Proceedings Article | 22 October 2010
KEYWORDS: Vegetation, Calibration, Reflectivity, Atmospheric corrections, Soil science, Data modeling, Atmospheric sensing, Ecosystems, Sensors, Remote sensing
Fraction of vegetation (Fv) plays an important role in ecosystems. Estimation of Fv is essential for drought monitoring,
natural resources studies, estimation of soil erosion volume etc. The aim of this study is to estimate Fv in an arid area in
Iran using ALOS Imagery (June 2008). In order to find the best index for estimation of Fv, Seventeen vegetation indices
(ARVI, DVI, EVI, GEMI, IPVI, MSAVI1, MSAVI2, NDVI, PVI, SAVI, SARVI, SARVI2, SR, TSAVI, WDVI) were
used. The canopy cover percentage of 52 sample plots (50m by 50m) was measured in the field in June 2009. Regression
models were used to assess the relationships between the field data and the calculated Fv. The 52 sample plots were
randomly divided two times to 30 calibrations and 22 validations, and to 35 and 17 samples. Results revealed that
selecting the calibration and validation data randomly leads to different results. Therefore, cross-validation method was
used to reduce random division effect. Results indicated that, among all indices, vegetation indices such as MSAVI1,
PVI, WDVI and TSAVI which are based on soil line have higher R2 and lower RMSE (R2 > 0.63, RMSE ≈ 3%). The
results confirm the dominant effect of soil reflectance in arid areas.