For Better Performance Please Use Chrome or Firefox Web Browser

SMA

This study aimed to compare the performance of pixel-based (vegetation indices, VIs) and subpixel-based (spectral mixture analysis, SMA) methods in mapping rangeland vegetation cover in Semirom region, Isfahan Province. First, the 2009 Landsat TM data was geometrically and radiametrically corrected. Then, the percentage of canopy cover was determined using step-point technique in radial direction in 30 sampling sites. The collected field data was correlated with different groups of VIs, including slop-based (NDVI), distance-based (PD54), orthogonal transformations (GVI and SBI) and plant-water sensitive (STVI-1) and also SMA in SPSS.16 software environment. To evaluate the performance of VIs and SMA techniques, vegetation cover data was classified in three percentage groups including 0-25, 25-50 and >50. In low vegetation cover group (0-25%), SMA in comparison with VIs had the highest relationship with field vegetation cover data (R2=0.52, p<0.05). Relationships between VIs and field cover data increased with increasing vegetation cover and all the VIs performed better than SMA where the percentage of vegetation cover was higher than 50%. In general, results presented here indicate that simple pixel-based techniques (e.g. NDVI) appear to be appropriate tools for assessing vegetation cover in medium and highly vegetated regions and SMA is a useful subpixel-based approach in areas with sparse vegetation cover. Therefore, this research suggests that SMA technique could be used as an alternative to VIs in arid and semi-arid environments of Iran.