International Journal of Applied Earth Observation and Geoinformation 18 (2012) 23–36 Contents lists available at SciVerse ScienceDirect International Journal of Applied Earth Observation and Geoinformation jo u r n al hom epage: www.elsevier.com/locate/jag Downscaling land surface temperatures with multi-spectral and multi-resolution images Wenfeng Zhan a , Yunhao Chen a,∗ , Jinfei Wang a,b , Ji Zhou c , Jinling Quan a , Wenyu Liu a , Jing Li a a State Key Laboratory of Earth Surface Processes and Resource Ecology (Beijing Normal University), College of Resources Science & Technology, Beijing Normal University, Beijing 100875, China b Department of Geography, The University of Western Ontario, London ON, N6A 5C2, Canada c Institute of Geo-Spatial Information Science and Technology, University of Electronic Science and Technology of China, Chengdu, Sichuan 610054, China a r t i c l e i n f o Article history: Received 2 September 2010 Received in revised form 30 December 2011 Accepted 2 January 2012 Keywords: Thermal remote sensing Sharpening Downscaling Land surface temperature Multi-resolution Multi-spectral a b s t r a c t Land surface temperature (LST) plays an important role in many fields. However, the limited spatial resolution of current thermal sensors impedes the utilization of LSTs. Based on a theoretical framework of thermal sharpening, this report presents an Enhanced Generalized Theoretical Framework (EGTF) to downscale LSTs using multi-spectral (MS) and multi-resolution images. MS proxy-sharpening and LST downscaling are combined under EGTF. Simulated images upscaled from Enhanced Thematic Mapper Plus (ETM+) and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data are produced for indirect validations. Validation of MS proxy-sharpening shows that EGTF is better than the Gram-Schmidt (GS) and the Principle Component (PC) methods, yielding a lower root mean square error (RMSE) and ERGAS (erreur relative globale adimensionnelle de synthèse) and, thus, maintaining higher spectral similarity. For LST downscaling, validations show that EGTF has a higher accuracy than the Unmixing-Based Image Fusion (UBIF) method and indicate that the proxy-sharpening process improves the accuracy of downscaled LSTs. Further discussions regarding the selection of the moving-window size (MWS) demonstrate that the MWS could be determined by the range in a semi-variance analysis of scaling factor images. © 2012 Elsevier B.V. All rights reserved. 1. Introduction Land surface temperature (LST) is of great significance for the estimation of radiation and energy budgets associated with land surface processes. However, the available satellite LST images have low spatial and temporal resolutions, which constrain their poten- tial applications. The downscaling of LST supplies LST products with higher spatial resolutions, which can therefore be used for the investigation of regional evapo-transpiration and other studies (Kustas et al., 2003). In the field of LST downscaling, the term downscale (Pardo- Igúzquiza et al., 2006; Liu and Pu, 2008) is often viewed as equivalent to terms such as sharpen (Agam et al., 2007; Jeganathan et al., 2010; Zhan et al., 2011) or disaggregate (Merlin et al., 2010). All three of these terms imply a similar process by which the LST is enhanced to a higher spatial resolution, while the thermal radi- ance is invariantly maintained. Although slight differences in the ∗ Corresponding author at: No. 19, Xinjiekouwai Street, Beijing 100875, PR China. Tel.: +86 10 58806098; fax: +86 10 58806098. E-mail addresses: zhanwenfeng1986@gmail.com (W. Zhan), cyh@bnu.edu.cn (Y. Chen), jfwang@uwo.ca (J. Wang), jzhou233@uestc.edu.cn (J. Zhou), quanjinlin @126.com (J. Quan), liuwenyusd@126.com (W. Liu), lijing@bnu.edu.cn (J. Li). meanings of these terms may exist, we do not distinguish among them herein. Many methods have been proposed for performing LST down- scaling. At the DN (Digital Number) level, the typical algorithms include the regressive window (Moran, 1990), unmixing-based image fusion (UBIF) (Zhukov et al., 1999), and Bayesian methods (Fasbender et al., 2008). However, the majority of studies are LST based or thermal radiance based. These involve using either the lin- ear relationship (Kustas et al., 2003; Agam et al., 2007; Liu and Pu, 2008; Nichol, 2009) or the nonlinear relationship between multi- spectral (or multispectral-derived) and thermal bands (Jeganathan et al., 2010; Dominguez et al., 2011). Most recently, Zhan et al. (2011) presented a generalized theoretical framework from an assimilation perspective (GTFAP) to downscale thermal images using semi-empirical regression and modulation. GTFAP, under which most sharpening algorithms can be considered as special cases, is based on the physical connections (e.g., LST is negatively linearly related to normalized difference vegetation index (NDVI)) between LSTs and the chosen scaling factors (Zhan et al., 2011). Modulation refers to a strategy aimed at maintaining thermal homo- geneity under the criterion of spectral radiance scale invariance (Liu and Moore, 1998). Taking ETM+ as an example, the criterion of scale invariance indicates that the average sharpened LST in a block of 16 pixels (each pixel has the resolution of 15 m) should be equal to 0303-2434/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.jag.2012.01.003