Disentangling the multi-scale effects of sea-surface temperatures on global precipitation: A coupled networks approach

Correlation change analysis and NDVI prediction in the Yellow River Basin of China using complex networks and GRNN-PSRLSTM

The Normalized Difference Vegetation Index (NDVI) is affected by various environmental factors, and its relationship with these factors is complex. In order to explore the complex relationship between NDVI and environmental factors, this paper adopts the complex network method to construct a correlation fluctuation network and analyze the interaction between them. It is found that temperature, precipitation, soil moisture, sunshine duration, and PM2.5 are all correlated with NDVI to varying degrees, and their combined correlation with NDVI varies over time. The correlation typically takes 3 to 6 months to change, and it tends to persist to some extent. Moreover, we fuse a generalized regression neural network (GRNN) with a long-short-term memory (LSTM) network combining phase space reconstruction (PSR) to propose a GRNN-PSRLSTM prediction model. The model achieves the prediction of monthly NDVI using the five environmental factors of the fluctuation network. The results indicate that the averages of root mean squared error (RMSE) and mean absolute percentage error (MAPE) predicted by the GRNN-PSRLSTM model in the nine provinces are 0.0232 and 0.0564 respectively. This model performs better in the assessment metrics for monthly NDVI forecasts. These findings are significant for evaluating vegetation changes and have some theoretical value for the ecological protection of the Yellow River Basin.

Study of interaction and complete merging of binary cyclones using complex networks

Cyclones are among the most hazardous extreme weather events on Earth. In certain scenarios, two co-rotating cyclones in close proximity to one another can drift closer and completely merge into a single cyclonic system. Identifying the dynamic transitions during such an interaction period of binary cyclones and predicting the complete merger (CM) event are challenging for weather forecasters. In this work, we suggest an innovative approach to understand the evolving vortical interactions between the cyclones during two such CM events (Noru-Kulap and Seroja-Odette) using time-evolving induced velocity-based unweighted directed networks. We find that network-based indicators, namely, in-degree and out-degree, quantify the changes in the interaction between the two cyclones and are excellent candidates to classify the interaction stages before a CM. The network indicators also help to identify the dominant cyclone during the period of interaction and quantify the variation of the strength of the dominating and merged cyclones. Finally, we show that the network measures also provide an early indication of the CM event well before its occurrence.

Multifractional Brownian motion characterization based on Hurst exponent estimation and statistical learning

This paper proposes an approach for the estimation of a time-varying Hurst exponent to allow accurate identification of multifractional Brownian motion (MFBM). The contribution provides a prescription for how to deal with the MFBM measurement data to solve regression and classification problems. Theoretical studies are supplemented with computer simulations and real-world examples. Those prove that the procedure proposed in this paper outperforms the best-in-class algorithm.

Eigen microstates and their evolution of global ozone at different geopotential heights

Studies on stratospheric ozone have attracted much attention due to its serious impacts on climate changes and its important role as a tracer of Earth's global circulation. Tropospheric ozone as a main atmospheric pollutant damages human health as well as the growth of vegetation. Yet, there is still a lack of a theoretical framework to fully describe the variation of ozone. To understand ozone's spatiotemporal variance, we introduce the eigen microstate method to analyze the global ozone mass mixing ratio between January 1, 1979 and June 30, 2020 at 37 pressure layers. We find that eigen microstates at different geopotential heights can capture different climate phenomena and modes. Without deseasonalization, the first eigen microstates capture the seasonal effect and reveal that the phase of the intra-annual cycle moves with the geopotential heights. After deseasonalization, by contrast, the collective patterns from the overall trend, El Niño-Southern Oscillation (ENSO), quasi-biennial oscillation, and tropopause pressure are identified by the first few significant eigen microstates. The theoretical framework proposed here can also be applied to other complex Earth systems.

Relationship between mutual information and cross-correlation time scale of observability as measures of connectivity strength

The task of identifying and characterizing network structures out of experimentally observed time series is tackled by implementing different solutions, ranging from entropy-based techniques to the evaluation of the significance of observed correlation estimators. Among the metrics that belong to the first class, mutual information is of major importance due to the relative simplicity of implementation and its relying on the crucial concept of entropy. With regard to the second class, a method that allows us to assess the connectivity strength of a link in terms of a time scale of its observability via the significance estimate of measured cross correlation was recently shown to provide a reliable tool to study network structures. In this paper, we investigate the relationship between this last metric and mutual information by simultaneously assessing both metrics on large sets of data extracted from three experimental contexts, human brain magnetoencephalography, human brain electroencephalography, and surface wind measurements carried out on a small regional scale, as well as on simulated coupled, auto-regressive processes. We show that the relationship is well described by a power law and provide a theoretical explanation based on a simple noise and signal model. Besides further upholding the reliability of cross-correlation time scale of observability, the results show that the combined use of this metric and mutual information can be used as a valuable tool to identify and characterize connectivity links in a wide range of experimental contexts.

Complex networks and deep learning for EEG signal analysis

Electroencephalogram (EEG) signals acquired from brain can provide an effective representation of the human's physiological and pathological states. Up to now, much work has been conducted to study and analyze the EEG signals, aiming at spying the current states or the evolution characteristics of the complex brain system. Considering the complex interactions between different structural and functional brain regions, brain network has received a lot of attention and has made great progress in brain mechanism research. In addition, characterized by autonomous, multi-layer and diversified feature extraction, deep learning has provided an effective and feasible solution for solving complex classification problems in many fields, including brain state research. Both of them show strong ability in EEG signal analysis, but the combination of these two theories to solve the difficult classification problems based on EEG signals is still in its infancy. We here review the application of these two theories in EEG signal research, mainly involving brain-computer interface, neurological disorders and cognitive analysis. Furthermore, we also develop a framework combining recurrence plots and convolutional neural network to achieve fatigue driving recognition. The results demonstrate that complex networks and deep learning can effectively implement functional complementarity for better feature extraction and classification, especially in EEG signal analysis.

Wavelet entropy-based evaluation of intrinsic predictability of time series

Intrinsic predictability is imperative to quantify inherent information contained in a time series and assists in evaluating the performance of different forecasting methods to get the best possible prediction. Model forecasting performance is the measure of the probability of success. Nevertheless, model performance or the model does not provide understanding for improvement in prediction. Intuitively, intrinsic predictability delivers the highest level of predictability for a time series and informative in unfolding whether the system is unpredictable or the chosen model is a poor choice. We introduce a novel measure, the Wavelet Entropy Energy Measure (WEEM), based on wavelet transformation and information entropy for quantification of intrinsic predictability of time series. To investigate the efficiency and reliability of the proposed measure, model forecast performance was evaluated via a wavelet networks approach. The proposed measure uses the wavelet energy distribution of a time series at different scales and compares it with the wavelet energy distribution of white noise to quantify a time series as deterministic or random. We test the WEEM using a wide variety of time series ranging from deterministic, non-stationary, and ones contaminated with white noise with different noise-signal ratios. Furthermore, a relationship is developed between the WEEM and Nash-Sutcliffe Efficiency, one of the widely known measures of forecast performance. The reliability of WEEM is demonstrated by exploring the relationship to logistic map and real-world data.

Generation of surrogate event sequences via joint distribution of successive inter-event intervals

The study of many dynamical systems relies on the analysis of experimentally-recorded sequences of events for which information is encoded in the sequence of interevent intervals. A correct interpretation of the results of the application of analytical techniques to these sequences requires the assessment of statistical significance. In most cases, the corresponding null-hypothesis distribution is unknown, thus forbidding an evaluation of the significance. An alternative solution, which is efficient in the case of continuous signals, is provided by the generation of surrogate data that share statistical and spectral properties with the original dataset. However, in the case of event sequences, the available algorithms for the generation of surrogate data can become cumbersome and computationally demanding. In this work, we present a new method for the generation of surrogate event sequences that relies on the joint distribution of successive interevent intervals. Our method, which was tested on both synthetic and experimental sequences, performs equally well or even better than conventional methods in terms of interevent interval distribution and autocorrelation while abating the computational time by at least one order of magnitude.