Maximum Entropy Spectral Analysis Research Articles

Maximum entropy spectral analysis: an application to gravitational waves data analysis

The maximum entropy spectral analysis (MESA) method, developed by Burg, offers a powerful tool for spectral estimation of a time-series. It relies on Jaynes’ maximum entropy principle, allowing the spectrum of a stochastic process to be inferred using the coefficients of an autoregressive process AR(p) of order p. A closed-form recursive solution provides estimates for both the autoregressive coefficients and the order p of the process. We provide a ready-to-use implementation of this algorithm in a Python package called memspectrum, characterized through power spectral density (PSD) analysis on synthetic data with known PSD and comparisons of different criteria for stopping the recursion. Additionally, we compare the performance of our implementation with the ubiquitous Welch algorithm, using synthetic data generated from the GW150914 strain spectrum released by the LIGO-Virgo-Kagra collaboration. Our findings indicate that Burg’s method provides PSD estimates with systematically lower variance and bias. This is particularly manifest in the case of a small (O(5000)) number of data points, making Burg’s method most suitable to work in this regime. Since this is close to the typical length of analysed gravitational waves data, improving the estimate of the PSD in this regime leads to more reliable posterior profiles for the system under study. We conclude our investigation by utilising MESA, and its particularly easy parametrisation where the only free parameter is the order p of the AR process, to marginalise over the interferometers noise PSD in conjunction with inferring the parameters of GW150914

A New Insight on the Upwelling along the Atlantic Iberian Coasts and Warm Water Outflow in the Gulf of Cadiz from Multiscale Ultrahigh Resolution Sea Surface Temperature Imagery

The ATLAZUL project is an Interreg effort among 18 partners from Spain and Portugal along the Atlantic Iberian coasts. One of its objectives is the development of new methods and data processing for oceanic information to produce useful products for private and public stakeholders. This study proposes a new insight on the sea surface dynamic of the ATLAZUL area based on almost two years of multiscale high resolution sea surface temperature imagery. The use of techniques such as the Karhunen–Loève transform (Empirical Orthogonal Function) and the Maximum Entropy Spectral Analysis were applied to study long- and short-term features in the sea surface temperature imagery. Mathematical Morphology and the Geometrical Theory of Measure are utilized to compute the Medial Axis Transform and the Hausdorff dimension. The results can be summarized as follows: (i) the tow upwelling areas are identified along the Galician–Portugal coast as indicated in the second and third modes of KLT/EOF analysis, and they are partially affected by wind; (ii) the tow warm water outflows from the Bay of Cádiz to the Gulf of Cádiz are identified as the second and third modes of KLT/EOF analysis, which are also influenced by wind; (iii) the skeletons of the surface signature of the upwelling and of the warmer water outflow, along with their fractal dimensions, indicate a chaotic pattern of spatial distribution and (iv) the harmonic prediction model should be combined with the wind prediction.

Maximum entropy spectral analysis of gamma ray logs for cyclostratigraphic analysis of the Late Albian- Early Turonian Sarvak Formation in the Anaran exploration block, southwestern Zagros, Iran

Maximum entropy spectral analysis of gamma ray logs for cyclostratigraphic analysis of the Late Albian- Early Turonian Sarvak Formation in the Anaran exploration block, southwestern Zagros, Iran

Affordable method for measuring fluorescence using Gaussian distributions and bounded MESE.

We present an accurate and low-cost method for measuring fluorescence in materials. Our method outputs an estimate of the material's Donaldson matrix, which is a commonly used two-dimensional spectral characterization of its fluorescence and reflectance properties. To find the estimate, only a few measurements of the material's reflectance under a few illuminants are needed, which we demonstrate using low-cost optical components. Internally, our algorithm is based on representing each Donaldson matrix with a multivariate Gaussian mixture model and its diagonal with a bounded MESE (maximum entropy spectral estimate). It parametrizes and constrains the estimate in a robust and simple way, allowing the use of gradient-descent optimization. We evaluate our algorithm on a combination of real and synthetic data, and four examples of distinct optical components. We reach significantly lower errors than the current state of the art on the exact same inputs, our estimates do not suffer from artifacts such as oscillations of the spectra, and they are stable and robust.

Application of the Wavelet Transform and INPEFA in Sequence Stratigraphy.

The strata of the Shan 2 Member of the Shanxi Formation of Permian(S2)in the LQ Gas Field in the northeast of the Ordos Basin restrict the study of sand body distribution and lateral distribution. The accuracy of sand body prediction is reduced, which greatly affects the exploration and development of oil and gas. In order to accurately characterize the transverse distribution of sand bodies and improve the prediction accuracy of sand bodies, S2 is divided by the sequence stratigraphy method. Research based on the traditional stratigraphic division method includes seismic, logging, and core data. The logging curve is processed by INPEFA and the wavelet transform to reduce artificial division interference. The range of the S2 section is limited by the seismic interpretation surface. The maximum entropy spectral analysis and the wavelet transform are used for fine division, and S2 is divided into 1 long-term base level cycle, 6 medium-term base level cycles, and 11 short-term base level cycles. The fine stratigraphic division improves the exploration accuracy of oil and gas reservoirs.

A RIC<sub>p</sub> criterion for periodicity analysis of hydrological time series

<p indent="0mm">Identification of periodicity in hydrological time series is important for understanding the complex variability of hydrological processes. Previous studies mainly focused on the comparison, improvement and development of different methods for the identification of periodicity. However, these methods have big subjectivity in the selection of significance levels, which may lead to pseudo-periodic components in the identification results. In this article, a new criterion called RIC_p<italic> </italic>(R for correlation coefficient, IC for information criterion, and subscript p for periodicity) is proposed to accurately identify periodic components in hydrological time series and evaluate their statistical significance. The RIC_p is based on the idea of Akaike information criterion, and it is established by consisting of a fitting accuracy index and an uncertainty measurement. The fitting accuracy index is based on the correlation coefficient, and the uncertainty measurement is based on the form of information entropy. Different synthetic time series, which are composed of multiple periodic components and Pearson-III noise data, are used to verify the superiority of the RIC_p method. The performance of RIC_p is compared with the moving correlation coefficient-based method for periodicity identification (MCCP), harmonic analysis method (HAM), and maximum entropy spectral analysis (MESA), by checking the accuracy and reliability of the identification results. The results suggest that in HAM and MCCP methods, more pseudo-periodic components are identified compared to other methods. Differently, the number of periodic components identified by MESA is less than the true value, as some significant periodic components can be missed. By comparison, the RIC_p performs the best among these four methods. The advantage of RIC_p is that it can describe the periodicity and also evaluate its statistical significance level. The RIC_p method is further applied to analyze the surface water resource time series in six secondary water resource regions in Southwest Rivers basin, by comparing with HAM and MESA. Results further verified the advantages of the RIC_p method developed. It also indicates that the runoff in the basin generally has a 9-year periodic pattern; the precipitation has the same periodicity as the runoff except Lancang River region, which can be attributed to the 11-year periodic cycle of the Indian Ocean monsoon.

EMD-Based Intermittent and Ground Clutter Removal in Radar Windprofilers

An Empirical Mode Decomposition (EMD) based intermittent and ground clutter removal algorithm for Radar Windprofilers is proposed. Intrinsic Mode Functions (IMFs) affected with intermittent clutter are identified and are separated into Amplitude Modulated (AM) and Frequency Modulated (FM) parts. The FM part is multiplied with the envelope of the non-amplitude modulated signal derived from the original amplitude modulated IMF to nullify the effect of intermittent clutter. The IMFs corresponding to ground clutter are identified from the Maximum Entropy Spectral Analysis (MESA) and are rejected prior to signal reconstruction in ground clutter removal. The proposed algorithm is applied on the radar returns of 205 MHz Windprofiler at Advanced Centre for Atmospheric Radar Research (ACARR) of Cochin University of Science and Technology(CUSAT), Cochin, Kerala, India (10.04°N, 76.33°E). The windprofiler signals corrupted with flight and helicopter echoes are retrieved and the ground clutter signatures, which were not removed with existing ground clutter removal methods are removed with the proposed algorithm.

Applicability Study of Hydrological Period Identification Methods: Application to Huayuankou and Lijin in the Yellow River Basin, China

Identifying implicit periodicities in hydrological data is significant for managing river–basin water resources and establishing flood forecasting systems. However, the complexity and randomness of hydrological systems make it difficult to detect hidden oscillatory characteristics. This study discusses the performance and applicability of five period identification methods, namely periodograms, autocorrelation analysis (AA), maximum entropy spectral analysis (MESA), wavelet analysis (WA), and the Hilbert–Huang transform (HHT). The annual and monthly runoff data are sampled from two stations (Huayuankou and Lijin on the Yellow River in China) in the years 1949–2015. The conclusions are as follows: (i) All methods identify the significant periods of 6 months, 12 months, and 18–19 months, which have relatively high energy of peaks; (ii) WA and HHT perform best when dealing with nonstationary time series, but they are ineffective for identifying large-scale periods; (iii) MESA has high resolution and stability but is prone to oscillate at small-scale periods when applied to monthly series; and (iv) periodograms and AA are relatively simple, but their results lack stability and are significantly affected by the data length—the resolution of AA is too low when applied to annual data, and periodograms can easily produce “false peaks”. Generally, it is better to apply multiple methods comprehensively than each method singularly, and this can be effective in reducing subjective influences.

Application of Spectral Extrapolation Technique to Stepped-Frequency RCS Measurement

“Time domain gating” used in the stepped-frequency radar cross section (RCS) measurement causes the inaccurate frequency domain data, especially at two ends of the band. This paper proposes a spectral extrapolation method for improving the measured RCS at two ends of the band more exactly. The core idea is: the measured frequency domain data are extrapolated to obtain the unknown value out of band with an auto-regressive model (AR model). The parameter in the AR model is calculated by the maximum entropy spectral estimation algorithm. Therefore, the span of the original band is extended, and both ends of frequency on the original band are inside the range of the new band. If the time domain gating is adding to the new band, the precision at two ends of the original band can be greatly improved. The simulation and experimental results show that more effective frequency domain data near the two ends of the band can be predicted by using the spectral extrapolation method, and the maximum error at the ends of the original band is less than 1dB after extrapolation, so it can ensure the accuracy of RCS measurement over the whole frequency band.

0323 Associations Between Sleep and Activity Patterns in the Mother-Child Dyad

Abstract Introduction Animal studies show links between light-dark patterns in gestating dams and offsprings’ sleep and circadian rhythms. Activity patterns between mothers and infants show synchrony as early as 12 weeks postpartum. Our goal was to investigate maternal sleep/activity patterns at two points in the perinatal period and assess associations with activity in their young children. Methods Participants were 20 mother-child dyads recruited from previous studies. Mothers (age±SD = 31.7±5.5 years, range 21-40 years) wore wrist actigraphs during their 33rd week of pregnancy and 2nd week postpartum. Children (age±SD = 2.13±1.36 years, range 8 months-4.6 years) were assessed for 5 days and nights. Circadian patterns were analyzed using Maximum Entropy Spectral Analysis (MESA) to estimate best-fitting circadian period, tau. We used cosinor analysis to calculate rhythm amplitude, Midline Statistic of Rhythm (MESOR, representing mean activity), and acrophase (time of peak amplitude). We used circadian quotient (CQ; amplitude÷MESOR) to assess rhythm strength while normalizing for intersubject variation in activity levels. Autocorrelation, or degree to which data is consistent for a particular period, was calculated to analyze regularity of activity patterns. Results Mothers’ activity pattern autocorrelation was significantly correlated at the two time points (r=.530, p=.016), such that women with inconsistent activity patterns in pregnancy also demonstrated more irregularity at postpartum week 2. Child CQ correlated with age, with older children showing greater rhythm strength (r=.530, p=0.016). We observed a moderate correlation between mothers’ CQs during pregnancy and children’s CQs (r=.413,p=.07). In mother-child dyads, longer tau in mothers during pregnancy predicted lower autocorrelation of the child’s rhythm to a 24-hr period (r=-.520, p=.019). Finally, later maternal acrophase at postpartum week 2 was associated with longer tau in children (r=.504, p=.024). Conclusion These data show that associations between mother-child sleep/activity patterns may begin during pregnancy and support the notion that mothers’ perinatal sleep patterns could affect the health of both mothers and their children. Support Supported by R34MH104377, K23MH086689, the Seleni Institute, the Depression and Bipolar Disorder Alternative Treatment Foundation, and a Karen T. Romer Undergraduate Teaching and Research Award from Brown University.

Phased evolution and variation of the South Asian monsoon, and resulting weathering and surface erosion in the Himalaya–Karakoram Mountains, since late Pliocene time using data from Arabian Sea core

AbstractWe investigate the phased evolution and variation of the South Asian monsoon and resulting weathering intensity and physical erosion in the Himalaya–Karakoram Mountains since late Pliocene time (c.3.4 Ma) using a comprehensive approach. Neodymium and strontium isotopic compositions and single-grain zircon U–Pb age spectra reveal the sources of the deposits in the east Arabian Sea, and show a combination of sources from the Himalaya and the Karakoram–Kohistan–Ladakh Mountains, with sediments from the Indian Peninsula such as the Deccan Traps or Craton. We interpret shifts in the sediment sources to have been forced by sea-level changes that correlate with South Asian monsoon rainfall variation since late Pliocene time. We collected 908 samples from the International Ocean Discovery Program Hole U1456A, which was drilled in the east Arabian Sea. Time series of hematite content and grain size of the sediments were examined downcore. We found South Asian monsoon precipitation and weathering intensity experienced three phases from late Pliocene time. Lower monsoon precipitation, with a lower variability and strong weathering intensity, occurred during 3.4–2.4 Ma; an increased and more variable South Asian monsoon rainfall, along with strengthened but fluctuating weathering intensity, occurred at 1.8–1.1 Ma; and a reduced rainfall with lower South Asian monsoon precipitation variability and moderate weathering intensity marked the period 1.1–0.1 Ma. Maximum entropy spectral analysis and wavelet transform show that there were orbital-dominated cycles of periodsc.100 andc.41 ka in these proxy-based time series. We propose that the monsoon, sea level, global temperature and insolation together forced the weathering and erosion in SW Asia.

Maximum entropy spectral analysis of uneven time series in the geosciences

Time series with a non-constant sampling interval (i.e., uneven time series) are ubiquitous in the geosciences. This is due to random sampling, gaps in sampling, missing data, hiatuses, or the transformation between a spatial scale and a temporal scale when, for example, the sedimentation rate is not constant. The preferred approach in the spectral analyses of these uneven sequences are interpolation-free spectral methods, and the Lomb-Scargle periodogram is a popular choice. In the work presented here, the maximum entropy spectral estimator, modified to deal with uneven time series, is proposed as an alternative to the periodogram. The appeal of this approach is that the maximum entropy spectral estimator is a high resolution estimator. The proposed methodology uses the equivalence between the maximum entropy and the autoregressive spectral estimator. The permutation test is used to assess the statistical confidence of the estimated power spectrum and real and simulated time series are used to illustrate the performance of the proposed methodology. This study shows that the maximum entropy spectral estimation of uneven time series avoids the side lobe problem that plagues the Lomb-Scargle periodogram whilst maintaining its high resolution for short time series. The maximum spectral estimator works well in cases where large proportions of the data are randomly missing, when there are gaps in the data, where there is one or more significant hiatus in the process that generates the data, and for time series with random sampling.

Multichannel maximum-entropy method for the Wigner-Ville distribution

The Wigner-Ville distribution is a powerful technique for the time-frequency spectral analysis of nonstationary seismic data. However, the Wigner-Ville distribution suffers from cross-term interference between different wave components in seismic data. To mitigate cross-term interference, we have developed a multichannel maximum-entropy method (MEM) to modify the Wigner-Ville kernel. The method is related to the conventional maximum-entropy spectral analysis (MESA) algorithm because both algorithms use Burg’s reflection coefficients for the calculation of the prediction-error filter (PEF). The MESA algorithm works on the standard autocorrelation sequence, but it does not work for the Wigner-Ville kernel, which is an instantaneous autocorrelation sequence. Our multichannel MEM algorithm uses the PEF to modify any single Wigner-Ville kernel sequence by exploiting multiple Wigner-Ville kernel sequences simultaneously. This multichannel implementation is capable of robustly determining the reflection coefficient and a minimum-phased PEF for the Wigner-Ville kernel sequence. The Wigner-Ville distribution and the multichannel MEM algorithm in conjunction with each other in turn can produce a high-resolution time-frequency spectrum by mitigating the cross-term interferences and suppressing the spurious energy in the spectrum.

A detailed investigation of low latitude tweek atmospherics observed by the WHU ELF/VLF receiver: I. Automatic detection and analysis method

As a dispersive wave mode produced by lightning strokes, tweek atmospherics provide important hints of lower ionospheric (i.e., D-region) electron density. Based on data accumulation from the WHU ELF/VLF receiver system, we develop an automatic detection module in terms of the maximum-entropy-spectral-estimation (MESE) method to identify unambiguous instances of low latitude tweeks. We justify the feasibility of our procedure through a detailed analysis of the data observed at the Suizhou Station (31.57°N, 113.32°E) on 17 February 2016. A total of 3961 tweeks were registered by visual inspection; the automatic detection method captured 4342 tweeks, of which 3361 were correct ones, producing a correctness percentage of 77.4% (= 3361/4342) and a false alarm rate of 22.6% (= 981/4342). A Short-Time Fourier Transformation (STFT) was also applied to trace the power spectral profiles of identified tweeks and to evaluate the tweek propagation distance. It is found that the fitting accuracy of the frequency–time curve and the relative difference of propagation distance between the two methods through the slope and through the intercept can be used to further improve the accuracy of automatic tweek identification. We suggest that our automatic tweek detection and analysis method therefore supplies a valuable means to investigate features of low latitude tweek atmospherics and associated ionospheric parameters comprehensively.

Utilizing Maximum Entropy Spectral Analysis (MESA) to identify Milankovitch cycles in Lower Member of Miocene Zhujiang Formation in north slope of Baiyun Sag, Pearl River Mouth Basin, South China Sea

Abstract Logs in the petroleum boreholes indirectly records the sedimentary cycles in the deep burial formation. In order to extract and understand the periodicity and cyclicity, it is necessary to process the data by digital signal analysis method. Taking the gamma ray (GR) log as the primary material, an identification approach of Milankovitch cycles in boreholes is proposed in this paper, which is based on the Maximum Entropy Spectral Analysis (MESA). The first stage chooses the appropriate windows for calculating the frequency spectral properties in a short section of the data. In each depth window, the second stage generates the two-dimension frequency spectrum utilizing the MESA. At each depth point, the third stage finds the potential Milankovitch cycles in the one-dimension frequency spectrum, in which the average amplitude spectrum peak would be matched to the ratio of Milankovitch period. According to the frequency and wavelength of the maximum amplitude in Milankovitch cycles, the fourth stage estimates the sedimentation rate controlled by cyclical factor. Finally, the Milankovitch cycles in Lower Member of Miocene Zhujiang Formation in north slope of Baiyun Sag, Pearl River Mouth Basin, are identified and the cyclical sedimentation rate is estimated. The results demonstrate that the proposed method is feasible and effective to identify Milankovitch cycles in boreholes, which may contribute to the other geological researches.

Moving correlation coefficient-based method for the identification of periodicities in hydrologic time series

Accurate identification of periodicities is an important issue for understanding the hydrological variability, medium- to long-term hydrologic simulation and forecasting, as well as water security assessment. Many methods have been developed presently for handling the issue, including the simple partial wave method (SPWM), harmonic analysis method (HAM), power spectrum method (PSM) and maximum entropy spectral analysis (MESA). However, the results of periodic identification are not good as expected due to the complexity of hydrological variability and defects of methods. In this article, a new moving correlation coefficient-based method is proposed for the identification of periodicities in hydrological time series. The specific steps of periodicity identification by the proposed methods is described as follows: (1) Generate a simulated periodic series by using the sinusoidal function; (2) use the correlation coefficient as an index to evaluate the fitting accuracy between the simulated periodic series and the original series, which generally includes the periodic components and stochastic components; (3) change the periodic value and the initial phase of the sinusoidal function, and repeat the above steps to calculate the correlation coefficient values; and (4) find the maximum magnitude of the correlation coefficient, and its corresponding simulated periodic series is regarded as the expected periodic component, and its significance is evaluated by doing hypothesis test. Theoretical analysis and formula derivation demonstrate that the index of correlation coefficient can reflect the significant degree of periodic variability of the original hydrological time series. Bigger correlation coefficient value corresponds to more significant periodic variability. Results of Monte-Carlo experiments confirm the advantages of the proposed method compared with other conventional methods such as HAM, PSM and MESA. The former has high identification efficiency and can keep its stability, even if encountering the influences of some advance factors, including the sample length, mean value, coefficient of variation, the semi amplitude, period length and initial phase. By applying the proposed method to analyze the annual precipitation series at 21 meteorological stations in the Lancang River Basin, results indicate that the annual precipitation presents periodic variations at multi-time scales; comparatively, annual precipitation series in the northern and central parts of the basin mainly presents a 2- to 5-year periodic variability, versus the 5- to 10-year periodic variability in the south part.

Using moments to represent bounded signals for spectral rendering

We present a compact and efficient representation of spectra for accurate rendering using more than three dimensions. While tristimulus color spaces are sufficient for color display, a spectral renderer has to simulate light transport per wavelength. Consequently, emission spectra and surface albedos need to be known at each wavelength. It is practical to store dense samples for emission spectra but for albedo textures, the memory requirements of this approach are unreasonable. Prior works that approximate dense spectra from tristimulus data introduce strong errors under illuminants with sharp peaks and in indirect illumination. We represent spectra by an arbitrary number of Fourier coefficients. However, we do not use a common truncated Fourier series because its ringing could lead to albedos below zero or above one. Instead, we present a novel approach for reconstruction of bounded densities based on the theory of moments. The core of our technique is our bounded maximum entropy spectral estimate. It uses an efficient closed form to compute a smooth signal between zero and one that matches the given Fourier coefficients exactly. Still, a ground truth that localizes all of its mass around a few wavelengths can be reconstructed adequately. Therefore, our representation covers the full gamut of valid reflectances. The resulting textures are compact because each coefficient can be stored in 10 bits. For compatibility with existing tristimulus assets, we implement a mapping from tristimulus color spaces to three Fourier coefficients. Using three coefficients, our technique gives state of the art results without some of the drawbacks of related work. With four to eight coefficients, our representation is superior to all existing representations. Our focus is on offline rendering but we also demonstrate that the technique is fast enough for real-time rendering.

On the Prediction of Upwelling Events at the Colombian Caribbean Coasts from Modis-SST Imagery.

The upwelling cores on the Caribbean Colombian coasts are mainly located at the Peninsula de la Guajira and Cabo de la Aguja. We used monthly averaged Moderate Resolution Imaging Spectroradiometer (MODIS) sea surface temperature as the only information to build up a prediction model for the upwelling events. This comprised two steps: (i) the reduction of the complexity by means of the Karhunen–Loève transform and (ii) a prediction model of time series. Two prediction models were considered: (a) a parametric autoregressive-moving average (ARMA) time series from the Box–Jenkins methodology and (b) a harmonic synthesis model. The harmonic synthesis also comprised of two steps: the maximum entropy spectral analysis and a least-squares harmonic analysis on the set of frequencies. The parametric ARMA time series model failed at the time of prediction with a very narrow range, and it was quite difficult to apply. The harmonic synthesis allowed prediction with a horizon of six months with a correlation of about 0.80. The results can be summarized using the time series of the weights of the different oscillation modes, their spatial structures with the nodal lines, and a high confidence model with a horizon of prediction of about four months.

Decomposition-ANN Methods for Long-Term Discharge Prediction Based on Fisher’s Ordered Clustering with MESA

Precise and reliable long-term streamflow prediction contributes to water resources planning and management. Artificial neural network (ANN) have shown its remarkable ability in forecasting non-linear hydrological processes without involvement of complex, dynamic, hydrological and hydro-climatologic physical process in the water shed. To improve its non-stationary responses, decomposition methods are adopted as pre-processing methods in this study including Empirical Mode Decomposition (EMD), Ensemble Empirical Mode Decomposition (EEMD) and Seasonal-Trend decomposition using Loess (STL). The original time sequence is decomposed to several components, which are then taken as the inputs of the ANN model. EMD and EEMD are data- adaptable methods, and thus the number of Intrinsic Mode Functions (IMFs) might differ for different sequences, leading to the discrepancy of the input number for ANN model in training and predicting. Fisher’s ordered clustering is thus used to classify the IMFs into a determined number of classes based on their frequency spectrum resulting from Maximum Entropy Spectral Analysis (MESA). The proposed methodology is applied on four important hydrological stations on the upper stream of the Yellow River and the Yangtze River in China, respectively, to forecast the streamflow of the next whole year with the historical daily data of the past 6 years. The Nash-Sutcliffe efficiencies of the monthly prediction are higher than 0.85 for all of the four cases, and various indicators indicates that the proposed hybrid method of STL-ANN performs better than other compared methods. The highlights of this study lies in that only historical daily streamflow data is used to derive an accurate long-term prediction by data mining based on decomposition technology and mapping relationships between the decomposed components and the original sequence in the future.

Spatio-temporal dynamics of phytoplankton functional groups in the South China Sea and their relative contributions to marine primary production

The sea-surface distribution of four phytoplankton functional groups in the South China Sea (SCS) namely Coccolithophores (Coc), Diatoms (Dia), Cyanobacteria (Cya) and Chlorophytes (Chlo) was studied across spatial and temporal scales. The time-series datasets derived from the National Aeronautic and Space Administration (NASA) Ocean Biogeochemical Model and retrieved via the Giovanni Portal were analysed statistically with well-known methods including Pearson’s correlation coefficient, ordinary least square regression (OLR), Maximum Entropy Spectral Analysis (MESA) and Principal Components Analysis (PCA). The data were analysed for relative abundance, seasonality, cyclicity, long-term trends, spatial variability, inter-relationships, and relative contributions of individual phytoplankton groups to primary production (PP, using Chl-a as a proxy). The results reveal that the numerically dominant phytoplankton group in terms of relative abundance is Cya, comprising about 57% of the planktonic biomass, followed closely by Coc (about 40%), while Dia constitutes almost 3% of the total abundance. Dia was found to have the strongest annual cycle, and hence displayed the highest rate of seasonal variability. Clear spatial segregation patterns have also been uncovered; Coc and Dia are strongly correlated in the Northern SCS (NSCS) while Cya dominate in Southern SCS (SSCS). Coc and Dia display a strong positive correlation with PP, whereas the seasonal relationships between Cya and Chlo with PP are non-linear, and hence appear statistically non-significant. The rise in the trend of Cya and decrease in Coc, since 2004, may be indicative of changes in successional patterns of phytoplankton functional groups in the SCS. This study, therefore, sets the precedence for more robust research to uncover the immediate and remote causes of the observed spatiotemporal changes in marine phytoplankton dynamics and to develop predictive models of marine primary productivity (MPP) and ecological health.