Direct radiative forcing and atmospheric absorption by boundary layer aerosols in the southeastern US: model estimates on the basis of new observations

In an eﬀort to reduce uncertainties in the quantiﬁcation of aerosol direct radiative forcing (ADRF) in the southeastern United States (US), a ﬁeld column experiment was conducted to measure aerosol radiative properties and eﬀects at Mt. Mitchell, North Carolina, and at an adjacent valley site. The experimental period was from June 1995 to mid-December 1995. The aerosol optical properties (single scattering albedo and asymmetry factor) needed to compute ADRF were obtained on the basis of a procedure involving a Mie code and a radiative transfer code in conjunction with the retrieved aerosol size distribution, aerosol optical depth, and diﬀuse-to-direct solar irradiance ratio. The regional values of ADRF at the surface and top of atmosphere (TOA), and atmospheric aerosol absorption are derived using the obtained aerosol optical properties as inputs to the column radiation model (CRM) of the community climate model (CCM3). The cloud-free instantaneous TOA ADRFs for highly polluted (HP), marine (M) and continental (C) air masses range from 20.3 to (cid:1) 24.8, 1.3 to (cid:1) 10.4, and 1.9 to (cid:1) 13.4 W m (cid:1) 2 , respectively. The mean cloud-free 24-h ADRFs at the TOA (at the surface) for HP, M, and C air masses are estimated to be (cid:1) 8 (cid:2) 4 ( (cid:1) 33 (cid:2) 16), (cid:1) 7 (cid:2) 4 ( (cid:1) 13 (cid:2) 8), and (cid:1) 0.14 (cid:2) 0.05 ( (cid:1) 8 (cid:2) 3) W m (cid:1) 2 , respectively. On the assumption that the fractional coverage of clouds is 0.61, the annual mean ADRFs at the TOA and the surface are (cid:1) 2 (cid:2) 1, and (cid:1) 7 (cid:2) 2 Wm (cid:1) 2 , respectively. This also implies that aerosols currently heat the atmosphere over the southeastern US by 5 (cid:2) 3Wm (cid:1) 2 on annual timescales due to the aerosol absorption in the troposphere. # 2001 Elsevier Science Ltd. All rights reserved.


Introduction
Eastern China, south central Europe and the eastern United States (US) are regions where the radiative forcing of anthropogenic sulfate aerosols supersedes that of anthropogenic greenhouse gases, manifesting itself in terms of cooling of the surface-air temperature (IPCC, 1995;Kiehl and Briegleb, 1993). Saxena et al. (1997) and Saxena and Yu (1998) have verified this model prediction in the southeastern US by finding the existence of aerosol cooling in the southeastern US in the temperature records of the surface air during 1949-1994. The direct aerosol effect refers to scattering and absorption of radiation by the aerosol particles themselves (Charlson et al., 1991;Penner et al., 1994;Kiehl and Briegleb, 1993;IPCC, 1995;Saxena and Yu, 1998). The conclusion of IPCC (1995) is that global mean direct forcing resulting from anthropogenic sulfate may range from À0.25 to À0.9 W m À2 , with substantial uncertainty; and biomass burning aerosol forcing may range from À0.05 to À0.6 W m À2 . The low confidence in the estimates of aerosol radiative forcing is due to highly non-uniform compositional, spatial, and temporal distributions of tropospheric aerosol on regional scales owing to their heterogeneous sources and short lifetimes (Schwartz and Andreae, 1996). The shortwave absorption in Earth's atmosphere due to the absorbing aerosols is also one of the least quantified properties of the climate system (Zender et al., 1997). The comparison of signals of regional aerosol-induced radiative forcing in eastern China and the southeastern US indicates that the high concentration of absorbing aerosol over eastern China was one of the major reasons for the warming trend during 1951-1994.
To reduce the uncertainty in the climate effect of tropospheric aerosols in the industrial pollution plumes that are transported from the east coast of the US over the Atlantic Ocean, the tropospheric aerosol radiative forcing observational experiment (TARFOX) was carried out on the eastern coast during July 10-31, 1996 Redemann et al., 2000). In two case studies, Redemann et al. (2000) found that the instantaneous shortwave aerosol radiative forcing was of the order of À36 W m À2 at the top of the atmosphere (TOA) and about À56 W m À2 at the surface. However, the TARFOX datasets do not overlap geographically with the southeastern US nor is it of sufficient length to estimate the regional annual average aerosol direct radiative forcing (ADRF). The objectives of this study are to use tropospheric aerosol measurements to estimate the instantaneous and annual mean ADRF at the surface and TOA, and aerosol absorption over the southeastern US with column radiation model (CRM) of the NCAR (National Center for Atmospheric Research) community climate model (CCM3) (Kiehl et al., 1998;Briegleb, 1992). The atmospheric shortwave absorption due to the aerosol layer was determined as the difference between the TOA ADRF and surface ADRF. Since the ADRF is not determined by surface aerosols only but by total aerosols within the atmospheric column, it is very useful to obtain the average aerosol radiative properties for the atmospheric column, for example, through a column experiment. One of our field sites (Mt. Mitchell) is at a remote and elevated location that is influenced by air masses arriving from marine, continental and polluted sectors. It stays frequently in the free troposphere due to decoupling with the surrounding inversion layer. The data sets are considered to be regionally representative of the southeastern US (Yu et al., 2000;Wenny et al., 1998). We also compare our in situ results with the estimates from other global models for the aerosol radiative forcing in the southeastern US.

Retrieval of aerosol radiative properties in the southeastern US
A detailed description of the instrumentation, methodology, data quality assurance and quality control of the data has been given by Wenny et al. (1998) and Yu et al. (2000). Here a brief description is presented. The research sites are a mountain top station located on the peak of Mt. Gibbes (35.788N, 82.298W, 2006 m MSL) Yu et al., 2000). The two sites are separated horizontally by 10 km and vertically by 1 km. The experimental period was from June 1995 to mid-December 1995. The aerosol optical depths (AOD) at the three operational wavelengths (415, 500 and 673 nm) are determined on the basis of the direct components of solar irradiance measured by the multi-filter rotating shadowband radiometer (MFRSR) (Yu et al., 2000). Table 1 lists the mean total AOD and standard deviation as measured from the valley and mountain sites and mean 1-km layer AOD between the two sites as a function of air mass type for cloud-free days during the experimental period. The clear-sky intervals were determined by a full-sky imaging camera. Total 34 cloud-free days of measuremental data were available because of the bad weather (cloud and rain) situation and instrumental problems during the field experiment period. Fig. 1 shows the time series of AOD at 500 nm for cloud-free periods during the experiment. The air masses are classified (Yu et al., 2000) as highly polluted (HP), marine (M) and continental (C) based on the SO 2 and NO x emission inventory of the US. Environmental Protection Agency. It is clear from Table 1 that the HP air masses exhibit the largest average AOD, compared to M and C air masses at the three operational wavelengths. The mean total AOD at 500 nm at the valley site was 0.68 AE 0.33, 0.29 AE 0.12, and 0.10 AE 0.04 for HP, M, and C air masses, respectively. As mentioned by Yu et al. (2000), there are two reasons that the mean AOD of marine air masses is higher than that of continental air masses at the research site. First, the marine classification for the sampling site does not imply pure marine air, but rather a modified marine air caused by additional influence from continental and polluted sources between the ocean and the site (the shortest distance between the site and the ocean is about 290 km). Second, the AOD is typically very low during winter months (Peterson et al., 1981). In this study period, the majority of the marine air mass cases occurred during September and early October and the majority of continental air mass cases occurred during late October and November. Overall, the ratios of mean 1-km layer AOD to total mean AOD from the valley site for HP air mass were 73%, 73% and 79% for 415, 500 and 673 nm, respectively. This indicates that the major portion of atmospheric aerosols at our sites, which make an important contribution to the AOD, is located in the lowest 1-km boundary layer of the troposphere. This is in reasonable agreement with that of Hegg et al. (1997), who found that AOD for the lowest 4 km of the troposphere constituted over 90% of the total column AOD off the mid-Atlantic coast of the US. Table 2 lists the experimental results of AOD, and retrieved columnar lognormal size distribution (number concentration (N), number median radius (geometric mean radius, r g ) and geometric standard deviation ðs g Þ) by the search-graph method. Table 2 also lists the aerosol radiative properties (imaginary part, asymmetry factor and singe scattering albedo ð$Þ) and ground albedo obtained by using a procedure involving a Mie code and a radiative transfer code in conjunction with the retrieved aerosol size distribution, AOD, and diffusedirect ratio (DDR) (Yu et al., 2000). A detailed description of the search-graph method and the procedure has been given by Yu et al. (2000). Table 2 shows that N, r g and s g are in the range of 12 to 6000 cm À3 , 0.03 to 0.54 mm and 1.12 to 2.95, respectively. The asymmetry factors at 500 nm are in the range of 0.61 to 0.77. $ was in the range of 0.72 to 0.97 and the imaginary part of refractive index was in the range of 0.005 to 0.051. Reported values of $ in the scientific literature show $ to be 1.0 for sulfate and pure marine aerosols and range from 0.5 to 0.9 for desert dust and Table 1 Mean total optical depth, and their standard deviation at 415, 500 and 673 nm as measured at the valley and mountain sites. The data for the intervening layer were obtained from the difference between the two sites. The sites were influenced by highly polluted (HP), continental (C) and marine (M) , 1995). Wenny et al. (1998) reported that $ at 312 nm varied from 0.75 to 0.93 for aerosols at our research site. Russell et al. (1999) found that the estimated column single scattering albedos for the ambient TARFOX aerosol was $0.9 at 550 nm. The values of $ from this study are consistent with these values given by other authors. Lacis and Mishchenko (1995) showed that the asymmetry factor of soot and large desert aerosols was about 0.9, but for sulfate, marine and smaller dust aerosols the asymmetry factor ranges from 0.65 to 0.8. The values of asymmetry factor reported here are close to those for sulfate, marine and smaller dust aerosols. Since the size distribution was inferred from the total AOD, the aerosol radiative properties derived by this method represent a weighted average over the entire column. Yu et al. (2000) found that ground albedo is 0.19 AE 0.10 at our research site. The mean ground albedo (0.19) was used in the calculation of this study.

Calculation of aerosol direct radiative forcing and aerosol absorption
The NCAR CCM3 CRM (Kiehl et al., 1998;Briegleb, 1992) is used to estimate the TOA and surface ADRF and aerosol absorption for the measurement sites. The CCM3 CRM uses an d-Eddington approximation with 19 spectral intervals (7 for O 3 , 2 for the visible, 7 for H 2 O, and 3 for CO 2 ) spanning the solar spectrum from 0.2 to 5.0 mm. A detailed description of all the physical and numerical methods used in CCM3 CRM is given in Kiehl et al. (1998) and Briegleb (1992). The atmospheric initial conditions of Kiehl and Briegleb (1993) for midlatitude summer vertical profiles of temperature, pressure, ozone and H 2 O vapor mixing ratios were used in the CRM. The mass extinction coefficient, single scattering albedo and asymmetry factor at 19 spectral interval wavelengths calculated by a Mie code with the retrieved columnar size distribution and refractive index, and ground albedo (0.19) were input into the CRM. The density of aerosol particles was assumed to be 1.86 g cm À3 . The ADRF is obtained as the difference in shortwave net radiative fluxes at the TOA (or at the surface) between CRM simulations with and without aerosol mass loading. We also infer the atmospheric absorption due to boundary layer aerosols as the difference between the TOA ADRF and surface ADRF. Table 2 lists the examples of the instantaneous ADRF for each cloud-free time during the experiment. The zenith angle, air mass types, AOD at the three wavelengths, size distribution, retrieved aerosol radiative properties and refractive indices are listed in Table 2. Hansen et al. (1997) indicated that the ADRF is very sensitive to single scattering albedo (or imaginary part of refractive index). Unfortunately, it is very difficult to retrieve mean single scattering albedo for the whole atmospheric column. As mentioned by Yu et al. (2000), our DDR method can only have solutions for the cases whose AOD at 500 nm was between 0.1 and 0.3. The averages of the imaginary part of refractive index for September, October and November are À0.017 AE 0.009, À0.027 AE 0.021, À0.050 AE 0.001 on the basis of available results in Table 2. The values of À0.017, À0.027 and À0.050 were used for cases for which imaginary part of refractive index is not available, between July and September, in October and in November, respectively, as indicated in Table 2. Since entirely cloud-free days were very rare, we assume that the aerosol radiative properties obtained at one cloud-free time were constant to study the diurnal variation of instantaneous cloudfree ADRF. Fig. 2 shows the diurnal variations of instantaneous surface and TOA ADRF for C, M, and HP air masses, calculated on the basis of measurements at Julian Day 321 (11/17/1995, 10:27), 272 (9/29/1995, 8:59) and 227 (8/15/1995, 9:50) (see Table 2). The diurnal behavior of the TOA aerosol direct radiative forcing (ADRF) is determined by the changing ratio of aerosol forward-scattered radiation (scattering angle less than 908) to aerosol up-scattered radiation (scattering up to space rather than down to surface) as the zenith angle changes (Wiscombe and Grams, 1976). When the sun is closer to the horizon much (up to half ) of the forwardscattered sunlight corresponds to up-scattered radiation that is reflected back to space and thus cools the climate system (hence ADRF minima (more negative) at twilight). When the sun is overhead, the forwardscattered solar radiation coincides with the downscattered radiation, i.e., it penetrates the aerosol layer where it may be absorbed by the absorbing aerosols or by the surface, warming the climate system (hence ADRF maxima (less negative) at noon). These results are in agreement with those of Russell et al. (1997). Fig. 2 also indicates that the TOA ADRF is much smaller than surface ADRF during the daytime (from sunrise to sunset), especially for highly absorbing aerosol (small single scattering albedo) and at the midday peak of incident sunlight. This is consistent with the finding of Satheesh and Ramananthan (2000) who reported the large difference in tropical aerosol forcing at TOA and the Earth's surface due largely to solar absorption by carbonaceous aerosol. Table 2 shows that the instantaneous TOA ADRF ranges from +20.3 to -24.8, +1.3 to À10.4, and +1.9 to -13.4 W m À2 for HP, M, and C air masses, respectively. The instantaneous surface ADRF ranges from -55.3 to À156.5, À16.7 to À24.3, and À9.2 to À30.2 W m À2 for HP, M, and C air masses, respectively. The atmospheric aerosol absorption ranges from 5.3 to 176.8 W m 2 , and is always positive. Clearly, the difference between ADRF at surface and TOA are very large, especially for highly absorbing aerosols. The inclusion of absorbing aerosols makes the ADRF at TOA less negative while making the ADRF at the surface more negative. This is in agreement with the results of Hignett et al. (1999) . Fig. 3 shows the surface and TOA ADRF, and absorption as functions of aerosol optical depth and single scattering albedo. As can be seen, the surface ADRF was generally more sensitive to AOD and single scattering albedo than TOA ADRF. Note that the zenith angle (or local time) is also an important factor for ADRF as indicated in Fig. 2. Figs. 3a and b also show that both surface and TOA ADRFs are more sensitive to the AOD than single scattering albedo on the basis of the correlation coefficients. The small single scattering albedo (more absorbing) and large optical depth will greatly increase the heating rate of the atmosphere due to aerosols as indicated in Fig. 3 and Fig. 2. The instantaneous aerosol direct radiative forcing (ADRF) at the top of the atmosphere (TOA) and the Sfc (surface) as a function of local time for three cases on the basis of the aerosol measurements on 11/17/1995 10:27 (continental air mass) (A), 9/29/1995 8:59 (marine air mass) (B), and 8/15/ 1995 9:50 (highly polluted air mass) (C). The calculations assume that the aerosol radiative properties measured are constant for the whole day. 'Sfc', and 'TOA' represent instantaneous aerosol forcing at the surface and at the TOA, respectively. Table 2. Fig. 4 shows the statistical summaries of ranges of the AOD, asymmetry factor, single scattering albedo, surface and TOA ADRF, and absorption with their frequencies of occurrences on the basis of Table 2. As can be seen, the most frequencies of occurrences for AOD, surface ADRF, TOA ADRF and absorption are in the ranges of 0.04 to 0.30, À9.2 to -50 W m À2 , 0 to -10 W m À2 , and +5 to +50 W m À2 , respectively.
A sensitivity test was performed by calculating the ADRF difference between the case with our aerosol radiative properties and that with the sulfate optical properties of Kiehl and Briegleb (1993), keeping the AOD same. It is found that the instantaneous surface ADRF is À27.2 W m À2 on the basis of measurement of aerosol radiative properties at 10:58 AM, 11/17/1995, while the surface ADRF is À10.0 W m À2 on the basis of sulfate aerosol radiative properties of Kiehl and Briegleb (1993). The surface ADRF from this study is much higher. This is reasonable because the aerosol of this study is total aerosol and the imaginary part of refractive index is also considered.
To estimate the 24 h and annual average ADRF for different air mass types for the whole year, it is ideal to Fig. 3. The aerosol direct radiative forcing (ADRF) at the TOA, surface (Sfc) and atmospheric absorption due to aerosol as functions of aerosol optical depth (a) and single scattering albedo (b) on the basis of the results of Table 2.  measure the aerosol radiative properties and calculate the aerosol forcing for each air mass and then average them during the cloud-free period for the whole year. Unfortunately, the information is very scarce. In this study, we select the typical aerosol radiative properties (mass extinction coefficient, single scattering albedo, asymmetry factor) of 321 (11/17/1995, 10:27 EST, C air mass), 272 (9/29/1995, 8:59 EST, M air mass) and 227 (8/15/1995, 9:50 EST, HP air mass) as representatives for the C, M, and HP air masses, respectively. These three cases were selected because their AOD was close to the mean AOD of corresponding air masses and most of continental, marine, and polluted air masses occurred during winter, fall and summer seasons, respectively as shown in Tables 1 and 2. Then, the mean optical depths in Table 1 for three air mass types are used in the calculation. It was found that the mean cloud-free 24-h ADRF at the TOA (and at the surface) for HP, M, and C air masses was estimated to be À8 AE 4 (À33 AE 16), À7 AE 4 (À13 AE 8), and -0.14 AE 0.05 (À8 AE 3) W m À2 , respectively, as listed in Table 3. To assess the annual average ADRF in the southeastern US, the relative contributions of three air masses are needed. On the basis of back trajectory analysis of air masses from June 1996 to October 1996, and from March 1997 to June 1997, Bahrmann and Saxena (1998) found that the percentages of air masses influencing our research site were 43.2, 22.4 and 34.4% for C, M, and HP masses, respectively. Using these percentages, the annual cloudfree, 24-h mean ADRF at the TOA (and at the surface) was estimated to be -5 AE 3 (À17 AE 6) W m À2 . If the mean fraction of the area with clear sky condition in the southeastern US is assumed to be 0.39, which is the globally average clear sky condition (Charlson et al., 1991), the mean ADRF at the TOA (and at the surface) will be -3.2 AE 1.5 (13 AE 6), À2.7 AE 1.5 (À5 AE 3), and -0.05 AE 0.02 (À3 AE 1) W m À2 for HP, M, and C air masses, respectively. The annual mean ADRF at the TOA (and at the surface) is À2 AE 1 (À7 AE 2) W m À2 . Note that we assume that the ADRF is zero when clouds are present. So our estimate of annual mean ADRF is a lower bound. The annual mean atmospheric absorption due to aerosol layer will be 5 AE 3 W m 2 . Table 3 lists the comparison of our estimates of ADRF with those of other investigators for the southeastern US or eastern US. The methods used by each author are also listed in Table 3. As can be seen, box models, GCM and column radiation models have been used to estimate the ADRF on the basis of the actual measurements or chemical simulation. The estimates of ADRF in the eastern US (or southeastern US) have a large variation because of different assumptions in each model and relative uncertainty of aerosol radiative properties. Our annual estimated TOA ADRF (À2 AE 1 W m À2 ) was close to those of Boucher and Anderson (1995), Schult et al. (1997) and Haywood and Ramaswamy (1998), but a little lower than those of Bergstrom and Russell (1999), Kiehl and Rodhe (1995). Bergstrom and Russell (1999) show that cloud-free, 24-h average ADRF is À9 W m À2 near the eastern US coast in summer. Since our calculations are for total aerosols (anthropogenic+natural) and for whole aerosol components (sulfate+organic), our estimated ADRF is consistent with the reported values. The major portion of atmospheric aerosol at our sites is located in the lowest 1-km boundary layer of the troposphere where it exerts a negligible longwave forcing. Thus we make the approximation that ADRF is due to shortwave forcing alone. It is noteworthy that although there was large negative ADRF at the surface, the atmospheric heating rate due to aerosols is large, especially for high polluted absorbing aerosols. As pointed out by Ackerman et al. (2000) and Hansen et al. (1997), absorbing aerosols were more effective than their radiative forcing due to their cloud-burning effects. Further observations on absorbing aerosols are obviously needed.

Conclusion
Using a column radiation model, the instantaneous and cloud-free 24-h average ADRF at both TOA and surface for C, M, and HP air masses in the southeastern US was estimated on the basis of measurements of aerosol radiative properties. The results show that the annual regional cloud-free mean ADRF at the TOA (and at the surface) is À5 AE 3 (-17 AE 6) W m À2 . Considering the effect of cloud fraction, we estimate that the annual regional mean ADRF at the TOA (and at the surface) is À2 AE 1 (-7 AE 2) W m À2 . This estimate of ADRF in the southeastern US is close to those of other investigators (see Table 3) when they considered both sulfate and black carbon. Since our study considers the total aerosol (anthropogenic+natural) instead of anthropogenic and whole aerosol (sulfate+organics) instead of sulfate, our estimated ADRF provide an upper bound on the direct effect of anthropogenic aerosols. This study confirms the existence of a cooling effect (negative forcing) due to the direct effect of aerosol at the surface in the southeastern US as indicated by Saxena and Yu (1998). However, the atmosphere of the troposphere above the ground is still heated with annual mean rate of 5 AE 3 W m À2 due to the aerosol layer. Since the aerosol field and composition are inhomogeneous in time and space, and our study also indicates the effect of long-range transport of pollution from other regions, the complexity of this aerosol puzzle seems to grow with each new study as pointed out by Kiehl (1999). Continuity in the long-term monitoring of the aerosol properties, especially for absorbing aerosol, is still crucial in improving the understanding of role of aerosols in the climate change.