The normalized difference yellow vegetation index (NDYVI): A new index for crop identification by using GaoFen-6 WFV data

Abstract

The yellowing morphologies of crops provide typical spectral characteristics for crop identification. However, this feature was generally neglected by most existing vegetation indices (VIs) focused on greening feature. Chinese GaoFen-6 satellite (GF-6) equipped with a wide field-of-view (WFV) camera has a refined spectral system within 0.40–0.89 μm, including the spectral bands that are sensitive to yellowing feature. This study proposes a new normalized difference yellow vegetation index (NDYVI) based on GF-6 image, by capitalizing on spectral reflectance feature of crops with yellowing morphologies, such as flowers and tassels. We used yellow and red-edge1 band to discriminate between crops within similar growing periods and incorporated NIR band to distinguish non-crop types. The performance of NDYVI was evaluated in two distinct classification scenarios involving different cropping systems: rapeseed with winter wheat in southern China, and maize with soybean in northeastern China. By calculating NDYVI and using Classification and Regression Tree (CART) algorithm, we generated classification maps in two scenarios. Additionally, the effectiveness of NDYVI was tested and compared with other six VIs, such as Normalized Difference Vegetation Index, Red-Edge Normalized Difference Vegetation Index and Normalized Difference Yellowness Index. The results demonstrated that NDYVI outperformed the other vegetation indices in both scenarios, achieving overall accuracies over 85 % (Kappa coefficient greater than 0.80) and each crop accuracy exceeding 80 %. Due to the higher reflectance in yellow band and red-edge1 band, NDYVI is more sensitive to canopies yellowness, which offers significant advantages in distinguishing crops during similar growing periods. Moreover, NDYVI is constructed from original spectral bands in GF-6 images, offering potential for significant flexibility in diverse classification scenarios. Consequently, NDYVI holds significant potential as a new vegetation index suitable for different remote sensing applications, including crop identification, growth monitoring and land cover classification.