347 research outputs found
Inorganic Carbon Pools and Their Drivers in Grassland and Desert Soils
No full textABSTRACT Inorganic carbon is an important component of soil carbon stocks, exerting a profound influence on climate change and ecosystem functioning. Drylands account for approximately 80% of the global soil inorganic carbon (SIC) pool within the top 200 cm. Despite its paramount importance, the components of SIC and their contributions to CO 2 fluxes have been largely overlooked, resulting in notable gaps in understanding its distribution, composition, and responses to environmental factors across ecosystems, especially in deserts and temperate grasslands. Utilizing a dataset of 6011 samples from 173 sites across 224 million hectares, the data revealed that deserts and grasslands in northwestern China contain 20 ± 2.5 and 5 ± 1.3 petagrams of SIC in the top 100 cm, representing 5.5 and 0.76 times the corresponding soil organic carbon stock, respectively. Pedogenic carbonates (PIC), formed by the dissolution and re‐precipitation of carbonates, dominated in grasslands, accounting for 60% of SIC with an area‐weighted density of 3.4 ± 0.4 kg C m −2 at 0–100 cm depth, while lithogenic carbonates (LIC), inherited from soil parent materials, prevailed in deserts, constituting 55% of SIC with an area‐weighted density of 7.1 ± 1.0 kg C m −2 . Soil parent materials and elevation determined the SIC stocks by regulating the formation and loss of LIC in deserts, whereas natural acidification, mainly induced by rhizosphere processes including cation uptake and H + release as well as precipitation, reduced SIC (mainly by PIC) in grasslands. Overall, the massive SIC pool underscores its irreplaceable role in maintaining the total carbon pool in drylands. This study sheds light on LIC and PIC and highlights the critical impact of natural acidification on SIC loss in grasslands.National Key Research and Development Program of China https://doi.org/10.13039/501100012166National Natural Science Foundation of China https://doi.org/10.13039/501100001809Natural Science Foundation of Gansu Province https://doi.org/10.13039/50110000477
LARGE DIMENSIONAL ANALYSIS AND OPTIMIZATION FOR MASSIVE MIMO WIRELESS NETWORKS
No full textThe Last decade has seen a massive growth of wireless devices. Demands for high capacity and massive connectivity always increases. To meet the these requirements, massive multiple-input multiple-output (MIMO) technology is proposed for the next generation wireless systems. With massive antenna arrays at the BS, the channel vectors between the users and the BS are asymptotically orthogonal, and the intercell interference can be eliminated with simple linear processing method. The effects of fast fading, and uncorrelated noise tend to disappear as the number of BS station antennas grows large. In this thesis, we mainly focus on the performance analysis and system optimization of massive MIMO cellular networks. We compare two cooperative multicell precoding methods, namely, centralized precoding with global CSI and distributed precoding with local individual CSI in the large dimensional regime. Two different massive MIMO scenarios are considered. When the number of antennas and that of users go large at the same rate, there is a constant gap between the two precoding cases. When the number of antennas goes large, while the number of users is fixed, the performances of both schemes are the same. This means that the impact of local individual CSI vanishes. In the large dimension limit, certain terms such as signal-to-interference-plus-noise ratio (SINR) can be approximated depending only on the statistical CSI. Based on this result, the power optimization problem does not need to adapt as frequently as the instantaneous channel state information. By this means, performance evaluation and optimization is more computationally efficient. Partial interference alignment in finite SNR is proposed, which captures the trade-off between interference avoidance at other users and spatial multiplexing at the intended user. We try to project the interference into partial subspace of the small cell user instead of the full subspace, so that extra improvement on the rate of macro cell can be offered in finite SNR. Large dimensional analysis is provided to obtain the the number of transmit dimensions for small cell in finite SNR. We show that the number of dimensions the small cell can transmit is determined by the singular value distribution of channel matrix and the transmit SNR of the system. In heterogeneous sensor networks, the sensing data may not be compatible with each other due to heterogeneous sensing modalities. We propose a probabilistic inference framework for fusing in formation from heterogeneous sensors. The inherent inter-sensor relationship is exploited to encode the original sensor data in a graph. The iterative belief propagation is used to fuse the local sensing belief. Then we consider the more general correlation case, in which the relation between two sensors is characterized by the correlation factor. The belief propagation provides intuitive insights as to how the local probabilistic update helps to reinforce beliefs when performing information fusion
Behavior Analysis and Enhancement of Robustness for Deep Neural Networks
No full textApart from the remarkable success of machine learning models utilizing deep neural networks in solving a variety of problems including image classifications, these models are highly vulnerable to small and carefully chosen modifications to inputs, known as adversarial examples. While such perturbations are often simple noise to a human or even imperceptible, the perturbations cause state-of-the-art models to misclassify input data with high confidence. To close this technological gap, we are primed to investigate the behaviors of deep neural networks, aiming to understand inner working mechanisms and to boost the classification accuracy of deep neural networks. In the first part of this dissertation, we study the neuron activation behaviors of a well-trained classification model. An information theoretical method is leveraged to examine the behavior of layer-wise neurons in deep neural networks. We discover that in a well-trained classification model, the randomness level of a neurons activation pattern is curtailed with the depth of fully connected layers. This finding suggests that the neuron activation patterns of deep layers are more stable than those of shallow layers.
In the second part of the dissertation, we advocate for an approach to incorporating a diversity of symmetries, such as rotation and scaling, into an existing CNN model to enhance the robustness of deep neural network models. We illustrate that the perturbation invariance property can be approximated by various symmetries incorporated in the models. Furthermore, it is guaranteed that a model - equipped with the symmetry enforcement - becomes generalizable to input data that are shifted, rotated, or scaled. Importantly, we explore the relationship between generalization and robustness of deep neural networks, thereby empirically shedding bright light on the impact of generalization on adversarial robustness. The symmetry operations not only multiply the efficiency of training data but also substantially strengthen the expressive capacity of a network, where the adversarial robustness is bolstered.
In the third part of the dissertation research, we delve in the development of robust target detection techniques against malicious attack based on representation learning. We construct a two-stage framework for fusing information from heterogeneous sensors. The representation learning stage, a core underpinning, is capable of transforming data into a unified data form. The nature encoded fusion - implemented in the newly designed framework - allows data from different modalities to be processed in a unified probabilistic space. Inherent inter-sensor relationships are exploited: such relationships are envisioned as a nature encoded sensing with heterogeneous sensors. Malicious data injection attacks may tamper with the sensors of a fusion center or a data acquisition system, thereby seriously downgrading the target detection performance of sensor fusion system. We demonstrate that the iterative belief propagation is slated to refine and fuse local individual beliefs to combat the malicious data attacks. Our findings confirm that the belief propagation method furnishes intuitive insights: probabilistic updates are capable of reinforcing beliefs with the help of a correlation factor
Data products for monthly GRACE derived groundwater storage changes (GWS) in North America from 2002 to 2017
No full textWe release three data products for monthly GRACE derived groundwater storage changes (GWS) in North America. In the first one, we provide an independent estimate for monthly GWS changes within North America in 1-degree-grids and their trends over the whole GRACE mission lifetime from April 2002 to June 2017. In the second one, we give the monthly GWS changes and the trends averaged for the 5 major GWS trend anomalies in around Saskatchewan, Nevada, California, Arizona and Texas, respectively. The third data product includes the monthly GWS changes and the trends averaged for the 14 states or provinces in the US and Canada, affected by the above GWS trend anomalies, i.e., for Saskatchewan, Montana, Nevada, California, Arizona, New Mexico, Texas, Oklahoma, Kansas, Alberta, North Dakota, Minnesota, Colorado and Chihuahuas, respectively. Our GWS estimates are derived from the release-6 version of GRACE monthly level-2 data, GNSS data, two land surface models of new version of GLDAS dataset (GLDAS 2.1) for soil moisture and snow water equivalent, and satellite altimetric lake level data. Especially, unlike previous studies, glacial isostatic adjustment (GIA) effects are eliminated by employing an independent separation approach with the aid of GNSS (Global Navigation Satellite System) vertical velocity data. We find a GWS anomaly in form of an increasing trend in Saskatchewan, which affects the Saskatchewan Province and the states of Montana, North Dakota and Minnesota, and 4 GWS anomalies with declining trends in Nevada, California, Arizona and Texas, respectively. The monthly changes of these GWS anomalies are validated by well level data
Comparison of surface water chemistry and weathering effects of two lake basins in the Changtang Nature Reserve, China
No full textThe geochemistry of natural waters in the Changtang Nature Reserve, northern Tibet, can help us understand the geology of catchments, and provide additional insight in surface processes that influence water chemistry such as rock weathering on the Qinghai-Tibet Plateau. However, severe natural conditions are responsible for a lack of scientific data for this area. This study represents the first investigation of the chemical composition of surface waters and weathering effects in two lake basins in the reserve (Lake Dogaicoring Qiangco and Lake Longwei Co). The results indicate that total dissolved solids (TDS) in the two lakes are significantly higher than in other gauged lakes on the Qinghai-Tibet Plateau, reaching 20-40 g/L, and that TDS of the tectonic lake (Lake Dogaicoring Qiangco) is significantly higher than that of the barrier lake (Lake Longwei Co). Na+ and Cl-are the dominant ions in the lake waters as well as in the glacier-fed lake inflows, with chemical compositions mainly affected by halite weathering. In contrast, ion contents of inflowing rivers fed by nearby runoff are lower and concentrations of dominant ions are not significant. Evaporite, silicate, and carbonate weathering has relatively equal effects on these rivers. Due to their limited scope, small streams near the lakes are less affected by carbonate than by silicate weathering. (C) 2015 The Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences. Published by Elsevier B.V
Low complexity optimization for user centric cellular networks via large dimensional analysis
No full textWater storage changes in the interior of the Tibetan Plateau, link to data underlying Fig. 1-3 and A1-A13
No full textWater storage changes in the interior of the Tibetan Plateau, link to data underlying Fig. 1-3 and A1-A1
Performance Analysis of Cooperative Multicell Precoding With Global CSI and Local Individual CSI in the Large Dimensional Regime
No full text
- …
