Intercomparison of GOME Ozonesonde and SAGEII Measurements of

Intercomparison of GOME, Ozonesonde, and SAGE-II Measurements of Ozone: Demonstration of the Need to Homogenize Available Ozonesonde Datasets K. Chance 1, X. Liu 1 , C. E. Sioris 1, 2, T. P. Kurosu 1, M. J. Newchurch 3 (kchance@cfa. harvard. edu) A 33 B-0903 1 Atomic and Molecular Physics Division, Harvard-Smithsonian Center for Astrophysics, Cambridge, MA, USA 2 Sioris Atmospheric Consulting, Brampton, ON, Canada 3 Atmospheric Science Department, University of Alabama in Huntsville, Alabama, USA Abstract We intercompare Stratospheric Column Ozone (SCO) and ozone profiles retrieved from GOME data with ozonesonde and SAGE-II data for 1996 -1999. GOME SCO over the altitude range ~15 -35 km usually agrees with SAGE-II SCO to within 2. 5 DU (1. 5%), without significant spatiotemporal dependence, but is systematically larger than ozonesonde SCO by 8 -20 DU (5 -10%) over all stations using the carbon iodine method and most stations between 30ºN-30ºS. Evaluation with SAGE-II, TOMS and Dobson data illustrates that those biases mainly originate from ozonesonde underestimates in the stratosphere. GOME retrievals also show large positive biases of 30 -60% at carbon iodine stations (except for Syowa) and of 20 -55% for all the stations between 30ºN-30ºS except for Paramaribo and Easter Island (<10%) over ~10 -20 km, while biases relative to SAGE-II data over ~15 -20 km are usually 10 -20%. The discrepancies over this altitude region reflect biases in GOME retrievals as well as ozonesonde measurements. In addition, GOME/ozonesonde biases vary from station to station and depend on ozonesonde technique, instrument type, sensor solution, and data processing, demonstrating the need to homogenize available ozonesonde datasets and standardize future operational procedures for reliable satellite validation. 1. Motivation q Validate retrieved ozone profiles (especially in the stratosphere) and stratospheric column ozone from GOME [Liu et al. , 2005, JGR] q Investigate the sources of large and systematic biases in the stratosphere between GOME and sonde observations at some stations q Approach: GOME vs. ozonesonde/SAGE-II Fig. 1 Locations of ozonesonde stations 2. GOME, Ozonesonde, SAGE-II q GOME: retrieved from UV spectra using optimal estimation v 11 layers (~5 -km thick, 2 -3 layers in the troposphere) v Vertical resolution: 7 -12 km in the stratosphere, 9 -16 km in the troposphere v Estimated uncertainties (random-noise errors and smoothing): 3 -10% in the middle and upper stratosphere, 15 -25% in the troposphere, ~1. 2% in SCO v Spatial resolution: normally 960 80 km 2. 3. Comparison Methodology q Ozonesonde: 33 stations as shown in Fig. 1 v Sources: WOUDC, CMDL, SHADOZ v Types: ECC, Carbon Iodine (CI), Brewer Mast (BM) v Precision: 3 -5% for ECC, 5 -15% for CI & BM v Accuracy: depends on techniques, sensor solution, ozonesonde model, and processing Fig. 2 Mean biases for GOME/SAGE-II SCO 4 -7 (~15 -35 km) GOME/ozonesonde SCO 4 -7 and SCO (tropopause-~30/35 km) at ozonesonde stations during 1996 -1999. q Time series (Fig. 3) v At Hohenpei enberg(a) and Lauder(e): good q GOME/sonde SCO 4 -7 or SCO biases (Fig. 2) agreement with no significant temporal drift v Usually similar to GOME/SAGE-II biases at v Systematic positive biases at Naha (b), mid- to high-latitudes, except: similarly at Tateno and Kagoshima v Large biases (10 -15 DU) at Tateno & v Transition at Ascension Island (c) in early Kagoshima (CI stations), opposite biases at 1998: corresponding to switch from En. Sci to Hohenpei enberg and Payerne mostly SPC, a bias increase of 7. 5 DU v In the tropics: large and varying biases (8 v Large bias change at American Samoa (d) 20 DU), 1 , and poorer correlation, except at and Tahiti (Table 2) in early 1998, Hilo and Paramaribo (within 1 -2 DU) corresponding to sensor solution change from Fig. 3 Comparison of GOME, a priori, ozonesonde 1%-bufferred to 2%-unbufferred. The bias SCO (tropopause to ~30/35 km): (left) Time series. changes are ~16 and ~11 DU, respectively. (b) differences. The black symbols in (d) show the v Smaller bias changes (3 -5 DU) at Boulder ozonesonde data with altitude-dependent correction and Hilo (Table 2) due to altitude-dependent and total ozone normalization at American Samoa. correction and total ozone normalization. v At Samoa, the bias change is reduced to 4 DU with a similar correction Table 2. GOME/sonde SCO comp. at 4 CMDL stations before & after switching sensor solution from 1%-bufferred (blue) to 2%-unbuffered (red). 5. GOME/Sonde SCO Comparison m q SAGE-II (version 6. 2): use O 3 profiles down to 15 km v ~1 km vertical resolution v Accuracy: ~5% over 20 -45 km v 10% mean agreement with ozonesonde measurements down to 15 km [Wang et al. , 2002, JGR] q GOME vs. SAGE-II v Coincidence: same day, ± 1. 5º lat. , ± 5. 0º lon. v SAGE-II data: integrated to GOME retrieval altitude grids, convolved with GOME retrieval AKs. v SCO 4 -7/SCO 8 -11: stratospheric column ozone within layers 4 -7/8 -11 or ~15 -35 km/35 -60 km are summed up from transformed profiles v. GOME/SAGE-II comparison over a station: use all coincidences within ± 5. 0º lat. , ± 40. 0º lon q GOME vs. ozonesonde v Use profiles extending above layer 6 (~30 km) v Coincidence: same day, ± 1. 5º lat, ± 600 km in lon. (relaxed to 12º at a few tropical stations to obtain enough collocations) v Ozonesonde profiles are similarly integrated and convolved. v Sonde SCO (ozone from tropopause to layer 6/7 or ~30/35 km)/SCO 4 -7: similarly summed up. Table 1. GOME/SAGE-II SCO comparison statistics (#, mean, 1 , R). 4. GOME/SAGE-II SCO Comparison q SCO 8 -11: -3– 6 DU with poor correl. (Table 1): v wavelength-dependent calibration error v no measurements below 289 nm are used q SCO 4 -7: biases <2. 5 DU (1. 5%) (Table 1) v good correlation: 0. 86 -0. 95 v 1 : varies from 4. 5 DU (2. 4%) in the tropics to 12 DU (5. 6%) at higher latitudes v negligible latitude dependence q SCO 4 -7: over ozonesonde stations (Fig. 2) v Mean biases: ± 2. 5 DU (1. 5%) except over three Northern Europe Stations (- 6 -7 DU). 2005 Fall 6. GOME/Ozonesonde Profile Comparison (Fig. 5) v GOME retrievals show similar altitudedependent biases, due to radiometric calibration error: usually within 3 DU in the bottom two layers (~0 -10 km), positive at layers 4 -5 (~15 -25 km), and negative from layers 6 or 7 on upward. v Absolute 1 s mainly a function of lat. v At mid- to high-latitudes (>30ºN/S), relative mean biases and 1 are typically within 20% except for layers 3 -4 (~10 -20 km) over carbon iodine (CI) stations. v Over CI stations, mean biases and 1 are within 30 -60%, usually larger than other stations at similar latitudes. v In the tropics, large positive biases of 20 -55% over ~10 -20 km, where ozone concentration is low except for Paramaribo and Easter Island v In the tropical and CI stations GOME/SAGE-II biases at ~15 -25 km are relatively homogeneous: 0 -5% for layer 5 Fig. 5. Absolute (columns 1 -2) and relative (columns 3 -4) and 8 -20% for layer 4; GOME/sonde mean biases and 1 between GOME and ozonesonde. shows larger and varying biases: 5 -20% Row (e) shows comparisons for 1%-buffered, 2%for layer 5 and 20 -55% for layer 4 (Fig. 6) unbuffered, and 200 m downward shift at American v At American Samoa and Tahiti, biases Samoa and Tahiti and corrected data at American Samoa. increases by 11 -15% over 0 -25 km and 68% over ~25 -35 km after switching to 2%unbuffered sensor solution (Fig 5, row e). v The altitude-dependent correction and total ozone normalization at American Samoa reduces the GOME/sonde biases mainly in 2%-unbuffered data. v Uncorrected altitude hysteresis in ozonesonde data introduces an error of 515% over ~10 -20 km (Fig 5, row e) v Inhomogeneity exists in ozonesonde datasets, making it difficult to perform a reliable and consistent satellite validation Fig. 6. Mean biases and 1 for comparisons of column considering ozonesonde ozone over GOME layer 4 (~15 -20 km) and 5 (~20 -25 km) without between GOME and SAGE-II/sonde in 1996 -May 1998. operational characteristics. Conclusions Fig. 4 GOME/other mean biases and 1 at 11 tropical q Comparison at 11 SHADOZ stations in 98 -99. SCO: tropopause to ~30/35 km; SCO 4 -7: during 1998 -1999 (Fig. 4). ~15 to 35 km; ICO: surface to ~30/35 km; TCO: tropospheric v Positive SCO biases from 4 DU(2. 7%) column ozone; TO: total column ozone. Shaded area shows to 16 DU(8 -13%) at Ascension Island, ± 1 GOME/SAGE-II spread. Tahiti, American Samoa except at Paramaribo (-2 DU). v Similar biases in TCO + SCO v Generally consistent with the TOMS / SHADOZ biases seen in Thompson et al. [2003, JGR] v Usually consistent with TOMS, Dobson, and SAGE-II to within 3 DU. v Those large SCO biases of 10 -16 DU, significant relative to 1 of various differences, mainly due to ozonesonde stratospheric underestimates. q GOME Stratospheric Column Ozone (SCO) over the altitude range ~15 -35 km usually agrees with SAGE-II SCO to within 2. 5 DU (1. 5%) without significant spatiotemporal dependence. q GOME SCO is systematically larger than ozonesonde SCO by 8 -20 DU (5 -10%) over Carbon Iodine (CI) stations and most stations within 30ºN-30ºS. q The large biases of 10 -20 DU mainly originates from sonde underestimates in the stratosphere. q GOME shows large positive biases of 30 -60% over CI stations (except Syowa) and of 20 -55% for all the stations in 30ºN-30ºS (except Eastern Island Paramaribo) over 10 -20 km q The discrepancies over this altitude regions reflect biases in both GOME and ozonesonde data. q GOME/sonde biases in SCO and profiles depends on sonde technique, instrument type, sensor solution, and data processing, demonstrating the need to homogenize available ozonesonde datasets and standardize future operational procedures for reliable satellite validation. Acknowledgements This study is supported by NASA and by the Smithsonian Institution. We thank the ESA for providing GOME level-1 data. We acknowledge WOUDC, NASA LRAB, SHADOZ, and CMDL for the ozonesonde and SAGE-II data. We acknowledge Dr. B. J. Johnson, Dr. S. J. Otlmans, and Dr. J. A. Logan for helpful discussion on the comparisons with ozonesondes.