A Comparison of OCI and Sea WiFS Satellite Imagery in the Waters Adjacent to Taiwan

With the successful launch of the first scientific satellite of the Repub­ lic of China (ROCSAT-1) on 27 January 1999, Taiwan enters the era as a satellite data providing country. There are three scientific payloads on the satellite. One of them is the Ocean Color Imager (OCI). OCI is a push­ broom reflective imager for monitoring ocean colors. OCI has taken im­ ages since March 1999. Because there is no on-board calibrator on OCI, a comparison with Sea-viewing Wide Field-of-view Sensor (SeaWiFS) data was performed in order to validate OCI data. Simultaneously in-situ ob­ servations of optical properties and chlorophyll a concentration were also collected in waters adjacent to Taiwan for vicarious calibration. We ap­ plied Sea WiFS atmospheric correction and bio-optical algorithms on OCI data to derive normalized water-leaving radiance and chlorophyll a con­ centration. Results show that the chlorophyll a concentration derived from OCI is generally larger than that derived from Sea WiFS. The correlation coefficient is about 0.60 with a root-mean-squared (RMS) of difference of chlorophyll a concentration of 0.10 mg/m3•

trum. Measurements of water-leaving radiance allow concentrations of chlorophyll a to be derived.
After the successful launch of SeaWiFS in August 1997 has provided a very good oppor tunity to perform in-flight calibration and validation for both sensors by the intercomparison method, especially for OCI because it has no on-board calibrator. For sensors without on-.
board calibration capabilities, indirect methods are the only ways available to monitor calibra tion coefficients while the instruments operate in orbit. The intercomparison work between OCI and Sea WiFS is also a part of the Sensor lntercomparison and Merger for Biological and Interdisciplinary Oceanic Studies (SIMBIOS) project, organized by the National Aeronautics and Space Administration (NASA).
For calibration and validation purposes, images are scheduled to be taken simultaneously when OCI tracks overlap with Sea WiFS, or the calibration buoy location or the location of the experimental cruises. To achieve radiance from the ocean, space-borne ocean color sensors require several levels of radiometric calibration. It includes gain-offset correction, which con verts digital counts into at-sensor radiance (Schowengerdt, 1997) and atmospheric correction, which transforms at-sensor radiance into water-leaving radiance. In this study, we applied the SeaWiFS atmospheric correction and bio-optical algorithms on OCI data, and then compared the results with quasi-simultaneous SeaWiFS data. The results help us to better understand the quality of OCI and its difference from SeaWiFS.

DESCRIPTION OF OCI
OCI is a push-broom reflective and nadir-viewing imager. It is designed to map reflected spectral radiance from ocean surfaces in seven visible and near-infrared (NIR) bands. The seven spectral barids actually cover six independent wavebands. The central wavelengths of the seven bands are 443 nm, 490 nm, 510 nm, 555 nm, 670 nm, 865 nm and 555 nm, and are denoted as Band 1, 2, 3, 4, 5, 6 and 7, respectively. The Band 7 is a redundancy of Band 4. It is used for inter-track radiance and calibration purposes. Since the constraints on total mass and envelope of ROCSAT-1, OCI has been designed and implemented with seven spectral bands to share four telescopes. Bands 1 and 3 get radiance from the same telescope. Bands 2 and 4 and Bands 5 and 6 are another two pairs. Band 7 has a stand-alone telescope. A compari son of characteristics between SeaWiFS and OCI is listed in Table 1. OCI f ocal planes are of Thomson's 1728-element linear charge-coupled device (CCD) .
sensors. This kind of sensor is designed to have a distinctive anti-blooming feature. With a nominal telescope f ocal length of 19 .5 mm and a photosite pitch of 13 µm square, an instanta neous field of view (IFOV) covers 2 CCD pixels. Therefore, an 800 by 800 square meter footprint on the ground is formed by the push-broom action. The swath of OCI is about 700 km. A detailed de�cription of OCI characteristics can be found in Lee et al. (1999).

VALIDATION
It is believed that the OCI calibration at each wavelength will change in some unpredict-  (Mueller and Austin, 1992). To validate OCI data, in situ mea surements of optical properties were carried out with a Tethered Spectral Radiometer Buoy (TSRB-II) and a SeaWiFS Profiling Multi-channel Radiometer (SPMR) made by Satlantic Company. TSRB-11 is an optical buoy, which can measure the in-water upwelling radiance just beneath the sea surface Lu (0-, ll) and the incident spectral irradiance above the sea sur face £3(/l). SPMR is a profiling radiometer that can measure the in-water upwelling radiance L,,(z,ll) and downwelling spectral irradiance E/z,ll) with depth z. Originally these instru ments were developed for SeaWiFS data validation. Since OCI has similar band characteris tics to SeaWiFS, we can also used these instruments for OCI data validation. All of these in situ measurement instruments have the same wavelength in visible bands as those of SeaWiFS and OCI.
To obtain Lu(z,ll) that is measured by the OCI, it is necessary to propagate Lu (O-,ll) upward through the sea surface as where p(A,6) and nw(A) are the Fresnel reflectance at the solar zenith angle e and the refractive index for seawater, respectively. In order to remove the influence of view angle, sun angle and the solar irradiance, the water-leaving radiance is generally transferred to the nor malized water-leaving radiance Lwn as (2) where F;; (ll) denotes the mean extraterrestrial solar irradiance (Neckel and Labs, 1984). For calculating the remote sensing reflectance just above the sea surface Rr. �(ll), the following equation was used (O'Reilly et al., 1998) The bio-optical algorithm for chlorophyll a concentration calculation from Rr s(ll) is given by where C is chlorophyll a concentration, and R is defined as R= log(Rr.1 .(4 9 0)/ Rr. /555)).

Sea WiFS Algorithms
The total upward radiance at the top of the ocean-atmosphere system, measured at a wave length I, can be written as where Lr(A) is the radiance from multiple scattering by air molecules, L0(1l) is the radiance from multiple scattering by aerosols, Lr,J:l.) is multiple interaction between molecules and aerosols, and 4c(ll) is the radiance at the sea surface that arises from sunlight and skylight reflecting from whitecaps on the surface (Gordon and Wang, 1994). The t(lt) is the diffuse transmittance which !lCcounts the effects of propagating water-leaving radiance and whitecap reflectance from the sea.surface to the top of the atmosphere. Since SeaWiFS has a tilt func-tion to avoid the effect of sun glitter, in Eq. (6) the surface sun glitter term has been ignored. For OCI, the term of sun glitter has to be taken into account, because it has no tilt function.
The goal of the atmospheric correction is to retrieve the Lw (ll ) accurately from the spec tral measurements ofradiance at the satellite. To relate the derived Lw(A.) to the ocean inher ent optical properties, the atmospheric effects on the . Lw(A.) must removed. The 4CA.) can be defined from Gordon and Clark (1981) where t(A.,8) is the atmospheric diffuse transmittance in the solar direction with the solar zenith angle. To determine the term of t(A.,8), the atmospheric correction algorithm for SeaWiFS was developed by the SeaWiFS team (Wang and Gordon, 1994). After doing the atmospheric correction on Sea WiFS data, the· Lwn (A.) measured from Sea WiFS was then de rived.

Match-up Data Set
There were six optical data and water samples taken in the waters adjacent to Taiwan from October 1998 to June 1999. These data were measured when Sea WiFS flights over the filed measurement areas within three hours. There are three data sets whose time lag is less than one hour. The information related to in-situ measurements and SeaWiFS images are listed in Table 2    measurements are high, especially for chlorophyll a concentration. The correlation coefficient of chlorophyll a concentration is as high as 0.994, when the difference in measuring time between SeaWiFS and in-situ is within one hour. Therefore we will compare the chlorophyll a concentration derived from OCI with that derived from SeaWiFS.

COMPARISON WITH SeaWiFS DATA
A way to validate OCI data products is by intercomparison with data from other space borne ocean color sensors. The difficulties of this method, however, are spectral matching, ground spatial resolution, and sun-sensor geometry (Che, 1991 of which have the same view angle and sun angle at the nadir point. We can not compare the total radiance received by the ocean color sensors. For intercomparison, we adopted quasi simultaneous images and assumed that parameters in the ocean do not change much within a short time. Thus, if we apply the SeaWiFS atmospheric correction algorithm and the bio optical algorithm on OCI, we can intercompare the chlorophyll a concentration, which is not affected by the sensor view angle and sun angle.
As we know, different ocean color sensors usually have different band spectral character izations. It is difficult to apply one atmospheric correction algorithm to other sensors. Accord ing to the evaluation of accuracy of the Sea WiFS atmospheric correction algorithm for various ocean color sensors (Wang, 1999), the SeaWiFS atmospheric correction can be applied to other color sensors for solar zenith angles less than 60°. Therefore, we can perform the SeaWiFS atmospheric correction algorithm on OCI data and have the same accuracy as SeaWiFS, if we have pixels on the OCI images in which sun zenith angles are less than 60°.    After performing geometrical correction and the Mercator projection, we produced the same spatial resolution for SeaWiFS and OCI images. We performed the SeaWiFS atmo spheric correction algorithm on OCI in order to do the intercomparison. Only pixels in which sun zenith angles were less than 60° were used for comparison. We then applied the SeaWiFS bio-optical algorithm to OCI data to derive the chlorophyll a concentration. There were four pairs of chlorophyll a concentration from OCI and SeaWiFS for intercomparison. These imag eries were taken in the waters adjacent to Taiwan. The range of chlorophyll a concentration is from 0.0 1 mg/m3 to 3.0 mg/m3.

DISCUSSION AND CONCLUSIONS
OCI is the first space-borne ocean color sensor in Taiwan. From the OCI images obtained, we found that the quality of OCI images is acceptable. Oceanic features on SeaWiFS images can also be found on OCI images. OCI data qualitatively bear comparison with SeaWiFS data. From the quantitative comparison we found that the chlorophyll a concentration derived from OCI using Sea WiFS atmospheric correction and bio-optical algorithms is larger than that de rived from SeaWiFS data. Generally speaking, the correlation coefficient of chlorophyll a concentration between OCI and SeaWiFS is 0.60. We believe that the sensor calibration and the difference of imaging time caused the difference between OCI and SeaWiFS. To calibrate and validate OCI data, in-situ measurements of optical properties and chlorophyll a concentra tion are still necessary. The comparison of OCI with Sea WiFS is also a good way for OCI data validation. The engineers of NSPO and the OCI science team are still working on the sensor