35017 Las Palmas de Gran Canaria, Spain Santa Cruz de Tenerife, Spain ABSTRACT

Transcription

1 Atmospheric correction models for high resolution WorldView-2 multispectral imagery: A case study in Canary Islands, Spain. J. Martin* a F. Eugenio a, J. Marcello a, A. Medina a, Juan A. Bermejo b a Institute for Oceanographic and Global Change, Campus Universitario de Tafira, 3517 Las Palmas de Gran Canaria, Spain b Fundación Observatorio Ambiental de Granadilla, Edificio Puerto-Ciudad 1B, 381 Santa Cruz de Tenerife, Spain ABSTRACT The emergence of high-resolution satellites with new spectral channels and the ability to change its viewing angle has highlighted the importance of modeling the atmospheric effects. So, atmospheric correction serves a critical role in the processing of remotely sensed image data, particularly with respect to identification of pixel content. Efficient and accurate realization of images in units of reflectance, rather than radiance, has proven to be a crucial point in the pre-processing of images in remote sensing applications, acquired under a variety of measurement conditions. However, reflectance of the objects recorded by satellite sensors is generally affected by atmospheric absorption and scattering, sensor-targetillumination geometry, and sensor calibration. These normally result in distortion of the actual reflectance of the objects that subsequently affects the extraction of information from images. The use of atmospheric models has significantly improved the results of the corrections. In this study we have proceeded to make the atmospheric correction of the eight multispectral bands of high resolution WorldView-2 satellite by three different atmospherics models (COST, DOS, 6S) defining the geometry of the satellite observation, viewing angle and setting the weather conditions more suited for the acquired images of the study area (Granadilla, Canary Islands). For this purpose, the reflectance obtained by COST, DOS and 6S atmospheric correction techniques are compared with the Top of Atmosphere (TOA) reflectance. Specifically, the 6S atmospheric correction model, based on radiative transfer theory, provides patterns which describe properly atmospheric conditions in this specific study area for monitoring turbid coastal environments. To check the proper functioning of the atmospheric correction comparison was performed between ground-based measurements and corresponding obtained by the eight multispectral satellite channels through the 6S atmospheric model, with similar date, weather and lighting conditions. Keywords: high resolution satellite images, WorldView-2, atmospheric correction models, COST, DOS, 6S, coastal environmental. 1. INTRODUCTION A major benefit of multitemporal remotely sensed images is their applicability to change detection over time. However, to maximize the usefulness of the data from a multitemporal point of view, an easy-to-use, cost-effective, and accurate radiometric calibration and atmospheric correction procedure is needed. The atmosphere affects the radiance received at the satellite by scattering, absorbing, and refracting light; corrections for these effects, as well as for sensor gains and offsets, solar irradiance, and solar zenith angles, must be included in radiometric and atmospheric corrections procedures that are used to convert satellite-recorded digital counts to ground reflectances. To generate acceptable atmospheric correction results, a model is required that typically uses in-situ atmospheric measurements and radiative transfer code (RTC) to correct for atmospheric effects. The main disadvantage of this type of correction procedure is that it requires in-situ field measurements during each satellite over flight. This is unacceptable for many applications and is often impossible, as when using historical data or when working in very remote or difficult access locations 1. *jmartin@teledeteccioncanarias.es; phone ; fax

2 The atmospheric correction has proven to be a crucial point in the pre-processing of high resolution images that can affect subsequent steps in remote sensing applications of satellite data. For instance, the need for an optimal atmospheric correction model for monitoring and managing of littoral zone environments for, i.e., water quality monitoring, benthic habitat mapping and remote bathymetry. The spectral information provided by the eight Worldview-2 bands within the visible and infrared spectrum increase the amount of spectral data available for our area under study, thereby improving the quality of coastal environmental products. The main goal of this study is to identify an optimal atmospheric correction model for multispectral WorldView-2 channels for estimating atmospheric reflectance and remove it effects from apparent reflectance leaving from the water surface and the sea floor. With this purpose we compare the capabilities of three well-known atmospheric correction models: The COST image-based algorithm, Dark Object Subtraction technique (DOS) and Second Simulation of a Satellite Signal in the Solar Spectrum (6S). In the following sections, the effects of these atmospheric correction methods applied to high resolution satellite remote sensing multispectral bands images obtained from WorldView2 sensor are analyzed. 2. STUDY AREA AND MATERIALS 2.1 Study Area The study area is in the south part of Tenerife Island, near Granadilla City (Figure 1). The Port of Granadilla Environmental Monitoring Programme was established in 21 in order to ensure sustained environmental quality across the wide range of natural and artificially created habitats within and immediately outside of the Port. So, turbid coastal environments and optically shallow waters need to be studied and regularly analyzed within a strategic plan for environmental monitoring. (a) (b) Figure 1. Study area.(a) Location (28.7 N, W) and, (b) Granadilla area WV2 image acquired on February 18, 212 superposed in goggle map.

3 2.2 In-situ measurements Granadilla area has a water quality monitoring network in place for two years. Water samples at fixed stations and samples collected by routine ships are analyzed weekly, monthly and seasonally now. The analysis result has provided water quality data regarding Chl-a, TSM, CDOM. To evaluate the results generated by the various atmospheric models, we used groundbased spectral data collected by the spectroradiometer Vis/NIR ASD FieldSpec 3 nearly coincident with WorldView-2 satellite over flight (see Figure 8, section 4). The spectroradiometer main characteristics are: spectral range: 35 nm 25 nm, spectral resolution: 3 7 nm, 3 21 nm and sample interval: 1,4 7 nm y 2 21nm. 2.3 WorldView-2 imagery The WorldView-2 high-resolution commercial imaging satellite was launched on October 8, 29, from Vandenberg AFB, and was declared to be operating at Full Capability on January 4, 21. The satellite is in a nearly circular, sun-synchronous orbit with a period of 1.2 minutes at an altitude of approximately 77 km, and with a descending nodal crossing time of approximately 1:3 a.m. WorldView-2 acquires 11-bit data in nine spectral bands covering panchromatic, coastal, blue, green, yellow, red, red edge, NIR1, and NIR2. Each sensor is focused to a particular part of the electromagnetic spectrum to be sensitive to a specific type of feature on the ground or property of the atmosphere 2. The spectral response and the central wave length of each band are shown in Figure 2 and Table 1, respectively. Figure 2. WorldView-2 relative spectral radiance response (nm). Table 1. WorldView2 band passes spectral bands (µm). Bands Center Wavelength 5% Band Pass 5% Band Pass Panchromatic Coastal Blue Green Yellow Red Red Edge NIR NIR

4 This work relied on Ortho Ready Standard Worldview-2 images. The images were captured monthly from August 211 (at the present time 12 images have been obtained). At nadir, the collected nominal ground sample distance is.46 m (panchromatic) and 1.84 m (multispectral), however, commercially available products are resampled to.5 m (panchromatic) and 2. m (multispectral). The nominal swath width is 16.4 km. In Figure 3, four high resolution Ortho Ready Standard images acquired by WorldView-2 satellite in different seasons are shown. (a) (b) (c) (d) Figure 3. WorldView-2 imagery acquired of Granadilla Port area in different seasons: (a) August 211, (b) October 211, (c) December 211 and (d) March COMPARISON OF ATMOSPHERIC CORRECTION METHODS The atmospheric correction algorithms for processing remotely sensed data from low resolution sensors (p.e. MODIS, SeaWiFS, MERIS) were primarily designed for retrieving water-leaving radiances in the visible spectral region over deep ocean areas where phytoplankton is the dominant water constituent ( Case 1 waters), where the water-leaving radiances are close to zero. For turbid coastal environments and optically shallow waters ( Case 2 waters), water-leaving radiances may be significantly greater than zero because of backscattering by suspended materials in the water and bottom reflectance 3. Hence, applications of the Case 1 algorithm to satellite imagery acquired over turbid coastal waters often result in negative waterleaving radiances over extended areas. Therefore, improved atmospheric correction algorithms must be developed for the remote sensing of Case 2 waters. The remote sensing from high-resolution satellites or airborne platform captures surface of land or sea in the visible and near infrared which is general affected by sensor-target illumination geometry and sensor calibration and strongly affected by the appearance of the atmosphere on the Sun-target-Sensor path. The qualitative and quantitative accuracy retrieved from remotely sensing images requires removing the effects by scattering and absorption in the atmosphere, reflection at the sea

5 surface and the measured top-of-atmosphere (TOA) radiances. Such procedures are called atmospheric corrections 4. The needs to correct the effect of the atmosphere are often a critical first step that can affect subsequent steps in the application of satellite data. As a result, we decided to compare the effects of three atmospheric correction methods, specifically, DOS, COST and 6S techniques. These three represented a range of levels of sophistication in correction algorithms and were found to be the most often recommended by researchers and analysts. 3.1 The DOS method The dark object subtraction technique (DOS) is an image-based model that has been proposed to simplify atmospheric correction 1,5. The DOS model is based on the assumption that dark objects exist within an image and have zero reflectance. Consequently, the radiance resulting from corresponding pixels is proportional to the atmospheric path radiance, and can be used to account for the additive effects of atmospheric scattering. The minimum pixel values are selected for each individual band with the histogram method and subtracted from all pixel values for the corresponding band across an image. Thus, the path radiances determined in this way are spectrally uncorrelated. To explain the dependency of atmospheric scattering on wavelength, an improved DOS technique was developed to estimate path radiances with selected relative atmospheric scattering models. The estimated path radiances for all spectral bands with the improved DOS technique are spectrally correlated. The following equation shows how the reflectance is calculated:,, (1) cos Where L sen,λ is the radiance sensed for the satellite, L haze,λ is the minimum radiance values obtained from the histogram of each band, d 2 is the squared distance between the surface and the satellite, E λ is the irradiance of the band, and is the incidence angle. 3.2 The COST method The COST method combines the assumption DOS method with the fact that very few objects on the earth's surface are quite dark. Thus, normally corresponds to 1% of the full reflectance image. The radiance of an object absolutely dark, when it is free of shade is as follows: %,. (2) Where %, is the 1% radiance of the dark object assumption. Then, the radiance is converted to reflectance of the objects at the Earth s surface using the following formula: 3.3 The 6S method,, %, (3) 6S (Second Simulation of a Satellite Signal in the Solar Spectrum) is an advanced radiative transfer code designed to simulate the reflection of solar radiation by a coupled atmosphere-surface system for a wide range of atmospheric, spectral and geometrical conditions 6. It belongs to the group of procedures called atmospheric correction for the process of removing the effects of the atmosphere on the reflectance values of images taken by satellite sensors. The code operates on the basis of an SOS (successive orders of scattering) method and accounts for the polarization of radiation in the atmosphere through the calculation of the Q and U components of the Stokes vector. 6S application uses FORTRAN programming language to model the atmospheric radiation transfer utilizing from the visible to short wave infrared range. This model predicts the reflectance ρ of objects at the top of atmosphere (TOA) using information about the surface reflectance and atmospheric conditions 7. This information is provided through a minimum of input data to the model and incorporated features. The TOA reflectance (, can be estimated using the following expression:

6 ,, (4) The minimum data set needed to run the 6S model is the meteorological visibility, type of sensor, sun zenith and azimuth angles, date and time of image acquisition, and latitude-longitude of scene center. In this study we have proceeded to correct the eight-band multispectral and panchromatic band of WV2 by 6S model, defining the geometry of the satellite observation and viewing angle. The pass filters of the 9 bands have been defined and the weather conditions more suited to the study area (South of Tenerife, Canary Islands) were properly set. Figure 4 shows the 6S input parameters configuration file for WorldView2 coastal channel, which defines the correction for sea surface directional effects (wind, salinity and pigments). Figure 4. 6S input file configuration. Using the input data and the embedded features, the model produces variables to assess the surface reflectance. The true reflectance value ρ λ is obtained from the model output by the following expression, (5) where ρ λ is the corrected reflectance, X a, X b, and X c are the coefficients obtained from the model (X a is the inverse of the transmittance, X b is the scattering term of the atmosphere and, X c is the reflectance of the atmosphere for isotropic light), and L λ is the observed radiance (w/m 2 *sr*µ m ). As mentioned, the purpose of the study is to compare these three atmospheric correction techniques on WorldView-2 image. For this purpose, the profiles obtained by COST, DOS and 6S, using the same atmospheric and aerosol conditions, are compared with the Top of Atmosphere (TOA) reflectance, as explained in the next section. TOA reflectance does not consider atmospheric effects at all, so it is useful to evaluate the performance of the models, especially in removing water vapor absorption effect in WorldView-2 image. 4. RESULTS In order to perform appropriate comparisons of the atmospherically corrected images, six representative targets locations were selected within the test site. A detailed sampling strategy is applied considering that all of the targets should be taken from uniform and well distributed points as much as it could be possible. Two of the targets (T1 and T2) are from land areas (ground and building roof). Targets T3, T4, T5 and T6 are collected from coastal areas (beach, port, turbid and clear water), respectively, as shown in Figure 5.

7 T3 T1 T2 T5 T4 T6 Figure 5. Location of land and coastal targets on WordView-2 imagery in Granadilla area. The results obtained by COST, DOS and 6S atmospheric correction techniques (% reflectivity units) on WorldView-2 image, compared with the Top of Atmosphere (TOA) reflectance, are presented in Figure 6. Although it appears to be little difference between the results of different atmospheric correction methods, in the case of the water reflectivity, there are large variations due to their low reflectivity, being essential an adequate atmospheric correction. The radiative modeling of 6S seawater enables better results than DOS and COST methods than usually underestimate the reflectivity of the dark elements like water. For this reason it was decided to use the 6S model Coastal WV2 Band TOA 4,97 16,67 15,46 17,88 16,77 16,25 DOS 26,71 2,41 1,2 3,62 2,5 1,99 COST 39,13 4,82 3,11 6,52 4,95 4,23 6S 38,73 7,45 5,77 9,11 7,58 6, Blue WV2 Band TOA 52,25 14,46 12,51 16,22 15,54 13,83 DOS 41,15 3,37 1,41 5,13 4,48 2,73 COST 59,52 6,17 3,41 8,65 7,69 5,27 6S 52,99 8,81 6,33 11,2 1,17 8,1

8 Green WV2 Band Yellow WV2 Band TOA 55,29 12,1 7,73 12,87 11,77 9, TOA 55,79 11,28 5,5 9,26 7,66 6,29 DOS 48,77 5,48 1,2 6,35 5,24 3,38 DOS 51,55 7,4,81 5,2 3,42 2,5 COST 7,28 9,16 3,11 1,38 8,82 6,18 COST 74,2 11,35 2,56 8,5 6,24 4,31 6S 59,37 9,34 4,3 1,4 9,4 6,74 6S 62,78 1,32 2,52 7,8 5,8 4, Red WV2 Band Red Edge WV2 Band TOA 58,46 11,52 4,71 7,43 7,12 5,27 TOA 51,93 11,51 2 5,25 5,18 3,72 DOS 55,18 8,25 1,43 4,15 2,84 1,99 DOS 49,36 9,5,53 2,79 2,71 1,25 COST 79,34 13,6 3,43 7,27 6,83 4,23 COST 71,11 14,19 2,16 5,35 5,25 3,18 6S 62,91 1,63 2,64 5,84 5,48 3,31 6S 61,85 12,38 1,46 4,37 4,28 2, NIR1 WV2 Band TOA 53,19 13,56 2,69 4,18 4,66 3,12 DOS 51,46 11,83,96 2,45 2,93 1,39 COST 74,8 18,12 2,77 4,87 5,55 3,38 6S 57,66 13,89 1,47 3,19 3,74 1, NIR2 WV2 Band TOA 38, ,88 2,71 3,54 2,36 DOS 37,67 1,73,61 1,44 2,27 1,9 COST 54,61 16,56 2,27 3,45 4,62 2,95 6S 52,84 15,62 1,23 2,43 3,62 1,92 Figure 6. Results of atmospheric corrections models COST, DOS, 6S and TOA reflectivity for land-sea targets in the area under study. As a case of study, in the Figure 7 can be observed how the change in turbidity of the water (targets T3, T4, T5 and T6) varies the reflectivity of the eight Worldview-2 channels. The variation of turbidity produced greater variations in the visible, especially in the blue and green channels. Furthermore, differences in the red and infrared (NIR1 and NIR2) are low.

9 S Water Surface Reflectance Coastal Blue Green Yellow Red Red Edge NIR1 T3 Clear Water 5,77 6,33 4,3 5,5 2,64 1,46 1,47 1,23 T4 Turbidity 1 9,11 11,2 1,4 9,26 5,84 4,37 3,19 2,43 T5 Turbidity 2 7,58 1,17 9,4 7,66 5,48 4,28 3,74 3,62 T6 Turbidity 3 6,87 8,1 6,74 6,29 3,31 2,39 1,97 1,92 NIR2 Figure 7. Variation of water surface reflectivity, obtained by 6S atmospheric model, due the turbidity in coastal waters. Finally, in order to check the proper functioning of the selected 6S atmospheric correction algorithm, ground-based reflectance measurements were performed for a variety of points (see section 2.1 for radiometer specifications), with similar weather and lighting conditions. Figure 8 (a) shows the WorldView-2 image of Granadilla area where in-situ radiometric test points were obtained. The results obtained by 6S atmospheric correction techniques (% reflectivity) on WorldView-2 image (bandwidths close to 5 nm, see Table 1), compared with ground-based reflectance measurements (radiometer step 1 nm), are presented in Figure 8 (b). As it can be observed, the results show a great correlation between the reflectivity values obtained by in-situ measurements and the corresponding obtained by the eight multispectral satellite channels through the 6S atmospheric model.. (a) Figure 8. (a) Location of in-situ test points on WordView-2 imagery in Granadilla area (February 212) and, (b) ground-based reflectance measurements (top) and corresponding WV2 multispectral 6S atmospheric correction reflectance (bottom). (b)

10 5. CONCLUSIONS Atmospheric correction is actually considered as the second step of the radiometric correction, following the radiance calibration. It has proven to be a crucial point in the pre-processing of images in remote sensing applications. The goal of this procedure is to transform the at-sensor radiance into ground/water-leaving radiance. During the study, it is proposed to compare and evaluate the success of three different atmospheric corrections for WorldView-2 multispectral data beyond the scope of turbid coastal environments mapping within a strategic plan for Granadilla Port environmental monitoring. We chose to compare the effects of DOS, COST and 6S atmospheric correction methods and it was found that all algorithms performed successful the overall evaluation but 6S is found to be a better corrector algorithm for turbid coastal environments and optically shallow waters. Also this comparison over arid and vegetated land cover types and over water surfaces have been carried out by Samadzadegan et al 8 obtaining in their study that 6S model gives better results over water surfaces than FLAASH (Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes) and Atmospheric CORrection (ATCOR) atmospheric correction algorithms which are also based on Radiative Transfer (RT) code. The most common method for evaluating atmospheric correction is to compare the retrieved reflectance from satellite images with ground-based measurements for a variety of targets. So, finally we compared the RT 6S model with coincident groundbased reflectance measurements in the area under study areas obtaining a very good correlation between the reflectivity values obtained by in-situ measurements and the corresponding acquired by atmospheric processing of the eight multispectral satellite channels. The favorable results from this study for optimal atmospheric correction model have spawned a follow-up project that consists on the upgrade of the developed open ocean oceanographic techniques ( Case 1 waters) to study water quality, benthic habitat and remote bathymetry for monitoring and managing of littoral zone ( Case 2 waters) environments with high resolution imagery. 6. ACKNOWLEDGEMENTS This work has been supported by the Programa Innova Canarias 22 of the Fundación Universitaria de Las Palmas. Ground-based measurements was supported by the project MICINN CGL C2. REFERENCES [1] Pat S. Chavez, Jr., "Image-Based Atmospheric Corrections. Revisited and Improved," Photogrammetric Engineering & Remote Sensing, 62(9), (1996). [2] Todd Updike, Chris Comp., "Radiometric Use of WorldView-2 Imagery. Technical Note," DigitalGlobe, (21). [3] Robert, A. Schowengerdt., [Remote Sensing. Models and Methods for Image Processing], Academic Press: Elsevier, London, (27). [4] Abdolrassoul, S. Mahiny, Brian, J. Turner, "A comparison of four common atmospheric correction methods," Photogrammetric Engineering & Remote Sensing, 73(4), ( 27). [5] Gilabert, M. A., Conese, C., Maselli, F., An atmospheric correction method for the automatic retrieval of surface reflectances from TM images, International Journal of Remote Sensing, 15(1), (1994). [6] Svetlana, Y. Kotchenova, Eric, F. Vermote, Raffaella, M., Frank, J. Klemm, Jr., Validation of vector version of 6s radiative transfer code for atmospheric correction of satellite data. Part I. Parth radiance, Applied Optics, 45(26), (26). [7] Vermote E., D. Tanré, J. L. Deuzé, M. Herman, J. J. Morcrette, S. Y. Kotchenova., Second Simulation of a Satellite Signal in the Solar Spectrum Vector (6SV), 6S User Guide Version 3, (26). [8] Farhad Samadzadegan, Seyed Hossein Seyed Pourazar, Mahdi Hasanlou, Comparative Study of Different Atmospheric Correction Models on Worldview-2 Imagery, XXIII Congress of the International Society for Photogrammetry and Remote Sensing, Melbourne (Australia), (212).