The study of vegetation phenology changes is important because it is a sensitive indicator of climate change, affecting the exchange of carbon, energy and water fluxes between the land and the atmosphere. Previous studies have focused on the effects of climatic factors among environmental factors on vegetation phenology, thus the effects of non-climatic factors among environmental factors have not been well quantified. This study endeavors to scrutinize the spatiotemporal inconsistency in the start-of-season (SOS) and the end-of-season (EOS) on the Tibetan Plateau (TP) and to quantify the effects of environmental factors on phenology. To this end, the Moderate-resolution Imaging Spectroradiomater (MODIS) Normalized Difference Vegetation Index (NDVI) data from 2001 to 2018 and four common used methods were employed to extract SOS and EOS, and the site data was used to select the most appropriate phenology results. The Geodetector model was used to assess and measure the explanatory power of different environmental factors. The research results indicate that temperature exerts a more substantial impact on phenology than precipitation on TP. non-climatic factors such as longitude, latitude, and elevation are more influential in determining the distribution of phenological trends than climatic factors. Among these non-climatic factors, latitude has the most prominent effect on the trends of SOS. Furthermore, non-climatic factors exhibit a stronger effect on SOS, whereas EOS is more susceptible to climatic factors and less influenced by non-climatic factors. These discoveries bear great significance in comprehending the intricate outcomes of regional changes on vegetation phenology and enhancing phenology models.
Hyperspectral image classification (HSI) faces challenges from high-dimensional redundancy and spatial heterogeneity. Existing methods struggle with computational inefficiency in classifier selection due to diverse architectures and hyperparameters. We propose DEDES, a dynamic ensemble framework that addresses this via spectral-spatial embedding and density-based clustering to partition features into homogeneous regions, enabling localized classifier activation. Experiments on Botswana and Purdue datasets demonstrate state-of-the-art accuracy (overall accuracy of 86.25% and 94.40%) compared to conventional methods. The framework eliminates exhaustive aggregation, demonstrating optimal accuracy through adaptive minimal classifier selection, offering a scalable solution for real-world HSI interpretation.