This paper summarizes the geometric and radiometric calibration results of the PALSAR achieved during the ALOS initial calibration phase, which covers five months between May 16, 2006, and October 23, 2006, and the half year of the operational phase. All the PALSAR modes, FBS (fine beam single), FBD (Fine beam dual), SCANSAR, DSN (band limited SAR), and PLR (Full polarimetry) were calibrated and validated using in-total 500 calibration points collected worldwide and distributed target data from the Amazon. Through the characterization of the PALSAR, antenna pattern determination, and polarimetric calibration, we performed the adjustments of the PALSAR radiometric and geometric model installed on the SAR processor (SIGMA-SAR). Using the reference points, we finally confirmed that the geometric accuracy of the FBS, FBD, DSN, and PLR modes is 9.3 m, that of SCANSAR is 70 m, and radiometric accuracy is 0.64 dB. Polarimetric calibration was successful that amplitude balance of VV/HH is 0.025 dB and the phase balance is 0.32 degrees.
The Advanced Land Observing Satellite (ALOS, nicknamed “Daichi”) was successfully launched on January 24, 2006. This paper introduces the calibration and validation plan and their initial results of two optical sensors, the Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) and the Advanced Visible and Near Infrared Radiometer type-2 (AVNIR-2) onboard ALOS, and accuracy assessments during one year after launch. PRISM consists of three independent panchromatic radiometers and is capable to derive a digital surface model (DSM) with high spatial resolution, which is an objective of the ALOS mission. Thus, geometric calibration is important in generating a highly accurate DSM with stereo pair images of PRISM. The geometric calibrations of both PRISM and AVNIR-2 were done by evaluating relative accuracies within one scene as relative calibration, and sensor alignments as absolute calibration. AVNIR-2 has four radiometric bands from blue to near infrared to be used for regional environment monitoring such as land-use and land-cover classification, disaster monitoring, etc. Relative radiometric calibrations of both PRISM and AVNIR-2 was carried out using acquired images over homogeneous targets such as ocean, deserts, ice and snow areas, and nighttime observation. For absolute radiometric calibration of AVNIR-2, cross-calibration methods using calibrated satellite images e.g. the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) onboard TERRA satellite was applied. The image quality is also important for many application fields. The absolute radiometric calibration of PRISM was carried out as cross-calibration with AVNIR-2. In this paper, we introduce the calibration and validation plans, results of initial calibration, and accuracy assessments of the standard products of PRISM and AVNIR-2.
This paper presents the polarimetric calibration results of the Phased Array type L-band Synthetic Aperture Radar (PALSAR) system installed in the Advanced Land Observing Satellite (ALOS) and discusses the stability of polarimetric calibration parameters during the ALOS calibration phase as well as the influence of Faraday rotation. Data acquired over the Amazon area was used to estimate the polarimetric calibration parameters such as channel imbalance and cross-talk because the Amazon is located in the vicinity of the equator where the effect of Faraday rotation is expected to be small. The Quegan method, which is one of the polarimetric calibration methods that cannot deal with Faraday rotation, was applied to the data. The Amazon data showed that the channel imbalance remained stable during the calibration phase and the cross-talk was very small regardless of the descending path (daytime observation) and the ascending path (nighttime observation). On the other hand, the data acquired over the Tomakomai area, where the effect of Faraday rotation is not expected to be small, was compared with the Amazon data and indicated that there was a correlation between the cross-talk and Faraday rotation. The cross-talk of the Tomakomai is higher than that of the Amazon. Therefore, it is shown that Amazon data is more suitable for estimating the polarimetric calibration parameters of PALSAR than Tomakomai data. The polarimetric calibration parameters derived from the Amazon data showed good calibration accuracy.
An effective cross-calibration scheme is proposed and used for evaluation of AVNIR-2 radiometric accuracy. The scheme uses top-of-atmosphere reflectance functions of satellite zenith angle at each sample point. Each function was made using MODIS 500 m observations at temporally and spatially stable ground sites over 16 days which includes an AVNIR-2 observation date. As the result, radiances of AVNIR-2 channels 1 (463 nm), 2 (560 nm) and 3 (652 nm) agreed well to the radiances of Aqua and Terra MODIS channels 3 (466 nm), 4 (554 nm) and 1 (646 nm) respectively within 5% accuracy. AVNIR-2 Channel 4 (821 nm) radiance was evaluated lower than that of MODIS channel 2 (856 nm) about 15% on average. If we considered influences of atmospheric absorption and spectral slope of the ground sites, the AVNIR-2 channel-4 difference against MODIS was estimated to be lass than half of the 15%. This cross-calibration scheme among similar orbit satellite sensors can provide many samples which enable us to analyze sensor response dependency on different observation conditions such as sensor-pointing angles.
The geometric accuracy of the Phased-Array L-band Synthetic Aperture Radar (PALSAR) on board of ALOS (Advanced Land Observing Satellite) was evaluated by using a digital elevation model (DEM). The PALSAR amplitude images and those simulated with a DEM were produced. Tie points between them were then automatically calculated by pattern matching. The automated co-registration procedure efficiently collected the favorable tie points, especially for rugged terrain. The coefficients of affine transformation calculated from the tie points indicated no significant range- or azimuth-dependent errors over the standard image size. Absolute geometric accuracy was then defined as offsets between the amplitude and DEM-simulated images and was estimated to be better than the specification requirement.
Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) carried at Advanced Land Observing Satellite (ALOS) is expected to generate worldwide topographic data in respects of its high resolution and stereoscopic observation. The algorithms for generating Digital Surface Model (DSM) and Ortho-Rectified Image (ORI) have been developed for those objectives in Earth Observation Research Center/Japan Aerospace Exploration Agency (EORC/JAXA). During first one year following the successful ALOS launch, the capabilities of the algorithm have been widely tested. In this paper, intermediate results of performance analysis of DSM and corresponding ORI processing are described. First, the geometric model analysis of PRISM sensor is presented with the experimental results of the orientation processing. Then, the performance analysis of DSM and ORI generated with the PRISM geometric model is presented. The accuracy assessment results of generated DSM are presented from the comparison with high accuracy and high resolution reference DSM data sets of LiDAR DSM and Aerial Photo DSM. The accuracy assessment results of generated ORI are presented from the comparison with GCP.
Debris flow sites at Kumamoto and Nagano, where the flows were induced by heavy rain, have been examined by using L-band SAR data. Several L-band SAR data taken with air-borne PiSAR (polarimetry mode) and PALSAR (HH polarization) are used and examined the detectability of the debris flow. The disaster sites are well detected by a three components scattering model and volume, double bounce, and surface scattering account for 62%, 31%, and 7%, respectively. But a wasteland shows resemble scattering rates and a land utilization map is needed to distinguish these two sites. We also analyzed two PALSAR data taken at Nagano before and after the debris flow with an off-nadir angle of 41.5 degree and with a polarization of HH. The σ0 value over debris flow site become larger after the disaster for both debris flow sites. But the absolute value of the σ0 in Nagano is 2 to 6 dB lower than that in Leyte. We use small perturbation model and conclude that the possible causes of this difference is due to the difference of incident angle.