Assessment of spectral vegetation indices for estimating vegetation cover in arid and semiarid shrublands

摘要

Forty four spectral vegetation indices based on Enhanced Thematic Mapper Plus (ETM+) data were compared for predicting vegetation cover in Nodushan arid and semi arid shrublands. The efficiency of the indices was evaluated based on calculating the critical error, power of prediction and the prediction interval for the error. MSAVI, TDVI, TVI, CTVI, NDVI, NRVI, IPVI, MND, TTVI, MSR and SRI provided the most accurate estimate of the vegetation cover with an error of less than 10%. MSAVI and TDVI had the lowest critical error (9.3%). The indices based on only near infrared and red bands gave the most accurate prediction of vegetation cover followed by the indices based on only near infrared and green bands (critical error of 10.7-11%), while the indices that use blue and/or shortwave infrared, or only visible (B-G-R) bands provided the least accurate estimate of the vegetation cover (critical error of 25-84%).