Relationship between errors in AVHRR-derived ... - Semantic Scholar

GEOPHYSICAL RESEARCH LETTERS, VOL. 28, NO. 10, PAGES 1989-1992, MAY 15, 2001

Relationship between errors in AVHRR-derived sea surface temperature and the TOMS Aerosol Index J. P. Diaz, M. Arbelo, F. J. Exp´osito Departamento de Fisica, Universidad de La Laguna, La Laguna, Islas Canarias, Spain

G. Podest´a, J. M. Prospero and R. Evans Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, Florida

Abstract. We investigate the effects of various types of atmospheric aerosols on satellite-derived sea surface temperature (SST) retrievals. The association between aerosol presence identified by the Earth Probe Total Ozone Mapping Spectrometer (TOMS) aerosol index (AI) and systematic errors in Advanced Very High Resolution Radiometer (AVHRR) Pathfinder SST retrievals is explored. We find a significant increase in systematic PFSST errors in the presence of dust aerosols. Average errors range from 0.34◦ C for AI values between 0.5 and 1.0 to 1.74◦ C for AI≥1.5. The bias is introduced by the AVHRR channel brightness temperature difference (T4 − T5 ) which is intended to correct for atmospheric absorption normally due primarily to water vapor. Our study shows that (T4 − T5 ) is sensitive to the dust aerosols (i.e., TOMS AI values). In contrast, smoke aerosols do not seem to have a significant effect on PFSST errors.

Data: Pathfinder SST and Earth Probe/TOMS Aerosol Index

Introduction The sea surface temperature (SST) is one of the most important parameters for understanding the global climate system. Satellite-derived measurements are ideal for monitoring SST due to their high-frequency global coverage, unattainable with other type of measurements. The most reliable global SST measurements are obtained from multichannel infrared radiometers operating in cloud free conditions [Barton, 1995]. Since 1981, the NOAA series of polarorbiting spacecraft have carried the Advanced Very High Resolution Radiometer/2 (AVHRR), an instrument with three infrared (IR) channels suitable for estimating SST. Several techniques have been proposed to correct for the atmospheric absorption of surface IR radiance to enable accurate retrievals of SST. Barton [1995] shows that the differential absorption of IR radiance at suitable pairs of wavelengths is exploited in all SST algorithms, and that there is a basic form for most algorithms. High loads of atmospheric aerosols (volcanic particles, dust, smoke, etc.) can introduce large errors in satellitederived SST retrievals (>1.0◦ C). For example, the influence of volcanic aerosols was evident in depressed AVHRRderived SSTs during most of 1982-1983 and in 1991-1992 due to the eruptions of El Chich´ on and Mount Pinatubo volcanoes respectively [Strong et al., 2000]. With regard Copyright 2001 by the American Geophysical Union. Paper number 2000GL012446. 0094-8276/01/2000GL012446$05.00

to dust aerosols, dust plumes from sources in North Africa are the most prominent and persistent and cover the largest ocean areas [Husar et al., 1997; Prospero, 1999]. May et al. [1992] proposed a method to correct for Saharan dust effects on satellite-derived SSTs. This methodology linked SST errors with an estimate of aerosol optical depth derived from the AVHRR’s channel 1. To our knowledge there have been no studies of satellite-derived SST errors caused by smoke aerosols, such as those introduced by biomass burning (naturally occurring or caused by agricultural practices), oil industry fires, or industrial smoke. The main objective of this paper is to evaluate errors in satellite-derived SST retrievals in areas with absorbing atmospheric aerosols as identified by the Total Ozone Mapping Spectrometer (TOMS) Aerosol Index (AI)[Torres et al., 1998].

SST values, both in situ and satellite-derived, have been obtained from the NASA/NOAA Pathfinder AVHRR/Ocean matchup database [Kilpatrick et al., 2000]. The matchup database includes co-temporal and co-located AVHRR and in situ observations. Matchups used in this work encompass the period 1996-1998 and correspond to the AVHRR onboard the NOAA-14 spacecraft. The AVHRR-derived SSTs were estimated using the Pathfinder algorithm described by Kilpatrick et al. [2000]. The matchups include two ancillary quantities that will be used in subsequent analyses: Reynolds/NCEP Optimally Interpolated SST (OISST) values [Reynolds and Smith, 1994] and integrated column water vapor (CWV). Global NCEP/ OISST fields are produced weekly on a 1◦ grid, using both in situ and AVHRR-derived SSTs. The CWV values were derived from the Special Sensor Microwave/Imager (SSM/I) and provided by the SSM/I Pathfinder Project. CWV was calculated with the algorithm proposed by Wentz [1997], which has a reported error of 0.12 g·cm−2 . The AI allows the identification of several aerosol types, ranging from non-absorbing to highly UV-absorbing aerosols. The AI calculation is based on the spectral contrast of the backscattered radiances at the wavelengths of 331 and 360 nm, sensed by the TOMS instrument aboard the Earth Probe (EP) [Torres et al., 1998]. The AI data have been obtained from the NASA/GSFC TOMS Ozone Processing Team (http://jwocky.gsfc.nasa.gov/ftpdata.html). Available EP/TOMS AI fields encompass the period from July 1996 to present. The daily AI values are mapped onto a grid

1989

DIAZ ET AL.: RELATIONSHIP BETWEEN ERRORS IN PFSST AND TOMS AI ≥ 1.5

≥ 1.5

a

1.0−1.5

AI range

of 1.25◦ in longitude by 1◦ in latitude. AI positive values are associated with UV-absorbing aerosols, mainly dust, smoke and volcanic aerosols. Chiapello et al. [1999] showed that AI values were highly correlated with dust measurements made at ground level at sites in the tropical North Atlantic and with Aerosol Optical Depth (AOD) measured in dusty regions of North Africa. Negative AI values are associated with non-absorbing aerosols (e.g. sulfate and sea-salt particles) from both natural and anthropogenic sources [Torres et al., 1998]. AI values from the corresponding daily file were appended to the Pathfinder matchups. To eliminate possible suspect matchups affected by cloud contamination, radiometer digitization errors, high water vapor loading, and slightly miscalibrated buoys, we applied a series of tests described in Kilpatrick et al. [2000]. Further, we require absolute differences between buoy SST and PFSST to be 1.5◦ C) positive bias (i.e., an undercorrection of atmospheric absorption) is observed in regions associated with high content of dust aerosols (e.g. the tropical eastern North Atlantic and the northern Arabian Sea), as estimated by the AI contours. This confirms the association between areas with potentially high dust loads and under-estimation of AVHRR-derived SST retrievals. The under-correction in SST retrievals is in agreement with the PFSST residuals found in this work for AI>1. The OISST fields provide a useful reference to detect problems in AVHRR retrievals. In fact, they are probably the only reference fields with a spatial and temporal Results resolution that will allow such comparisons. There may be To explore the effects of atmospheric aerosols on satellite- concerns about comparing these fields, as the OISST fields derived SST, we computed PFSST residuals as (in situ SST include AVHRR data and thus are potentially affected by - satellite SST), where in situ SST is the bulk temperature problems in AVHRR SST retrievals. However, OISST fields measured by a buoy. For the case of dust over the Atlantic include a bias correction derived from in situ data. Even Ocean, boxplots of PFSST residuals are plotted for five AI the absence of in situ data in some region of a given OISST ranges, from AI0.5, (T4 − T5 ) values are lower than expected for a given value of CWV in the absence of aerosols (see curve for AI