiation balance, consequently on the atmospheric structures such as air temperature, air pressure, wind speeds and planetary boundary layer (PBL) circulation and cloud fields, etc, and the atmospheric changes may affect dust emissions, depositions and transport in turn [Haywood et al., 2005; Ahn et al., 2007; Heinold et al., 2008; Wang et al., 2006]. The dust forecast with the online integrated radiation forcing showed significant improvements in the modeling capability (Perez et al., 2006; Wang et al., 2010), and this improvement highly depends on the accuracy of modeled dust concentrations or loadings. Data assimilation is an important way

H. Wang, T. Niu / Atmospheric Environment 64 (2013) 208e218

to decrease the bias of dust concentration simulations. A three dimensional variational (3DV) data assimilation method was used for the sand and dust storm data assimilation system (SDS-DAS) in China (Niu et al., 2008). The operational measurements of visibility and weather phenomena for dust aerosols from surface meteorological stations and the spatial coverage observed by the Chinese geostationary satellite FY-2C have been used in the SDS-DAS in China. The SDS-DAS has been applied in an operational dust forecasting system MM5/CUACE-dust in East Asia providing the corrections to both under- and over-predictions of East Asian dust aerosols (Zhou et al., 2008). However, the dust aerosol assimilation alone cannot improve modeling meteorological fields for lacking of the linking bridge, dust radiation, between the improved forecast dust concentrations and model dynamical core. In this study, the SDS-DAS is introduced into the mesoscale twoway dust model GRAPES/CUACE_Dust and thereby enables the online interaction of dust forecast-assimilation-direct radiative effects. The assimilated dust concentration alters the meteorological fields via reforming radiation balance and finally results in improving of the model predictions of meteorological fields and dust concentrations themselves. 2. Study method By running the mesoscale dust aerosol model GRAPES/CUACE_Dust, four modeling experiments are conducted to evaluate the interactions of dust forecast-assimilation-radiation effects: one control test (CTL) excluding 3DV and RAD as well as three sensitive experiments respectively with 3DV, RAD, 3DV_RAD of the integrated 3DV and RAD. 2.1. Introduction of GRAPES/CUACE_Dust

209

transport of Aeolian Dust and it’s Effects on Climate (ADEC)” has been proved to be the most representative optical data for East Asia dust (Wang et al., 2006, 2010). It is used to calculate dust AOD, single-scattering albedo (SSA), and asymmetry factor (ASY) as inputs for the CLIRAD-SW and CLIRAD-LW. 2.3. Dust data assimilation system The SDS-DAS was developed for CUACE/Dust model. It is based on a three dimensional variational method (Zhang et al., 2004; Zhuang et al., 2005) with the measurements of surface visibility, dust loading derived from the Chinese geostationary satellite FY-2C (Hu et al., 2008) and the modeled dust concentrations for the initial date. The background error covariance matrix is important to the analysis system, which controls how the information from the observation influences the value of model grids nearby the observational position. A statistical harmonious correction is given via the background error covariance matrix to the model grid nearby observational position in order to make sure the dynamical harmony of model variables. However, the background error covariance cannot be calculated accurately because the true situation of the atmosphere cannot be known. The observation method or Hollingsworth and Lonnberg (1986) method, NMC method (Parrish and Derber, 1992) and analysis ensemble method (Fisher, 2001) are used to solve this problem, which has been described in detail in the previous paper of Niu et al. (2008). As the SDS-DAS had been found to provide the reasonable corrections to both under- and overestimates of dust concentration forecast in MM5/ CUACE-Dust (Zhou et al., 2008), it is introduced in GRAPES/CUACE_Dust model to study the improvements of the modeling performance through dust radiation feedbacks.

The GRAPES/CUACE_Dust is a mesoscale operational dust forecasting system developed by the Chinese Academy of meteorological Science (CAMS) (Wang et al., 2009, 2010.). The model top is set about 30 km with 31 vertical layers, and the model domain covers the East Asian region (15w60 N, 70w145 E) with the horizontal resolution of 0.25  0.25 in all studies. The dust particles are divided into 12 size bins respectively with the diameter ranges of 0.01e40.96 mm: 0.01e0.02 mm, 0.02e0.04 mm, 0.04e 0.08 mm, 0.08e0.16 mm, 0.16e0.32 mm, 0.32e0.64 mm, 0.64e 1.28 mm, 1.28e2.56 mm, 2.56e5.12 mm, 5.12e10.24 mm, 10.24e 20.48 mm and 20.48e40.96 mm (Gong et al., 2003) according to the measurements of soil dust size distribution in Chinese desert regions during 1994e2001 (Zhang et al., 2003). The model initial time is at 00 UTC, and the assimilation time is at 03 UTC of the start days in April 2006.

In order to evaluate the impacts of dust radiation, assimilation as well as their composite impacts on model capability, the PM10 (particles with diameter less than 10 mm) forecast bias of CTL, 3DV and 3DV_RAD experiments are analyzed. The model bias of air temperature, wind speed, air pressure and geopotential height (GPH) of CTL, RAD and 3DV_RAD experiments are contrastively studied in the following sections. The spring 2006 is observed as one of the most frequent SDS season in the recent 10 years with the SDS-peak in April in northeast Asia (Yang et al., 2008). In the sensitivity modeling studies, the typical severe dust storms occurred in April 2006 over East Asia are discussed in the following sections.

2.2. Radiative transfer model and aerosol parameterization

3.1. The impacts on surface dust concentration predictions

The shortwave and long wave radiative transfer model (CLIRADSW and CLIRAD-LW) developed by the Climate and Radiation Branch, NASA/Goddard Space Flight Center (Chou and Suarez, 1999), for a wide variety of weather and climate applications, is introduced into GRAPES/CUACE_Dust. The CLIRAD-SW and -LW include the absorption due to water vapor, O3, O2, CO2, clouds and aerosols. The interactions among absorption and scattering by clouds, aerosols as well as Rayleigh scattering are considered. The entire solar spectrum is divided into 11 bands, from 0.175 to 10 mm and thermal infrared spectrum into 9 bands from 3.333 to 40 mm. The absorption and scattering of solar and thermal radiations due to dust are considered with this radiative transfer model in this paper. The dust refractive index data collected in Chinese deserts during the international project “studies on the origin and

Using the PM10 observation data of six stations in CMA SDS monitoring network (Wang et al., 2008), the daily averaged PM10 bias of CTL, 3DV and 3DV_RAD experiments from April 1 to 30, 2006 are calculated as displayed in Fig. 1. The station identification number STID, latitude, longitude of the six stations are listed in Table 1. At Tazhong station (Fig. 1a), two dust storms occurred on April 9e12 and April 23e25. It can be seen that the modeled PM10 bias of 3DV_RAD is smallest among the three experiments for the two dust events, which indicates effective reduction of model bias by the interaction of dust assimilation and radiation over Taklimakan desert. The modeled PM10 bias of 3DV experiment is also obviously smaller than that of CTL experiment. Four dust events were observed in Minqin on April 5, 9e11, 14e17 and 23e25 (Fig. 1b). The 3DV_RAD experiment shows the best performance

3. The impacts of dust radiation and assimilation on dust modeling

210

H. Wang, T. Niu / Atmospheric Environment 64 (2013) 208e218

Fig. 1. Modeled daily averaged PM10 bias of CTL, 3DV and 3DV_RAD experiments (mgm (f) Xilinhaote.

among the three experiments and exhibits better capability than 3DV experiment by correcting the positive bias of SDS concentrations occurred during April 9e11 and the negative bias during 14e17 and 23e25. For Wulatezhongqi station (Fig. 1d), the 3DV experiment shows a better performance than the CTL by reducing its positive and negative PM10 bias and the 3DV_RAD experiment seems to reduce the PM10 bias of 3DV experiment slightly. At Zhurihe (Fig. 1c) and Zhangbei (Fig. 1e) stations, the 3DV experiment shows the corrections of CTL experiment, but the 3DV_RAD experiment don’t show obvious improvements further. For Xilinhaote station (Fig. 1f), the 3DV experiment displays less PM10

3

) April 2006 for (a) Tazhong, (b) Minqin, (c) Zhurihe, (d) Wulatezhongqi, (e) Zhangbei, and

Table 1 Observation stations for the dust concentration comparisons. Name Tazhong Minqin Zhurihe

STID

Lat (N)

Lon (E)

Representing deserts and regions

51747 39.00 83.67 Taklimakan desert 52681 38.63 103.08 Tengger and Badanjinlin desert 53276 42.40 112.90 Middle, east of Inner Mongolia and Mongolia deserts Wulatezhongqi 53336 41.57 108.52 Sandy deserts over Sino-Mongolia border Zhangbei 53339 41.25 114.70 South of Hunshandake desert Xilinhote 54102 43.95 116.07 Hunshandake desert

H. Wang, T. Niu / Atmospheric Environment 64 (2013) 208e218 Table 2 Surface (S) and radiosonde (H) meteorology comparison stations. Name

STID

Lat (N)

Lon (E)

S/H

Representing deserts and regions

1a 1b 2 3 4 5 6 7 8 9 10 11 12 13 14

52203 52313 44354 52267 53505 54102 54511 54292 47095 47426 44292 52652 53336 47090 47590

42.82 41.53 44.90 41.95 39.08 43.95 39.80 42.88 38.08 42.17 47.93 38.93 41.57 38.20 38.27

93.52 94.66 110.12 101.07 105.38 116.07 116.47 129.47 127.82 142.78 106.98 100.43 108.52 128.60 140.90

S,H S S S S S,H S,H S,H S S H H H H H

Taklimakan Taklimakan Mongolia West Inner Mongolia Middle Inner Mongolia East Inner Mongolia East China North China Korea Japan Mongolia West Inner Mongolia Middle Inner Mongolia Korea Korea

bias than CTL experiment generally and 3DV_RAD experiment shows a better performance for the strong dust storms occurred on April 5e7 than the 3DV experiment by decreasing the negative PM10 bias.

211

Generally, it can be seen form the above discussion, the 3DV experiment shows obvious corrections of modeled 0e24 h averaged PM10 bias for the six sites, especially for the strong and severe dust storms. Comparing to the 3DV experiment, the 3DV_RAD experiment achieves more reasonable dust concentration predictions during strong dust storms at four sites. For weak dust events, the 3DV_RAD experiment doesn’t show better capability of PM10 forecasting than the 3DV experiment. 3DV and 3DV_RAD experiments show the desirable revision on the dust concentrations during the severe dust storm of April 9e11 at most sites. This dust storm is selected as the case to study the impacts on meteorological fields due to dust data assimilation and radiation in the following sections. 3.2. The impacts on meteorological fields The 3DV alone has no influence on the model meteorological field predictions because there is no linkage between the improved dust concentrations and meteorological fields. The simulation results of CTL, RAD and 3DV_RAD experiments are performed to study the impacts of dust radiation and assimilation on meteorological fields.

Fig. 2. T2m Bias within 72 h modeled by CTL, RAD, and RAD_3DV experiments of the 9 stations (aei).

212

H. Wang, T. Niu / Atmospheric Environment 64 (2013) 208e218

Considering the geography locations of East Asian deserts and the main downwind regions affected by dust storms, nine surface meteorological stations (Table 2) representing Taklimakan, Mongolia, western Inner Mongolia, middle Inner Mongolia, east Inner Mongolia, east China, northeast China, Korea, and Japan are selected to study the impacts of dust radiation and assimilation on meteorological fields. The model bias of temperature at 2 m above surface (T2m in Fig. 2), sea level pressure (Psl, Fig. 3) and surface wind speed (SWS, Fig. 4) of CTL, RAD and 3DV_RAD experiments of the nine sites are analyzed. It can be seen from Fig. 2 that the RAD experiment presents a remarkable reduction of T2m errors compared with the CTL experiment due to the inclusion of the dust absorption and scattering on solar and infrared radiation interactively in the model. The positive T2m errors at nine sites (Fig. 2aei) are almost all corrected to a certain degree during the dust storm events. At the sites located in the source and near downwind areas, such as for the stations with STID 44354 (a), 52313 (b), 52267 (c), 53505 (d), 54102 (e), 54511 (f) and 54292 (g) in Fig. 2, the T2m error corrections reach as high as 5 K or even more. The results are consistent with the previous study on the Asian dust storm of April 16e18, 2006 (Wang et al., 2010). Even for the stations located as far

as in Korea (STID: 47095) and Japan (STID: 47426), the corrections of T2m can be observed clearly for certain period of time. Furthermore, the RAD experiment also reduces the negative T2m errors from the CTL experiment during local night time (12 UTC April 9 to 00 UTC April 10) at sites 52267 (c), 53305 (d) and 54102 (e) due to the online calculation of dust infrared radiative effects in this study, which was not be found in the previous study (Wang et al., 2010). It should be attributed to the long wave radiation emitting by the dust layers near the surface, which may heat the earth surface and thereby lead to the increase of T2m during night time. This result indicates although the dust infrared radiative effect is smaller than the solar radiative effects during daytime, it is important during night time and may result in opposite impacts on modeling T2m. When the dust infrared radiation is calculated online, the negative bias of modeled T2m during night time can be corrected. Compared with the RAD experiment, the T2m predicted by the 3DV_RAD experiment don’t exhibit the obvious improvement during the dust storms at the sites except for the site 54511 (f) and 54292 (g). The assimilation scheme used in this study is done only once a day at 03 UTC in the model running. For the sites located in

Fig. 3. Psl Bias within 72 h modeled by CTL, RAD, and RAD_3DV experiments of the 9 stations (aei).

H. Wang, T. Niu / Atmospheric Environment 64 (2013) 208e218

213

Fig. 4. Surface wind speed Bias within 72 h modeled by CTL, RAD, and RAD_3DV experiments of the 9 stations (aei).

or near desert sources such as sites 44354 (a), 52313 (b), 52267 (c), 53505 (d), and 54102 (e), the modeled dust concentrations are mainly affected by the dust emission sources. It is easy for the data assimilation to revise the dust concentrations at 03 UTC, but it is hard to continuously correct the dust concentrations over the 72 h forecasting time. Consequently, the 3DV_RAD does not further improve the T2m forecast at these stations. At the stations located in downwind area in Korea and Japan (h and i), the 3DV_RAD experiment also fails to forecast T2m more reasonably than the RAD experiment. T2m by the 3DV_RAD experiment shows better results than RAD experiment during 0e36 h forecasts at the stations located in downwind area as east and northeast China (f and g), especially within 0e12 h at 54511 station (f). Fig. 3 shows that the RAD experiment revises the Psl negative bias of CTL experiment evidently, which suggests that, corresponding with T2m decreasing, the RAD experiment increased Psl at all the nine sites (aei) during almost the whole dust storm process with the more reasonable simulations. The RAD_3DV experiment doesn’t show any further improvement over the RAD experiment. This may be because Psl as a relative stable physical parameter can be not changed so easily and rapidly with the

instant changes of dust concentrations and air temperature by the SDS-DAS. An accurate and quantitative prediction of SWS is difficult for most mesoscale models and this is also a big challenge for dust models. The changes of the SWS due to dust radiative effects eventually determine whether the radiative feedbacks on dust storm are positive or negative (Wang et al., 2010). Fig. 4 displays the SWS errors for the CTL, RAD, and 3DV_RAD experiments. It is found that the RAD experiment obtains better results at almost all sites except for site 47095 (h). Comparing study of SWS errors of the CTL and RAD experiments during daytime (from 00 to 12 UTC) with that during night time (from 12 to 00 UTC), it is found that both the positive SWS errors during daytime and the negative SWS bias during night are corrected to some extent by the RAD experiment. This means that the dust solar radiation effects can decrease the SWS during daytime, suggesting negative radiative feedbacks on dust emission, while infrared effects can increase the SWS during night time, suggesting positive radiative feedbacks. The dust’s infrared effects are usually concealed by its solar radiation effects. This can lead to a misunderstanding of dust radiative feedback on dust emission during night.

214

H. Wang, T. Niu / Atmospheric Environment 64 (2013) 208e218

Generally, the bias analysis on T2m, Psl, and SWS from the CTL, RAD, and 3DV_RAD experiments shows that the dust radiation could continuously improve surface meteorological prediction with the more accurate simulations, but its improvement on surface meteorological fields are also influenced by the dynamical characteristics of these parameters themselves and the error structure. It should be also noted that the dust radiative

feedbacks may be different or opposite during day and night time over East Asia. The data assimilation is not as effective as expected on the improving of surface meteorological field simulation, probably because it is difficult for the instant dust concentration corrections to revise surface meteorological forecast distinctly in the model. On the other hand, SDS-DAS’s improvement on dust concentration is also restricted by both the

Fig. 5. Vertical profiles of temperature bias modeled by CTL, RAD and RAD_3DV experiment from surface to 100 hPa (aei).

H. Wang, T. Niu / Atmospheric Environment 64 (2013) 208e218

dust emission processes of the model and the observation data used in the SDS-DAS. 3.3. The impacts on vertical structures of meteorological fields Dust radiation and SDS-DAS can modify the meteorological fields at surface and also their vertical structures. Previous studies

215

(Perez et al., 2006; Wang et al., 2010) indicated that the impacts of dust on the atmosphere following vertical altitudes are different or even opposite. As mentioned in Section 3.2, nine radiosonde meteorological stations (Table 2) are selected representing Taklimakan, Mongolia, western Inner Mongolia, middle Inner Mongolia, east Inner Mongolia, East China, Northeast China, Korea, and Japan to study the impacts of SDS-DAS and dust radiation on the air

Fig. 6. Vertical profiles of geopotential height bias modeled by CTL, RAD and RAD_3DV experiments from surface to 100 hPa (aei).

216

H. Wang, T. Niu / Atmospheric Environment 64 (2013) 208e218

temperature, pressure and wind speed fields from surface to 100 hPa. Figs. 5e7 present the vertical profiles of air temperature, geopotential height (GPH), and wind speed bias of CTL, RAD and 3DV_RAD experiments at the nine stations (aei). At present, the regular meteorological radiosonde stations can only offer observation data at 00 and 12 UTC (08 am in the morning and 08 pm in the evening in local time) and the results of Figs. 5e7 are all at these

two moments of April 9e11. It can be seen from Fig. 5 that the positive bias in upper air temperature and negative errors in middle and low air temperature of RAD experiment are obviously less than those of the CTL experiment. This means that the dust radiation cools the low and middle air but heats the upper air at the same time, which leads to better modeling of temperature from surface to 100 hPa. The altitudes where the dust radiation effects change

Fig. 7. Vertical profiles of wind speed bias modeled by CTL, RAD and RAD_3DV experiments from surface to 100 hPa (aei).

H. Wang, T. Niu / Atmospheric Environment 64 (2013) 208e218

from heating to cooling atmosphere alter from 300 hPa to 500 hPa at different sites. A further improvement of air temperature by 3DV_RAD experiment can not be seen as expected in Fig. 5. SDSDAS introduced at 03 UTC weakens its effect on temperature after 12 h (Fig. 2a). Therefore, neither 00 nor 12 UTC is the best time to show the impacts by 3DV. If the air temperature bias at 06 UTC or 09 UTC is available, possibly better revision on temperature by the 3DV_RAD than that by RAD experiment might be seen in Fig. 5. GPH is usually used to describe the atmosphere pressure patterns at the vertical constant-pressure layers above ground for weatherprognostics and it is an important parameter describing atmospheric circulation. Fig. 6 indicates that both GPH negative errors below 600e700 hPa and the positive errors above 600e700 hPa in the RAD experiment are less than that of CTL experiment. This suggests the low level air pressure might be increased by the dust radiation, as the upper air pressure might be decreased. The more reasonable GPH and air pressure prediction from ground to 100 hPa result from the online calculating of dust radiation. It is notable that 3DV_RAD experiment shows further reduction of GPH negative bias of middle and upper air (Fig. 6aei). The changes are maintained in GPH during the modeling period because GPH is more stable compared to air temperature. The impact of dust radiation on wind is a little more complex than that on temperature and GPH. However, Fig. 7 shows the positive bias of wind speed of upper air in the RAD are less than those in the CTL experiment at sites 44292 (Fig. 7b), 52652 (Fig. 7c), 54511 (Fig. 7f), 54292 (Fig. 7g) and 47090 (Fig. 7h), indicating a decrease of wind speed by dust radiation. The negative bias of wind speed in upper air is also decreased by dust radiation at other sites such as 53336 (Fig. 7d) and 47590 (Fig. 7i). The RAD experiment provides an improvement on the wind speed forecast in upper air. The 3DV_RAD experiment shows a better performance on reducing the wind speed bias in upper air than RAD experiment at sites 44292 (Fig. 7b), 52652 (Fig. 7c), 54511 (Fig. 7f), 54292 (Fig. 7g), and 47090 (Fig. 7h). 4. Concluding remarks Dust direct radiation and data assimilation are online integrated into the mesoscale dust model GRAPES/CUACE_Dust to study their impacts on the regional circulation in East Asia during springtime dust storms with four sensitive experiments, i.e. CTL, RAD, 3DV and 3DV_RAD. Comparing study of modeled daily averaged PM10 bias of CTL, 3DV and 3DV_RAD experiments in April 2006 shows that the 3DV experiment can improve modeling dust concentration distinctly and the 3DV_RAD experiment may further corrects the modeling capacity when strong dust storms occurred. A case study on the dust storm during April 9e11 2006 show dust radiation cools the lower atmosphere and heats the upper atmosphere, which leads to an increase of air pressure in lower atmosphere and decrease of air pressure in upper atmosphere. The impacts of dust radiation also lead to the continuous improvement in the surface air temperature and pressure prediction during the 72 h modeling interval, and the surface wind speed is also corrected to some extent. It is noteworthy that dust radiation may decrease the surface temperature during daytime while increase it during night time at some stations, which is associated with the surface wind speed drop at daytime and rise at night time. This indicates a negative feedback during daytime and positive feedback during night on dust emission. Based on the RAD experiment, the 3DV_RAD experiment further improves the air temperature and pressure simulations over 12 h after the dust data assimilation at certain stations. 3DV_RAD experiment seems to be more efficiency in enhancing model prediction ability in upper air meteorological fields than these in lower and surface layers.

217

More detailed study and cases are needed to further illustrate the different radiative feedbacks of dust aerosols between day and night, which requires observational data at a specific time, i.e. at noon, for the comparisons. This kind of data could be also used to further evaluate the influence of 3DV_RAD on the model forecasting ability. The responses of different meteorological fields to the interaction of dust radiation and assimilation should be studied based on their dynamic and physical characteristics in future work. Acknowledgments The authors wish to thank for the financial supports from the National Basic Research Program (973) (2011CB403404), the National Natural Science Foundation of China (Grant Nos. 41130104), Ministry of Science and Technology of China (2010DFA22770), the open project of Key Laboratory of Semi-Arid Climate Change, Lanzhou University, Ministry of Education, China and CAMS key project (2009z001). References Ahn, H.J., Park, S.U., Chang, L.S., 2007. Effect of direct radiative forcing of Asian dust on the meteorological fields in East Asia during an Asian dust event period. J. Appl. Meteorol. Climatol. 46, 1655e1681. Barker, D.M., Huang, W., Guo, Y.R., Bourgeois, A.J., Xiao, Q.N., 2004. A threedimensional variational data assimilation system for MM5: Implementation and initial results. Mon. Weather Rev. 132, 197e914. Chou, M.D., Suarez, M.J., 1999. A solar radiation parameterization for atmospheric studies. NASA Tech. Memo 15, 10460. http://climate.gsfc.nasa.gov/chou. Fisher, M., 2001. Assimilation Techniques(1), 3D-Var, ECMWF Meteorological Training Course Lecture Series, 220 pp. Gong, S.L., Zhang, X.Y., Zhao, T.L., McKendry, I.G., Jaffe, D.A., Lu, N.M., 2003. Characterization of soil dust aerosol in China and its transport and distribution during 2001 ACE-Asia: 2. Model simulation and validation. J. Geophys. Res. 108 (D9), 4262. http://dx.doi.org/10.1029/2002JD002633. Gong, S.L., Zhang, X.Y., 2008. CUACE/dust e an integrated system of observation and modeling systems for operational dust forecasting in Asia. Atmos. Chem. Phys. 8, 2,333e82,340. Haywood, J.M., Allan, R.P., Culverwell, I., Slingo, T., Milton, S., Edwards, J., Clerbaux, N., 2005. Can desert dust explain the outgoing longwave radiation anomaly over the Sahara during July 2003? J. Geophys. Res. 110 (D05105). http://dx.doi.org/10.1029/2004JD005232 Heinold, B., Tegen, I., Schepanski, K., Hellmuth, O., 2008. Dust radiative feedback on Saharan boundary layer dynamics and dust mobilization. Geophys. Res. Lett. 35, L20817. http://dx.doi.org/10.1029/2008GL035319. Hollingsworth, A., Lonnberg, P., 1986. The statistical structure of short-range forecast errors as determined from radio sonde data. Part I: the wind field. Tellus, 111e136. Hu, X.Q., Lu, N.M., Niu, T., Zhang, P., 2008. Operational retrieval of Asian dust storm from FY-2C Geostationary Meteorological Satellite and its application to real time forecast in Asia. Atmos. Chem. Phys. 8, 1649e1659. Kotamarthi, V.R., Carmichael, G.R., 1993. A modeling study of the long range transport of Kosa using particles trajectory methods. Tellus Ser. B 45, 426e441. Liu, Y., Zhou, M.Y., 1997. A numerical simulation of dust storm and its long rang transport process. J. Nanjing Inst. Meteorol. 20, 511e517. Niu, T., Gong, S.L., Zhu, G.F., Liu, H.L., Hu, X.Q., Zhou, C.H., Wang, Y.Q., 2008. Data assimilation of dust aerosol observations for CUACE/Dust forecasting system. Atmos. Chem. Phys. 8, 3473e3482. Parrish, D.F., Derber, J.D., 1992. The national meteorological center spectral statistical interpolation analysis system. Mon. Weather. Rev. 120, 1747e1763. Perez, C., Nickovic, S., Pejanovic, G., Maldasano, J.M., Ozsoy, E., 2006. Interactive dust-radiation modeling: a step to improve weather forecast. J. Geophys. Res. 111 (D16206). http://dx.doi.org/10.1029/2005JD006717 Shao, Y., 2001. A model for mineral dust emission. J. Geophys. Res. 106, 20,239e20,254. Shao, Y., 2004. Simplification of a dust emission scheme and comparison with data. J. Geophys. Res. 109, D10202. http://dx.doi.org/10.1029/2003JD004372. Uno, I., Hiroyasu, A., Emori, S., Kinoshita, K., Matsui, I., Sugimoto, N., 2001. TransPacific yellow sand transport observed in April 1998: a numerical simulation. J. Geophys. Res. 106 (D22), 18,331e18,344. Uno, I., Satake, S., Gregory, R.C., Tang, Y.H., Wang, Z.F., Takemura, T., Sugimoto, N., Shimizu, A., Murayama, T., Cahill, T.A., Cliff, S., Uematsu, M., Ohta, S., Quin, P.K., Bates, T.S., 2004. Numerical study of Asian dust transport during the springtime of 2001 simulated with the Chemical Weather Forecasting System (CFORS) model. J. Geophys. Res. 109, D19S24. http://dx.doi.org/10.1029/2003JD004222. Uno, I., Wang, Z., Chiba, M., Chun, Y.S., Gong, S.L., Hara, Y., Jung, E., Lee, S.-S., Liu, M., Mikami, M., Music, S., Nickovic, S., Satake, S., Shao, Y., Song, Z., Sugimoto, N., Tanaka, T., Westphal, D.L., 2006. Dust model intercomparison (DMIP) study over Asia: overview. J. Geophys. Res. 111, D12213. http://dx.doi.org/10.1029/ 2005JD006575.

218

H. Wang, T. Niu / Atmospheric Environment 64 (2013) 208e218

Wang, H., Shi, G.Y., Li, W., Wang, B., 2006. The impacts of optical properties on radiative forcing due to dust aerosol. Adv. Atmos. Sci. 23 (2), 431e441. Wang, H., Gong, S.L., Zhang, H.L., Chen, Y., Shen, X.S., Chen, D.H., Xue, J.S., Shen, Y.F., Wu, X.J., Jin, Z.Y., 2009. A new-generation sand and dust storm forecasting system GRAPES/CUACE_Dust: model development, verification and numerical simulation. Chin. Sci. Bull.. http://dx.doi.org/10.1007/s11434-009-0481-z Wang, H., Zhang, X.Y., Gong, S.L., Chen, Y., Shi, G.Y., Li, W., 2010. Radiative feedback of dust aerosols on the East Asian dust storms. J. Geophys. Res. 115, D23214. http://dx.doi.org/10.1029/2009JD013430. Wang, Y.Q., Zhang, X.Y., Gong, S.L., Zhou, C.H., Hu, X.Q., Liu, H.L., Niu, T., Yang, Y.Q., 2008. Surface observation of sand and dust storm in east Asia and its application in CUACE/Dust forecasting system. Atmos. Chem. Phys. 8, 545e553. Wang, Z.F., Hiromasa, U., Huang, M.Y., 2000. A deflation module for use in modeling long-rang transport of yellow sand over East Asia. J. Geophys. Res. 105 (D22), 26,947e26,959.

Xiao, H., Carmichael, G.R., Durchenwald, J., 1997. Long-range transport of Sox and dust in East Asia during the PEM B experiment. J. Geophys. Res. 102, 28,589e28,612. Yang, Y.Q., Hou, Q., Zhou, C.H., Liu, H.L., Wang, Y.Q., Niu, T., 2008. A study on sand/ dust storms over northeast Asia and associated large-scale circulations in spring 2006. Atmos. Chem. Phys. 8, 25e33. Zhang, H., Xue, J.S., Zhuang, S.Y., Zhu, G.F., Zhu, Z.S., 2004. Idea experiments of GRAPES three dimensional variational data assimilation system. ACTA. Meteor. Sin. 62, 31e41. Zhang, X.Y., Gong, S.L., Zhao, T.L., et al., 2003. Sources of Asian dust and role of climate change versus desertification Asian dust emission. J. Geophys. Res. 30 (24), 2272. http://dx.doi.org/10.1029/2003 GL018206,20. Zhou, C.H., Gong, S.L., Zhang, X.Y., Wang, Y.Q., Niu, T., Liu, H.L., Zhao, T.L., Yang, Y.Q., Hou, Q., 2008. Development and evaluation of an operational SDS forecasting system for east Asia: CUACE/Dust. Atmos. Chem. Phys. 8, 787e798. Zhuang, S.Y., Xue, J.S., Zhu, G.F., Zhao, J., Zhu, Z.S., 2005. GRAPES global 3D-Var system-Basic scheme design and single observation test. Chin. J. Atmos. Sci. 29, 872e884.