No abstract

Convolutional neural networks provide an ideal solution for hyperspectral image (HSI) classification. However, the classification effect is not satisfactory when limited training samples are available. Focused on “small sample” hyperspectral classification, we proposed a novel 3D-2D-convolutional neural network (CNN) model named AD-HybridSN (Attention-Dense-HybridSN). In our proposed model, a dense block was used to reuse shallow features and aimed at better exploiting hierarchical spatial–spectral features. Subsequent depth separable convolutional layers were used to discriminate the spatial information. Further refinement of spatial–spectral features was realized by the channel attention method and spatial attention method, which were performed behind every 3D convolutional layer and every 2D convolutional layer, respectively. Experiment results indicate that our proposed model can learn more discriminative spatial–spectral features using very few training data. In Indian Pines, Salinas and the University of Pavia, AD-HybridSN obtain 97.02%, 99.59% and 98.32% overall accuracy using only 5%, 1% and 1% labeled data for training, respectively, which are far better than all the contrast models.

Exploring the variations in the water use efficiency (WUE) is helpful in gaining an in-depth understanding of the regional carbon and water cycles on the Chinese Loess Plateau (CLP). Here, we employed the spatial variations in the WUE and the quantitative contributions of the influencing factors, including the precipitation (P), temperature (Temp), vapor pressure deficit (VPD), sunshine duration (SD), and leaf area index (LAI), with the drought index varying over the last two decades. Results showed that the multiyear average WUE decreased significantly as the drought index increased for all of the vegetation types. Per-pixel interannual variability of WUE trend was 0.024 gC·m−2·mm−1·yr−1. As the drought index increased, the WUE initially increased and then decreased for the forests, grassland, and shrubland, and their peaks occurred at drought index values of 2.60–3.10. Among the influencing factors, the WUE was predominantly controlled by the LAI, with an impact and relative contribution of 0.014 gC·m−2·mm−1·yr−1 and 58.3%, respectively. The P and SD contributed the least to the trend in WUE, and impact and relative contribution of both were 0.001 gC·m−2·mm−1·yr−1 and 4.17%. Our study also demonstrated that the LAI was the dominant factor affecting the WUE trends for grassland and the Yan River due to the structural parameters and geographical location. In addition, the impact and relative contribution of the residual factors on the WUE trend were 0.004 gC·m−2·mm−1·yr−1 and 16.7%. Our findings suggested that comprehensive effects such as micro-geomorphic changes and nitrogen deposition could not be ignored except for vegetation and climate change. This study will clarify the spatial and temporal evolution of WUE and its influence mechanism.

No abstract

No abstract