Estimation and validation of biomass of a mountainous agroecosystem by means of sampling,spectral data and QuickBird satellite image

Biomass estimation in agroecosystems (AESs) is important to understand their role in carbon exchange for a sustainable environment. We used field spectra and sampled biomass of an AES including cultivated and abandoned croplands to develop a simple biomass estimation model. The digital number (DN) of a QuickBird (QB) satellite image was converted to a reflectance factor using the dark object subtraction method and the spectral reflectance of asphalt. The relationship between the reflectance factor of field-based spectra and the QB image obtained in early July 2007 was insignificant in the blue (R ² = 0.15) and green (R ² = 0.18) bands but was significant (p < 0.05) in the red (R ² = 0.57) and near-infrared (NIR, R ² = 0.45) bands in the AES. Better correlations were obtained between field-based and QB-based vegetation indices (VIs). The best correlations were obtained with the normalized difference vegetation index (NDVI) (R ² = 0.97, p < 0.001) and the ratio vegetation index (RVI) (R ² = 0.99, p < 0.001). Biomass was significantly correlated with both field-based NDVI and RVI (R ² = 0.79 and 0.72, respectively, p < 0.001). Although RVI saturated at higher biomass densities (>600 g m^?2), NDVI showed a linear relationship. Other field-based VIs showed poorer correlations with biomass. The model was evaluated by incorporating it into high-resolution QB images to obtain the observed biomass. The relationship between field-estimated and QB-observed biomass appeared to be a one-to-one linear relationship (R ² = 0.79). Thus, models using field spectra and sampled biomass can be applied to QB images for remote estimation of biomass in an AES.