process takes place, molecules turn into an excited state. These transitions can be classified into electronic, vibrational or rotational. Due to their structure, each molecule (and atom) will have specific spectral absorption bands that can be used to characterise the chemical species with spectroscopic methods. The reverse process of absorption of radiation will, evidently, result in emission of photons. The molecule is photolysed when the energy of the absorbed photon is high enough to compensate the energy balance from initial state and end product.
Important examples of atmospheric absorption of radiation were already acknowledged: UV sunlight absorbed by ozone; NO2 being photolysed by sunlight; greenhouse gases absorbing shortwave and IR radiation contributing to climate change.
Beer–Lambert’s law
Beer-Lambert’s law (also known as Beer’s law or Beer–Lambert–Bouguer law) states that light passing through a layer of thickness l will be partially absorbed and the intensity I0 reduced to I. Such a decline of intensity is proportional to the absorption cross section (units of cm2/molecule), the number density (units of molecules/cm3) and s the path length (units of cm). This can be represented by the following equation:
0
exp
I I N s
(2.31).According to Beer-Lambert’s law, the effect of light absorption by atmospheric constituents on the intensity and wavelength distribution of sunlight at the Earth’s surface depends on the nature and concentration of the gases and particles present, as well as the path length. The latter is a function of solar zenith angle (SZA), which can be understood as the angle between the local zenith and the line of sight to the Sun. Larger SZAs correspond to longer paths and, hence, the reduction in solar intensity by absorption and scattering processes will be larger.
In section 2.4 scattering by particulate matter was introduced. This redistribution of the radiation depends on several properties of the particles (size, composition, and amount) and can be represented by phase functions. Light scattering can be divided into two types: inelastic - where the energy of the photon changes; or elastic - where the energy remains unchanged. Raman scattering (rotational or vibrational) is an example of the first, where scattering will change the state of excitation of the molecule that absorbed the photon. For the elastic scattering, three domains can be defined depending on the size (diameter d) of the scatterer relative to the wavelength: d << Rayleigh scattering; d ≈
Mie scattering; and d >> geometric scattering. Within the visible wavelength range, the size of the majority of atmospheric aerosol corresponds to the Mie scattering regime, while for gases this will be Rayleigh scattering. This last category describes the scattering on very large particles that can be described by classical optics, based on the principles for reflection, refraction and diffraction.
Rayleigh scattering
Named after Lord Rayleigh (Rayleigh, 1899), this solution addresses the (elastic) scattering of radiation by molecules much smaller than the wavelength of incident light. The shape of the scattering object can be neglected in the calculations, which will then result in a relatively simple phase function (PRay) for unpolarised light, with as scattering angle:
2
P ( )
Ray 3 4 1 cos ( )
(2.32).Since the scatterers that follow this regime are relatively small, the scattering of radiation in forward and backwards directions is symmetric (see Figure 2.13). The cross section of Rayleigh particles can be described by the following equation:
5 2
4
128
Ray
3
x
(2.33),where, x is the size parameter as defined above. This equation highlights the inverse relation between the Rayleigh scattering efficiency with the fourth power of , which implies that shorter wavelengths will be scattered more efficiently. This relation is the explanation for the blue (shorter wavelength) skies and reddish sunsets (longer light path through the air and increased scattering). In an ideal case, an atmosphere clear of particles would be the perfect example for Rayleigh scattering.
Mie theory
The Mie theory (Mie, 1908), also called Lorenz-Mie theory, provides a solution to Maxwell’s equations, describing the optics of homogenous spherical particles, i.e., determines the scattering and absorption of light by spheres at different wavelengths. Often, for simplicity of calculations and reduction of computation time, this theory is used for all species of aerosol although it is known that particulate matter is usually non-spherical (see Figure 2.9). Still, adaptations have been developed for other shapes, such as, spheroids or layered concentric spheres, among others. Experimental results have shown that errors arise when this theory is used for non-spherical particles (Mishchenko et al., 1995). The full derivation of equations for the Mie theory is quite extensive and a more complete explanation of the mathematical details and descriptions of Mie scattering can be found at Bohren and Huffman (1983) and van de Hulst (1981). Relevant to mention are some of the differences with regard to the scattering of small particles. The phase functions for Mie scattering (equation (2.25)) do not correspond to symmetric scattering (as for Rayleigh), but suggest a predominant forward scattering, i.e., more radiation scattered in the direction of smallest scattering angles (see some examples in Figure 2.13). Nevertheless, the Mie scattering will match the Rayleigh regime for smallest particles.
This is an indication of the stronger dependence of the Mie scattering on the size of the particles. In addition, for an ensemble of particles, the amount of total radiation scattered is not linearly proportional to the amount of particles but rather dependent on the size distribution of the aerosol.
Figure 2.13 Scattering phase functions for particles in the Mie regime (spherical) with different radii (r). For comparison also the phase function in the Rayleigh regime is presented (black dashed line).
2.5.2 Radiative transfer equation
Atmospheric radiative transfer (RT) is a relevant process for many scientific questions. Radiative transfer models (RTMs, that solve the RT equations) determine what happens to radiation when it is traversing the atmosphere (or another medium), accounting for absorption, scattering and emission by all atmospheric constituents and the surface. Basically, information is provided regarding the gain and loss of radiation at a certain point in the atmosphere. Several examples were given above that depend on this computation, e.g., quantification of greenhouse effect and the changes of the radiative balance of the Earth-atmosphere system by aerosol, an important process for climate change. Furthermore, a practical application of this estimation, especially within the scope of the study presented here, is that it simulates the light path in the atmosphere, essential information for the retrieval of trace gases and aerosol columns from measurements performed with remote sensing techniques. In a way, Beer-Lambert’s law provides a very simplified solution for the radiative transfer equation, for a highly simplified scenario. With that in mind, the general form of the equation can be written as
( )
ext( ) ( ) ( )
dI k I J
ds
(2.34),where I is the intensity, s the length of the light path, J represents the gain and kext (extinction coefficient) the loss of radiation, for a given wavelength . These terms can be further specified in different contributions:
the increase from elastic and inelastic scattering processes;
the loss from either absorbing or scattering processes, i.e., extinction.
Combining all, the overall RT equation is
( )
abs( )
scat( ) ( )
th( , )
s( )
dI k k I I T I
ds
(2.35),where kabs and kscat are the absorption and scattering coefficients, respectively, and the last terms represent the intensities Ith from thermal emission, and Is from incoming radiation.
In addition, it is important to remember that the radiation arriving at a certain point (say, a molecule of gas) is not only coming from the direct source (the Sun for example) but also from diffuse radiation.
The full explanation of the solution for the radiative transfer equation can be found, for example, in Liou (2002).
2.5.3 SCIATRAN – radiative transfer model
The radiative transfer model used for the calculations of the work presented in this thesis was the SCIATRAN model (Rozanov et al., 2005; Rozanov et al., to be submitted, 2011). SCIATRAN developed at IUP-Bremen is in fact a follow up of GOMETRAN (Rozanov et al., 1997) that simulated the radiation measured by GOME in the 240 – 800 nm wavelength range, in nadir viewing geometry. With the launch of SCIAMACHY satellite instrument it became necessary to include a broader range of wavelengths (0.2 – 40 μm in the latest version) and other viewing geometries (limb, occultation) in such simulations, and, hence, SCIATRAN was created. Several versions have existed in the past decades as frequent improvements are implemented incorporating more features and broaden the potential applications, as it is the case of different viewing geometries: limb, nadir, off-nadir, zenith, or off-axis measured by space-, air- and balloon-borne, or ground-based instruments.
The analysis presented here was carried out with different versions of the model: 2.2 for the sensitivity study performed in chapter 3, and version 3.1 for the case study introduced in chapter 4. The main differences between these two consist of (A. Rozanov, personal communication, Feb. 2011):
a new option to solve the RT equation in vector mode;
polarised radiative transfer with the discrete ordinate method;
further extension of spectral range to 40 μm;
rotational Raman scattering included;
surface reflection, previously considered as Lambertian surface, can now be described by the bidirectional reflectance distribution function (BRDF);
a coupled mode for ocean-atmosphere system was implemented, including some underwater calculations.
None of the additional features was used in this research but some bug fixes and additions to the user interface were also performed in between versions. This RT code resolves the equation and principles mentioned in the previous section(s) providing radiances at the top-of-atmosphere (TOA) or ground but also other quantities such as weighting functions and airmass factors (AMFs), which is this the most relevant output for the work of the present thesis. The mathematical solving of the RT equation can be done in several manners, from more basic assumptions to rather complex and complete scenarios. Calculations in the plane-parallel mode do not consider the spherical shape of the Earth’s atmosphere and the equation can be solved with both the discrete-ordinates method and the finite differences. This method is limited to SZAs smaller than 90°. The pseudo-spherical mode, on the other hand, determines the light paths for the direct solar beam in a spherical atmosphere, but solves the RT equation for diffuse radiation in plane-parallel mode. Finally, a fully spherical atmosphere can also be considered. It is also important to mention that SCIATRAN vertical profiles of trace gas, pressure, temperature, and aerosol parameters are defined for a certain altitude grid but these will be linearly interpolated in between the levels. Furthermore, when using the Legendre expansion coefficients to define the aerosol phase functions, a delta-M approximation is used to accelerate the performance. According to Rozanov et al. (to be submitted, 2011), when determining the intensity at the TOA with a RT model, the highest levels of accuracy are obtained when this approximation is combined with the single scattering correction technique. In the case study presented in chapter 4, this set-up was essential to achieve results in reasonable computational time and compatible to the available resources. Some of the phase functions used for that analysis had more than 500 Legendre coefficients in the phase matrix requiring an extremely high number of streams to be used. This would have severely slowed down the performance of the model. Therefore, it was necessary to use less streams in the RT calculations. The resulting error on the AMFs from this approximation is rather small, in the order of 0.03%.
A more detailed description of SCIATRAN can be found at Rozanov et al. (to be submitted, 2011).