Search

Performing remote sensing scene classification (RSSC) directly on satellites can alleviate data downlink burdens and reduce latency. Compared to convolutional neural networks (CNNs), the all-adder neural network (A2NN) is a novel basic neural network that is more suitable for onboard RSSC, enabling lower computational overhead by eliminating multiplication operations in convolutional layers. However, the extensive floating-point data and operations in A2NNs still lead to significant storage overhead and power consumption during hardware deployment. In this article, a shared scaling factor-based de-biasing quantization (SSDQ) method tailored for the quantization of A2NNs is proposed to address this issue, including a powers-of-two (POT)-based shared scaling factor quantization scheme and a multi-dimensional de-biasing (MDD) quantization strategy. Specifically, the POT-based shared scaling factor quantization scheme converts the adder filters in A2NNs to quantized adder filters with hardware-friendly integer input activations, weights, and operations. Thus, quantized A2NNs (Q-A2NNs) composed of quantized adder filters have lower computational and memory overheads than A2NNs, increasing their utility in hardware deployment. Although low-bit-width Q-A2NNs exhibit significantly reduced RSSC accuracy compared to A2NNs, this issue can be alleviated by employing the proposed MDD quantization strategy, which combines a weight-debiasing (WD) strategy, which reduces performance degradation due to deviations in the quantized weights, with a feature-debiasing (FD) strategy, which enhances the classification performance of Q-A2NNs through minimizing deviations among the output features of each layer. Extensive experiments and analyses demonstrate that the proposed SSDQ method can efficiently quantize A2NNs to obtain Q-A2NNs with low computational and memory overheads while maintaining comparable performance to A2NNs, thus having high potential for onboard RSSC. [ABSTRACT FROM AUTHOR]

The interaction between soil moisture (SM) and vegetation dynamics has been proven in previous studies. In situ measurements have provided reliable data to investigate and validate the time effect in different zones, which is important in the hydrology and agriculture fields. There were 845 SM in situ monitoring measurements utilized with the correlation between SM and vegetation across various soil depths and climate zones in China. The impact of climate and teleconnection factors on SM and the leaf area index (LAI) are also discussed. The results indicate that SM increases from northwest to southeast in China. The time lag responses of SM to temperature, precipitation, relative humidity, and sunshine duration are 0–3 days, 3–7 days, 1–3 days, and 3–15 days, respectively. The LAI is most strongly correlated with the climate of the current month. When the LAI leads SM, a negative correlation is observed, whereas a positive correlation is observed when SM leads the LAI. This proves that vegetation growth restricts the increase in SM, and soil drying further restricts the growth of vegetation. There was a response time of 2–4 months between the LAI and SM. The effect of vegetation and deeper SM was significant in the arid zone, while they were coupled with shallow SM in the humid zone. Additionally, the El Niño–Southern Oscillation (ENSO) showed a significant positive correlation with SM in 2015–2016 with signals of 9–14 months. The results provide support for balancing the contradiction between future vegetation restoration and water resource scarcity. [ABSTRACT FROM AUTHOR]

Terrestrial water storage (TWS) is of great importance to the global water and energy budget, which modulates the hydrological cycle and then determines the spatiotemporal distributions of water resources availability. The Tibetan Plateau is the birthplace of the Yangtze, Yellow, and Lancang–Mekong River, where the water resources are directly related to the life of the Eastern and Southeastern Asian people. Based on multi-source datasets during the period 1981–2015, the long-term spatiotemporal variabilities of the TWS over the Tibetan Plateau were investigated by the Sen's slope and Mann–Kendall test trend analysis methods; the changing mechanisms were explored from two perspectives of components analysis and the hydrological cycle. The water conservation capacity of vegetation in the alpine mountainous areas was also discussed by geostatistical methods such as correlation analysis, extracted by attributes and zonal statistics. The results show that the TWS of the Tibetan Plateau increased with the speed of 0.7 mm/yr as the precipitation accumulated and the glaciers melted during the period 1981–2015. The TWS values were low and generally present a trend of obvious accumulation over the northern Tibetan Plateau, while the high and decreasing values were distributed in the south of Tibetan Plateau. The results of the components analysis indicate that the TWS mainly consisted of soil moisture at one-fourth layers, which are 0–200 cm underground in most areas of the Tibetan Plateau. The precipitation is mainly lost through evapotranspiration over the northern Tibetan Plateau, while in the northwestern corner of the Tibetan Plateau, the Himalayas, and northeastern Yarlung Zangbo River basin, the runoff coefficients were larger than 1.0 due to the influence of snow melting. In the alpine mountains, different climate and vegetation conditions have complex effects on water resources. The results are helpful for understanding the changing mechanism of water storage over the Tibetan Plateau and have scientific meaning for the development, utilization, and protection of regional water resources. [ABSTRACT FROM AUTHOR]