6.3.2 Remote Sensing from the Ground and Aircraft  6.3 Measurements of Chemical Composition  6.3 Measurements of Chemical Composition

6.3.1 Remote Sensing Techniques Using Satellites

Satellite observations provide a powerful way to study the global distributions of these chemical constituents in the stratosphere. To understand such changes, one needs to distinguish between natural changes, due to seasonal variations in dynamics and chemistry, and the perturbations due to the addition of chlorine, nitrogen, and bromine to the atmosphere. Interest in stratospheric composition has focused during the last several decades upon changes in ozone amount.

This section will review some of the basic principles which govern global remote sensing observations. The process by which the observations are transformed into global fields of chemical composition will be discussed. An overview of an error analysis formalism, which is applicable to many observation systems, is also presented. Finally, an extremely brief overview of the progress in global remote sensing of the stratosphere is given.

Several different types of stratospheric measurements are made by satellites. Satellite instruments make occultation measurements of the Sun (or bright stars), observe the atmospheric emission itself, or look at light which has been absorbed and scattered by the atmosphere. Viewing geometries look downward (the nadir view), along ray paths which are horizontal (the limb view, see Figure 6.3), or at an intermediate angle.

 

Figure 6.3:  Limb view geometry for occultation and atmospheric emission measurements.

The fundamental quantity measured in all these observations is the optical depth $\tau (s)$ (defined in terms of the Beer-Lambert Law). 

$$\tau (s) = \Sigma \tau_i(s)$$ (6.13)

 
where s is the distance from the observation instrument to point s along the observation ray path. The summation is over all absorbing species. At any given wavelength, the optical depth is made up of the contributions from the overlap of numerous molecular spectral lines. The optical depth for species i at wavelength $\lambda$ is given by

$$\tau (s) = \int \Sigma l_m \left(\lambda - \lambda_{om}\right) d_i (s)ds$$ (6.14)

 
l_m is the spectral line shape for the m^th spectral line. The line shape is a function of line position $\lambda_{om}$ and wavelength $\lambda$, pressure, and temperature. d_i(s) is the number density of species i at point s along the observation ray path. See Goody and Yung (1989) for a thorough discussion.

In the occultation (absorption) experiment, the observed intensity $I(\lambda)$ is given by 

$$I(\lambda ) = I_o (\lambda )\exp (-\int d\tau )$$ (6.15)

 where $I_o(\lambda)$ is the solar intensity at the top of the atmosphere. This expression is equivalent to the Beer-Lambert Law in the "Absorption Cross Sections" portion of the text. For the emission experiment in the infrared, the atmospheric source function (e.g., the Planck function $B(\lambda ,T)$ for the case of local thermodynamic equilibrium, abbreviated LTE) is related to the observed spectral radiance $R(\lambda )$ by 

$$R(\lambda ) = \int B(\lambda , T)\exp (-\tau ) d\tau$$ (6.16)

 where T is the local temperature at positions s along the ray. When the population levels of the atmospheric molecules deviate from Boltzmann statistics, the source function is calculated by non-LTE techniques (see Goody and Yung, 1989).

Optical instruments have finite resolution, both in the spectral (i.e., wavelength) domain, and in the spatial domain. Convolution of I (or R) with a specific instrument function yields instrument averaged values of intensity (or radiance) which can be compared to observation. A convenient way to graph the spatial resolution of an instrument is to calculate the weighting function W(s),
 

$$W(s) = d(\exp (-\tau ))/ds$$ (6.17)

 The weighting function specifies the relative contribution to the observed signal, as a function of position along a ray path. Typical nadir views have weighting functions with a vertical width on the order of a pressure scale height. Limb viewing geometries yield weighting functions with a vertical width of 3 km. The book by Houghton et al. (1984) illustrates weighting functions for several instruments.

Since satellite measurements sample various heights in the atmosphere simultaneously, it is necessary to find some way to relate the total optical depth to the chemical concentrations. To convert instrumental observations to vertical mixing ratio profiles, one applies inversion techniques, i.e., mathematically find the pressure-temperature (P, T) and mixing ratio (f) vertical profiles which best represent the observations. Two inversion techniques, onion peeling, and matrix inversion, are described below, using the limb viewing emission experiment as an example.

Inversion techniques require a way to calculate observed radiances, based upon specified vertical profiles of pressure-temperature and mixing ratio. (The mixing ratio for a species, at height z, is given by the ratio of the number density of that species, and the total atmospheric number density). The calculation of spectra is based upon line-by-line calculations, which utilize extensive spectral line parameter databases, such as the HITRAN 91 compilation (HITRAN 91). Since line-by-line calculations of atmospheric spectra are computer-intensive, most operational satellite instruments use a parameterized form of radiative transfer. For the following discussion, pressure-temperature profiles are retrieved from CO₂ spectral features, assuming a CO₂ vertical mixing ratio profile derived from in-situ measurements.

In the onion peeling technique, the atmosphere is represented as a set of homogeneous spherical shells. The ray paths of the observations have tangent points throughout the stratosphere. One first solves for the P, T, and f values for the topmost shell. These values are fixed in the next step, which solves for the P, T, and f values in the next lowest shell. The process continues downward into the atmosphere. One can improve upon the solution by iterating, improving the boundary solution at the topmost altitude by using the estimated vertical gradients to update the solution at the topmost altitude.

In the matrix approach to inversion, one specifies a model atmosphere, calculates spectral radiance values R, compares these radiances to the observed radiances, takes the difference (denoted by $\Delta R$), and forms the matrix equation 

$$\Delta R = A \Delta V$$ (6.18)

 The matrix A is comprised of derivatives, i.e., $\partial R/\partial T$, $\partial R/\partial lnP$, $\partial R/\partial f$, evaluated for each segment along each ray path. $\Delta V$ is a perturbation vector which predicts deviations $\Delta T$, $\Delta lnP$, and$\Delta f$ from the model P, T, and f values. One then uses matrix inversion algebra to solve for $\Delta V$. The $\Delta V$ vector is used to update the model, a new $\Delta R$ is calculated, and one iterates until $\Delta R$ approaches a desired degree of convergence.

To avoid numerical instability, matrix inversion usually incorporates mathematical smoothing techniques (e.g., one applies a constrained linear inversion). Mathematical forms such as 

$$\Delta V = (A^* A + \gamma H)^{-1} A^* \Delta R$$ (6.19)

 are encountered, where $\gamma$ is a Lagrangian multiplier, H is a smoothing matrix, and * denotes the matrix transpose. Good discussions of these methods are given in Twomey (1977). Houghton et al. (1984) describe other techniques (maximum likelihood, minimum variance, sequential estimation, and Backus-Gilbert methods).

The assignment of error bars can be an involved process. Many factors contribute to the overall error analysis. Instrument noise and calibration accuracy, uncertainties in spectral line parameters, instrument field of view and weighting function limitations, and temporal sampling limitations all contribute to the error bars assigned to global fields. A particularly useful error analysis scheme for instrument systems has recently been developed by Rodgers (1990). In this analysis, multidimensional error bars are calculated, based upon knowledge of atmospheric emission model uncertainties, weighting function considerations, model bias, and measurement errors. The Rodgers analysis can be applied to many instrument systems. Readers not interested in global observations should at least take note of the error analysis presented by Rodgers. Another reference of general applicability is the work of Menke (1984).

Having provided some background information on global remote sensing, a partial overview of progress in stratospheric remote sensing - past, present, and future - is given. Emphasis, due to necessary brevity, is focused upon the Limb Infrared Monitor of the Stratosphere (LIMS) experiment, the Upper Atmosphere Research Satellite (UARS), and the Earth Observing System (EOS). The time frame for these projects are 1979, 1991-present, and 1996-next century.

The LIMS experiment (Gille and Russell, 1984) flew on the Nimbus 7 spacecraft. Radiometer channels observed six infrared regions, each typically 150 cm^-1 wide. Pressure-temperature, O₃, NO₂, H₂O, and HNO₃ global fields between 15 and 60 km altitude were observed for a nine month period, between October 25, 1978, and May 28, 1979. This data set provided input to many theoretical studies of the interplay of stratospheric dynamics and chemistry.

The Upper Atmosphere Research Satellite (UARS), launched in 1991, focuses primarily on observations of the stratosphere. Ten experiments measure a suite of variables: the solar UV and particle flux which enters the Earth's atmosphere, stratospheric and mesospheric winds, temperature-composition fields, and chemical composition fields (O₃, ClO, H₂O, CH₄, CO, HCl, HF, CFC-12, NO₂, NO, HNO₃, and ClONO₂). Latitude coverage is from 80 North to 80 South, and a full range of hour angle is observed in 35 days (due to the orbital precession of the satellite). UARS instruments will observe over a time period of 20 months to several years. An example of UARS results is given in Figure 6.4 (Waters, 1993a), in which Microwave Limb Sounder (MLS) maps of column O₃ and ClO for the 20km altitude level over Antarctica in September of 1991 and 1992 reveal the Antarctic Ozone Hole. (See Waters et al., 1993b, for further discussion). Heterogeneous chemistry on polar stratospheric clouds, followed by the return of Sunlight in Spring, converts stratospheric chlorine from inert forms to the reactive form of chlorine (chlorine monoxide, ClO), which then decreases ozone concentrations (Solomon et al., 1986). These, and other composition fields from UARS, will place important constraints upon detailed 3-d transport-chemistry models.

 

Figure 6.4:  MLS column ClO and O₃ over Antarctica for 21 September 1991 and 20 September 1992. Light and dark  shadings correspond to low and high column amounts.

The Earth Observing System (EOS), still in the development phase, will focus upon processes and changes in the stratosphere, troposphere and the biosphere-ocean system, with emphasis upon scientific problems dealing with the biosphere and hydrosphere (EOS, 1991). The instruments will be expected to operate over 5 years' time. A long time period of observation is needed to obtain observational understanding of changes in vegetation patterns, changes in rainfall distribution, cloud effects on global climate, and changes in surface and ocean temperature, ocean circulation, sea ice and glacier flows, and atmospheric composition.

Future progress in remote sensing will include the identification of additional trace chemical species which play important roles in the complicated gas phase and heterogeneous chemical transformations of the stratosphere. Detection of long term trends, viewed by satellite systems and by ground based networks such as the Network for the Detection of Stratospheric Change (NDSC) will also be an important aspect of future work. The NDSC is an internationally-supported network of ground-based observations intended to monitor and detect changes in stratospheric composition and dynamics for the next twenty years. Instrumented sites are currently being established at latitudes covering the Earth. Since global changes are punctuated by transient events, e.g., by the recent injection of aerosol into the stratosphere by the Mt. Pinatubo volcano, there exists an interpretive challenge to distinguish between the transient and longer-term changes. 

  
 6.3.2 Remote Sensing from the Ground and Aircraft  6.3 Measurements of Chemical Composition  6.3 Measurements of Chemical Composition 
http://www.asp.ucar.edu