2918 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 43, NO. 12, DECEMBER 2005 Potential of Getis Statistics to Characterize the Radiometric Uniformity and Stability of Test Sites Used for the Calibration of Earth Observation Sensors Abderrazak Bannari, K. Omari, P. M. Teillet, and G. Fedosejevs Abstract—The calibration of airborne and satellite remote sensing sensors is a fundamental step for the rigorous validation of products derived from satellite data. Because of the inaccessibility of Earth Observation Satellites on orbit, the direct calibration method based on a test site with ground reference data is often considered necessary. However, the problem of radiometric spa- tial uniformity and temporal stability of test sites constitutes an important issue in the accuracy achieved in calibration operations and the long-term characterization of satellite sensor radiometry. Generally, the coefficient of variation and semivariograms are the most widely used tools for evaluating the radiometric uniformity and stability of a calibration site. In this study, we analyze for the first time the potential of Getis statistics compared to the coefficient of variation for the study of the radiometric spatial uni- formity and temporal stability of the Lunar Lake Playa, Nevada (LLPN) test site. The results obtained show the potential and the importance of the synergy generated by these two methods for analyzing the radiometric temporal stability of the LLPN site. Getis statistics provide an excellent spatial analysis of the site while the coefficient of variation provides complementary information on the temporal evolution of the site. Index Terms—Coefficient of variation (CV), Getis statistics, Lunar Lake Playa, Nevada (LLPN), optical sensor, radiometric calibration, test sites. I. INTRODUCTION D URING the last three decades, the demand for remote sensing products has increased tremendously, particularly for the management of natural resources and more generally for environmental. Moreover, the surveillance of the Earth’s en- vironment at the local, regional, continental, or global scales using various sensors requires adequate radiometric calibration in order to have accurate and reproducible geophysical and bio- physical surveys through time [4]. Consequently, significant er- rors can spread through all subsequent image processing oper- Manuscript received October 11, 2004; revised April 20, 2005. This work was supported by the Natural Sciences and Engineering Research Council of Canada. A. Bannari and K. Omari are with the Remote Sensing and Geomatics of En- vironment Laboratory, Department of Geography, University of Ottawa, Ottawa, ON K1N 6N5, Canada (e-mail: abannari@uottawa.ca). P. M. Teillet is with the Remote Sensing and Geomatics of Environment Labo- ratory, Department of Geography, University of Ottawa, Ottawa, ON K1N 6N5, Canada and also with the Canada Centre for Remote Sensing, Ottawa, ON K1A 0Y7, Canada. G. Fedosejevs is with the Canada Centre for Remote Sensing, Ottawa, ON K1A 0Y7, Canada. Digital Object Identifier 10.1109/TGRS.2005.857913 ations, including spatial and multitemporal analyses, thematic classifications, and the generation of vegetation indexes [15], [18], [29], [36], [42]. To get the maximum from satellite and airborne data-derived products, sensors must constantly be cal- ibrated, the data validated, and the stability and quality of data ensured [3], [42]. The radiometric performances of Earth Observation Satel- lite sensors change between calibration in the laboratory before launch and on orbit operations [9], [35], [21], [24]. The spectral response characteristics of sensor bandpasses, mirror surfaces, and optical elements can also change postlaunch and over the lifetime of the mission [11], [21], [37]. Therefore, it is normal to view with suspicion any postlaunch change in the relative and absolute sensor calibration parameters and to question the quality of any onboard calibration systems (lamps, solar sen- sors, etc.). With or without change, spectral response character- istics are an important consideration in radiometric cross-cali- bration between sensors [38], [40]. In general, the calibration of instruments dedicated to Earth observation is not an easy task. To increase the accuracy of this operation, it is advisable to use several independent methods [9]. Different methods have been used for the relative and absolute calibration of optical sensors: calibration in laboratory before launch in a well-controlled environment; onboard calibration using a lamp, a sphere, a solar diffusion panel, or a solar sensor; calibration through lunar observation; calibration using ground sites with simultaneous ground reference data; calibration using pseudoinvariant features on the ground without independent reference data, interinstrument, and interband calibration [16], [19], [20], [35], [42], [44]. Because of the inaccessibility of the satellite on orbit, the vicarious calibration method based on a ground site using simultaneous ground reference data is often considered an essential step to ensure the best “accuracy versus investment” compromise [16], [35]. The method has the advantage of reproducing the real conditions of image data acquisition. Its accuracy depends closely on the radiometric stability of the calibration site, site reflectance measurements, and knowledge of the atmospheric parameters at the time of image acquisition. In the best conditions of site and measure- ment stability, it ensures a calibration accuracy in the range of to [9], [46], [47]. It should also be noted that these operations concern the space agencies or the organizations responsible for the dissemination of the remote sensing data 0196-2892/$20.00 © 2005 IEEE