In: Wagner W„ Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, ¡APRS, Vol. XXXVIII, Part 7B 
407 
EFFICIENCY OF TEXTURE MEASUREMENT FROM TWO OPTICAL SENSORS FOR 
IMPROVED BIOMASS ESTIMATION 
J. E. Nichol, M. L. R. Sarker 
Department of Land Surveying and Geo-Informatics, The Hong Kong Polytechnic University 
Kowloon, Hong Kong - lsjanet@polyu.edu.hk 
KEY WORDS: AVNIR-2, SPOT-5, texture measurement, biomass estimation, image processing 
ABSTRACT: 
No technique has so far been developed to quantify biomass carbon sources and sinks over large areas. Among the remote sensing 
techniques tested, the use of multisensors, and spatial as well as spectral characteristics of data have demonstrated strong potential 
for biomass estimation. However, the use of multisensor data accompanied by spatial data processing has not been fully investigated 
because of the unavailability of appropriate data sets and the complexity of image processing techniques for combining multisensor 
data with the analysis of spatial characteristics. This research investigates the texture parameters of two high (10m) resolution optical 
sensors AVNIR-2 and SPOT-5 in different processing combinations for biomass estimation. Multiple regression models are 
developed between image parameters extracted from the different stages of image processing and the biomass of 50 field plots, 
which was estimated using a newly developed “Allometric Model” for the study region. 
The results demonstrate a clear improvement in biomass estimation using the texture parameters of a single sensor (r 2 =0.854 and 
RMSE=38.54) compared to the highest accuracy obtained from simple spectral reflectance (r 2 =0.494) and simple spectral band ratios 
(^=0.59). This accuracy was further improved, to obtain a very promising accuracy using texture parameters of both sensors together 
(r 2 =0.897 and RMSE=32.38), the texture parameters from the PCA of both sensors (^=0.851 and RMSE=38.80) and the texture 
parameters from the averaging of both sensors (r 2 =0.911 and RMSE=30.10). Improved accuracy was also observed using the simple 
ratio of texture parameters of AVNIR-2 (r 2 =0.899 and RMSEA32.04) and SPOT-5 (r 2 =0.916) and finally a surprisingly high 
accuracy (r 2 =0.939 and RMSE=24.77) was achieved using the ratios of the texture parameter of both sensors together. 
1. INTRODUCTION 
Remote sensing is the most promising technique to estimate 
biomass at local, regional and global scales, thereby helping to 
reduce the uncertainties associated with the role of forests in 
key environmental issues (Brown et al, 1989; Rosenqvist et al 
2003). A number of studies has been carried out using different 
types of sensors including optical (Mukkonen and Heiskannen, 
2005; Fuchs et al 2009; Foody et al, 2003; Dong et al, 2003) 
SAR (Santos et al, 2003; Kuplich et al, 2005), and Lidar 
sensors (Zhao et al 2009) for biomass/forest parameter 
estimation. Apart from the use of a single sensor, combining 
information from multiple sensors has yielded promising results 
for the estimation of forest parameters/biomass (Rosenqvist et 
al, 2003; Hyde et al, 2006; Boyd and Danson, 2005. 
Although vegetation indices, have been successfully used in 
temperate forests Zheng et al, 2004; Rahman et al, 2005), they 
have shown less potential in tropical and subtropical regions 
where biomass levels are high, the forest canopy is closed with 
multiple layering, and great diversity of species is present 
(Foody et al, 2001, 2003; Boyd et al, 1996; Lu, 2005). On the 
other hand, the spatial characteristics of images have such as 
texture have been found particularly useful in fine spatial 
resolution imagery (Franklin et al, 2001; Boyd and Danson, 
2005), and capable of identifying different aspects of forest 
stand structure, including age, density and leaf area index 
(Champion et al, 2008; Wulder et al, 1996). Indeed, texture has 
shown potential for biomass estimation with both optical 
(Franklin et al, 2001; Lu, 2005; Fuchs et al, 2009) and SAR 
data (Santos et al, 2003; Lu, 2005; Kuplich et al, 2005) 
Moreover, although most previous biomass estimation projects 
used Landsat TM data with a 30m spatial resolution (Lu, 2006), 
texture is expected to be more effective with finer spatial 
resolution imagery since finer structural details can be 
distinguished (Kuplich et al, 2005; Boyd and Danson, 2005; 
Franklin et al, 2001). This research investigates texture 
processing for biomass estimation using data from two high 
resolution optical sensors ANVIR-2 and SPOT-5 along with 
raw spectral processing and some simple band ratios. The 
overall objective of the study is to explore the potential of 
texture processing combined with multisensor capability for the 
improvement of biomass estimation using data from two high 
resolution optical sensors. 
The study area for this research is the Hong Kong Special 
Administrative Region (Fig. 1) which lies on the southeast coast 
of China, just south of the Tropic of Cancer. Approximately 
40% of Hong Kong is designated as Country Parks which are 
reserved for forest succession. The native sub-tropical 
evergreen broad leaf forest has been replaced by a complex 
patchwork of regenerating secondary forest in various stages of 
development, and plantations. Forest grades into woodland, 
shrubland then grassland at higher elevations.