ASPRS 2014 Annual Conference Louisville, Kentucky ♦ March 23-28, 2014 A NOVEL CONFIDENCE METRIC APPROACH FOR A LANDSAT LAND SURFACE TEMPERATURE PRODUCT Monica J. Cook and Dr. John R. Schott Rochester Institute of Technology Chester F. Carlson Center for Imaging Science Digital Imaging and Remote Sensing Laboratory 54 Lomb Memorial Dr. Rochester, NY 14623 mxc7441@rit.edu ABSTRACT The Landsat series is the longest body of continuously acquired, moderate resolution satellite imagery. The spatial and temporal resolution and coverage of Landsat make it an intriguing instrument for a land surface temperature product, which is an important earth system data record for a number of fields including climate, weather, and agriculture. Because current archived Landsat imagery has only a single thermal band, generation of a land surface temperature product requires an emissivity estimation and atmospheric compensation. This work, assuming imagery from a characterized and calibrated sensor and integration with ASTER derived emissivity data, focuses on the atmospheric compensation component by using reanalysis data and radiative transfer code to generate estimates of radiative transfer parameters. Along with an estimation of land surface temperature, the goal is to provide a confidence estimation for every pixel in a scene. Using water temperatures from buoy data, actual temperatures have been compared to predicted temperatures as validation of performance. These comparisons have shown acceptable performance when the atmosphere is well characterized, but larger errors when the atmosphere is not as well understood. The reanalysis data, radiative transfer code, bulk to skin temperature conversion, and lack of knowledge of atmospheric variation all complicate traditional error analysis. Various methods have been attempted, including error propagation based on perturbed atmospheres, regressions between metrics and error values, and thresholds based on atmospheric variables. Because of complications not faced by other large-scale products, a novel approach to error analysis will be developed by combining multiple approaches and data sources. Keywords: Landsat, land surface temperature, thermal INTRODUCTION Land surface temperature (LST) is an independently useful Earth system data record, with applications focused mostly around environmental endeavors, including agriculture, climate studies and meteorological research among other fields. It is difficult to directly measure the temperature of a surface without altering that temperature, making remote sensing an optimal method to measure large scale surface temperatures. When there are multiple thermal bands available, it is common to use a split window algorithm that utilizes differential absorption in adjacent thermal bands (Wan and Dozier, 1996). However, with a single thermal band, retrieval of LST requires both an atmospheric compensation and an emissivity component. Landsat is the longest body of continuously acquired, moderate resolution satellite imagery, with an archive that contains over 30 years of thermal imagery. However, until the launch of Landsat 8 in 2013, historical Landsat satellites have captured a single thermal band. The archived data is calibrated and characterized, but provided as radiance values, which are not immediately intuitively applied. Development of a land surface temperature product for the Landsat archive, from the available radiance data, requires atmospheric compensation and emissivity. Recently, the Advanced Spaceborne Thermal Emission and Reflection (ASTER) radiometer has been used to develop a high spatial resolution (100 m) surface emissivity product. Currently available for North America, known as the North American ASTER Land Surface Emissivity Database (NAALSED), plans are underway to extend this dataset to global coverage (Hulley and Hook, 2009). Assuming the ability to combine with this product from the Jet Propulsion Laboratory, this work focuses on the generation and error analysis of the atmospheric compensation component.