## Spectral Extension

The three particle properties of interest in radiative transfer simulations are AOD, ro0, and g, which is estimated from AnP. All three properties are spectrally dependent. Thus, for solar and infrared broadband simulations, which are required in climate applications, the merged monthly global fields of the new climatology for the mid-visible spectral region must be extended to other wavelengths of the solar and infrared spectrum. This spectral extension involves the following strategy:

First, assume bimodality to stratify AOD in accumulation and coarse mode fractions. The spectral extension takes advantage of observational evidence that almost all particle size distributions are bimodal in shape, with a concentration minimum at radii of 0.5 pm, separating smaller accumulation mode sizes from larger coarse mode sizes. Coarse mode (predominantly natural aerosol) particles are assumed to be large enough so that they do not display any significant spectral AOD dependence in the solar spectrum. Thus, for coarse size mode particles, the Angstrom parameter, AnPC (based on extinctions at 0.44 pm and 0.87 pm), is assumed to be zero (AnPC = 0). For particles of the accumulation mode, the associated Angstrom parameter, AnPA, is prescribed but allowed to

Aerosol fields Angstrom (440/870)

Aerosol fields Angstrom (440/870)

0.0000 0.5000 1.0000 1.5000

Figure 3.5 Mid-visible Angstrom parameter combining 440 and 870 nm wavelength data. Comparison between annual maps from global modeling (M), the merged product (X), and sun-photometer gridded samples (A). Deviations to model simulations suggest size overestimates in modeling.

### 0.0000 0.5000 1.0000 1.5000

Figure 3.5 Mid-visible Angstrom parameter combining 440 and 870 nm wavelength data. Comparison between annual maps from global modeling (M), the merged product (X), and sun-photometer gridded samples (A). Deviations to model simulations suggest size overestimates in modeling.

vary with ambient humidity. At larger ambient relative humidity, particles are expected to increase in size, thus lowering AnPA. Based on statistics with truncated (accumulation mode only) size distributions of AERONET, AnPA varies between 2.2 and 1.6. It is assumed that the larger value, 2.2, refers to completely dry conditions and that as the ambient relative humidity increases, this value decreases until, at completely wet conditions, a value of 1.6 is reached. Data on ambient relative humidity are approximated by the scaled low-level cloud cover,

1 cmid chigh by applying cloud cover data (CLOW, CMID, CHIGH) of the ISCCP cloud climatology (Rossow et al. 1993). This defines the locally applicable value for AnPA (AnPA = 2.2 - 0.6 x CLOW, s). If the local AnP of the new climatology that needs to be matched falls below the local threshold AnP, this indicates not only the presence of coarse mode particles but also quantifies the AOD fractions attributed to each mode. For a local AnP outside the (two) coarse and accumulation threshold values, the AOD would thus be assigned entirely to the coarse size mode (if AnP < 0) or to the accumulation size mode (if AnP > AnPA, whereby AnPA then adopts the value of AnP).

Second, apply ro0 to determine the composition and size of the coarse mode. Coarse mode particles are assumed to be sea salt, dust, or a combination of both. The startup configuration assumes dust over land and sea salt over water. Both aerosol types are defined by log-normal size distributions with effective radii (ref is the ratio of the third and second moment of the particle size distribution) of 1.25 pm for dust and of 2.5 pm for sea salt. Both components are assumed to be associated with rather wide distributions as a standard deviation (<r) of 2.0 was assumed. The adopted refractive indices for sea salt (Nilsson et al. 1979) and dust (Sokolik, pers. comm.) translated for the mid-visible (spectral) region in no absorption (= 1) for sea salt and in weak absorption for dust. The initially assumed composition is then modified to match the local value for (mid-visible) ro0. If (over land) the dust-associated single-scattering albedo, ®0 du, is larger (less absorption) than the local ro0, then part of the absorbing dust is replaced by non-absorbing sea salt. A possible compensation by a less-absorbing fine mode is not permitted, because it is assumed that the absorption potential of the accumulation size mode (1 - ro0A) cannot be smaller than the absorption potential of the coarse size mode (1 - ro0 c). If the local ro0 is smaller than the suggested coarse mode background value (which is quite common), then excess absorption is assumed to have been caused by the smaller (accumulation mode) aerosol. However, to avoid unrealistic low values for ro0 a, a minimum ro0Amin is defined as a function of both coarse and accumulation mode AOD (k>0 amin = 0.80 x AODa + 0.95 x AODC). In case the required k>0 a falls below that threshold (ro0 amin), the dust size is doubled (now rff = 2.5 pm), because increasing the size of an absorbing aerosol (here coarse dust) is an alternate way to lower the overall ro0. Working now with a smaller ro0C of larger dust sizes, ro0 A is recalculated. If the required ro0 A continues to falls below the revised ro0 Amin threshold, the doubling of the dust size is repeated to reg = 5 pm and if necessary even to rff = 10 pm. Thus, the ro0 constraint not only defines the coarse mode type, it also defines the single-scattering albedo of the accumulation mode ro0 A and provides information on the dust size.

Third, solar extension of accumulation mode and AnP conversion into g. Overall spectral aerosol properties are defined by combining spectral properties of coarse size mode and accumulation size mode. Precalculated spectral properties for the relevant dust size and/or sea salt are applied for the coarse mode. For the accumulation mode, the spectral dependence of the AOD is defined by AnPA, which, as explained above, varies between 2.2 and 1.6 depending on local low cloud cover. Even for the smallest possible AnPA value of 1.6 (at wet conditions), the associated effective radius of 0.27 pm remains small enough, so that far-infrared impacts (at wavelengths > 4 pm) can be neglected. Thus, accumulation mode properties of ro0A and gA need only to be defined for the solar spectrum. The mid-visible single-scattering albedo of the accumulation mode ro0 F is assumed to apply for all UV and visible wavelength. However, with size parameters significantly smaller than one in the near-infrared, the single-scattering albedo is reduced. The asymmetry factor gA is based on the accumulation mode AnPA using a relationship derived from AERONET statistics. Using truncated AERONET size distributions (no coarse size mode concentrations), scatter plots at wavelengths of 0.44, 0.55, and 1.0 pm suggest an anticorrelation between gA and AnPA which increases at longer wavelengths gA = 0.7 - 0.4 x (AnPA - 1) x ln(w05 + 0.5), when wavelengths (w) are between 0.25 and 3.0 pm. This quite general relationship is associated with significant scatter, but it is certainly better than assuming the mid-visible value gA (0.58 for dry and 0.64 for wet conditions) for the entire solar spectrum.

Finally, apply mixing rules to combine coarse mode and accumulation mode properties. The single-scattering properties of the coarse and accumulation size modes are combined via the common mixing rules, whereby AOD is additive, ro0 needs to apply an AOD weight, and g is weighted by the product of AOD and ro0. Figure 3.6 presents samples of annual maps at four wavelengths for AOD, ro0, and g.

## Post a comment