Vertical dependence of black carbon, sulphate and biomass burning aerosol radiative forcing

Transcription

1 GEOPHYSICAL RESEARCH LETTERS, VOL. 38,, doi: /2011gl049697, 2011 Vertical dependence of black carbon, sulphate and biomass burning aerosol radiative forcing Bjørn H. Samset 1 and Gunnar Myhre 1 Received 15 September 2011; revised 1 November 2011; accepted 6 November 2011; published 16 December [1] A global radiative transfer model is used to calculate the vertical profile of shortwave radiative forcing from a prescribed amount of aerosols. We study black carbon (BC), sulphate (SO 4 ) and a black and organic carbon mixture typical of biomass burning (BIO), by prescribing aerosol burdens in layers between 1000 hpa and 20 hpa and calculating the resulting direct radiative forcing divided by the burden (NDRF). We find a strong sensitivity in the NDRF for BC with altitude, with a tenfold increase between BC close to the surface and the lower part of the stratosphere. Clouds are a major contributor to this dependence with altitude, but other factors also contribute. We break down and explain the different physical contributors to this strong sensitivity. The results show a modest regional dependence of the altitudinal dependence of BC NDRF between industrial regions, while for regions with properties deviating from the global mean NDRF variability is significant. Variations due to seasons and interannual changes in cloud conditions are found to be small. We explore the effect that large altitudinal variation in NDRF may have on model estimates of BC radiative forcing when vertical aerosol distributions are insufficiently constrained, and discuss possible applications of the present results for reducing inter-model differences. Citation: Samset,B.H.,andG.Myhre(2011),Verticaldependence of black carbon, sulphate and biomass burning aerosol radiative forcing, Geophys. Res. Lett., 38,, doi: / 2011GL Introduction [2] Atmospheric aerosols affect the energy balance of the Earth through scattering and absorption of solar and thermal radiation. Important examples are black carbon (BC), which both absorbs and scatters solar radiation, and sulphate compounds (SO 4 ) which only scatter. Their total effect on the top-of-atmosphere radiation balance is however influenced by external factors such as surface albedo, absolute humidity, the presence of other aerosols, and of clouds above or below the aerosol layer [Haywood and Shine, 1997]. It is thought that such factors are major contributors to the variation in total anthropogenic radiative forcing from aerosols found by different climate models [Schulz et al., 2006]. [3] The radiative forcing (RF) induced by a given atmospheric load of aerosols also varies with altitude. However aerosol vertical profiles are not well measured, and their descriptions vary significantly among global aerosol climate models [Textor et al., 2006, 2007; B. Koffi et al. Application 1 Center for International Climate and Environmental Research - Oslo (CICERO), Oslo, Norway. Copyright 2011 by the American Geophysical Union /11/2011GL of the CALIOP Layer Product to evaluate the vertical distribution of aerosols estimated by global models. Part 1: AeroCom phase I results, submitted to Journal of Geophysical Research, 2011]. Recently the climate impact of black carbon as function of altitude has received increased attention, e.g., in the work by Zarzycki and Bond [2010] which uses a column model, and that by Ban-Weiss et al. [2011] which studies dynamical climate response in a full GCM. State-of-the-art model descriptions of aerosol vertical distributions are summarized and discussed by Schwarz et al. [2010]. Radiative forcing uncertainties associated with sulphate profiles is discussed by Goto et al. [2011]. The present study aims to better understand the factors that contribute to changes in normalized direct radiative forcing (NDRF) with altitude, for a set of absorbing and scattering aerosol species. [4] NDRF is a measure of the amount of energy scattered or absorbed by the atmosphere per unit weight. While it is not always a good predictor of surface temperature change, in particular for absorbing aerosols such as BC [Hansen et al., 2005; Lohmann et al., 2010], it can be an excellent diagnostic for understanding global RF results from threedimensional aerosol models. The global and regional NDRF curves and vertical sensitivity figures given here can be taken together with the vertical aerosol distributions from models, and used to determine what fraction of model differences they cause. 2. Method and Inputs [5] In this section we present the models and methods used, and introduce a measure for NDRF vertical sensitivity. We use a state of the art 3-D four band, eight stream radiative transfer model [Myhre et al., 2007] based on the discrete ordinate method [Stamnes et al., 1988], in T42 (2.8 ) horizontal resolution and 3 hour time steps. For the annual mean results in the present study, we run the model for one day per calendar month to save computation time. We have verified that the uncertainty introduced by this approximation is 1% for the final NDRF profiles. A background of aerosols (notably sulphate, organic and black carbon from fossil fuel and biomass burning, mineral dust and sea salt) are provided by the OsloCTM2 global aerosol and chemical transport model [Myhre et al., 2009], using emissions based on those of Lamarque et al. [2010]. OsloCTM2 is among the global aerosol models with vertical profiles closest to the CALIPSO satellite retrieved aerosol extinction profiles (Koffi et al., submitted manuscript, 2011) and close to the mean of the global aerosol models in terms of BC vertical profiles in remote regions [Schwarz et al., 2010]. Meteorological data, including clouds, were generated through the Integrated Forecast System at the ECMWS. For details, see Myhre et al. [2009]. The main study uses year of5

2 Figure 1. Annual mean BC aerosol burden (in mg/m 2 ) for the background fields used, and regions selected for the regional study (boxes). meteorology, while years were used to study cloud sensitivity. [6] Three aerosol species were studied, each of which have distinct parametrizations in the model. In the following BC refers to black carbon particles from fossil fuel burning, SO 4 refers to sulphate compounds resulting from SO 2 emissions, and BIO refers to a mixture of black and organic carbon from biomass burning. The aerosol optical properties are calculated with Mie theory. For BC we increase the total absorption by 50% for aged (hydrophilic) BC to take into account internal mixing with other aerosols as suggested by Bond et al. [2006], whereas for fresh (hydrophobic) BC we use the simulated absorption coefficient of around 7.5 m 2 /g (at 550 nm) in accordance with measurements [Bond and Bergstrom, 2006]. Biomass burning aerosols are typically a mixture of absorbing black carbon and strongly scattering organic carbon compounds. In addition they are to a large extent internally mixed, and have a lower light absorption coefficient than soot particles emitted from higher temperature combustion processes. Hygroscopicity is treated separately for each species. BC is assumed to be in a 50/50 mixture of hydrophilic and hydrophobic states. For SO 4, the relative humidity is important due to its effect on the radiative properties through hygroscopic growth. We performed two experiments one where relative humidity was kept constant at 80% for all pressure levels, and one where we used the OsloCTM2 ambient relative humidity. For a complete discussion of size distributions, optical properties and other parametrizations, see Myhre et al. [2007]. [7] To calculate radiative forcing for a given species, an atmospheric pressure was chosen and a prescribed amount of aerosol was introduced here. The model was run twice, with and without the added aerosol, and we define instantaneous direct radiative forcing (DRF) as the difference in outgoing top-of-atmosphere shortwave flux between these two model runs. This procedure is repeated for 20 levels in the vertical. Pressures of (1000, 950, 900, 100, 50, 20)hPa were used. Finally, NDRF is calculated as the DRF divided by the prescribed aerosol load. The amounts prescribed were 0.3 mg/m 2 for BC, 3.0 mg/m 2 for SO 4 and 0.15 mg/m 2 for BIO. A detailed study of NDRF variability with the prescribed load is beyond the scope of the present work, but has been found in simple tests to be small (not shown). [8] Figure 1 shows the annual mean aerosol burden from the fossil fuel BC background field, and indicates some regions that will be used below. Blue boxes show indicative regions with properties that are expected to deviate from the global mean, e.g., the Pacific or the Arctic, while red boxes show industrial regions with high emissions of one or more aerosol species. [9] Figure 2 shows the annual mean vertical distribution of cloud fraction (Figure 2, left), ice (Figure 2, middle) and liquid water content (Figure 2, right). Solid lines show global means, while conditions over the regions selected in Figure 1 are indicated with various dashed lines. The global distribution can be seen to be composed of liquid water clouds concentrated at lower altitudes, and ice water clouds spread out over higher altitudes up to 20 hpa which is the top point of the present study. Figure 2. Global, annual mean vertical distributions of (left) cloud fraction, (middle) cloud liquid water content, (right) cloud ice water content. Global, annual means, year 2006, from ECMWF reanalysis data. 2of5

3 Figure 3. Global annual mean vertical NDRF distributions broken down into contributing components, for (a) BC, (c, d) SO4 and (e) BIO aerosols. (b) Variability with seasons and cloud year. Figure 3c shows SO4 with hygroscopic growth based on ambient relative humidity, Figure 3d with relative humidity prescribed to 80% (right). The vertical axis shows the pressure level where an aerosol load was introduced. [10] For the discussion below, we define vertical sensitivity as the derivative of NDRF with respect to atmospheric pressure: s v ¼ dndrf dp ½W = ðg hpaþš A high s V will indicate a rapidly changing NDRF with altitude. The minus sign is included to make a positive s V indicate that NDRF strengthens with increasing altitude. 3. Results [11] Here we present global and regional NDRF profiles for BC, SO 4 and BIO aerosols, and discuss changes with underlying factors such as cloud cover, season and region. [12] Figure 3 shows global annual mean NDRF vertical profiles, calculated as described in the previous section. For each species (BC, SO 4, BIO) we ran a baseline simulation where only the scattering and absorption properties of the aerosol itself were active in the model. We then progressively turned on modelling of Rayleigh scattering, atmospheric ozone, water vapour, of a background of other aerosols, and finally of clouds. [13] In Figure 3a the thick, solid line shows NDRF for BC including all effects. A variation is seen from 380 W/g at 1000 hpa, up to 3800 W/g at 20 hpa. The aerosol only simulation (red dashed line) shows a small strengthening near the surface, up to a constant value of 950 W/g below 800 hpa. This increase is due to BC being both scattering and absorbing, leading to the possibility of competing extinction processes near the surface. Purely absorbing BC has a fully constant NDRF profile (not shown). [14] Turning on simulation of Rayleigh scattering (Figure 3, orange dot-dashed line) strengthens the NDRF notably with altitude. Ozone and water vapour effects (yellow and blue dashed lines) weaken the NDRF due to absorption of solar radiation by these gases rather than the aerosol, while the background of other aerosol species strengthens the NDRF again due to scattering effects. Ozone, water vapour and background aerosols do not significantly alter the vertical NDRF profile (see discussion below). [15] Finally adding cloud effects we see a strong effect on the NDRF, as expected from previous column model studies [e.g., Zarzycki and Bond, 2010]. The global cloud profile (see Figure 2a) shows a peak in cloud fraction above 800 hpa, followed by a slow increase down to 200 hpa and subsequent drop to zero cloud fraction. These features are clearly reflected in the NDRF distribution for BC. Similar to introduction of Rayleigh scattering, clouds increase the NDRF with altitude, and the difference in the NDRF for BC in the lower stratosphere or upper troposphere to close to the surface is a factor of ten. [16] Figures 3c and 3d show similar simulations for sulphate aerosols. The general trends seen from BC are still present, but the final vertical sensitivities are much lower. For the SO 4 experiment using ambient relative humidity (Figure 3c), the bare aerosol NDRF changes significantly at low altitudes. Progressively adding the other terms, NDRF is strongly weakened at all altitudes, and the sensitivity seen at low altitudes is masked. The SO 4 experiment with constant relative humidity (Figure 3d) illustrates this weakening more clearly, from about 800 W/g for bare sulphate aerosols to about 500 W/g for clear sky conditions. Adding clouds further weakens the NDRF to 250 W/g close to surface, with modest vertical sensitivity to around 350 W/g in the upper troposphere. This shows that even though the sensitivity is low, treatment of scattering effects, relative humidity and clouds significantly affect SO 4 NDRF and hence also modelled sulphate direct radiative forcing. [17] Figure 3e shows BIO aerosol NDRF. Here we see a shift from a strong and negative NDRF for the bare aerosol, Figure 4. Vertical NDRF sensitivities for BC, SO4 and BIO aerosols. Dashed lines show aerosol-only simulations, solid lines show simulations with all effects included. For BC clear sky simulations are also shown. The vertical axis shows the pressure level where an aerosol load was introduced. 3of5

4 Figure 5. Vertical NDRF distribution for BC, for selected regions. (left) Illustrative regions (Sahara, Pacific, Arctic). (right) Industrial regions with varying degrees of BC emissions (North America, Europe, Africa, China). Results for global and annual mean are shown in both panels (thick black line). The vertical axis shows the pressure level where an aerosol load was introduced. to a vertical profile that goes from negative to positive at around 700 hpa when all effects are included. [18] Figure 4 shows the vertical sensitivity s V, as defined above, for bare aerosols (dashed lines) and all effects included (solid lines). For BC the clear sky sensitivity is also shown (dotted line). For all species sensitivities are highest at low altitudes, and change little above 700 hpa. While BIO and SO 4 do have significant, non-zero sensitivities, the BC sensitivities are far greater in magnitude. The correlation of BC vertical sensitivity with cloud fraction can clearly be seen, and in the region with high water cloud fraction we find sensitivities of up to 10 W/(g hpa). For this reason, we will focus on BC in the following discussion. [19] Several factors may influence the global mean NDRF distributions, including the meteorology for the chosen year, seasonal variations, and most prominently regional differences. To study the effects of varying cloud distributions, we have run the BC simulations for the years Also, seasonal variations were studied by extracting BC NDRF for four three-month periods rather than a full year. Figure 3b shows the result from one cloud year divided by a chosen baseline year (black lines), and the same for one season (three-month period) divided by a chosen baseline season (green lines). We assume these to be uncorrelated errors, and indicate their total effect (one standard deviation) on the global annual mean profiles as a grey band in Figure 3a. Of the two, seasonal variations can be seen to be of greatest magnitude. [20] Figure 5 shows the variability of BC NDRF with region. In Figure 5 (left) we have chosen regions with significantly different conditions the Sahara desert, the Pacific Ocean and the Arctic. Unlike for clouds and seasonal variations, we here see significant changes in NDRF. Despite a large fraction of high cloud cover over the Pacific, BC causes lower forcing due to the low albedo of the ocean. Over the Arctic, with its high albedo and high solar zenith angle, the impact of BC is conversely very high. Above the Sahara, where there are very few clouds (see Figure 2) but Figure 6. Vertical distribution of RF from fossil fuel BC in OsloCTM2. (left) BC NDRF (present work) interpolated to the model resolution of 61 levels (solid line), and global, annual mean fossil fuel BC concentration (dotted green line). (middle) Fossil fuel BC RF per height, calculated from the NDRF in the present work (solid blue line), and multiplied by the OsloCTM2 layer heights to show the actual RF per model layer (dashed black line). (right) The integrated RF in OsloCTM2 as function of pressure. 4of5

5 high surface albedo, the NDRF values are close to the global mean but the vertical sensitivity is lower as expected from an almost cloud-free sky. Figure 5 (right) shows the BC NDRF for industrial regions with high aerosol emissions North America, Europe, Southern Africa and China (see Figure 1a for region definitions). Here the variability is less pronounced than for the indicative regions, but still significantly higher than for changes in cloud cover or season. Notably, at pressures above 800 hpa, where BC concentrations are highest for realistic vertical profiles (see, e.g., Figure 6), all industrial regions show lower NDRF than the global mean. Similar statements hold for BIO and SO 4 aerosols (not shown). This shows that the variability in NDRF on regional scales can affect the RF uncertainty on a global scale, and one should therefore be careful using globally averaged NDRF values to study regional effects. 4. Application to Model Results [21] As shown, e.g., by Koffi et al. (submitted manuscript, 2011) and Schwarz et al. [2010], the actual vertical distribution of aerosols is poorly determined and varies significantly between aerosol models. The NDRF calculations in the previous section can be used to extract the contributions to the total aerosol forcing from different altitudes within a model, and thus to quantify the difference introduced by poor understanding of the vertical distributions. [22] In Figure 6, we have taken recent output from the OsloCTM2 model [Skeie et al., 2011] and combined with the present results. The solid line in Figure 6 (left) shows the global, annual mean BC NDRF distribution, interpolated to the full resolution of OsloCTM2 (61 layers) and assuming a constant NDRF for pressures below 20 hpa. The dotted line shows the global and annual mean BC concentration from OsloCTM2. In Figure 6 (middle), the solid line shows the RF per height as calculated from the present analysis, while the dashed line shows the actual BC direct radiative forcing vertical profile when multiplied by the model layer heights. We note a sharp rise from sea level and up to about 800 hpa, followed by a slow decline up to around 200 hpa where the BC RF rapidly vanishes. Hence for most of the vertical profile the declining BC concentration is counteracted by the strengthening NDRF. In Figure 6 (right) we show the normalized integrated RF from BC (dot-dashed line). More than half of the BC RF from this model arises from atmospheric pressures lower than 500 hpa, while only about 30% of the aerosol mass is found in this region. This shows that uncertainties in BC profiles higher in the atmosphere can have a significant impact on the modelled forcing, even if the loads are low, due to the strong vertical sensitivity of BC aerosol forcing documented above. [23] While studying a single model is interesting, the full potential of the present results becomes clear when discussing intermodal comparisons. The AeroCom direct aerosol experiment [Schulz et al., 2006] found significant variations between models, both for total aerosol forcing and for individual components. Schwarz et al. [2010] show that one reason for the variations in BC could be the wide range of vertical profiles. For the ongoing Phase 2 of this global aerosol model intercomparison, the vertical mass-mixingratio distributions provided by the model groups will be combined with NDRF distributions like the ones reported above, to quantify the differences in model RF due to this variability. For such an exercise model differences in NDRF and vertical sensitivities would also be useful to know. We therefore encourage other model groups to repeat the exercise reported here and publish their NDRF vertical distributions and sensitivities. [24] Acknowledgments. We acknowledge the support of the Research Council of Norway through the SLAC programme. Computations were run via a NOTUR grant on the TITAN cluster. We thank the two anonymous reviewers for thorough comments that have improved the quality of the paper. [25] The Editor thanks two anonymous reviewers for their assistance in evaluating this paper. References Ban-Weiss, G. A., L. Cao, G. Bala, and K. Caldeira (2011), Dependence of climate forcing and response on the altitude of black carbon aerosols, Clim. Dyn., doi: /s y. Bond, T. C., and R. W. Bergstrom (2006), Light absorption by carbonaceous particles: An investigative review, Aerosol Sci. Technol., 40(1), 27 67, doi: / Bond, T. C., G. Habib, and R. W. Bergstrom (2006), Limitations in the enhancement of visible light absorption due to mixing state, J. Geophys. Res., 111, D20211, doi: /2006jd Goto, D., T. Nakajima, T. Takemura, and K. Sudo (2011), A study of uncertainties in the sulfate distribution and its radiative forcing associated with sulfur chemistry in a global aerosol model, Atmos. Chem. Phys. Discuss., 11, 12,269 12,322, doi: /acpd Hansen, J., et al. (2005), Efficacy of climate forcings, J. Geophys. Res., 110, D18104, doi: /2005jd Haywood, J. M., and K. P. Shine (1997), Multi-spectral calculations of the direct radiative forcing of tropospheric sulphate and soot aerosols using a column model, Q. J. R. Meteorol. Soc., 123(543), , doi: /qj Lamarque, J. F., et al. (2010), Historical ( ) gridded anthropogenic and biomass burning emissions of reactive gases and aerosols: Methodology and application, Atmos. Chem. Phys., 10(15), , doi: /acp Lohmann, U., L. Rotstayn, T. Storelvmo, A. Jones, S. Menon, J. Quaas, A. M. L. Ekman, D. Koch, and R. Ruedy (2010), Total aerosol effect: Radiative forcing or radiative flux perturbation?, Atmos. Chem. Phys., 10(7), , doi: /acp Myhre, G., et al. (2007), Comparison of the radiative properties and direct radiative effect of aerosols from a global aerosol model and remote sensing data over ocean, Tellus, Ser. B, 59(1), , doi: /j x. Myhre, G., et al. (2009), Modelled radiative forcing of the direct aerosol effect with multi-observation evaluation, Atmos. Chem. Phys., 9(4), , doi: /acp Schulz, M., et al. (2006), Radiative forcing by aerosols as derived from the AeroCom present-day and pre-industrial simulations, Atmos. Chem. Phys., 6(12), , doi: /acp Schwarz, J. P., J. R. Spackman, R. S. Gao, L. A. Watts, P. Stier, M. Schulz, S. M. Davis, S. C. Wofsy, and D. W. Fahey (2010), Global-scale black carbon profiles observed in the remote atmosphere and compared to models, Geophys. Res. Lett., 37, L18812, doi: /2010gl Skeie, R. B., T. Berntsen, G. Myhre, C. A. Pedersen, J. Strom, S. Gerland, and J. A. Ogren (2011), Black carbon in the atmosphere and snow, from pre-industrial times until present, Atmos. Chem. Phys., 11(14), , doi: /acp Stamnes, K., S. C. Tsay, W. Wiscombe, and K. Jayaweera (1988), Numerically stable algorithm for discrete-ordinate-method radiative-transfer in multiple-scattering and emitting layered media, Appl. Opt., 27(12), , doi: /ao Textor, C., et al. (2006), Analysis and quantification of the diversities of aerosol life cycles within AeroCom, Atmos. Chem. Phys., 6(7), , doi: /acp Textor, C., et al. (2007), The effect of harmonized emissions on aerosol properties in global models An AeroCom experiment, Atmos. Chem. Phys., 7(17), , doi: /acp Zarzycki, C. M., and T. C. Bond (2010), How much can the vertical distribution of black carbon affect its global direct radiative forcing?, Geophys. Res. Lett., 37, L20807, doi: /2010gl G. Myhre and B. H. Samset, Center for International Climate and Environmental Research - Oslo (CICERO), PO Box 1129 Blindern, N-0318 Oslo, Norway. (b.h.samset@cicero.uio.no) 5of5