Abstract

High-quality vegetation index time series are crucial for timely and accurate phenological mapping. However, vegetation index time series with high temporal resolution (such as daily resolution), can be influenced by cloudiness, shadows, and atmospheric effect and affects the interpretation of vegetation growth. We proposed a method for computing a daily 250m two-band enhanced vegetation index (EVI2) time series using MOD09GQ surface reflectance products through a self-supervised neural network autoencoders model. We tested the consistency of EVI2 with conventional three-band EVI, its temporal stability, and sensitivity to aerosol. The results indicated this method is suitable to be applied to areas where atmospheric effects are not prominent. (tel: +86 15229367141, e-mail: rsliuxk@whu.edu.cn).

© 2023 The Author(s)