Showing papers by "Nanjing University of Information Science and Technology published in 2014"

Strong contributions of local background climate to urban heat islands

Abstract: Climate modelling is used to show that for cities across North America, geographic variations in daytime urban heat islands—that is, the temperature differences between urban and adjacent rural areas—are largely explained by variations in the efficiency with which those areas convect heat to the lower atmosphere. It is often warmer in a city than in the surrounding rural areas, sometimes by up to a few degrees. This urban heat island effect is commonly explained as a consequence of a lower rate of evaporative cooling in urban areas. But here Xuhui Lee and colleagues use climate modelling to show that for cities across North America, the daytime urban heat island effect varies with the efficiency of heat convection between the land surface and the lower atmosphere. The convection effect varies with climate regime, causing significant urban warming in wet climates but cooling in dry climates. Aerodynamics also play a part, and if urban areas are aerodynamically smoother than surrounding rural areas, urban heat dissipation is less efficient and warming occurs. The health impact of heatwaves means that mitigation of the heat island effect may be beneficial. The authors suggest that aerodynamic spoilers — a city-wide increase in building height for instance — may be impractical. But efforts to increase urban albedo, by installing reflective roofs for instance, might be worth pursuing. The urban heat island (UHI), a common phenomenon in which surface temperatures are higher in urban areas than in surrounding rural areas, represents one of the most significant human-induced changes to Earth’s surface climate1,2. Even though they are localized hotspots in the landscape, UHIs have a profound impact on the lives of urban residents, who comprise more than half of the world’s population3. A barrier to UHI mitigation is the lack of quantitative attribution of the various contributions to UHI intensity4 (expressed as the temperature difference between urban and rural areas, ΔT). A common perception is that reduction in evaporative cooling in urban land is the dominant driver of ΔT (ref. 5). Here we use a climate model to show that, for cities across North America, geographic variations in daytime ΔT are largely explained by variations in the efficiency with which urban and rural areas convect heat to the lower atmosphere. If urban areas are aerodynamically smoother than surrounding rural areas, urban heat dissipation is relatively less efficient and urban warming occurs (and vice versa). This convection effect depends on the local background climate, increasing daytime ΔT by 3.0 ± 0.3 kelvin (mean and standard error) in humid climates but decreasing ΔT by 1.5 ± 0.2 kelvin in dry climates. In the humid eastern United States, there is evidence of higher ΔT in drier years. These relationships imply that UHIs will exacerbate heatwave stress on human health in wet climates where high temperature effects are already compounded by high air humidity6,7 and in drier years when positive temperature anomalies may be reinforced by a precipitation–temperature feedback8. Our results support albedo management as a viable means of reducing ΔT on large scales9,10.

Fast Visual Tracking via Dense Spatio-temporal Context Learning

Abstract: In this paper, we present a simple yet fast and robust algorithm which exploits the dense spatio-temporal context for visual tracking. Our approach formulates the spatio-temporal relationships between the object of interest and its locally dense contexts in a Bayesian framework, which models the statistical correlation between the simple low-level features (i.e., image intensity and position) from the target and its surrounding regions. The tracking problem is then posed by computing a confidence map which takes into account the prior information of the target location and thereby alleviates target location ambiguity effectively. We further propose a novel explicit scale adaptation scheme, which is able to deal with target scale variations efficiently and effectively. The Fast Fourier Transform (FFT) is adopted for fast learning and detection in this work, which only needs 4 FFT operations. Implemented in MATLAB without code optimization, the proposed tracker runs at 350 frames per second on an i7 machine. Extensive experimental results show that the proposed algorithm performs favorably against state-of-the-art methods in terms of efficiency, accuracy and robustness.

Investigation of the sources and evolution processes of severe haze pollution in Beijing in January 2013

Abstract: China experienced severe haze pollution in January 2013. Here we have a detailed characterization of the sources and evolution mechanisms of this haze pollution with a focus on four haze episodes that occurred during 10–14 January in Beijing. The main source of data analyzed is from submicron aerosol measurements by an Aerodyne Aerosol Chemical Speciation Monitor. The average PM1 mass concentration during the four haze episodes ranged from 144 to 300 µg m−3, which was more than 10 times higher than that observed during clean periods. All submicron aerosol species showed substantial increases during haze episodes with sulfate being the largest. Secondary inorganic species played enhanced roles in the haze formation as suggested by their elevated contributions during haze episodes. Positive matrix factorization analysis resolved six organic aerosol (OA) factors including three primary OA (POA) factors from traffic, cooking, and coal combustion emissions, respectively, and three secondary OA (SOA) factors. Overall, SOA contributed 41–59% of OA with the rest being POA. Coal combustion OA (CCOA) was the largest primary source, on average accounting for 20–32% of OA, and showed the most significant enhancement during haze episodes. A regional SOA (RSOA) was resolved for the first time which showed a pronounced peak only during the record-breaking haze episode (Ep3) on 12–13 January. The regional contributions estimated based on the steep evolution of air pollutants were found to play dominant roles for the formation of Ep3, on average accounting for 66% of PM1 during the peak of Ep3 with sulfate, CCOA, and RSOA being the largest fractions (> ~ 75%). Our results suggest that stagnant meteorological conditions, coal combustion, secondary production, and regional transport are four main factors driving the formation and evolution of haze pollution in Beijing during wintertime.

Fast Compressive Tracking

Abstract: It is a challenging task to develop effective and efficient appearance models for robust object tracking due to factors such as pose variation, illumination change, occlusion, and motion blur. Existing online tracking algorithms often update models with samples from observations in recent frames. Despite much success has been demonstrated, numerous issues remain to be addressed. First, while these adaptive appearance models are data-dependent, there does not exist sufficient amount of data for online algorithms to learn at the outset. Second, online tracking algorithms often encounter the drift problems. As a result of self-taught learning, misaligned samples are likely to be added and degrade the appearance models. In this paper, we propose a simple yet effective and efficient tracking algorithm with an appearance model based on features extracted from a multiscale image feature space with data-independent basis. The proposed appearance model employs non-adaptive random projections that preserve the structure of the image feature space of objects. A very sparse measurement matrix is constructed to efficiently extract the features for the appearance model. We compress sample images of the foreground target and the background using the same sparse measurement matrix. The tracking task is formulated as a binary classification via a naive Bayes classifier with online update in the compressed domain. A coarse-to-fine search strategy is adopted to further reduce the computational complexity in the detection procedure. The proposed compressive tracking algorithm runs in real-time and performs favorably against state-of-the-art methods on challenging sequences in terms of efficiency, accuracy and robustness.

From Heuristic Optimization to Dictionary Learning: A Review and Comprehensive Comparison of Image Denoising Algorithms

Abstract: Image denoising is a well explored topic in the field of image processing. In the past several decades, the progress made in image denoising has benefited from the improved modeling of natural images. In this paper, we introduce a new taxonomy based on image representations for a better understanding of state-of-the-art image denoising techniques. Within each category, several representative algorithms are selected for evaluation and comparison. The experimental results are discussed and analyzed to determine the overall advantages and disadvantages of each category. In general, the nonlocal methods within each category produce better denoising results than local ones. In addition, methods based on overcomplete representations using learned dictionaries perform better than others. The comprehensive study in this paper would serve as a good reference and stimulate new research ideas in image denoising.

Construction of Tree Network with Limited Delivery Latency in Homogeneous Wireless Sensor Networks

Abstract: In this paper we propose an approximation algorithm, which is called ADCMCST (algorithm with the minimum number of child nodes when the depth is restricted), to construct a tree network for homogeneous wireless sensor network, so as to reduce and balance the payload of each node, and consequently prolong the network lifetime. When the monitoring node obtains the neighbor graph, ADCMCST tries to find a tree topology with a minimum number of child nodes, and then broadcast the topology to every node, and finally a tree network is constructed. Simulation results show that ADCMCST could greatly reduce the topology formation time, and achieve good approximation results; when the compression ratio is less than 70 %, the network lifetime of ADCMCST will be larger than that of energy driven tree construction.

Widespread decline of Congo rainforest greenness in the past decade

Abstract: Tropical forests are global epicentres of biodiversity and important modulators of climate change, and are mainly constrained by rainfall patterns. The severe short-term droughts that occurred recently in Amazonia have drawn attention to the vulnerability of tropical forests to climatic disturbances. The central African rainforests, the second-largest on Earth, have experienced a long-term drying trend whose impacts on vegetation dynamics remain mostly unknown because in situ observations are very limited. The Congolese forest, with its drier conditions and higher percentage of semi-evergreen trees, may be more tolerant to short-term rainfall reduction than are wetter tropical forests, but for a long-term drought there may be critical thresholds of water availability below which higher-biomass, closed-canopy forests transition to more open, lower-biomass forests. Here we present observational evidence for a widespread decline in forest greenness over the past decade based on analyses of satellite data (optical, thermal, microwave and gravity) from several independent sensors over the Congo basin. This decline in vegetation greenness, particularly in the northern Congolese forest, is generally consistent with decreases in rainfall, terrestrial water storage, water content in aboveground woody and leaf biomass, and the canopy backscatter anomaly caused by changes in structure and moisture in upper forest layers. It is also consistent with increases in photosynthetically active radiation and land surface temperature. These multiple lines of evidence indicate that this large-scale vegetation browning, or loss of photosynthetic capacity, may be partially attributable to the long-term drying trend. Our results suggest that a continued gradual decline of photosynthetic capacity and moisture content driven by the persistent drying trend could alter the composition and structure of the Congolese forest to favour the spread of drought-tolerant species.

Accuracy assessment of global barotropic ocean tide models

Abstract: The accuracy of state-of-the-art global barotropic tide models is assessed using bottom pressure data, coastal tide gauges, satellite altimetry, various geodetic data on Antarctic ice shelves, and independent tracked satellite orbit perturbations. Tide models under review include empirical, purely hydrodynamic (“forward”), and assimilative dynamical, i.e., constrained by observations. Ten dominant tidal constituents in the diurnal, semidiurnal, and quarter-diurnal bands are considered. Since the last major model comparison project in 1997, models have improved markedly, especially in shallow-water regions and also in the deep ocean. The root-sum-square differences between tide observations and the best models for eight major constituents are approximately 0.9, 5.0, and 6.5 cm for pelagic, shelf, and coastal conditions, respectively. Large intermodel discrepancies occur in high latitudes, but testing in those regions is impeded by the paucity of high-quality in situ tide records. Long-wavelength components of models tested by analyzing satellite laser ranging measurements suggest that several models are comparably accurate for use in precise orbit determination, but analyses of GRACE intersatellite ranging data show that all models are still imperfect on basin and subbasin scales, especially near Antarctica. For the M2 constituent, errors in purely hydrodynamic models are now almost comparable to the 1980-era Schwiderski empirical solution, indicating marked advancement in dynamical modeling. Assessing model accuracy using tidal currents remains problematic owing to uncertainties in in situ current meter estimates and the inability to isolate the barotropic mode. Velocity tests against both acoustic tomography and current meters do confirm that assimilative models perform better than purely hydrodynamic models.

Global heat and salt transports by eddy movement

Abstract: Oceanic mesoscale eddies contribute important horizontal heat and salt transports on a global scale. Here we show that eddy transports are mainly due to individual eddy movements. Theoretical and observational analyses indicate that cyclonic and anticyclonic eddies move westwards, and they also move polewards and equatorwards, respectively, owing to the β of Earth's rotation. Temperature and salinity (T/S) anomalies inside individual eddies tend to move with eddies because of advective trapping of interior water parcels, so eddy movement causes heat and salt transports. Satellite altimeter sea surface height anomaly data are used to track individual eddies, and vertical profiles from co-located Argo floats are used to calculate T/S anomalies. The estimated meridional heat transport by eddy movement is similar in magnitude and spatial structure to previously published eddy covariance estimates from models, and the eddy heat and salt transports both are a sizeable fraction of their respective total transports.

Steganalysis of least significant bit matching using multi-order differences

Abstract: This paper presents a learning-based steganalysis/detection method to attack spatial domain least significant bit LSB matching steganography in grayscale images, which is the antetype of many sophisticated steganographic methods. We model the message embedded by LSB matching as the independent noise to the image, and theoretically prove that LSB matching smoothes the histogram of multi-order differences. Because of the dependency among neighboring pixels, histogram of low order differences can be approximated by Laplace distribution. The smoothness caused by LSB matching is especially apparent at the peak of the histogram. Consequently, the low order differences of image pixels are calculated. The co-occurrence matrix is utilized to model the differences with the small absolute value in order to extract features. Finally, support vector machine classifiers are trained with the features so as to identify a test image either an original or a stego image. The proposed method is evaluated by LSB matching and its improved version "Hugo". In addition, the proposed method is compared with state-of-the-art steganalytic methods. The experimental results demonstrate the reliability of the new detector. Copyright © 2013 John Wiley & Sons, Ltd.

Feature Learning for Image Classification Via Multiobjective Genetic Programming

Abstract: Feature extraction is the first and most critical step in image classification. Most existing image classification methods use hand-crafted features, which are not adaptive for different image domains. In this paper, we develop an evolutionary learning methodology to automatically generate domain-adaptive global feature descriptors for image classification using multiobjective genetic programming (MOGP). In our architecture, a set of primitive 2-D operators are randomly combined to construct feature descriptors through the MOGP evolving and then evaluated by two objective fitness criteria, i.e., the classification error and the tree complexity. After the entire evolution procedure finishes, the best-so-far solution selected by the MOGP is regarded as the (near-)optimal feature descriptor obtained. To evaluate its performance, the proposed approach is systematically tested on the Caltech-101, the MIT urban and nature scene, the CMU PIE, and Jochen Triesch Static Hand Posture II data sets, respectively. Experimental results verify that our method significantly outperforms many state-of-the-art hand-designed features and two feature learning techniques in terms of classification accuracy.

Projected Changes in Temperature and Precipitation Extremes in China by the CMIP5 Multimodel Ensembles

Abstract: This paper presents projected changes in temperature and precipitation extremes in China by the end of the twenty-first century based on the Coupled Model Intercomparison Project phase 5 (CMIP5) simulations. The temporal changes and their spatial patterns in the Expert Team on Climate Change Detection and Indices (ETCCDI) indices under the RCP4.5 and RCP8.5 emission scenarios are analyzed. Compared to the reference period 1986–2005, substantial changes are projected in temperature and precipitation extremes under both emission scenarios. These changes include a decrease in cold extremes, an increase in warm extremes, and an intensification of precipitation extremes. The intermodel spread in the projection increases with time, with wider spread under RCP8.5 than RCP4.5 for most indices, especially at the subregional scale. The difference in the projected changes under the two RCPs begins to emerge in the 2040s. Analyses based on the mixed-effects analysis of variance (ANOVA) model indicate that by ...

An enhanced fall detection system for elderly person monitoring using consumer home networks

Abstract: Various fall-detection solutions have been previously proposed to create a reliable surveillance system for elderly people with high requirements on accuracy, sensitivity and specificity. In this paper, an enhanced fall detection system is proposed for elderly person monitoring that is based on smart sensors worn on the body and operating through consumer home networks. With treble thresholds, accidental falls can be detected in the home healthcare environment. By utilizing information gathered from an accelerometer, cardiotachometer and smart sensors, the impacts of falls can be logged and distinguished from normal daily activities. The proposed system has been deployed in a prototype system as detailed in this paper. From a test group of 30 healthy participants, it was found that the proposed fall detection system can achieve a high detection accuracy of 97.5%, while the sensitivity and specificity are 96.8% and 98.1% respectively. Therefore, this system can reliably be developed and deployed into a consumer product for use as an elderly person monitoring device with high accuracy and a low false positive rate.

Spatio-Temporal Laplacian Pyramid Coding for Action Recognition

Abstract: We present a novel descriptor, called spatio-temporal Laplacian pyramid coding (STLPC), for holistic representation of human actions. In contrast to sparse representations based on detected local interest points, STLPC regards a video sequence as a whole with spatio-temporal features directly extracted from it, which prevents the loss of information in sparse representations. Through decomposing each sequence into a set of band-pass-filtered components, the proposed pyramid model localizes features residing at different scales, and therefore is able to effectively encode the motion information of actions. To make features further invariant and resistant to distortions as well as noise, a bank of 3-D Gabor filters is applied to each level of the Laplacian pyramid, followed by max pooling within filter bands and over spatio-temporal neighborhoods. Since the convolving and pooling are performed spatio-temporally, the coding model can capture structural and motion information simultaneously and provide an informative representation of actions. The proposed method achieves superb recognition rates on the KTH, the multiview IXMAS, the challenging UCF Sports, and the newly released HMDB51 datasets. It outperforms state of the art methods showing its great potential on action recognition.

Forecasting wind speed using empirical mode decomposition and Elman neural network

Abstract: Because of the chaotic nature and intrinsic complexity of wind speed, it is difficult to describe the moving tendency of wind speed and accurately forecast it. In our study, a novel EMD-ENN approach, a hybrid of empirical mode decomposition (EMD) and Elman neural network (ENN), is proposed to forecast wind speed. First, the original wind speed datasets are decomposed into a collection of intrinsic mode functions (IMFs) and a residue by EMD, yielding relatively stationary sub-series that can be readily modeled by neural networks. Second, both IMF components and residue are applied to establish the corresponding ENN models. Then, each sub-series is predicted using the corresponding ENN. Finally, the prediction values of the original wind speed datasets are calculated by the sum of the forecasting values of every sub-series. Moreover, in the ENN modeling process, the neuron number of the input layer is determined by a partial autocorrelation function. Four prediction cases of wind speed are used to test the performance of the proposed hybrid approach. Compared with the persistent model, back-propagation neural network, and ENN, the simulation results show that the proposed EMD-ENN model consistently has the minimum statistical error of the mean absolute error, mean square error, and mean absolute percentage error. Thus, it is concluded that the proposed approach is suitable for wind speed prediction.

Weakly-Supervised Cross-Domain Dictionary Learning for Visual Recognition

Abstract: We address the visual categorization problem and present a method that utilizes weakly labeled data from other visual domains as the auxiliary source data for enhancing the original learning system. The proposed method aims to expand the intra-class diversity of original training data through the collaboration with the source data. In order to bring the original target domain data and the auxiliary source domain data into the same feature space, we introduce a weakly-supervised cross-domain dictionary learning method, which learns a reconstructive, discriminative and domain-adaptive dictionary pair and the corresponding classifier parameters without using any prior information. Such a method operates at a high level, and it can be applied to different cross-domain applications. To build up the auxiliary domain data, we manually collect images from Web pages, and select human actions of specific categories from a different dataset. The proposed method is evaluated for human action recognition, image classification and event recognition tasks on the UCF YouTube dataset, the Caltech101/256 datasets and the Kodak dataset, respectively, achieving outstanding results.

Response of inland lake dynamics over the Tibetan Plateau to climate change

Abstract: The water balance of inland lakes on the Tibetan Plateau (TP) involves complex hydrological processes; their dynamics over recent decades is a good indicator of changes in water cycle under rapid global warming. Based on satellite images and extensive field investigations, we demonstrate that a coherent lake growth on the TP interior (TPI) has occurred since the late 1990s in response to a significant global climate change. Closed lakes on the TPI varied heterogeneously during 1976-1999, but expanded coherently and signifi- cantly in both lake area and water depth during 1999-2010. Although the decreased potential evaporation and glacier mass loss may contribute to the lake growth since the late 1990s, the significant water surplus is mainly attributed to increased regional precipitation, which, in turn, may be related to changes in large-scale atmospheric circulation, including the intensified Northern Hemisphere summer monsoon (NHSM) circulation and the poleward shift of the Eastern Asian westerlies jet stream.

Unraveling the cause-effect relation between time series.

Abstract: Given two time series, can one faithfully tell, in a rigorous and quantitative way, the cause and effect between them? Based on a recently rigorized physical notion, namely, information flow, we solve an inverse problem and give this important and challenging question, which is of interest in a wide variety of disciplines, a positive answer. Here causality is measured by the time rate of information flowing from one series to the other. The resulting formula is tight in form, involving only commonly used statistics, namely, sample covariances; an immediate corollary is that causation implies correlation, but correlation does not imply causation. It has been validated with touchstone linear and nonlinear series, purportedly generated with one-way causality that evades the traditional approaches. It has also been applied successfully to the investigation of real-world problems; an example presented here is the cause-and-effect relation between the two climate modes, El Nino and the Indian Ocean Dipole (IOD), which have been linked to hazards in far-flung regions of the globe. In general, the two modes are mutually causal, but the causality is asymmetric: El Nino tends to stabilize IOD, while IOD functions to make El Nino more uncertain. To El Nino, the information flowing from IOD manifests itself as a propagation of uncertainty from the Indian Ocean.

State of the Climate in 2013

Abstract: Editors note: For easy download the posted pdf of the State of the Climate for 2013 is a very low-resolution file. A high-resolution copy of the report is available by clicking here. Please be patient as it may take a few minutes for the high-resolution file to download.

Satellite evidence for significant biophysical consequences of the “Grain for Green” Program on the Loess Plateau in China

Abstract: Afforestation has been implemented worldwide as regional and national policies to address environmental problems and to improve ecosystem services. China's central government launched the “Grain for Green” Program (GGP) in 1999 to increase forest cover and to control soil erosion by converting agricultural lands on steep slopes to forests and grasslands. Here a variety of satellite data products from the Moderate Resolution Imaging Spectroradiometer were used to assess the biophysical consequences of the GGP for the Loess Plateau, the pilot region of the program. The average tree cover of the plateau substantially increased because of the GGP, with a relative increase of 41.0%. The GGP led to significant increases in enhanced vegetation index (EVI), leaf area index, and the fraction of photosynthetically active radiation absorbed by canopies. The increase in forest productivity as approximated by EVI was not driven by elevated air temperature, changing precipitation, or rising atmospheric carbon dioxide concentrations. Moreover, the afforestation significantly reduced surface albedo, leading to a positive radiative forcing and a warming effect on the climate. The GGP also led to a significant decline in daytime land surface temperature and exerted a cooling effect on the climate. The GGP therefore has significant biophysical consequences by altering carbon cycling, hydrologic processes, and surface energy exchange and has significant feedbacks to the regional climate. The net radiative forcing on the climate depends on the offsetting of the negative forcing from carbon sequestration and higher evapotranspiration and the positive forcing from lower albedo.

A climatology of aerosol optical depth over China from recent 10 years of MODIS remote sensing data

Abstract: Using the recent 10-year (March 2000 to February 2010) MODIS data of aerosol optical depth (AOD), the distributions of annual and seasonal mean AOD over China are presented, and the trends and seasonal variations in AOD over 10 regions in China are analysed. The spatial pattern of annual mean AOD is characterized generally with two low centres and two high centres over China. Two low AOD centres are located in the areas with a high vegetation cover and a sparse population in (1) the high-latitude region in Northeast China with AOD of about 0.2 and (2) the high-altitude region in Southwest China with AOD from 0.1 to 0.2. These two low AOD centres are connected by a low AOD zone (0.2–0.3) in a northeast–southwest direction across China. Demarcated by this low AOD zone, two high centres with AOD of about 0.8 are situated in (1) the most densely populated and industrialized regions in China with high anthropogenic aerosols from North China Plain, Yangtze River areas covering Sichuan Basin, Central China and Yangtze River Delta to South China with Pearl River Delta region and (2) Tarim Basin in Northwest China with high natural aerosols dominated with desert dust. The spatial AOD patterns over China keep seasonally unchanged, but the strengths of the AOD-centres vary from season to season. The wintertime AOD is lowest in China. The monthly AOD variations from March to September in Southern China correspond with high AOD before, after the rain periods and low AOD during the rain periods of Asian summer monsoon. Asian summer monsoons also make a notable impact on the seasonal cycle of aerosols in China. The AOD in Northern China changes monthly with a single peak between April and June and a low in the winter months. The positive trends in AOD occur mostly in the aerosol source regions with higher annual mean AOD (>0.25), while the negative trends are found in the regions with lower annual mean AOD (<0.25) over China.

Multistability in the Lorenz System: A Broken Butterfly

Abstract: In this paper, the dynamical behavior of the Lorenz system is examined in a previously unexplored region of parameter space, in particular, where r is zero and b is negative. For certain values of the parameters, the classic butterfly attractor is broken into a symmetric pair of strange attractors, or it shrinks into a small attractor basin intermingled with the basins of a symmetric pair of limit cycles, which means that the system is bistable or tristable under certain conditions. Although the resulting system is no longer a plausible model of fluid convection, it may have application to other physical systems.

Increasing autumn drought over southern China associated with ENSO regime shift

Abstract: In the two most recent decades, more frequent drought struck southern China during autumn, causing an unprecedented water crisis. We found that the increasing autumn drought is largely attributed to an ENSO regime shift. Compared to traditional eastern-Pacific (EP) El Nino, central-Pacific (CP) El Nino events have occurred more frequently, with maximum sea surface temperature anomalies located near the dateline. Southern China usually experiences precipitation surplus during the autumn of EP El Nino years, while the CP El Nino tends to produce precipitation deficits. Since the CP El Nino has occurred more frequently while EP El Nino has become less common after the early 1990s, there has been a significant increase in the frequency of autumn drought. This has implications for increasing precipitation shortages over southern China in a warming world, in which CP El Nino events have been suggested to become more common.

Learning deep and wide: a spectral method for learning deep networks.

Abstract: Building intelligent systems that are capable of extracting high-level representations from high-dimensional sensory data lies at the core of solving many computer vision-related tasks. We propose the multispectral neural networks (MSNN) to learn features from multicolumn deep neural networks and embed the penultimate hierarchical discriminative manifolds into a compact representation. The low-dimensional embedding explores the complementary property of different views wherein the distribution of each view is sufficiently smooth and hence achieves robustness, given few labeled training data. Our experiments show that spectrally embedding several deep neural networks can explore the optimum output from the multicolumn networks and consistently decrease the error rate compared with a single deep network.