Results

Results of the extinction coefficient retrieval are provided in Figs. 2, 3,4, and 5. For each channel, we have processed five batches of randomly noisy data. The vertical distributions of the extinction coefficient retrieved (plain lines) are compared to the initial data (dot-dash lines). Results for Zr channel suffer from a higher level of noise (10-4 instead of 10-5).

For each LYRA channel, we have then determined the active components by comparing their contribution to the total extinction (see Table 1)).

Fig. 2 Extinction coefficient retrieved for Zr channel as a function of altitude from noisy simulated data

Fig. 2 Extinction coefficient retrieved for Zr channel as a function of altitude from noisy simulated data

Fig. 3 Extinction coefficient retrieved for Al channel as a function of altitude from noisy simulated data
Fig. 5 Extinction coefficient retrieved for Hz channel as a function of altitude from noisy simulated data
Table 1 Active components in each Lyra channel

Channel Active components Altitude range (in km)

Zr O,O2,N2 154-292

Ly O2 116-189

Hz O2,O3 45-74

In an ideal world, the extinction coefficient could have been retrieved over the same range of altitudes for all the channels. This would have meant that Eq. (9) is a well-determined system with four equations and four unknowns. But this is unfortunately not the case. In each channel, the extinction coefficient has proven to be retrievable over a range of optical thicknesses varying from 0.1 to 10. This limitation is induced by the Levenberg-Marquardt algorithm and could maybe be improved to a limited extend by using a priori constraints. Nevertheless, regions where the optical thickness is not between 0.1 and 10 are regions where the attenuation is either very small, either so high that nearly no signal is detected. Such regions will always be affected by a significant error. Depending on the channel, the retrieved interval of optical thicknesses corresponds to very different ranges of altitudes (the altitudes corresponding to optical thicknesses from 0.1 to 10 are provided in Table 1). The above system is therefore under-determined and will not be auto-sufficient to separate the densities of constituents. Nevertheless the fact that we could retrieve the extinction coefficient distribution is considered as a positive result in itself.

0 0

Post a comment