Multi-step forecasting of chlorophyll-a concentration in coastal waters through Wavelet Dense Attention Transformer model.

Accurate estimation of chlorophyll-a concentration in turbid coastal waters by means of remote sensing is challenging due to the optical complexity of these waters. We have developed a four-band quasi-analytical algorithm for assessment of chlorophyll-a concentration in coastal waters. The objectives of this study are to validate the applicability of three-band semi-analytical algorithm, quasi-analytical algorithm, and four-band quasi-analytical algorithm in estimating chlorophyll-a concentration in turbid coastal waters for MODIS sensor. These three algorithms are calibrated and evaluated against coastal evaluation datasets provided by SeaWiFS Bio-optical Archive and Storage System. The algorithm validation results indicate that the four-band quasi-analytical algorithm produces a superior performance to both three-band semi-analytical algorithm and quasi-analytical algorithm. By comparison, using four-band quasi-analytical algorithm produces 21.61 % uncertainty in estimating chlorophyll-a concentration from turbid coastal waters, lower than 77.90 % for three-band semi-analytical algorithm and 74.31 % for quasi-analytical algorithm, respectively. The significantly reduced uncertainty in chlorophyll-a concentration assessment is due to effectively removal of pigment-package effects and particle overlapping effects on the chlorophyll-a absorption estimation using a optical classification method. These findings imply that, provided that an atmospheric correction scheme for visible and near-infrared bands is available, the database of MODIS imagery could be used for quantitative monitoring of chlorophyll-a concentration in turbid coastal waters by four-band quasi-analytical algorithm.

MODIS ocean color product downscaling via spatio-temporal fusion and regression: The case of chlorophyll-a in coastal waters

With the development of quantitative ocean color remote sensing, estimation of chlorophyll-a concentration in the coastal waters has aroused increasing attention from researchers. Currently, researches are confronted with difficulty in improving the accuracy of chlorophyll-a concentration estimation for turbid waters. Atmospheric correction, chlorophyll-a concentration modeling, and scale effect have already been identified as three critical factors affecting coastal water remote sensing. The in-depth exploration of them will accelerate the research progress of ocean color remote sensing. The ultimate objective of atmospheric correction and scale effect correction is to accurately estimate active constituents of turbid coastal waters in an optical way. Accordingly, the chlorophyll-a concentration modeling is a basic problem to be resolved, while atmospheric correction is the essential one. The scale effect problem arises during the modeling procedure where unrealistic homogeneous assumption is taken to measure chlorophyll-a concentration from the realistic non-homogeneous pixel. In the coastal remote sensing field, these three problems have become the most important topics in the current researches, and they will remain be the hot topics in the future.

Accurate assessment of phytoplankton chlorophyll-a (chl-a) concentration in turbid waters by means of remote sensing is challenging because of the optical complexity of case 2 waters. We applied a bio-optical model of the form [R–1(λ1) – R–1(λ2)](λ3), where R(λi) is the remote-sensing reflectance at wavelength λi, to estimate chl-a concentration in coastal waters. The objectives of this article are (1) to validate the three-band bio-optical model using a data set collected in coastal waters, (2) to evaluate the extent to which the three-band bio-optical model could be applied to the spectral radiometer (SR) ISI921VF-512T data and the hyperspectral imager (HSI) data on board the Chinese HJ-1A satellite, (3) to evaluate the application prospects of HJ-1A HSI data in case 2 waters chl-a concentration mapping. The three-band model was calibrated using three SR spectral bands (λ1 = 664.9 nm, λ2 = 706.54 nm, and λ3 = 737.33 nm) and three HJ-1A HSI spectral bands (λ1 = 637.725 nm, λ2 = 711.495 nm, and λ3 = 753.750 nm). We assessed the accuracy of chl-a prediction with 21 in situ sample plots. Chl-a predicted by SR data was strongly correlated with observed chl-a (R2 = 0.93, root mean square error (RMSE) = 0.48 mg m–3, coefficient of variation (CV) (RMSE/mean(chl-amea)) = 3.72%). Chl-a predicted by HJ-1A HSI data was also closely correlated with observed chl-a (R2 = 0.78, RMSE = 0.45 mg m–3, CV (RMSE/mean(chl-amea)) = 7.51%). These findings demonstrate that the HJ-1A HSI data are promising for quantitative monitoring of chl-a in coastal case-2 waters.

Keeping Tabs on HABs: new tools for detecting, monitoring, and preventing harmful algal blooms.

Recently proposed ordinal networks not only afford novel methods of nonlinear time series analysis but also constitute stochastic approximations of the deterministic flow time series from which the network models are constructed. In this paper, we construct ordinal networks from discrete sampled continuous chaotic time series and then regenerate new time series by taking random walks on the ordinal network. We then investigate the extent to which the dynamics of the original time series are encoded in the ordinal networks and retained through the process of regenerating new time series by using several distinct quantitative approaches. First, we use recurrence quantification analysis on traditional recurrence plots and order recurrence plots to compare the temporal structure of the original time series with random walk surrogate time series. Second, we estimate the largest Lyapunov exponent from the original time series and investigate the extent to which this invariant measure can be estimated from the surrogate time series. Finally, estimates of correlation dimension are computed to compare the topological properties of the original and surrogate time series dynamics. Our findings show that ordinal networks constructed from univariate time series data constitute stochastic models which approximate important dynamical properties of the original systems.

The stock market is a chaotic, complex, and dynamic financial market. The prediction of future stock prices is a concern and controversial research issue for researchers. More and more analysis and prediction methods are proposed by researchers. We proposed a hybrid method for the prediction of future stock prices using LSTM and ensemble EMD in this paper. We use comprehensive EMD to decompose the complex original stock price time series into several subsequences which are smoother, more regular and stable than the original time series. Then, we use the LSTM method to train and predict each subsequence. Finally, we obtained the prediction values of the original stock price time series by fused the prediction values of several subsequences. In the experiment, we selected five data to fully test the performance of the method. The comparison results with the other four prediction methods show that the predicted values show higher accuracy. The hybrid prediction method we proposed is effective and accurate in future stock price prediction. Hence, the hybrid prediction method has practical application and reference value.

Time series forecasting based on kernel mapping and high-order fuzzy cognitive maps

Time series forecasting is one of the most important issues in numerous applications in real life. The objective of this study was to propose a hybrid neural network model based on wavelet transform (WT) and feature extraction for time series forecasting. The motivation of the proposed model, which is called PCA-WCCNN, is to establish a single simplified model with shorter training time and satisfactory forecasting performance. This model combines the principal component analysis (PCA) and WT with artificial neural networks (ANNs). Given a forecasting sequence, order of the original forecasting model is determined firstly. Secondly, the original time series is decomposed into approximation and detail components by employing WT technique. Then, instead of using all the components as inputs, feature inputs are extracted from all the sub-series obtained from the above step. Finally, based on the extracted features and all the sub-series, a famous neural network construction method called cascade-correlation algorithm is applied to train neural network model to learn the dynamics. As an illustration, the proposed model is compared with two classical models and two hybrid models, respectively. They are the traditional cascade-correlation neural network, back-propagation neural network, wavelet-based cascade-correlation network using all the wavelet components as inputs to establish one model (WCCNN) and wavelet-based cascade-correlation network with combination of each sub-series model (WCCNN multi-models). Results obtained from this study indicate that the proposed method improves the accuracy of ANN and can yield better efficiency than other four neural network models.

Chlorophyll-a is the pigment presented in living plants or phytoplankton responsible for the photosynthesis, and it is a very important ecological and environmental parameter of waters, not only used for estimation of ocean primary productivity, but also for detection of red tides and for water quality. The chlorophyll-a concentration in coastal waters is generally overestimated from ocean color satellite data with common algorithm. In order to improve the remotely-sensed estimation of chlorophyll-a, many efforts have been made. In this study, six cruises for in situ data collection were conducted in lower reaches of the Pearl River and its estuary, the Lingdingyang, from 2003 to 2006. And the variables such as temperature, salinity, chlorophyll-a, total suspended matters (TSM), nutrients and gelbstoff absorption coefficient (Ag) were collected. The in situ remote sensing reflectance (Rrs) were measured from above waters with an ocean optics USB2000 spectrometer, which covers wavelength from 200 to 850 nm with spectral resolution of 0.38 nm. The original in situ Rrs data were processed to EO-1/hyperion bands (at resolution about 10 nm), and the derivative spectra with the different spectral resolutions were analyzed. The statistic analysis shows that the relative coefficient between derivative spectrum and chlorophyll-a is higher than that between original spectrum and chlorophyll-a. The band with highest relative coefficient to chlorophyll-a was employed for development of chlorophyll-a retrieval algorithm. A derivative spectrum algorithm was developed, and then applied to EO-1/hyperion data, which were acquired in December 16, 2006. The distribution image maps of chlorophyll-a concentration retrieved from EO-1/hyperion data were obtained. It shows that the derivative spectral method is an alternative approach for detection of water quality in coastal waters.

In various industrial applications of bubbling fluidized beds, defluidizing parts of the bed or even of the complete bed can occur as a result of agglomeration and sintering of particles due to unintentional maloperation of the bed, changes in operating conditions, or variations in gas or solids feed. Defluidization may be prevented by increasing the gas velocity or changing the solids feed, if the change in the “quality” of the fluidized state of particles is detected early enough. An analysis method is proposed that uses the short‐term predictability of time series of (local) pressure fluctuations in the fluidized bed to detect changes early in the hydrodynamic state of the bed. It is based on the comparison of an original time series of pressure fluctuations with successive time series measured during operation of a fluidized bed. The comparison is based on a discriminating statistic computed for the original time series as well as for each successive series. The null hypothesis that the original and successive time series are similar is rejected if the discriminating statistics of both time series significantly differ. Experimental application of the method is illustrated for fluidization at elevated temperatures (ca. 120°C) of agglomerating plastic particles in a 5‐cm‐ID laboratory fluidized bed. The method recognizes the change in the hydrodynamics due to the incipient agglomeration of the particles. In this particular case the time period between the moment of detection of a significant change in the hydrodynamics and the end of the experiment when the bed becomes defluidized seems sufficiently large to take preventive measures. The average bed‐pressure drop is not a sensitive quantity to detect changes early in the fluidization behavior.

The surrogate data method was applied in order to test the existence of nonlinear components in the blood pressure (BP) time series, recorded from a one month old conscious normotensive rat (WKY). The surrogate data was generated by producing five new data sets from the original BP time series. The correlation dimension of each of these time series was calculated. Two scaling regions were found in the original time series, whereas, in the surrogate data, one of these two regions was consistently missing. As for the second scaling region, a significant difference was found between the correlation dimension computed for the original time series and that of the different surrogate data sets. The authors' results thus strongly indicate that there are nonlinear components in the BP time series taken from a WKY rat. >

The modeling and predicting of vegetation Leaf area index (LAI) is an important component of land surface model and assimilation of remote sensing data. The MODIS LAI product (i.e. MOD15A2) is one of the most widely used LAI data sources. However, the time series of MODIS LAI contains some data of low quality. For example, because of the influence of the cloud, aerosol, etc., the MODIS LAI presents the characteristics of the discontinuous in time and space. In fact, the time series of MODIS LAI include both linear and nonlinear components, which cannot be accurately modeled and predicted by either linear method or nonlinear method alone. In this paper, the original LAI time series data were first smoothed with Savitzky-Golay (SG) filtration and linear interpolation; SARIMA, BP neural network and a hybrid method of SARIMA-BP neural network were then used for modeling and predicting MODIS LAI time series. The SARIMA-BP neural network combined both SARIMA and BP neural network, which could model the linear and the nonlinear component of MODIS LAI time series respectively. That is, the final result of SARIMA-BP neural network was the sum of results of the two methods. Experiments showed that the time series of MODIS LAI that were smoothed with the SG filtration and linear interpolation were more smooth than original time series, with a determination coefficient up to 0.981, closer to 1 than that of SARIMA (0.941) and BP neural network (0.884); the correlation coefficient between SARIMA-BP neural network and the observation is 0.991, higher than that of between SARIMA (0.971) or BP neural network (0.942) SARIMA and the observation. Thus, it can be concluded that, the proposed SARIMA-BP neural network method can better adapt to the LAI time series, and it outperforms the SARIMA and BP neural network methods.

Marine harmful algal blooms (HAB) have been implicated in marine mammal die-offs; but the relationship between sub-lethal algal toxicity and marine mammal vulnerability to human activities has not been assessed. HAB toxins can result in compromised neurological or muscular systems and we posit these conditions can expose marine mammals to increased likelihood of entanglement in commercial fishing gear or ship strike. To investigate whether HABs and large whale injuries and deaths were associated, we assessed the spatiotemporal co-occurrence of HAB events and large whale mortalities/injuries in U.S. east (from 2000-2021) and west (2007-2021) coastal waters. The number of mortalities/injuries was frequently higher in years with large-scale or severe HABs. We found statistically significant relationships between the occurrence of HABs and whale mortalities/injuries in west coast waters – at least three additional whale deaths/injuries were detected near an active HAB than in areas where a HAB was not reported. This relationship was similarly positive but weaker for east coast waters, a difference that may be attributable to differing oceanographic features, or approaches used in whale data collection, between coasts. Saxitoxin-producing Alexandrium was the most common causative agent on both the east (64.1%) and west (57.8%) coasts; and domoic acid-producing Pseudo-nitzschia was more common along the west (33.3%) than the east coast (8.7%). Algal toxins can be entrained in marine ecosystems, including in whale prey, and can chronically persist in marine mammals. Given many whale deaths/injuries result from fishing gear entanglement and vessel strikes, algal-induced morbidities may diminish whale capacities to detect or avoid fishing gear and approaching vessels. While there was much interannual variability, general increasing trends were observed in both whale death/injury and HAB datasets which may be attributable to increased monitoring or rising ocean temperatures. HAB prediction modeling has become increasingly sophisticated and could be used as a tool to reduce whale mortality by limiting human activities (e.g., curtailing fishing operations) when HABs, whale occurrence, and maritime activities are expected to overlap. Additional systematic data collection is needed to track and model mechanisms underlying relationships between HABs and incidental whale mortality.

More than 1,000 manmade satellites currently orbit our planet.1 Some are near the edge of the Earth’s atmosphere just a few hundred kilometers up. Others are tens of thousands of kilometers above us.2 They aid in communication, navigation, defense, and science. A small number3,4 play a critical and quickly expanding role: monitoring the Earth’s surface and atmosphere to track environmental conditions that are intimately tied to human health. A number of new Earth-observing missions are planned for the next decade, including Sentinel-5 aboard the European Space Agency’s MetOp Second Generation satellites (pictured).48 In the meantime researchers are finding new uses for the satellite ... Researchers and government agencies worldwide already use satellite data to monitor air pollutants, infectious disease epidemics, harmful algal blooms (HABs), climate change, and more. But as current research indicates, that’s only the beginning of what we can do with the technology, broadly referred to as “remote sensing.” In the coming years, new satellites will offer higher-resolution imagery in conjunction with more robust and precise algorithms to process the data they deliver. As a result, researchers expect to dramatically expand their ability to view and understand Earth’s land, water, and air, from its remotest ocean waters to its largest cities. The National Aeronautics and Space Administration (NASA) launched its first satellite in 1958,5 and TIROS-1, the country’s first meteorological satellite, came 2 years later.6 Within a few decades members of the epidemiological and public health communities began actively looking at satellite data, says John Haynes, program manager of the NASA Applied Sciences Health and Air Quality Applications Program. In recent years interest in remote-sensing data has soared, with newer avenues being developed and fine-tuned, including air-quality measurements and vector-borne disease projections. “There’s really been a paradigm shift in the use of remote sensing for public health issues,” Haynes says. “Every year there seems to be more and more interest.” Indeed, by March 2015 NASA will have launched 6 Earth-observing missions in 12 months,7 more than in any year in at least a decade.8 New launches include a “global precipitation observatory” that will make frequent global measurements of rain and snowfall, plus one satellite designed to measure soil moisture and another that will measure how carbon moves through the Earth’s atmosphere, land, and oceans. In addition, the International Space Station will receive three new instruments, one that will observe how winds behave around the world, one that will measure clouds and aerosols (particles suspended in the atmosphere)—two variables that remain difficult to predict in climate-change models—and one that will take global, long-term measurements of key components of the Earth’s atmosphere, including aerosols and ozone.9 The momentum will carry through at least the next 8 or so years, with NASA and other space agencies in Europe and Asia planning to launch new satellites that will provide even higher-resolution snapshots of the Earth. Along with technological and scientific advances, a third development is leading to new and improved applications of satellite data: NASA and the National Oceanic and Atmospheric Administration (NOAA) have made their satellite data available free of charge, Haynes says, while the European Space Agency (ESA) has reduced prices and promised to provide free access to data from its next generation of instruments. “More people use the data, and you get more out of it than when you try to restrict it,” says Raphael Kudela, an oceanographer at the University of California, Santa Cruz, who uses satellite imagery to study HABs. This free sharing of data has been instrumental in his field, allowing researchers at institutions around the world to study HABs from above and to improve systems to track and predict them.