Biomass assessment in the Cameroon savanna using ALOS PALSAR data Stéphane Mermoz a, ⁎, Thuy Le Toan a , Ludovic Villard a , Maxime Réjou-Méchain b , Joerg Seifert-Granzin c a Centre d'Etudes Spatiales de la BIOsphère, UMR CNRS 5126, University of Paul Sabatier, Toulouse, France b Laboratoire Evolution et Diversité Biologique, UMR CNRS 5174, University of Paul Sabatier, Toulouse, France c Mesa-Consult, Konstanz, Germany abstract article info Article history: Received 26 October 2012 Received in revised form 19 September 2013 Accepted 2 January 2014 Available online xxxx Keywords: Biomass mapping Carbon assessment Savanna ALOS PALSAR Cameroon In this paper, ALOS PALSAR data have been used to map above ground biomass (AGB) in savanna ecosystems in Cameroon. The study has been motivated by the need to have estimates of carbon in African savannas. The L-band PALSAR mosaic data are suitable for the retrieval of savanna biomass (typically less than 100 Mg.ha -1 ) at national and continental scales. The retrieval methods have been developed using the following steps a) collec- tion of in situ data and estimate of AGB and its uncertainties, b) pre-processing of SAR data, with an emphasis on the reduction of uncertainties due to speckle, while preserving the SAR resolution (25 m), c) development of a regression model with a reduced number of fitting parameters. The methodology is developed in a representative study area where in situ data are collected, using standard PALSAR Fine Beam Dual polarisation data, and d) pixel-to-pixel mapping of AGB over 259 228 km 2 of Cameroon savanna using PALSAR mosaic data. The dense tropical forest has been masked using the GlobCover 2009 land cover map. A value of AGB and its uncertainty has been assigned to each pixel. The results indicate a total AGB of 1.25 ± 0.04 Pg or 0.63 ± 0.02 PgC of above- ground carbon in the Cameroon savanna. © 2014 Elsevier Inc. All rights reserved. 1. Introduction Africa was recently recognised as a major source of interannual var- iability in global atmospheric CO 2 (Williams et al., 2007), although Africa contributes less than 4% of the global fossil fuel emissions (Canadell, Raupach, Houghton, et al., 2009). High uncertainties in the estimates of the net long-term carbon balance of African ecosystems have been pointed out in Ciais et al. (2011): the carbon balance of African ecosystems varies currently from a sink of approximately 3.2 PgC.yr -1 to a small source of approximately 0.03 PgC.yr -1 . Although African forest ecosystems appear to be vulnerable, they have been spared from massive deforestation in tropical rain forests (Ciais et al., 2011). For example, according to the FAO (2010), the annual change rate in Democratic Republic of the Congo and in Gabon during the 2005–2010 period was -0.2% and 0% respectively. For the same period, deforestation of savanna woodlands in Africa was found to be rapid, with for example an annual change rate of -2.19% and -2.72% reported in Ghana and in Uganda respectively. Although different definitions of savanna exist, a savanna can be defined as a continuum of physiognomic types, the most important of which is the wooded grassland (Menaut, Barbault, Lavelle, & Lepage, 1985). In Africa, sa- vannas cover approximately 50% of the continent and the main savanna vegetation types are woodlands and tree and shrub savannas. Forest sa- vanna mosaics are common in transitional zones. In the context of the Reduced Emissions from Deforestation and forest Degradation (REDD+) programmes in Africa, emphasis has been put on the biomass change in tropical rain forests, particularly in the Congo basin (Haeusler et al., 2012). However, Ciais et al. (2011) indicate a need to consider the contribution of savanna to the overall carbon budget. Remotely sensed optical and radar data can help estimate changes in carbon in the savan- na. Vegetation indices (e.g., Normalised Difference Vegetation Index (NDVI)) can be derived from optical systems such as MODIS, MERIS and SPOT Vegetation to follow the inter-annual variability of the Net Primary Productivity (NPP) of the savanna (Fensholt, Sandholt, & Stisen, 2006). Radar systems are also expected to provide estimates of the above-ground carbon stocks and to monitor their change over time. Many studies have shown that long-wavelength radar data are sensitive to forest biomass. Successful mapping of above-ground bio- mass (AGB) has been demonstrated using P- and L-band backscatter data. Whereas P-band data are expected to be sensitive to a larger range of AGB values (Le Toan et al., 2011), only spaceborne L-band data from ALOS PALSAR have been available for studies until now. As a result, the L-band has been extensively assessed for estimating AGB (Cartus, Santoro, & Kellndorfer, 2012; Saatchi, Marlier, Chazdon, Clark, & Russell, 2011; Santoro et al., 2009; Watanabe et al., 2006). The literature suggests that there is an AGB saturation level (usually in the range of 75–150 Mg.ha -1 ) above which the L-band backscatter intensity does not increase with AGB (Dobson et al., 1992; Hoekman & Quinones, 2000; Saatchi, Halligan, Despain, & Crabtree, 2007). As a result, PALSAR data cannot be used to derive the AGB of Remote Sensing of Environment xxx (2014) xxx–xxx ⁎ Corresponding author. RSE-09018; No of Pages 11 http://dx.doi.org/10.1016/j.rse.2014.01.029 0034-4257/© 2014 Elsevier Inc. All rights reserved. Contents lists available at ScienceDirect Remote Sensing of Environment journal homepage: www.elsevier.com/locate/rse Please cite this article as: Mermoz, S., et al., Biomass assessment in the Cameroon savanna using ALOS PALSAR data, Remote Sensing of Environment (2014), http://dx.doi.org/10.1016/j.rse.2014.01.029