Ozone

Results from three flights, 26 April, 6 May and 21 September 1997, are presented in Table 6.7. Below these are vertically integrated ozonesonde April-May averages from five stations and vertically integrated in-situ measurements from portions of the 6 May and 21 September flights. Mixing ratios from the UV-absorption photometer were converted into number density and integrated through the dive, ascent, and descent portions of these flights. The integrated VCDs from the ascent from and decent into Fairbanks during the 6 May flight agree quite well but there is variability throughout the dives. The effect of latitude is clear from the 21 September flight as both the 12-16 km layer (C₂) and the 16- $z_{\rm ER2}$ km layer (C₃) decreased by 30 DU between the ascent from Fairbanks and the descent into Hawaii.

 

Table 6.7: Summary of ozone limb retrievals, average April-May ozonesonde data, and integrated ER-2 in-situ ozone measurements (Proffitt et al., 1989) from 6 May and 21 September 1997 (see text for details).

Flight	Time	Location	$z_{\rm ER2}$	Retrieved O3 VCD (DU)
Date	(UTC)	($^{\circ }$N,$^{\circ }$W)	(km)	0-12 km	12-16 km	16- $z_{\rm ER2}$ km
6 May	2048	76.2,114.5	19.5	51.6	65.3	72.1
	2248	70.1,132.8	20.1	57.4	49.8	77.9
26 April	1932	87.0,148.0	19.1	91.3	71.0	47.6
	2106	83.1,148.0	19.4	74.3	74.8	70.9
	2130	80.1,148.0	19.6	51.2	68.8	51.0
21 September	2017	53.1,153,1	19.5	39.4	44.4	66.2
	2150	42.1,154.5	20.1	22.2	27.3	60.5
	2327	31.4,155.4	20.7	51.9	25.2	52.8
(April-May) Overall Mean Ozonesonde				65.4	58.7	73.4
Fairbanks (5)a		64.5,147.9		68.1	51.2	70.0
Alert (2)		82.5,62.3		58.2	65.8	73.6
Resolute Bay (3)		74.4,94.6		62.5	45.5	68.2
Stoneyplain (3)		53.3,114.1		62.1	72.1	79.2
Eureka (2)		80.0,85.6		70.7	65.6	80.5
6 May	1817	66.6,147.3	Ab	-	43.1	69.5
	2007	78.5,133.8	B	-	-	75.3
	2026	77.6,123.9	C	-	-	68.7
	2128	73.3,104.1	D	-	-	85.3
	2147	72.3,111.8	E	-	-	77.4
	0011c	65.9,149.5	F	-	43.6	70.2
21 September	1845	62.8,149.1	G	-	39.8	53.5
	2212	39.9,154.7	H	-	-	41.6
	2215	37.7,154.9	I	-	-	39.5
	0051d	23.2,157.4	J	39.3	8.8	21.2

^a Number refers to number of ozonesonde profiles used for the average.
^b Columns derived from integrating in-situ measurements over segments of the flight: A Ascent from Fairbanks; B descent during first dive; C ascent after first dive; D descent during second dive; E ascent after second dive; F descent into Fairbanks; G ascent from Fairbanks; H decent during dive; I ascent after dive; J descent into Hawaii.
^c 7 May 1997.
^d 22 September 1997.

Retrievals were performed on two limb scans from the 6 May flight. Overall the values retrieved at all layers were comparable to those observed from the ozonesondes and the integrated in-situ measurements. If anything, the ozone retrieved in the lowest layer was smaller than from either other VCD source. The 26 April retrievals were more variable. The first and third scan show marked decreases in C₃. This can be traced back to the ACDs as for elevations angles just below the limb these two scans produced ACDs about 30-40% smaller than that from the middle scan. For steps in the scan below this, all three have similar values. This is especially interesting as all three scans were made within two hours. The first two scans were made inside this region of increased nadir-derived VCD ozone as discussed in Section 6.5.1, which is the reason for the large amount of ozone in C₁, as was the second. The third scan was made just on the edge of this region.

Results from the 21 September flight generally follow the expected trend of decreased ozone in every layer with decreasing latitude, although they are consistently larger than the in-situ values. In particular, C₃ shows only a modest 14 DU decrease over the three scans (over a 22$^{\circ }$ change in latitude) while the in-situ values dropped by 32 DU (over a 39$^{\circ }$ change in latitude). Ozone in the lowest layer was substantially smaller than that observed from the other two flights, due in part to the ER-2 flying over the ocean where less tropospheric ozone is present. The exception is the final scan in which 52 DU was retrieved in this layer. This is a direct result of the larger nadir-derived VCD retrieved below the ER-2 during the second half of this flight (see Figure 6.20).

 

Table 6.8: Sensitivity of retrieved ozone from 2048 UTC 6 May 1997 (first retrieval from Table 6.7) to changes in several parameters. `Retrieval parameters' refers to quantities to which the air mass factor matrix may be sensitive and yet are not known exactly while `model parameters' affect the accuracy of the air mass factor matrix elements. ( ${\rm C_1+C_2+C_3}$ is the VCD below the ER-2; ${\rm C_4}$ is the VCD above the ER-2.)

	Parameter	Altered	Retrieved O3 VCD (DU)
	(Initial Value)	Value	0-12 km	12-16 km	16- $z_{\rm ER2}$ km
Standard (from Table 6.7)			51.6	65.3	72.1
Retrieval Parameters
1	Albedo	0.0	46.4	65.1	77.5
	(0.6)	0.3	49.1	65.3	74.6
		0.9	53.8	65.2	70.0
2	Aerosol Optical	0.01	45.0	60.3	83.7
	Deptha (0.015)	0.02	64.1	68.9	56.0
3	$z_{\rm ER2}$	19.3 km	48.0	68.9	72.1
	(19.5 km)	19.7 km	55.2	62.1	71.7
4	SZA	63.6$^{\circ }$	51.9	64.8	72.3
	(60.6$^{\circ }$)	57.6$^{\circ }$	51.4	65.7	71.9
5	${\rm C_1+C_2+C_3}$	$+10\%$	74.7	63.2	70.6
	(189 DU)	$-10\%$	28.4	67.5	73.6
		UCb	132.0	52.0	62.7
6	${\rm C_4}$	$+10\%$	52.0	65.5	71.5
	(197 DU)	$-10\%$	51.1	65.2	72.7
		UCb	66.1	68.2	62.7
7	Retrieval Wavelength	510 nm	55.2	62.0	71.7
	(500 nm)	490 nm	46.8	68.9	73.3
Model Parameters
8	Vertical Grid	0.25 km	51.2	65.6	72.2
	Spacing (0.5 km)	1.00 km	53.4	62.6	73.0
9	Scattering Orders	1	44.1	64.1	80.8
	(MSc)	2	49.1	65.3	74.4
10	Polarization (no)	yes	51.7	65.3	71.9

^a Optical depth refers to that at 750 nm.
^b UC=Unconstrained
^c MS=Multiple Scattering

The sensitivity of the retrieved ozone VCDs to several retrieval parameters, in particular that from 2048 UTC 6 May 1997 (from Table 6.7), was assessed with the results presented in Table 6.8. Each of these parameters either affect the elements of the air mass factor matrix or are constraints on the retrieval. Overall, the most sensitive layer was C₁ and the least sensitive was C₃. The impact of the different parameters is discussed below.

Albedo was found to have little influence on the VCDs. This is because for all elevation angles there is either no direct surface component or the slant optical thickness along the line-of-sight is much larger than unity. Also, albedo acts to increase (or decrease) all elements of the AMF matrix together and the tightly constrained nature of the retrieval (i.e. VCD below ER-2 held constant) will not allow large changes under these circumstances. In varying the albedo, ozone was essentially shuffled between C₁ and C₃ with C₂ remaining nearly constant. The largest relative changes were in C₁ which decreased by 10% when using an albedo of 0.

Aerosols had a much larger impact on the retrieved ozone. Increasing the amount of stratospheric aerosol acts to increase the AMF at higher altitudes due to increased scattering and decrease the AMF at lower altitudes as relatively less light penetrates to these layers. However, for background levels of stratospheric aerosol, the individual AMFs are not particularly sensitive. It is the tight constraint together with the different responses of the three layers which amplifies the effect of aerosols. Where this constraint acted to help minimize the change with respect to albedo, it increases the change with respect to aerosol. Increasing the total optical depth of the aerosol layer by 0.005, C₁ was increased (as $\delta_{i1}$ decreased) by 25%, C₂ was increased by 5%, and C₃ was decreased by 20% (as $\delta_{i3}$ decreased). Comparable changes in the opposite direction were observed when the aerosol optical depth was decreased by 0.005. No attempt was made to determine what the effect of varying the shape of the aerosol profile has on the retrieved ozone amounts, such as where the peak occurs, although this too is likely to be important.

Values for the next two parameters, the altitude of the ER-2 and the solar zenith angle, are accurately known. However, over the course of a scan, both can vary, sometimes substantially. Changing the altitude at which the AMFs are calculated by $\pm0.2$ km did not greatly impact the VCDs, although increasing the altitude had the larger effect. Of the three layers, the lowest layer was the most sensitive, changing by 3.6 DU or 7%. Varying the SZA by $\pm 3^{\circ}$ had an even smaller impact as the largest observed change was by 0.5 DU. This is because the contribution to the total slant path enhancement by the direct solar beam is small. Also, in the visible for a relatively high sun, the average scattering height varies slowly with SZA. Larger changes are expected near a SZA of 90$^{\circ }$.

The effects of the two retrieval constraints, holding the VCDs above and below the ER-2 constant, are now investigated. Increasing or decreasing the VCD below the aircraft by 10% had a small effect on C₂ and C₃, about 2% and 1%, respectively. The change in VCD was almost entirely incorporated into the lowest layer. Upon eliminating this constraint, C₂ and C₃ decreased by 15-20% while C₁ more than doubled. Clearly, the DOAS ACDs contain little information about ozone near the tropopause and below although at this juncture, it is not known why too much (as opposed to too little) ozone is retrieved in C₁ without this constraint. This indicates that while the constraint is necessary, any errors in the VCD below the aircraft will only impact C₁. The effects of constraining the VCD above the ER-2 were less severe. Varying this by $\pm10\%$ results in changes no larger than 0.5 DU. Eliminating this constraint forced larger errors, 15 DU in C₁ and 10 DU in C₃. Again, while this constraint is necessary, moderate errors in its value will not significantly effect the retrieved ozone below the aircraft.

Changing the wavelength at which the AMFs are calculated by $\pm 10$ nm had an impact of about 4, 2, and 0.5 DU on C₁, C₂, and C₃, respectively. This is larger than expected based on the simulations using synthetic spectra but the conclusions concerning the average wavelength being the best are likely valid. This problem can be avoided entirely by calculating air mass factors not at a single-wavelength, but over the same spectral range used in the DOAS fit. This could be done by generating synthetic spectra and taking the ratio of ACD to the model input VCD. This would also be a more effective means of determining the validity of single wavelength air mass factors.

The effects of some model parameters are now discussed. Decreasing the vertical grid spacing to 0.25 km had little effect on the retrievals but increasing it to 1 km resulted in differences up to 1.8 DU. Thus, 0.5 km is both necessary and sufficient. Multiple scattering is also necessary as single scattering placed more ozone near the ER-2 and less in the troposphere. An extra scattering order helped but differences of 2 DU remained. As expected, the inclusion of polarization had no significant impact on the retrievals.