Investigations on the Geometric Quality of AVHRR Level 1B Imagery Aboard MetOp-A


Bas S., Debaecker V., KOCAMAN GÖKÇEOĞLU S. , Saunier S., Garcia K., Just D.

PFG-JOURNAL OF PHOTOGRAMMETRY REMOTE SENSING AND GEOINFORMATION SCIENCE, 2021 (Journal Indexed in SCI) identifier identifier

  • Publication Type: Article / Article
  • Publication Date: 2021
  • Doi Number: 10.1007/s41064-021-00161-0
  • Title of Journal : PFG-JOURNAL OF PHOTOGRAMMETRY REMOTE SENSING AND GEOINFORMATION SCIENCE
  • Keywords: Geometric quality assessment, AVHRR, MetOp-A, Absolute accuracy, Interband accuracy, Image matching, ACCURACY

Abstract

The products of Advanced Very High Resolution Radiometer/3 (AVHRR/3) sensors operating aboard the National Oceanic and Atmospheric Administration (NOAA) and MetOp-A satellites are frequently used in meteorological applications. The geometric accuracy and stability of the sensors are regularly monitored by European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT). As existing methods for evaluation resulted in a small number of check points only, a scientific project was initiated by EUMETSAT. The project aimed at the development of methods for improving the density and distribution of automatically generated check points for evaluating data having various spectral bands. The absolute, multi-temporal, and interband registration accuracy assessments of various meteorological satellites were carried out within the project including AVHRR Level 1B images. In this study, the geometric quality of AVHRR Level 1B products acquired from MetOp-A over Europe and North Africa regions was investigated. Global image mosaics with higher spatial resolutions were utilized as reference. The pre-processing methods applied to the images include warping to the same projection grid, image enhancement with Laplacian filter, and extraction of validity masks, which were used to eliminate cloud and water areas from processing. A 2D assessment approach was carried out using an area-based least squares matching method. The results show that the developed algorithm can produce dense and reliable check points. Biases ranging between 0.7 and 1.6 pixels (approximately 0.8-1.8 km) in x direction and 0.2-0.9 pixels (approximately 0.2-1.0 km) in y direction were obtained from the evaluated datasets. The standard deviations range are between 0.15 and 0.30 pixels. The multi-temporal geometric accuracy was found stable and the interband accuracy demonstrated high coherence of different spectral bands.