3.3 Carbon Monoxide (CO) Instrumentation

3.3.1 Correlation Spectroscopy of Carbon Monoxide

Carbon monoxide fulfils these requirements: it may be contained in a system for extended periods as it is a relatively unreactive gas. Some precautions must be taken with materials and pressure maintenance, but these are not serious. Spectrally it has a very simple spectrum, as would be expected from a diatomic molecule. The fundamental vibration- rotation band (see Figure 3.4) is centred at 2140/cm and in practice that band as observed is a composite of the fundamental C12O16 band, isotopic bands and "hot" bands. The band has been extensively studied and tabulated both for line strength (Rothman et al. (1983)) and collision broadened widths for a variety of gases including O2, N2 and the self-broadening widths (Nakazawa & Tanaka, 1982). In fact it is one of the most well-measured spectra of all molecules. The first overtone band occurs at 4260/cm and is very similar in structure.

The fundamental band is situated at the short-wave end of the thermal infrared and therefore the energy from that band will be weak and temperature-dependent. The first overtone band is beyond the thermal region and any signal from that band will be primarily scattered sunlight or some other, more complex, scattering.

The line profiles in the troposphere will be almost entirely collision broadened (Lorentz) in shape. Therefore the collision-broadened half widths are very important to our understanding of the radiative transfer problem. A technique which has recently been employed on a pressure modulator which will assist in all aspects of the characterization of the instrument is that of diode laser scanning. In this technique the absorption line profiles are scanned whilst the modulator is operating and thus a direct measure of the modulation can be obtained. Although the technique has currently only been applied to single lines in the carbon monoxide spectrum (Drummond and Ashton (1988), Berman et al. (1993)) it is expected that in the future it can be extended to the whole region in almost real time. Although essentially only confirmatory data of instrument operation, these data can be very valuable for critical tests of performance.

3.3.2 Carbon Monoxide in the Atmosphere

In addition to the emission from carbon monoxide, there will also be emission from other gases including carbon dioxide and water. Since there are comparatively large concentrations of these interfering species in the troposphere, the optical passband filter must possess strong out of band rejection. The required filter performance is achievable at the 4.7µm wavelength.

Any instrument which views the troposphere will, almost of necessity, view the ground as well. Thus any scheme to measure the concentration of carbon monoxide must account for the surface effects explicitly. At this point it is advantageous to write down the radiative transfer equations for the average and difference signals for a two-state correlation spectrometer (PMR or LMR) in nadir view including the ground radiance term. The average and difference radiances are:

where and are the electronic gains of the average and difference channel respectively, the factor f allows for the efficiency of the modulator as compared to an ideal square-wave case, I(v) is the ground radiance, B(v,T(z)) is the atmospheric emission at frequency v and temperature T(z), z is the atmospheric altitude, ç(v,z,) is the atmospheric transmission from level z to the spacecraft, and is the instantaneous cell transmission at amount u.

As may be seen directly from the above equations, the sensitivity of the instrument to carbon monoxide is governed by the difference between the surface and atmospheric radiance. In fact, in the unlikely event that the entire atmosphere and the surface are at the same temperature, no information is received. A knowledge of both the surface radiance and the atmospheric temperature profile is therefore essential to the data interpretation.

3.3.3 Carbon Monoxide Signal Functions

contain information about the height distribution of the CO signal. Derived from the total difference signal by the subtraction of the integrated ground radiance,

their magnitude varies with between a linear and square-root law as the CO concentration varies - the exact dependence being a function of the cell pressure. The signal functions are therefore more complex than the "weighting functions" of a conventional temperature sounder. Although the signal functions are not even first-order invariant for concentration changes, they do allow some consideration of the potential vertical resolution obtainable. The use of linear combinations of these functions will allow higher resolution to be obtained. Four specimen functions are shown in Figure 3.5, Figure 3.6 which demonstrate peaks at various levels. In extreme cases of pollution with a heavily-loaded boundary layer underlying a "clean" troposphere, a secondary peak in the lower signal functions is formed at the top of the boundary layer. It is not possible to produce peaks in the signal functions much below that of the lowest function shown, since the temperature contrast at low level is tending to zero.

Notice that the two lowest signal functions corresponding to a 40kPa and an 80kPa modulator are very similar indicating that the retention of both channels in the final instrument will probably not add to the overall data recovery. However the absolute magnitude of the 80kPa system is less and this places more stress on the signal performance.

The corresponding average signals for the above difference signals carry considerable information about the ground radiance as well as the interference effects from other species. It should therefore the possible, using a fairly simple scheme, to deduce the ground emission term (I(v) in the above equations) without using data from other sources.

3.3.4 CO Profile Measurements - Performance Requirements

Calculations of the performance of the correlation cells for various pressures are given in Table 3.1 for a 100ppbv atmosphere with a temperature profile appropriate to June at 40°N and a 290.8K surface temperature. The surface (1st) term of the difference signal is separated from the integrated signal function (2nd) term of the radiative transfer equation. The information about the atmosphere is primarily in the second term.

The correlation cell pressures required to generate signal functions which peak at appropriate levels in the troposphere are 2.5/5, 5/10, 20 and 80kPa (approximately). The lower pressure pair being assumed to be PMR channels and the highest pressures LMR channels with length ratio of 10/2mm used in double-pass configuration.

The instrument performance is limited by photon statistics at the detector (assuming no digitiser or other limitations). The instrument performance is evaluated in Table 3.2. In this table (and subsequent similar tables) some "reasonable" assumptions have been made about the overall characteristics of an instrument. However the numbers do not correspond to any particular instrument design requirement. The intent is to demonstrate the feasibility of this approach.

The sensitivity of these signals to a 10% and a 1% change in amount at 100ppbv is now evaluated and the "noise equivalent change in CO per pixel" in percentage terms is calculated in Table 3.3.

The figures show that a dynamic range (defined as the ratio of the total average signal to the sensitivity) of about 5 E+3 is required. These figures are of necessity approximate since the surface and atmospheric conditions vary over a wide range depending upon the conditions.

The performance figures indicate that the sensitivity to CO changes is adequate to measure CO profiles and it seems likely that other error sources, such as the vertical width of the signal functions will dominate. Studies conducted as part of the Phase B study indicate that this will contribute about a 10% error to the solution.

3.3.5 Total Column Measurements

The shorter wavelength of the CO channel allows the use of reflected sunlight because both the surface reflectivity and the solar intensity (usually) increase with decreasing wavelength. A study of the available literature suggests that reflectivity values of the order of 20% are possible in this wavelength region (see for example Deering (1989)), although lower values are also possible over certain regions, particularly water which has a notably low reflectivity.

The radiative transfer situation for a reflection measurement in this wavelength region is somewhat different from that for the thermal emission case as the surface emission is (to first order) negligible compared to the reflected radiation. This leads to the transfer equations:

Average:

Difference:

The quantity represents the Lambertian surface reflectivity and is the solar zenith angle at the time of the measurement. is the solar emission and the vertical transmission has been approximated using a plane parallel atmosphere approximation and a height-dependent absorption coefficient k(z) and density . Note that in this case it is the total column of CO that affects the result far more than the distribution.

3.3.6 CO Column Measurements - Performance Requirements

Using a surface reflectivity of 10%, a Lambertian surface characteristic and a 100ppbv constant CO profile, the signal values for the MOPITT instrument are evaluated in Table 3.4. The sensitivity is also significant in this case because the signal, although roughly linearly dependent upon the column amount of CO, has a large offset. The negative sign expresses the fact that the difference signal declines with increasing CO.

The instrument performance is limited by the available detector performance (assuming no digitiser or other limitations). The instrument performance is evaluated in Table 3.5.

The "noise equivalent change in CO column per pixel" in percentage terms is calculated in Table 3.6.

The figures show that a dynamic range (defined as the ratio of the total average signal to the sensitivity) of about 2 E+5 is required for the lowest pressure PMC and a 10% measurement, or about 2 E+4 for the highest pressure LMC in double-pass mode.

These performance figures indicate that with a single pass system integration over multiple pixels will be required to achieve a reasonable column measurement. For example a precision of approximately 2.0 E+17mol/cm² (10% of column amount) could be obtained with integration over a 10 pixels using a 40kPa LMC or over 42 pixels using a 5-10kPa PMC.

However with a double-pass system, there is an increased sensitivity which gives the possibility of single pixel resolution.

The filter specification of this channel is narrow, which adversely affects the signal to noise ratio, but yields a signal which has considerably less contamination from other substances, such as water vapour, which adversely affect the measurement precision.