ASAR Product Consistency and Geolocation Accuracy

The ability to geolocate ENVISAT ASAR image products and transform them into a map projection is a critical step required to enable not only overlays with other sources of information (e.g. DEM, GIS layers), but even with other ASAR products, especially those acquired with a different track, beam, or incidence angle. This paper summarises geolocation validation results evaluating both the relative (i.e. consistency within product families) and absolute geolocation accuracy inherent to ASAR products generated by the ESA operational processor (PF-ASAR). A set of slant-range complex IMS/APS/WSS-, ground-range detected IMP/APP-, IMM/APM/WSM-, and ellipsoidgeocoded detected IMG/APG products were ordered for a set of data takes over the Netherlands (IM & AP) and Switzerland (WS products). Subswath detection and mosaicking was performed on the WS-SLC (WSS) products to enable their comparison with WSM products derived from the same raw data. All products were then ellipsoid-geocoded to permit direct intercomparisons. To quantify the small relative shifts between the product types, a dense image matching algorithm was used to compare geocoded ellipsoid-corrected (GEC) product pairs. Statistical analysis (mean, standard deviation) of the resulting disparity fields revealed processor biases, and enabled improvement of the PF-ASAR processor. WSS/WSM comparisons revealed slight beam-dependent relative azimuth shifts of up to 19 m, well below the native WSM sample spacing of 75 m. Radiometric differences between WSS and WSM products were also evaluated point by point in the ellipsoid-geocoded map geometry, and revealed a possible error in the AGP calibration for WSS products. For some data sets, the absolute geolocation accuracy was measured using transponders with well-known locations, prominently visible in the ASAR products. Both range-Dopplerand geolocation grid-based geolocation was performed for the in-scene transponders. A given predicted position was then compared to the measured position obtained through direct observation of a strong peak response in the neighbourhood of the prediction. ASAR products were generally found to be extremely consistent with respect to each other, with measured relative differences all at the sub-sample level. An intermittent range bias is the one remaining source of large geolocation error, although it has not yet been observed in images acquired since 2006. ASAR’s absolute geolocation accuracy has set a new standard for spaceborne SAR sensors. ASAR was the first operational spaceborne SAR imager to provide data that can (generally) be geolocated without the use of tie-pointing.