Limb Results

Upon using the retrieval algorithm outlined in Section 6.4 it became apparent that the ACDs were insensitive to the lowest layer (0-12 km). The result of the noise in the measurements combined with the small air mass factors for this layer meant that an unrealistically large amount of absorber was placed here. To remedy this, after each iteration, the VCD in this layer was constrained to,

$$\rm C_1 = \rm\frac{F_{nadir}}{\delta_{nadir}} - C_2 - C_3$$ (11.35)

or the difference in the nadir-retrieved VCD below the aircraft and that retrieved in the other two layers. Recall that C$_{\rm j}$ is the VCD in layer j. Another problem which arose concerned the top layer ( $z_{\rm ER-2}$-top of atmosphere) where the retrieved VCDs were also too large. The impact of this could be minimized in two ways. The first is to exclude the uplooking steps in the scan from the retrieval so that the column above the aircraft is decoupled from that below. The second is to constrain the VCD in the top layer. In the end, the top three steps in the scan were omitted as they generally were the noisiest and the VCD above the aircraft was constrained to either the VCD derived from the CPFM using the Brewer method (ozone) or an estimate (NO₂ and BrO).

The criteria for selecting a scan to perform the retrieval on was the following: the ER-2 had to maintain a nearly constant altitude over the course of the scan and the 745 nm radiances for this scan and the ones immediately preceding and succeeding it must appear cloud free.

When initially performing the ozone retrievals it was discovered that the elevation angle calibration was not correct. Small additional shifts of about 0.05-0.1$^{\circ }$ were sometimes necessary and these varied from scan to scan (although the variability between scans made during the same flight was smaller). In an effort to keep the ozone limb retrievals as unbiased as possible by these necessary extra adjustment, they were made in 0.05$^{\circ }$ increments. In general, the retrieved ozone profiles remained fairly stable over this range and so if the initial $\Delta{\rm EA}$ produced obviously unrealistic results, it was shifted by 0.05$^{\circ }$. In the future it may be necessary for each scan to be tuned independently due to the required accuracy (to better than 0.05$^{\circ }$) and this process should probably include the simultaneous retrieval of aerosols. Recall that a change in EA of 0.05$^{\circ }$ at EA $=-4^{\circ}$ is equivalent to a change in tangent height of about 0.4 km. The same adjustments determined in retrieving the ozone were used in the retrieval of NO₂ and BrO.

There was no attempt to perform the retrievals on different altitude grids. Future work may reveal that a different division of the atmosphere below the ER-2 may be better suited for retrievals or that it is feasible to retrieve at four (or more) layers. Also, the standard aerosol profile of Chapter 5 is used scaled to the number density which gives an optical depth of 0.015 at 750 nm, as worked best for the initial calibration. Note that the optical depth will be slightly larger at the retrieval wavelengths.