The first processing step derives Top Of Atmosphere (TOA) reflectances and brightness temperatures from the raw HRTP data. These data are radiometrically corrected (NOAA OSPO). The pixels are still in orbit geometry, appending latitude and longitude layer define the spatial registration. An additional calibration process retrieves calibration correction factors, which are applied by the subsequent scientific data processors. The factors are derived for each sensor on a daily base using a shifting window of 30 days with recorded reflectances over the CEOS site Lybia-4. The band references for the correction are average values from a selection of NOAA OSPO calibrated NOAA-19 reflectances.
Due to minor satellite position and orbit uncertainties, the geolocation of AVHRR data is known to be imprecise. This can cause geolocation errors of up to 10km per pixel - a major obstacle for necessary compositing/mosaicking of images for time series analysis. In TIMELINE, a first guess of latitude and longitude positions is made by the TeraScan software developed by SeaSpace by adjusting the earth location parameters in order to achieve the best land/water boundary fit based on a given land/water boundary dataset. As this procedure still leaves some geolocation flaws, an additional chip matching is introduced. This method is based on two techniques. Both of them similarly attempt to find the best match between small image chips taken from a reference water mask in the first, and from a median Normalized Difference Vegetation Index (NDVI) mask in the second round. The latitude and longitude layer are updated to the position of the best matches using a third order polynomial, taking into account that location shifts may vary across the image. The chip matching procedure is described in detail in Dietz et al. (2017).
AVHRR reflectance over Europe
Dietz, A; Frey, C; Ruppert, T; Bachmann, M; Künzer, C & Dech, S (2017): Automated Improvement of Geolocation Accuracy in AVHRR Data Using a Two-Step Chip Matching Approach — A Part of the TIMELINE Preprocessor. Remote Sensing 9 (4): 303.