International Journal of Applied Earth Observation and Geoinformation 35 (2015) 320–328 Contents lists available at ScienceDirect International Journal of Applied Earth Observation and Geoinformation jo ur nal home p age: www.elsevier.com/locate/ jag The effect of atmospheric and topographic correction on pixel-based image composites: Improved forest cover detection in mountain environments Steven Vanonckelen ∗ , Stef Lhermitte, Anton Van Rompaey Division of Geography, Katholieke Universiteit Leuven, Celestijnenlaan 200E, BE-3001 Heverlee, Belgium a r t i c l e i n f o Article history: Received 9 June 2014 Accepted 9 October 2014 Keywords: Forest cover mapping Classification accuracy assessment Topographic correction Landsat Pixel-based compositing Mountain areas a b s t r a c t Quantification of forest cover is essential as a tool to stimulate forest management and conservation. Image compositing techniques that sample the most suited pixel from multi-temporal image acquisitions, provide an important tool for forest cover detection as they provide alternatives for missing data due to cloud cover and data discontinuities. At present, however, it is not clear to which extent forest cover detection based on compositing can be improved if the source imagery is firstly corrected for topographic distortions on a pixel-basis. In this study, the results of a pixel compositing algorithm with and without preprocessing topographic correction are compared for a study area covering 9 Landsat footprints in the Romanian Carpathians based on two different classifiers: Maximum Likelihood (ML) and Support Vector Machine (SVM). Results show that classifier selection has a stronger impact on the classification accuracy than topographic correction. Finally, application of the optimal method (SVM classifier with topographic correction) on the Romanian Carpathian Ecoregion between 1985, 1995 and 2010 shows a steady greening due to more afforestation than deforestation. © 2014 Elsevier B.V. All rights reserved. Introduction The Millennium Development Goals Report (2013) stated that accelerated progress and actions are needed for forest conserva- tion. Forest cover changes affect crucial ecosystem services, such as water supply, biodiversity, carbon storage, and climate regula- tion (Foley et al., 2005). Assessing the rate and spatial pattern of forest cover change is challenging since large forests are present in rather inaccessible and rough mountain areas (Lambin and Geist, 2006). Multiple efforts have been made to quantify global forest cover changes and forest transitions (Hansen and DeFries, 2004; Meyfroidt and Lambin, 2011; FAO, 2012). These global inventories were either sample-based or based on coarse spatial resolution data (Hansen et al., 2013). Moreover, time-series analysis of for- est cover change based on high resolution satellite data have been performed in different countries, e.g. Indonesia (Broich et al., 2011), United States of America (Kennedy et al., 2010; Hansen et al., 2011), Democratic Republic of Congo (Potapov et al., 2012), and Romania (Griffiths et al., 2013b). In contrast, Hansen et al. (2013) presented a global forest cover change inventory based on high resolution ∗ Corresponding author. Tel.: +32 496 366342. E-mail address: steven.vanonckelen@lne.vlaanderen.be (S. Vanonckelen). satellite data. Remote sensing techniques are privileged monitoring tools and yet suffer from methodological challenges that need to be resolved by correction methods (Lhermitte et al., 2011a; Balthazar et al., 2012). The opening of the Landsat archive provides opportunities to reconstruct forest cover changes for large areas on a 30–60 m spa- tial scale (Loveland and Dwyer, 2012; Giri et al., 2013; Hansen et al., 2013). The use of the Landsat archive with 185 km × 185 km footprint size and a 16-day repeat cycle, however, poses sev- eral challenges, ranging from image mosaicking over large areas that cover more than one footprint, to optimal data selection due to cloud cover (Ju and Roy, 2008; Lhermitte et al., 2011b; Griffiths et al., 2013a) and data discontinuities due to sensor or data related errors (e.g. the failure of scan line correction in Landsat 7; Arvidson et al., 2006). Moreover, optimal processing is essential to obtain consistent reflectance values for each Landsat image, where processing methods range from cloud/shadow/water screening and quality assessment, and image normalization for atmospheric con- ditions and surface anisotropy (Potapov et al., 2011, 2012; Hansen et al., 2013) to physically based atmospheric and topographic cor- rection methods (Minnaert, 1941; Teillet et al., 1982; Berk et al., 1998; Meyer et al., 1993; Jensen, 1996; Richter, 1996; Vermote et al., 1997; Veraverbeke et al., 2010). More recent, comparisons in the performance of different combinations of atmospheric and http://dx.doi.org/10.1016/j.jag.2014.10.006 0303-2434/© 2014 Elsevier B.V. All rights reserved.