Integration of airborne lidar and vegetation types derived from aerial photography for mapping aboveground live biomass Qi Chen a, ⁎, Gaia Vaglio Laurin b, c , John J. Battles d , David Saah e, f a Department of Geography, University of Hawai'i at Manoa, 422 Saunders Hall, 2424 Maile Way, Honolulu, HI 96822, USA b Department of Computer, System and Production Engineering, University of Tor Vergata, Rome 00133, Italy c CMCC - Centro Mediterraneo per i Cambiamenti Climatici, via Augusto Imperatore (Euro-Mediterranean Center for Climate Change), Lecce 73100, Italy d Department of Environmental Science, Policy, and Management, 137 Mulford Hall, University of California at Berkeley, Berkeley, CA 94720, USA e Spatial Informatics Group, LLC, 3248 Northampton Ct., Pleasanton, CA 94588, USA f College of Arts and Sciences, Environmental Science, University of San Francisco, San Francisco, CA 94117, USA abstract article info Article history: Received 3 August 2011 Received in revised form 29 December 2011 Accepted 25 January 2012 Available online xxxx Keywords: Biomass Mixed-effects model Airborne lidar Aerial photos Vegetation type The relationship between lidar-derived metrics and biomass could vary across different vegetation types. However, in many studies, there are usually a limited number of field plots associated with each vegetation type, making it difficult to fit reliable statistical models for each vegetation type. To address this problem, this study used mixed-effects modeling to integrate airborne lidar data and vegetation types derived from aerial photographs for biomass mapping over a forest site in the Sierra Nevada mountain range in California, USA. It was found that the incorporation of vegetation types via mixed-effects models can improve biomass estimation from sparse samples. Compared to the use of lidar data alone in multiplicative models, the mixed-effects models could increase the R 2 from 0.77 to 0.83 with RMSE (root mean square error) reduced by 10% (from 80.8 to 72.2 Mg/ha) when the lidar metrics derived from all returns were used. It was also found that the SAF (Society of American Forest) cover types are as powerful as the NVC (National Vegetation Classification) alliance-level vegetation types in the mixed-effects modeling of biomass, implying that the future mapping of vegetation classes could focus on dominant species. This research can be extended to investigate the synergistic use of high spatial resolution satellite imagery, digital image classification, and airborne lidar data for more automatic mapping of vegetation types, biomass, and carbon. © 2012 Elsevier Inc. All rights reserved. 1. Introduction Vegetation biomass, the weight of plant materials that exist over an area, is a critical measure of ecosystem structure and productivity that informs a range of applications such as fire emission calculations (e.g., De Santis et al., 2010), wildlife habitat analysis (e.g., Morris et al., 2009), hydrological modeling (e.g., Ursino, 2007), and greenhouse gas accounting (e.g., De Jong et al., 2010). In particular, accurate estimates of biomass are needed in order to inform national policies and interna- tional treaties regarding forest management and carbon sequestration (Malmsheimer et al., 2011). Lidar is a state-of-the-art remote sensing technology with a proven ability to map aboveground biomass (AGB). The accuracy and sensitivity of the metrics derived from optical and radar imagery (such as NDVI and backscatter coefficient) decline with increasing AGB (Waring et al., 1995). In contrast, vegetation height metrics derived from lidar have been found to be highly correlated to biomass even when the biomass density is very high (Gonzalez et al. 2010, Means et al., 1999). In the past, much research has been done to estimate AGB using airborne discrete-return lidar (e.g., Asner et al., 2009; Banskota et al., 2011; Lim et al., 2003), airborne profiling lidar (e.g., Nelson et al., 2009, 1988; Stahl et al., 2011), airborne waveform lidar (e.g., Dubayah et al. 2010; Lefsky et al., 1999; Ni-Meister et al., 2010), satellite lidar (e.g., Boudreau et al., 2008; Guo et al., 2010; Nelson et al., 2009), and ground-based lidar (e.g., Loudermilk et al., 2009; Ni-Meister et al., 2010). In these applications, statistical models were used to quantify the relationship between biomass measurements and vegetation struc- ture metrics derived from lidar for a number of forest plots or stands. Their performance varies depending on the vegetation conditions, the density of field observations, and the approach used for statistical modeling. Most of these existing studies have focused on the use of lidar- derived canopy structure metrics, such as height and canopy cover, for biomass estimation. However, studies of plant allometry sug- gested that biomass at the individual tree level is determined not only by canopy structure but also by factors such as trunk taper and wood density (Chave et al., 2006; Niklas, 1995), which are closely re- lated to the floristic characteristics of the plants. As a result, biomass should be related to vegetation types. For example, Drake et al. (2003) examined the relationships between lidar metrics from an Remote Sensing of Environment 121 (2012) 108–117 ⁎ Corresponding author. Tel.: + 1 808 956 3524; fax: + 1 808 956 3512. E-mail address: qichen@hawaii.edu (Q. Chen). 0034-4257/$ – see front matter © 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.rse.2012.01.021 Contents lists available at SciVerse ScienceDirect Remote Sensing of Environment journal homepage: www.elsevier.com/locate/rse