Assessing post-fire ground cover in Mediterranean shrublands with field spectrometry and digital photography Raquel Montorio Llovería a,c,⇑ , Fernando Pérez-Cabello a,c , Alberto García-Martín b,c a Department of Geography and Spatial Management, University of Zaragoza, C/ Pedro Cerbuna 12, 50009 Zaragoza, Spain b Centro Universitario de la Defensa de Zaragoza, Academia General Militar, Ctra. Huesca s/n, 50090 Zaragoza, Spain c Environmental Sciences Institute (IUCA), University of Zaragoza, C/ Pedro Cerbuna 12, 50009 Zaragoza, Spain article info Article history: Received 10 December 2015 Received in revised form 1 April 2016 Accepted 3 June 2016 Keywords: Fire severity Empirical models Hyperspectral VNIR data Spectral properties First derivative spectra abstract Fire severity can be assessed by identifying and quantifying the fractional abundance of post-fire ground cover types, an approach with great capacity to predict ecosystem response. Focused on shrubland for- mations of Mediterranean-type ecosystems, three burned areas (Ibieca and Zuera wildfires and Peñaflor experimental fire) were sampled in the summers of 2006 and 2007. Two different ground mea- surements were made for each of the 356 plots: (i) 3-band high spatial resolution photography (HSRP) and (ii) the hemispherical-conical reflectance factor (HCRF) in the visible to near-infrared spectral range (VNIR, 400–900 nm). Stepwise multiple lineal regression (SMLR) models were fitted to spectral variables (HCRF, first derivative spectra or FDS, and four absorption indices) to estimate the fractional cover of seven post-fire ground cover types (vegetation and soil – unburned and charred components – and ash – char and ash, individually and as a combined category). Models were developed and validated at the Peñaflor site (training, n = 217; validation, n = 88) and applied to the samples from the Ibieca and Zuera sites (n = 51). The best results were observed for the abundance estimations of green vegetation (R 2 adj: 0.70–0.90), unburned soil (R 2 adj: 0.40–0.75), and the combination of ashes (R 2 adj: 0.65–0.80). In com- parison of spectral data, FDS outperforms reflectance or absorption data because of its higher accuracy levels and, importantly, its greater capacity to yield generalizable models. Future efforts should be made to improve the estimation of intermediate severity levels and upscaling the developed models. In the con- text of fire severity assessment, our study demonstrates the potential of hyperspectral data to estimate in a quick and objective manner post-fire ground cover fractions and thus provide valuable information to guide management responses. Ó 2016 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier B.V. All rights reserved. 1. Introduction Immediate or first-order effects of fire on an environment, referred to as fire severity following the terminology of Lentile et al. (2006), Veraverbeke et al. (2010) and others, is of interest to forest management primarily because it is presumed to be an indicator of long-term ecosystem response (also referred to as burn severity or second-order effects). Many researchers have already shown its influence on vegetation recovery (Díaz-Delgado et al., 2003), changes in below-ground flora and fauna (Neary et al., 1999), seedling germination after fire (De Luís et al., 2005), species richness (Keeley et al., 2005), changes in soil structure (Mataix- Solera and Doerr, 2004), and runoff and erosion processes (Doerr et al., 2006; Moody et al., 2013), among others. An adequate remote sensing assessment of fire severity is of great importance, especially in a region such as the Mediterranean, where forest fire size and frequency are increasing (Tedim et al., 2013) and higher intensity and severity levels are being observed (Chuvieco et al., 2008). Traditional assessment has been based on spectral indices derived from multispectral satellite imagery, mostly from Landsat TM or ETM+ sensors (a.o. Epting et al., 2005; Picotte and Robertson, 2011; van Wagtendonk et al., 2004). The normalized ratio of near-infrared (NIR, band 4) and short-wave infrared (SWIR, band 7), known as the Normalized Burn Ratio (NBR, López and Caselles, 1991), and its delta or relative delta versions (dNBR and RdNBR, Key and Benson, 2006; Miller and Thode, 2007) have been widely applied empirically to estimate the field severity index Composite Burn Index (CBI, Key and Benson, http://dx.doi.org/10.1016/j.isprsjprs.2016.06.002 0924-2716/Ó 2016 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier B.V. All rights reserved. ⇑ Corresponding author at: Department of Geography and Spatial Management, University of Zaragoza, C/ Pedro Cerbuna 12, 50009 Zaragoza, Spain. Tel.: +34 876 55 38 51. E-mail addresses: montorio@unizar.es (R. Montorio Llovería), fcabello@unizar.es (F. Pérez-Cabello), algarcia@unizar.es (A. García-Martín). ISPRS Journal of Photogrammetry and Remote Sensing 119 (2016) 187–197 Contents lists available at ScienceDirect ISPRS Journal of Photogrammetry and Remote Sensing journal homepage: www.elsevier.com/locate/isprsjprs