|
MISR Level 1 Products Quality Statement |
This statement applies to MISR Level 1 Products with a version number of 0022. See the Versioning Page for an in-depth explanation of the differences between various MISR product versions. Quality statements covering earlier time periods may be accessed through links at the bottom of this page.
The MISR Level 1 software which generated these products is believed to be functioning quite well except where noted below. This statement lists known problems with Level 1 Products and clarifies issues which have confused some users.
There are no known problems with the current release of PGE7 software. Analysis of isolated case studies indicates that the software is meeting all of its requirements.
The Geometric Parameters exhibit one algorithmic quirk which has surprised some users. Solar zenith and azimuth angles near the swath edge occasionally appear to jump around. This inconsistency is the result of an intentional choice of algorithm whereby solar angles are computed at the mean time at which MISR cameras viewed the ground point in question. Adjacent points are not always visible to the same set of cameras. This can cause a bias in solar angle towards cameras which acquired that point.
This portion of the list is lengthy , so the sub-headings are listed for quick reference.
The MISR calibration team has been involved in an on-going effort to both validate and make incremental improvements to the radiometric accuracy of the Level 1B data products. Because the uncertainties in these products are well understood, those versioned 0012 or greater, are given the quality designation of "Stage 1 Validated."
Update: The MISR band-relative radiometric calibration has been improved for Level 1 products of version 0022 or greater. Red band radiances have been reduced by 3% for all cameras, and NIR band radiances have been reduced by 1%. The need for a change was discovered during analysis of data from several MISR vicarious calibration experiments. Results from MISR vs MODIS, MISR vs MERIS and Terra lunar calibration studies supported the decision to make this change.
Radiometric accuracy has been improved for products versioned 0015 or greater. The improvment was achieved by discovering and correcting an error in the code used to derive the radiometric calibration coefficients. A-Nadir camera data produced before this date may have had absolute errors as large as 10% (too bright) at the left/western edge of the scene, and -5% (too dim) at the right/eastern edge. Radiometry at the swath center, and for other cameras has been in error to a lesser degree. Descriptions of calibration uncertainties for older products can be found on the Calibration Page. The uncertainties at this time are listed below.
| PSF effects refer to scene-contrast reductions due to local scene inhomogeneity. The radiometry of one pixel is affected in proportion to the contrast difference and proximity of another pixel. This is a camera optical effect, and is measurable for objects that are within 6 km crosstrack distance of each other. The downtrack PSF effect is believed to be negligible. A deconvolution algorithm is now exercised in the software to minimize PSF effects. |
| Ghost-image refers to the presence of a secondary image, created as a reflection of a given target through the lens optical axis. This secondary ghost has been measured to be 0.3% of the primary image, and thus results in a negligible error except where the image of a bright target falls on a very dark scene. |
| L1B2 resampling is implemented by bilinear interpolation, and thus errors in this process are a function only of a sample's radiance value as compared to that of its immediate neighbors. |
| Illumination-level dependent errors are attributable to the goodness-of-fit of camera response data to a mathematical equation. The MISR cameras are described as having a linear relationship between incident radiance and camera output. For radiance levels less than 2% in equivalent reflectance, this assumption is valid to within 5% uncertainty. The error is considered negligible for larger input signals. |
| Detector uniformity of response errors occur when a set of detector elements are non-uniform in response (10% non-uniformity or greater), are image inhomogeneous scenes, and are DN-averaged as part of the on-board data compression (Global Mode) algorithm. Only a dozen detector regions (out of 13,000 such pixel blocks) are non-uniform, and these are identified by data quality indicators in the products. For conditions where bright scenes are adjacent to dark scenes, an additional radiometric error of 6% may result in pixel regions where the Data Quality Indicator level is given a value of 2. |
Update: MISR Level 1B2 products exhibit acceptable georectification and coregistration accuracy. In the nominal case, the expected mean geolocation error for eight out of nine cameras is below 50 meters. Standard deviations range between 60 meters (A-Nadir camera) and 100 meters (D-forward camera). In its raw state, the georectification of the D-aft (Da) camera is often significantly worse than the georectification of the other eight cameras. However, the raw Da data is corrected through the use of Reference Orbit Imagery (ROI). Even when corrections are not available for the current orbit, they are borrowed from a prior orbit. The average estimated geolocation error for corrected Da image data is better than 150 m.
Also, all camera data acquired during occasional spacecraft maneuvers is of degraded accuracy. Data products generated during these times contain an Orbit Quality flag which indicates this problem. The nominal value of the Orbit Quality flag in the File Metadata (Global Attributes) is 0.0. Degraded data has an Orbit Quality flag value of -1.0. See the Georectification Page for more details, including a link to the list of degraded orbits.
The Nadir, single-band L1A browse product has been replaced with a new Ellipsoid-based color product. The new browse product is generated for all 9 cameras at 2.2 km resolution (sub-sampled). MISR Red, Green and Blue bands are used to create the color image, which is intentionally clipped and gamma-stretched in order to make cloud, ocean and land features visible. The jpeg compression is performed at 75% quality, which means that compression artifacts are occasionally visible.
The MISR Level 1A product is a reformatted version of the raw L0 data packet stream from the spacecraft. In this format, the CCD Data Numbers (DNs) may be viewed as an unregistered image with data quality indicators occupying the two low-order bits. By design, L1A does not differ greatly from the raw data except that gaps are filled in with appropriate fill values.
The MISR Level 1B1 product has been radiometrically corrected, but it has not been registered. The quality of L1B1 radiances is equivalent to that of L1B2 radiances discussed above, except that L1B1 pixels correspond directly to instrument CCD detector samples.
The L1B1 Local Mode product consists of the L1B1 output acquired when the MISR instrument is put into Local Mode in which all nine cameras view a scene at 275 m resolution in all four bands.
The cloud mask produced during Level 1 processing is called the RCCM (Radiometric Camera-by-camera Cloud Mask). It is one of three independent cloud masks generated from MISR data. The other two cloud masks are produced at Level 2 and are called the ASCM (Angular Signature Cloud Mask) and the SDCM (Stereo Derived Cloud Mask).
The RCCM algorithm applies traditional spectral and spatial measures to data from each MISR camera in order to produce separate cloud masks for each camera. These measures and the threshold procedures are completely different for the two processing paths: ocean and land. For this reason, ocean and land may carry different quality statements. Another field, the Glitter Mask, is included in the RCCM product for the sake of convenience.
Over ocean, the RCCM employs a static threshold procedure. The static thresholds are a function of the sun-view geometry and have been fine-tuned several times since launch. The RCCM over ice-free ocean, version F02_0016 and greater, has been Validated to Stage 1 after the completion of the following tests:
Although overall performance looks good for the RCCM over ice-free ocean, it does suffer from the traditional problems encountered with spectral/spatial cloud masks:
The RCCM over land is designed to use dynamic thresholds. However, at this time, the dynamic threshold procedure is not in place. Instead, a set of seasonal thresholds is being used. This static threshold dataset was derived from MISR data, but it is only a placeholder for the thresholds generated using the dynamic threshold procedure. Its overall performance is not suitable for scientific use, except where the underlying surface is covered with moderate to dense vegetation. Detailed visual inspection of 12 orbits (3 for each season falling over Europe, Africa, Asia, and North America) revealed a classification success rate of 85% over vegetated land cover, with misclassification largely composed of very thin cirrus designated as clear. The classification success rate drops for surfaces containing sparse vegetation cover. Therefore, the RCCM product over moderate to heavy vegetation cover is of Provisional Quality for version F02_0016 or greater, and remains of Beta Quality over all other land surface cover.
The Glitter Mask indicates regions of the data that may contain sun-glint. As of February 5, 2002,(version 0011) the sun-glint cone angle was increased from 30 degrees to 40 degrees in order to mask some of the weaker glint that was observed in the imagery.
