Hohai University

A convolutional neural network (CNN) is one of the most significant networks in the deep learning field. Since CNN made impressive achievements in many areas, including but not limited to computer vision and natural language processing, it attracted much attention from both industry and academia in the past few years. The existing reviews mainly focus on CNN’s applications in different scenarios without considering CNN from a general perspective, and some novel ideas proposed recently are not covered. In this review, we aim to provide some novel ideas and prospects in this fast-growing field. Besides, not only 2-D convolution but also 1-D and multidimensional ones are involved. First, this review introduces the history of CNN. Second, we provide an overview of various convolutions. Third, some classic and advanced CNN models are introduced; especially those key points making them reach state-of-the-art results. Fourth, through experimental analysis, we draw some conclusions and provide several rules of thumb for functions and hyperparameter selection. Fifth, the applications of 1-D, 2-D, and multidimensional convolution are covered. Finally, some open issues and promising directions for CNN are discussed as guidelines for future work.

Cyanobacteria can form dense and sometimes toxic blooms in freshwater and marine environments, which threaten ecosystem functioning and degrade water quality for recreation, drinking water, fisheries and human health. Here, we review evidence indicating that cyanobacterial blooms are increasing in frequency, magnitude and duration globally. We highlight species traits and environmental conditions that enable cyanobacteria to thrive and explain why eutrophication and climate change catalyse the global expansion of cyanobacterial blooms. Finally, we discuss management strategies, including nutrient load reductions, changes in hydrodynamics and chemical and biological controls, that can help to prevent or mitigate the proliferation of cyanobacterial blooms. Cyanobacteria can form large blooms that threaten the water quality of lakes and seas. In this Review, Huisman and colleagues discuss bloom formation, the impact of eutrophication and climate change, and measures to prevent and control cyanobacterial blooms.

Understanding and quantifying the global methane (CH4) budget is important for assessing realistic pathways to mitigate climate change. Atmospheric emissions and concentrations of CH4 continue to increase, making CH4 the second most important human-influenced greenhouse gas in terms of climate forcing, after carbon dioxide (CO2). The relative importance of CH4 compared to CO2 depends on its shorter atmospheric\nlifetime, stronger warming potential, and variations in atmospheric growth rate over the past decade, the causes of which are still debated. Two major challenges in reducing uncertainties in the atmospheric growth rate arise from the variety of geographically overlapping CH4 sources and from the destruction of CH4 by short-lived hydroxyl radicals (OH). To address these challenges, we have established a consortium of multidisciplinary scientists under the umbrella of the Global Carbon Project to synthesize and stimulate new research aimed at improving and regularly updating the global methane budget. Following Saunois et al. (2016), we present here the second version of the living review paper dedicated to the decadal methane budget, integrating results of top-down studies (atmospheric observations within an atmospheric inverse-modelling framework) and bottom-up estimates (including process-based models for estimating land surface emissions and atmospheric chemistry, inventories of anthropogenic emissions, and data-driven extrapolations).\nFor the 2008–2017 decade, global methane emissions are estimated by atmospheric inversions (a top-down approach) to be 576 TgCH4 yr-1 (range 550–594, corresponding to the minimum and maximum estimates of the model ensemble). Of this total, 359 TgCH4 yr-1 or 60% is attributed to anthropogenic sources, that is emissions caused by direct human activity (i.e. anthropogenic emissions; range 336–376 TgCH4 yr-1 or 50 %–65 %). The mean annual total emission for the new decade (2008–2017) is 29 TgCH4 yr-1 larger than our estimate for the previous decade (2000–2009), and 24 TgCH4 yr-1 larger than the one reported in the previous budget for 2003–2012 (Saunois et al., 2016). Since 2012, global CH4 emissions have been tracking the warmest scenarios assessed by the Intergovernmental Panel on Climate Change. Bottom-up methods suggest almost 30% larger global emissions (737 TgCH4 yr-1, range 594–881) than top-down inversion methods. Indeed, bottom-up estimates for natural sources such as natural wetlands, other inland water systems, and geological sources are higher than top-down estimates. The atmospheric constraints on the top-down budget suggest that at least some of these bottom-up emissions are overestimated. The latitudinal distribution of atmospheric observation-based emissions indicates a predominance of tropical emissions (∼65% of the global budget, <30◦N) compared to mid-latitudes (∼30 %, 30–60◦ N) and high northern latitudes (∼4 %, 60–90◦N). The most important source of uncertainty in the methane budget is attributable to natural emissions, especially those from wetlands and other inland waters.\nSome of our global source estimates are smaller than those in previously published budgets (Saunois et al., 2016; Kirschke et al., 2013). In particular wetland emissions are about 35 TgCH4 yr-1 lower due to improved partition wetlands and other inland waters. Emissions from geological sources and wild animals are also found to be smaller by 7 TgCH4 yr-1 by 8 TgCH4 yr-1, respectively. However, the overall discrepancy between bottom-up and top-down estimates has been reduced by only 5% compared to Saunois et al. (2016), due to a higher estimate of emissions from inland waters, highlighting the need for more detailed research on emissions factors. Priorities for improving the methane budget include (i) a global, high-resolution map of water-saturated soils and inundated areas emitting methane based on a robust classification of different types of emitting habitats; (ii) further development of process-based models for inland-water emissions; (iii) intensification of methane observations at local scales (e.g., FLUXNET-CH4 measurements) and urban-scale monitoring to constrain bottom-up land surface models, and at regional scales (surface networks and satellites) to constrain atmospheric inversions; (iv) improvements of transport models and the representation of photochemical sinks in top-down inversions; and (v) development of a 3D variational inversion system using isotopic and/or co-emitted species such as ethane to improve source partitioning.\nThe data presented here can be downloaded from https://doi.org/10.18160/GCP-CH4-2019 (Saunois et al.,\n2020) and from the Global Carbon Project

This paper presents a methodology for modeling and analyzing the load demand in a distribution system due to electric vehicle (EV) battery charging. Following a brief introduction to the common types of EV batteries and their charging characteristics, an analytical solution for predicting the EV charging load is developed. The method is stochastically formulated so as to account for the stochastic nature of the start time of individual battery charging and the initial battery state-of-charge. A comparative study is carried out by simulating four EV charging scenarios, i.e., uncontrolled domestic charging, uncontrolled off-peak domestic charging, “smart” domestic charging and uncontrolled public charging-commuters capable of recharging at the workplace. The proposed four EVs charging scenarios take into account the expected future changes to the electricity tariffs in the electricity market place and appropriate regulation of EVs battery charging loads. A typical U.K. distribution system is adopted as an example. The time-series data of EV charging loads is taken from two commercially available EV batteries: lead-acid and lithium-ion. Results show that a 10% market penetration of EVs in the studied system would result in an increase in daily peak demand by up to 17.9%, while a 20% level of EV penetration would lead to a 35.8% increase in peak load, for the scenario of uncontrolled domestic charging-the “worst-case” scenario.

This paper is the outcome of a community initiative to identify major unsolved scientific problems in hydrology motivated by a need for stronger harmonisation of research efforts. The procedure involved a public consultation through online media, followed by two workshops through which a large number of potential science questions were collated, prioritised, and synthesised. In spite of the diversity of the participants (230 scientists in total), the process revealed much about community priorities and the state of our science: a preference for continuity in research questions rather than radical departures or redirections from past and current work. Questions remain focused on the process-based understanding of hydrological variability and causality at all space and time scales. Increased attention to environmental change drives a new emphasis on understanding how change propagates across interfaces within the hydrological system and across disciplinary boundaries. In particular, the expansion of the human footprint raises a new set of questions related to human interactions with nature and water cycle feedbacks in the context of complex water management problems. We hope that this reflection and synthesis of the 23 unsolved problems in hydrology will help guide research efforts for some years to come.

The new Scientific Decade 2013-2022 of IAHS, entitled Panta RheiEverything Flows, is dedicated to research activities on change in hydrology and society. The purpose of Panta Rhei is to reach an improved interpretation of the processes governing the water cycle by focusing on their changing dynamics in connection with rapidly changing human systems. The practical aim is to improve our capability to make predictions of water resources dynamics to support sustainable societal development in a changing environment. The concept implies a focus on hydrological systems as a changing interface between environment and society, whose dynamics are essential to determine water security, human safety and development, and to set priorities for environmental management. The Scientific Decade 2013-2022 will devise innovative theoretical blueprints for the representation of processes including change and will focus on advanced monitoring and data analysis techniques. Interdisciplinarity will be sought by increased efforts to connect with the socio-economic sciences and geosciences in general. This paper presents a summary of the Science Plan of Panta Rhei, its targets, research questions and expected outcomes.

A central topic in single-atom catalysis is building strong interactions between single atoms and the support for stabilization. Herein we report the preparation of stabilized single-atom catalysts via a simultaneous self-reduction stabilization process at room temperature using ultrathin two-dimensional Ti3–xC2TyMXene nanosheets characterized by abundant Ti-deficit vacancy defects and a high reducing capability. The single atoms therein form strong metal–carbon bonds with the Ti3–xC2Ty support and are therefore stabilized onto the sites previously occupied by Ti. Pt-based single-atom catalyst (SAC) Pt1/Ti3–xC2Ty offers a green route to utilizing greenhouse gas CO2, via the formylation of amines, as a C1 source in organic synthesis. DFT calculations reveal that, compared to Pt nanoparticles, the single Pt atoms on Ti3–xC2Ty support feature partial positive charges and atomic dispersion, which helps to significantly decrease the adsorption energy and activation energy of silane, CO2, and aniline, thereby boosting catalytic performance. We believe that these results would open up new opportunities for the fabrication of SACs and the applications of MXenes in organic synthesis.

It is of great interest in exploiting texture information for classification of hyperspectral imagery (HSI) at high spatial resolution. In this paper, a classification paradigm to exploit rich texture information of HSI is proposed. The proposed framework employs local binary patterns (LBPs) to extract local image features, such as edges, corners, and spots. Two levels of fusion (i.e., feature-level fusion and decision-level fusion) are applied to the extracted LBP features along with global Gabor features and original spectral features, where feature-level fusion involves concatenation of multiple features before the pattern classification process while decision-level fusion performs on probability outputs of each individual classification pipeline and soft-decision fusion rule is adopted to merge results from the classifier ensemble. Moreover, the efficient extreme learning machine with a very simple structure is employed as the classifier. Experimental results on several HSI data sets demonstrate that the proposed framework is superior to some traditional alternatives.

Evapotranspiration is a major component of the global water cycle and provides a critical nexus between terrestrial water, carbon and surface energy exchanges. Evapotranspiration is inherently difficult to measure and predict especially at large spatial scales. Remote sensing provides a cost‐effective method to estimate evapotranspiration at regional to global scales. In the past three decades a large number of studies on remote sensing based evapotranspiration estimation have emerged. This review summarizes the basic theories underpinning current remote sensing based evapotranspiration estimation methods. It also lays out the development history of these methods and compares their advantages and limitations. Several key directions for further study are identified and discussed, including identification of uncertainty sources in remote sensing evapotranspiration models, merging of different remote sensing methods, application of data assimilation and fusion techniques in producing robust evapotranspiration estimates, and utilization of multi‐source remote sensing data and latest sensor technologies. Further advances in the remote sensing of evapotranspiration will enhance capabilities for monitoring of the global water and energy cycles, including water availability and ecosystem responses and feedbacks to climate change and human impacts. WIREs Water 2016, 3:834–853. doi: 10.1002/wat2.1168 This article is categorized under: Science of Water > Hydrological Processes Science of Water > Methods

Limited population-based cancer registry data available in China until now has hampered efforts to inform cancer control policy. Following extensive efforts to improve the systematic cancer surveillance in this country, we report on the largest pooled analysis of cancer survival data in China to date. Of 21 population-based cancer registries, data from 17 registries (n = 138,852 cancer records) were included in the final analysis. Cases were diagnosed in 2003-2005 and followed until the end of 2010. Age-standardized relative survival was calculated using region-specific life tables for all cancers combined and 26 individual cancers. Estimates were further stratified by sex and geographical area. The age-standardized 5-year relative survival for all cancers was 30.9% (95% confidence intervals: 30.6%-31.2%). Female breast cancer had high survival (73.0%) followed by cancers of the colorectum (47.2%), stomach (27.4%), esophagus (20.9%), with lung and liver cancer having poor survival (16.1% and 10.1%), respectively. Survival for women was generally higher than for men. Survival for rural patients was about half that of their urban counterparts for all cancers combined (21.8% vs. 39.5%); the pattern was similar for individual major cancers except esophageal cancer. The poor population survival rates in China emphasize the urgent need for government policy changes and investment to improve health services. While the causes for the striking urban-rural disparities observed are not fully understood, increasing access of health service in rural areas and providing basic health-care to the disadvantaged populations will be essential for reducing this disparity in the future.

ABSTRACT The Standardized Precipitation Evapotranspiration Index ( SPEI ) was computed based on the monthly precipitation and air temperature values at 609 locations over China during the period 1951–2010.Various characteristics of drought across China were examined including: long‐term trends, percentage of area affected, intensity, duration, and drought frequency. The results revealed that severe and extreme droughts have become more serious since late 1990s for all of China (with dry area increasing by ∼3.72% per decade); and persistent multi‐year severe droughts were more frequent in North China, Northeast China, and western Northwest China; significant drying trends occurred over North China, the southwest region of Northeast China, central and eastern regions of Northwest China, the central and southwestern parts of Southwest China and southwestern and northeastern parts of western Northwest mainly due to a decrease in precipitation coupled with a general increase in temperature. In addition, North China, the western Northwest China, and the Southwest China had their longest drought durations during the 1990s and 2000s. Droughts also affected western Northwest, eastern Northwest, North, and Northeast regions of China more frequently during the recent three decades. The results of this article could provide certain references and triggers for establishing a drought early warning system in China. © 2013 Royal Meteorological Society

Internet of Things (IoT) generates large amounts of data at the network edge. Machine learning models are often built on these data, to enable the detection, classification, and prediction of the future events. Due to network bandwidth, storage, and especially privacy concerns, it is often impossible to send all the IoT data to the data center for centralized model training. To address these issues, federated learning has been proposed to let nodes use the local data to train models, which are then aggregated to synthesize a global model. Most of the existing work has focused on designing learning algorithms with provable convergence time, but other issues, such as incentive mechanism, are unexplored. Although incentive mechanisms have been extensively studied in network and computation resource allocation, yet they cannot be applied to federated learning directly due to the unique challenges of information unsharing and difficulties of contribution evaluation. In this article, we study the incentive mechanism for federated learning to motivate edge nodes to contribute model training. Specifically, a deep reinforcement learning-based (DRL) incentive mechanism has been designed to determine the optimal pricing strategy for the parameter server and the optimal training strategies for edge nodes. Finally, numerical experiments have been implemented to evaluate the efficiency of the proposed DRL-based incentive mechanism.

Sensitivity analysis (SA) is en route to becoming an integral part of mathematical modeling. The tremendous potential benefits of SA are, however, yet to be fully realized, both for advancing mechanistic and data-driven modeling of human and natural systems, and in support of decision making. In this perspective paper, a multidisciplinary group of researchers and practitioners revisit the current status of SA, and outline research challenges in regard to both theoretical frameworks and their applications to solve real-world problems. Six areas are discussed that warrant further attention, including (1) structuring and standardizing SA as a discipline, (2) realizing the untapped potential of SA for systems modeling, (3) addressing the computational burden of SA, (4) progressing SA in the context of machine learning, (5) clarifying the relationship and role of SA to uncertainty quantification, and (6) evolving the use of SA in support of decision making. An outlook for the future of SA is provided that underlines how SA must underpin a wide variety of activities to better serve science and society.

International audience

Abstract. Anticipation and preparedness for large-scale flood events have a key role in mitigating their impact and optimizing the strategic planning of water resources. Although several developed countries have well-established systems for river monitoring and flood early warning, figures of populations affected every year by floods in developing countries are unsettling. This paper presents the Global Flood Awareness System (GloFAS), which has been set up to provide an overview on upcoming floods in large world river basins. GloFAS is based on distributed hydrological simulation of numerical ensemble weather predictions with global coverage. Streamflow forecasts are compared statistically to climatological simulations to detect probabilistic exceedance of warning thresholds. In this article, the system setup is described, together with an evaluation of its performance over a two-year test period and a qualitative analysis of a case study for the Pakistan flood, in summer 2010. It is shown that hazardous events in large river basins can be skilfully detected with a forecast horizon of up to 1 month. In addition, results suggest that an accurate simulation of initial model conditions and an improved parameterization of the hydrological model are key components to reproduce accurately the streamflow variability in the many different runoff regimes of the earth.

In social network group decision making (SNGDM), the consensus reaching process (CRP) is used to help decision makers with social relationships reach consensus. Many CRP studies have been conducted in SNGDM until now. This paper provides a review of CRPs in SNGDM, and as a result it classifies them into two paradigms: (i) the CRP paradigm based on trust relationships, and (ii) the CRP paradigm based on opinion evolution. Furthermore, identified research challenges are put forward to advance this area of research.

Abstract The performance of 24 GCMs available in the fifth phase of the Coupled Model Intercomparison Project (CMIP5) is evaluated over the eastern Tibetan Plateau (TP) by comparing the model outputs with ground observations for the period 1961–2005. The twenty-first century trends of precipitation and temperature based on the GCMs’ projections over the TP are also analyzed. The results suggest that for temperature most GCMs reasonably capture the climatological patterns and spatial variations of the observed climate. However, the majority of the models have cold biases, with a mean underestimation of 1.1°–2.5°C for the months December–May, and less than 1°C for June–October. For precipitation, the simulations of all models overestimate the observations in climatological annual means by 62.0%–183.0%, and only half of the 24 GCMs are able to reproduce the observed seasonal pattern, which demonstrates a critical need to improve precipitation-related processes in these models. All models produce a warming trend in the twenty-first century under the Representative Concentration Pathway 8.5 (rcp8.5) scenario; in contrast, the rcp2.6 scenario predicts a lower average warming rate for the near term, and a small cooling trend in the long-term period with the decreasing radiative forcing. In the near term, the projected precipitation change is about 3.2% higher than the 1961–2005 annual mean, whereas in the long term the precipitation is projected to increase 6.0% under rcp2.6 and 12.0% under the rcp8.5 scenario. Relative to the 1961–2005 mean, the annual temperature is projected to increase by 1.2°–1.3°C in the short term; the warmings under the rcp2.6 and rcp8.5 scenarios are 1.8° and 4.1°C, respectively, for the long term.

Adsorption is considered to be one of the most effective technologies widely used in global environmental protection areas. Modeling of experimental adsorption isotherm data is an essential way for predicting the mechanisms of adsorption, which will lead to an improvement in the area of adsorption science. In this paper, we employed three isotherm models, namely: Langmuir, Freundlich, and Dubinin-Radushkevich to correlate four sets of experimental adsorption isotherm data, which were obtained by batch tests in lab. The linearized and non-linearized isotherm models were compared and discussed. In order to determine the best fit isotherm model, the correlation coefficient (r2) and standard errors (S.E.) for each parameter were used to evaluate the data. The modeling results showed that non-linear Langmuir model could fit the data better than others, with relatively higher r2 values and smaller S.E. The linear Langmuir model had the highest value of r2, however, the maximum adsorption capacities estimated from linear Langmuir model were deviated from the experimental data.

Recent years have witnessed the widespread popularity of Internet of things (IoT). By providing sufficient data for model training and inference, IoT has promoted the development of artificial intelligence (AI) to a great extent. Under this background and trend, the traditional cloud computing model may nevertheless encounter many problems in independently tackling the massive data generated by IoT and meeting corresponding practical needs. In response, a new computing model called edge computing (EC) has drawn extensive attention from both industry and academia. With the continuous deepening of the research on EC, however, scholars have found that traditional (non-AI) methods have their limitations in enhancing the performance of EC. Seeing the successful application of AI in various fields, EC researchers start to set their sights on AI, especially from a perspective of machine learning, a branch of AI that has gained increased popularity in the past decades. In this article, we first explain the formal definition of EC and the reasons why EC has become a favorable computing model. Then, we discuss the problems of interest in EC. We summarize the traditional solutions and hightlight their limitations. By explaining the research results of using AI to optimize EC and applying AI to other fields under the EC architecture, this article can serve as a guide to explore new research ideas in these two aspects while enjoying the mutually beneficial relationship between AI and EC.

Depth camera such as Microsoft Kinect, is much cheaper than conventional 3D scanning devices, and thus it can be acquired for everyday users easily. However, the depth data captured by Kinect over a certain distance is of extreme low quality. In this paper, we present a novel scanning system for capturing 3D full human body models by using multiple Kinects. To avoid the interference phenomena, we use two Kinects to capture the upper part and lower part of a human body respectively without overlapping region. A third Kinect is used to capture the middle part of the human body from the opposite direction. We propose a practical approach for registering the various body parts of different views under non-rigid deformation. First, a rough mesh template is constructed and used to deform successive frames pairwisely. Second, global alignment is performed to distribute errors in the deformation space, which can solve the loop closure problem efficiently. Misalignment caused by complex occlusion can also be handled reasonably by our global alignment algorithm. The experimental results have shown the efficiency and applicability of our system. Our system obtains impressive results in a few minutes with low price devices, thus is practically useful for generating personalized avatars for everyday users. Our system has been used for 3D human animation and virtual try on, and can further facilitate a range of home–oriented virtual reality (VR) applications.