Leaf age is one of the important parameters for describing crop growth stages. In actual production, cultivation methods can be adjusted in a timely manner according to leaf age, which can promote rapid seedling transformation and lay a solid foundation for achieving high yields of winter wheat. However, most existing image segmentation methods focus on leaf recognition and lack consideration for the impact of growing degree days (GDD) on winter wheat growth. This study takes Shandong province as the research area. Firstly, based on MOD09Q1 images, the normalized difference vegetation index (NDVI) time series was reconstructed using cubic spline interpolation and Savitzky-Golay (S-G) filter. Secondly, the dynamic threshold method was used to extract the winter wheat emergence date, the NDVI slope (β) was calculated from the emergence date to the monitoring date, and the GDD during this period was calculated based on ERA5-Land reanalysis temperature data. Finally, a GDD-based leaf age (LA GDD ) model and a Random Forest-based leaf age (LA RF ) model were constructed separately to monitor winter wheat leaf age at the start of the overwintering stage. The research results indicate that compared with the phenological observation data in Shandong province, the winter wheat emergence date extracted using the dynamic threshold method has high accuracy (bias = 0.47 days (d), R 2 = 0.77, and RMSE = 3.93 d). Compared with the LA GDD model ( R 2 = 0.59, RMSE = 0.68 leaves), the LA RF model significantly improves leaf age monitoring accuracy and performs well in both the training and validation sets, showing strong correlation with measured data (training set: R 2 = 0.85, RMSE = 0.39 leaves, validation set: R 2 = 0.67, RMSE = 0.56 leaves). In terms of spatial distribution, the leaf age in the southern and northern regions of Shandong province is relatively stable during the pre-wintering growth stage, with an overall higher leaf age of over 5 leaves in most years. However, the Jiaodong Peninsula has relatively lower leaf ages, but the seedlings are also developing well. In terms of temporal distribution, the leaf age is generally lower in 2021, while the leaf age in other years is moderate, and the region has favorable seedling conditions. This study integrates remote sensing phenological information, vegetation dynamic parameters, and meteorological factors to monitor winter wheat leaf age on a regional scale. The proposed model demonstrates high accuracy and stability, providing scientific support for seedling assessment and precise field management at the pre-overwintering stage.
Zhang et al. (Fri,) studied this question.