AVHRR 10-day Mosaic Composite Image Data Sets for Asian Region

Transcription

1 AVHRR 10-day Mosaic Composite Image Data Sets for Asian Region Ryuzo Yokoyama *, Liping Lei **, Ts. Purevdorj ** * Asian Center for Research on Remote Sensing (ACRoRS),Asian Institute of Technology P. O. Box 4, Klong Luang, Pathumthai 12120, Thailand yokoyama@ait.ac.th ** Basic Engineering Co., LTD., Japan Takasaki East Tower 5F, Sakae-cho, Takasaki, Gunma , Japan l-lei@basic-hq.co.jp, t-purevdorj@basic-hq.co.jp Abstract The 10-day AVHRR composite image data sets for Asian region have been generated from the daytime AVHRR HPRT data received at Tokyo, Kuroshima, Ulaanbaatar and Bangkok. A Global AVHRR Image Compositing System (GAVICS) developed by Iwate University, Japan, was applied to the image processing. The GAVICS is an automatic processing system that produces single or composite AVHRR data from the raw HPRT data. The image processing chain for Asian data sets includes radiometric calibration, geometric registration, image compositing using the optimal pixels extracted from different passes and atmospheric correction. Each 10-day composite data set is consisted of 13 layers which include NDVI and SST, reflectance of channels 1 and 2, albedos of channels 1 and 2, brightness temperatures of channels 3, 4 and 5, suntarget-sensor geometry of the sensor zenith, the solar zenith and the relative azimuth angles and measurement date of each pixel selected. The datasets for 1998 and 1999 are available. Keywords: AVHRR, Composite, Asian Data Set Introduction The availability of daily Advanced Very High Resolution Radiometer (AVHRR) imagery which cover very large regions has been shown to be useful for monitoring global environmental conditions. The use of daily AVHRR data, however, has difficulties because of obtaining sufficient cloud free images. The major advantages of AVHRR image such as high frequent overpass and wide scanning provides the chance of producing cloud free images by compositing AVHRR images collected in regular intervals. The compositing procedure has been a widely accepted technique to produce cloud free composite image, and the composite AVHRR data is particularly important for continual monitoring at regional and continental scales. It has been difficult for users to collect preprocess and produce multi-temporal AVHRR data set. The global 10 days composite AVHRR data sets were produced by EROS (Vermote, et al., 1997), but it is limited for the years of On the research base, a number of composite AVHRR data sets were produced for North America, Sahelian Africa, South America, Africa and European, but there are little trials for making composite image data set for Asian region. The AVHRR multi-temporal composite data set for Asian region is important for contributing to fundamental understanding the regional environment. In this paper, we describe the Asian 10-day AVHRR composite image data sets produced by Iwate University, Japan. HRPT Data Acquisition and Coverage AVHRR High Resolution Picture Transmission (HRPT) raw data of NOAA-14 afternoon pass were daily collected from four receiving stations of Tokyo (Japan), Kuroshima (Japan), Bangkok (Thailand) and Ulaanbaatar (Mongolia). The managing organization of four receiving stations is respectively Ulaanbaatar: Ministry of Environment, Mongolia and Chiba Univ., Japan; Tokyo: University of Tokyo, Japan; Kuroshima: National Institute of Environmental, Japan; Bangkok: Asian Institute of Technology, Thailand. Figure 1 illustrated the coverage area of the AVHRR/HRPT data received at the four locations. Mosaic Composite Data Set The specification of data sets and list of available data sets are shown in Table 1 and Table 2. Data Processing The data processing for the mosaic composite data sets included: a) radiometric calibration; b) geometric correction and registration; c) image composition; d) atmospheric correction; e) calculating of NDVI and SST; f) integration of image data set. These processing procedures have been packaged into an automatic processing system - GAVICS developed by the 144 Yokoyama, R.; Lei, L. and Purevdorj, T.

2 Table 1 Description features of Asian AVHRR 10-day Composite data sets Ulaanbaatar Bangkok Tokyo Kuroshima Figure 1 Coverage and location of receiving stations. Raw AVHRR data for 10 days collected by four stations Radiometric Calibration Geometric correction and registration Image composition Atmospheric correction Calculation of NDVI and SST Integration of image data set Asian AVHRR 10-day composite data sets Figure 2 The key implementation flow of image processing for creating the Asian AVHRR 10-day composite data sets. Geographic coverage: the upper left :80 o N, 50 o E Lat./Lon. the lower right :20 o S, 180 o E Lat./Lon. Size: rows/colums Projection: Plate Caree (Latitude/Longitude grid) Resolution: Data Type: 2-byte, but NDVI and SST is 1-byte Data Format: BSQ (layer sequential) Data Layer: 13 (1) Reflectance in Channel 1 [%] (2) Reflectance in Channel 2 [%] (3) Albedo in Channel 1 [%] (4) Albedo in Channel 2 [%] (5) Brightness temperature in Channel 3 [] (6) Brightness temperature in Channel 4 [] (7) Brightness temperature in Channel 5 [] (8) NDVI [unitless] (9) Sea Surface Temperature (SST) [] (10) Solar zenith angle [degree] (11) Sensor zenith angle [degree] (12) Relative azimuth angle [degree] (13) Julian day [day] Compositing Period: approximately 10 days created by month Availability Timeline: 1998, 1999 Auxiliary Data: 1) Digital elevation data 2) Ocean lines and Lake lines 3) Sea and land mask image Table 2 List of Asian AVHRR 10-day Composite Data sets * January February - - NaN March * April NaN NaN May June July August September - NaN October November December NaN - NaN means available; NaN means unavailable; 30* means changed to 28 or 31 by month Yokoyama, R.; Lei, L. and Purevdorj, T. 145

3 Laboratory of Remote Sensing Data Analysis, Iwate University. The GAVICS is an automatic processing system that produces single produces single or composite AVHRR data from a raw HPRT data. The processing flow for the composite data set is shown in Figure 2. The major implementation features of the image processing for the data sets were as follows. a)radiometric calibration The percent albedo measured by the sensor in the channels 1 and 2 is computed as linear function of digital number (DN) as follows. A i = S i *DN+ I i Where A i is the percent albedo of channel i, S i and I i are the scaled slope and intercept values for channel i respectively. The pre-launch slope and intercept values for AVHRR channels 1 and 2 were used from source calibration information of NOAA/NESDIS. The calibration process for channels 3, 4, and 5 converts raw DN to linear radiance using following equation. RLIN= M i *DN+L i Where RLIN is the linear radiance, M i and L i are gain and intercept coefficients of channel i respectively. The quantity M i, L i (in units of radiance/count) is calculated for each channel from onboard blackbody and deep space reference. The linear radiance calculated was corrected the non-linearity as follows: R=A i * RLIN+B i * RLIN * RLIN+C i Where R is corrected radiance, A i, B i and C i are correction coefficients for channel i. Then the radiance value was converted to brightness temperature using sensor spectral sensitivity weighted method. b)geometric correction and registration Firstly, systematic geometric correction is implemented by the GCP matching and correction for terrain elevation (Purevdorj and Yokoyama, 2002). A GCP data set including about 2,129 points extracted from hydrological feature of Digital Chart of the World (DCW), is referred. The control points and coastline image were warped into the satellite projection for the correlation process. The ground control matching uses automatic correlation of image segments of the AVHRR and then integrates the adjustments derived from the correlation with the orbital model. The digital elevation data used for the correction is GTOPO30 (Global 30 Arc Second Elevation Data, USGS). The image was resampled to image of Plate Caree projection with the grid size of using the nearest neighbor interpolation method. The resulting image includes all AVHRR five channels and sun-target-sensor geometry angles data. A systematic correction using satellite model alone can not achieve required multi-temporal accuracy in the composite image. The systematic correction result showed that there is still an insufficient accuracy in the output images and it causes blurring in the composite images. Therefore, in order to prevent blurring in the compositing process, an image registration procedure was done. The image registration process is performed by cross-correlation matching binary template image from coastline image with image segment derived from the AVHRR data. Here the line matching for coastline and pattern matching for rivers was applied. The image segment for the line matching is created in the following way. Step 1: an image of 64 by 64 data is extracted from the AVHRR channels 1 and 2 centered at the ground control points (GCP) position Step 2: a binary image of land and water area is created from the image by threshold of AVHRR channels 1 and 2, and Normalized Difference Vegetation Index (NDVI) value. The pixels of the binary image have a value 1 for land area and 0 for water area. Step 3: the edge image generated from the binary image by edge extraction was correlated with template image. The image segment for pattern matching is produced as previous but edge line was not extracted. Step 4: the GCP matching result of the template and edge image was selected according to threshold matching value. Step 5: resampling is performed using a quadratic equation from selected GCP and the nearest neighbor algorithm. The template image of 32 by 32 data for the matching is generated from the coastline image centered GCP positions. The coast and shoreline data was collected from DCW and hydrological feature of rivers was generated from water mask data (EROS DATA Center). Then coastal data was transformed to coastline image with same size of the output image. c) Image composition At first, the length of the compositing period of 10 days was used, which has the advantage of a common reporting period for the agronomic and biophysical characteristics. The different compositing techniques for land and sea area are respectively proposed. The compositing image for land area is commonly produced using the maximum value compositing (MVC) technique based on NDVI (Holben, 1986). When the MVC approach was originally proposed, it was argued that the method would preferentially select near-nadir views over larger scan angles. The work that led to that recommendation was based on a model using an assumption of Lambertian reflectance from the earth surface. Numerous studies have found that off-nadir viewing can produce greater NDVI values than at nadir viewing angles (Cihlar et. al., 1994). We have observed too, that the MVC algorithm preferentially selected forescattering views over nadir for a variety of cover types in the north-east region of Asia, specially when 146 Yokoyama, R.; Lei, L. and Purevdorj, T.

4 using atmospherically corrective AVHRR data (Kawada et. al., 2000). And the cloud pixels were selected only using NDVI in the desert area. Therefore we propose to use the multiple criteria compositing technique by using the maximum brightness temperature followed by maximum NDVI and minimum scan angles (MaTNiS) (Lei and Yokoyama, 2001). The composite technique was performed in the following steps as follows. Step 1: Pixels retained among the candidate pixels if (MAX[BT4(l,p,d)] - BT4(l,p,d) < 12 [centigrade units], where is the brightness temperature of channel 4 for pixel (l, p) on date d Step 2: Pixels retained among the retained pixels by Step 1 if NDVI(l,p,d) > 0.80*MAX{NDVI(l,p,d)}, where the threshold of 0.80 indicated the variation that the NDVI depends upon sensor scan angles and was specified by trial and investigation for the AVHRR data of Asian region. NDVI (l,p,d) is calculated from apparent reflectance of Channels 1 and 2 for pixel (l,p) on date d. Step 3: Pixel with minimum scan angle over the retained pixels by Step 2 is selected as the optimal pixel. The result from this multiple criteria compositing techniques is compared with the other four compositing techniques, which are maximum brightness temperature of Channel 4 (Cihlar et. al., 1994), MVC, maximum NDVI followed by minimum scan angle (Cihlar et. al., 1994) and multiple-object composite method (Stoms et al., 1997). As result, MaTNiS technique is directed to preferentially to select the nearest nadir pixel over larger scan angle. Moreover, a compositing technique for sea surface area is developed using AVHRR Channel 2 and Channel 4. At first, the sunlit pixels were removed using threshold value (18%) of apparent reflectance of Channel 2. Then the pixel with maximum brightness temperature of Channel 4, which is less than the threshold value of Channel 2, was selected for the composite. d) Atmospheric correction. The impact of atmospheric effects on AVHRR composite imagery of Channels 1 and 2 still is significant, although the compositing has been used to remove cloud contamination and atmospheric effects. An atmospheric correction algorithm for AVHRR imagery in the operational mode is used (Lei and Yokoyama, 1997). The algorithm works with look-up tables (LUTs) of the atmospheric corrective parameters derived from 6S code (Vermote et al., 1997) and maintained in a production environment. However this algorithm is validity when solar zenith angle and sensor zenith angle is less than 60 and 50 degree respectively in the input geometrical parameters. For atmospheric profile models, a standard climatology with latitudinal, seasonal dependence is used, and the aerosol model was taken to be continental. The elevation data corresponding to imagery area was taken from GTOPO30. The output of atmospheric correction for every pixel is the surface reflectance of Channels 1 and 2. For the atmospheric conditions, Subarctic summer and winter, Midlatitude summer and winter, and Tropical atmospheric profile are used for the pixels located between 80 o N and 50 o N, 50 o N and 23 o N, and between 23 o N and 20 o S respectively. The aerosol concentration is taken in visibility 60km (assumed very clear atmosphere). e) Calculation of NDVI and SST NDVI is computed by using the atmospherically corrected reflectance of channels 1 and 2. SST is computed from the brightness temperature of channels 4 and 5 by using the split window method. f) Integration of image data set The image implemented above the processing is lastly integrated into a data set including 13 layers as shown in Table 1. Figure 3-4 give the examples of the color composite image of channels 1 and 2, NDVI and SST for Aug , 1999 of data sets. The processing time for single data set was about 42 hours when the composite image was automatically generated from 80 scenes of a raw HRPT data using a workstation HP9000 (Model-C240). The datasets for 1998 and 1999 (Table 2) are available now through Satellite Environment Database of Tokyo University. Conclusion The data sets of Asian 1-km composite imagery have been created from a raw 1-km AVHRR data received from ground receiving stations of Tokyo University, Kuroshima (Japan), AIT (Thailand) and Ulaan-baatar (Mongolia). The composite images will contribute to research activity in environmental analysis of Asian region. During the development of AVHRR composite data set, it has been showed difficulties in the data processing because of frequent cloud coverage over large areas and geometric distortion at large scan angle associated with the edges of the across-track scan. Despite the number of difficulties, the improvement of the data processing method should lead to increase a quality of the composite images. Acknowledgments The research is partially supported by Research and Development Applying Advanced Computational Science and Technology, Japan Science and Technology Corporation and Telecommunications Advanced Organization of Japan. Yokoyama, R.; Lei, L. and Purevdorj, T. 147

5 References B.N.Holben, 1986, Characteristics of maximun-value composite images from temporal AVHRR data, INT.J. REMORTE SENSING, 7 (11), pp David M.Stoms, MiChael J.Bueno,and Frank W.Davis, 1997, Viewing geometry of AVHRR image composites derived using mulutiple criteria, Photogrammetric Engineering & Remote Sensing, 63 (6),pp J.Cihlar,D.Mnak,and M.D Iorio, Evaluation of Compositing Algorithms for AVHRR Data Over Land, IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 32 (2), pp J. Cihlar and F. Huang,, 1994, Effect of atmospheric correction and viewing angle restriction on AVHRR data composites. Canadian Journal of Remote Sensing, 20(2), pp Figure 3 Asian AVHRR 10-day composite image (ch.1:ch.2:ch.1=r:g:b) for August, 1999 Liping Lei, Mituo Koide, Yoshikazu Iikura, and Ryuzo Yokoyama, 1998, Correction of Atmospheric effects on AVHRR Imagery by 6S code, Journal of Remote Sensing Society of Japan, 18 (2), pp Liping Lei, and Ryuzo Yokoyama, 2001, A composite algorithm of NOAA/AVHRR Mosaic image for the total Asian Region (Part 2): Evaluation of algorithms for extracting the optimal observing pixel, Journal of the remoste sensing Society of Japan, Vol.21, No.1, pp Rie Kawada, Liping Lei, and Ryuzo Yokoyama, 2000, A composite algorithm of NOAA/AVHRR Mosaic image for the total Asian Region (Part 1): Evaluation of bidirectional reflectance and atmosphric effects, Journal of the remoste sensing Society of Japan, Vol.20, No.4, pp Figure 4 Asian AVHRR 10-day composite image (NDVI) for August, 1999 Teillet P. M.et al., 2000, An evaluation of the global 1- km AVHRR land dataset, International of Remote Sensing, Vol.21, No.10, pp Ts.Purevdorj and R.Yokoyama, 2002, An approach to automatic detection of GCP for AVHRR imagery, Journal of the Japan Society of Photogrammetry and Remote Sensing, Vol. 41, No.1, pp Vermote, E. F., et al., 1997, Second simulation of the Satellite Signal in the Solar Spectrum, 6S: an overview, IEEE Trans. Geosci. Remote Sensing, Vol. 35, No. 3, pp Figure 5 Asian AVHRR 10-day composite image (SST) for August, Yokoyama, R.; Lei, L. and Purevdorj, T.