Continuous Series Research Articles

Electricity supply chain hybrid long‐term demand forecasting approach based on deep learning—a case study of basic metals industry

AbstractDemand forecasting is a key parameter to achieve optimal supply chain management at different levels. The basic metals industry is one of the most energy‐intensive industries in the electricity supply chain. The impossibility of large‐scale energy storage, reservation constraints, and its high costs, limitations on electricity transmission lines capacity, real‐time response to high priority strategic demands, and variety of energy rates at different hours and seasons are issues that challenge the electricity supply chain. A hybrid approach is presented in this paper to improve the accuracy of long‐term demand forecasting in the electricity supply chain. The proposed approach uses wavelet decomposition and long short‐term memory (LSTM) neural networks to produce new predictors for demand forecasting in case of defective data. The data used in this study consists of the recorded hourly demand of Espidan Iron Stone (EIS) company and data of Mobarakeh Steel (MS) company in Isfahan Province. The understudy time series includes a lot of interruptions due to non‐production of the factory or power outages and spikes which cause uncertainty that makes the forecasting more difficult in comparison with continuous time series. Multiple machine learning models including decision trees (DTs), boosted and bagged trees (BGTs), support vector regression (SVR) models, extreme learning machines (ELMs), and a LSTM neural network model are employed to evaluate the effectiveness of the proposed approach in this paper. A comparison of the results obtained using the hybrid proposed approach and training machine learning (ML) methods shows a notable reduction in forecasting error.

189 Novel Wireless Battery-Free Fully Passive Flexible Brain/Spine Implant for Improved Recovery of Function Through Epidural Stimulation

INTRODUCTION: Acquired motor impairment from diseases such as stroke, trauma, epilepsy, and cancer remains an important cause of disability. Current treatments provide insufficient brain/spine recovery, with most patients failing to recover adequate motor function with rehabilitation alone. Brain stimulation can lead to enhanced synaptic density and synaptic response after long-term stimulation and enhanced motor function recovery. However, non-invasive options remain inadequate and invasive approaches are associated with unacceptable risks. METHODS: A flexible and transparent silicone elastomer substrate was used through microfabrication processes, with an external antenna near the implant to control the stimulation. In vitro validation was accomplished by wireless stimulation of human-induced pluripotent stem cell derived cardiomyocytes (hiPSC-CMs). In vivo validation was performed by motor cortex stimulation on an anesthetized rat. The corresponding hindlimb motion was captured by video. To separate implant-evoked movement from spontaneous movement, we compared the peak hindlimb velocity during 0.5 sec post-stimulation periods vs. the 0.5 sec pre-stimulation periods, across a continuous series of N = 43 stimulation pulses. RESULTS: Cultured hiPSC-CMs cells demonstrated well-controlled beating behavior under 0.5, 1, and 2Hz wireless stimulation. Statistical analysis further confirmed the change in beating rate corresponded with stimulation frequency (p < 0.001). The recorded hindlimb movement under wireless motor cortex stimulation demonstrated synchronization with the delivered voltage pulse trains. Average peak hindlimb velocity during pre-stimulation periods was 0.8 ± 0.4 mm/sec (± S.D.); during post-stimulation periods, 47 ± 13 mm/sec. A two-sample t-test confirmed the applied wireless power directly evoked movement, (p < 0.001). CONCLUSIONS: We have developed a wireless, battery-less implant with demonstrated efficacy in vitro and in vivo. Efforts at long-term safety evaluation are ongoing in a smaller and larger animal model as a bridge to phase 0/1 clinical trials.

Overcoming gaps in a seasonal time series of Japanese anchovy abundance to analyse interannual trends

Monitoring and understanding the level of exploitation and stock status of target and bycatch species are essential for effective fishery management and conserving biological diversity. Pursuit of these objectives is often challenging where yield statistics are unavailable, especially when scientific surveys are spatially and seasonally unbalanced. To address this issue, we attempt to combine recently developed methods to construct an assessment framework using Japanese anchovy (“anchovy”; Engraulis japonicus) in the Bohai Sea as an example. First, based on data from surveys conducted in spring, summer and autumn in the Bohai Sea during 2009–2019, we developed a seasonal spatio-temporal model for anchovy to interpolate missing observations thereby obtaining a continuous time series of abundance. Then, we fitted the abundance maximum sustainable yield (AMSY) model using the spring, summer, autumn, and seasonal average indices of abundance to obtain long-term trends in the stock status of anchovy. The model indicated that at the end of the survey series in 2019 there was a 17.6%–22.6% probability that the anchovy stock was in a sustainable state, 51.1%–69.3% probability it was in a period of recovery and 10.3%–29.8% probability it was declining. The relative stock size gained significant recovery throughout the study period, reaching a healthy level during 2016–2018. Our assessment framework enabled the estimation of anchovy stock status in the Bohai Sea in the absence of yield data, a process which will facilitate the assessment of other fishery resources in waters without yield statistics to provide estimates of local stock status for managers to take measures to avoid local depletion of those resources.

Ag8GeS6(Se6)–Ag8GeTe6 SYSTEMS: PHASE RELATIONS, SYNTHESIS, AND CHARACTERIZATION OF SOLID SOLUTIONS

Phase equilibria in the Ag8GeS6-Ag8GeTe6 and Ag8GeSe6-Ag8GeTe6 systems were studied using differential thermal analysis and powder X-ray diffraction technique. The T-x phase diagrams were constructed. It has been found that both systems are partially quasi-binary and are characterized by the formation of continuous series of substitutional solid solutions between Ag8GeTe6 and high-temperature cubic modifications of Ag8GeS6 and Ag8GeSe6. When solid solutions are formed, the temperatures of polymorphic phase transitions of these compounds decrease. This leads to the stabilization of the ion-conducting cubic phase in the 40 mol% Ag8GeTe6 range of compositions at room temperature and below. The regions of homogeneity based on Ag8GeS6 and Ag8GeSe6 are 15 and 10 mol.%, respectively. The crystal lattice parameters of the identified solid solutions were calculated from the XRD data, and their linear dependence on the composition is shown

Harmonization of Meteosat First and Second Generation Datasets for Fog and Low Stratus Studies

Operational weather satellites, dating back to 1970s, currently provide the best basis for climatological investigations, such as an analysis of changes in the cloud cover. Because clouds are highly dynamic in time, temporally high-resolution data from the geostationary orbit are preferred in order to take variations in the diurnal cycles into account. For such studies, a consistent dataset in space and time is mandatory, but not yet available. Ground-based point measurements of various cloud parameters, such as ceiling, visibility, and cloud type are often sparsely spread and inconsistent, making it difficult to derive reliable spatio-temporal information over large areas. The Meteosat program has generally provided suitable data from over Europe since 1977, but different spatial, spectral, and radiometric resolution of the instruments of the individual satellites, including early-years calibration uncertainties, makes harmonization necessary to finally derive a time series applicable to any kind of climatological study. In this study, a machine learning-based approach has been employed to generate a long-term consistent dataset with high spatio-temporal resolution and extensive coverage over Europe by the harmonization of Meteosat First Generation (MFG) and Meteosat Second Generation (MSG) satellite datasets (1991–2020). A random forest (RF) regressor is trained on the overlap period (2004–2006), where datasets of both satellite generation (MFG and MSG) are available to predict MFG Water Vapour (WV) and Infrared (IR) channels brightness temperature (BT) values based on MSG channels. The aim of the study is to synthesize MFG MVIRI data from MSG SEVIRI to generate a consistent MFG time series. The results indicate a good match of MFG synthesized data with the original MFG data with a mean absolute error of 0.7 K for the WV model and 1.6 K for the IR model, and an out-of-bag (OOB) R² score of 0.98 for both the models. Based on the trained models, the MFG scenes are synthesized from the MSG scenes for the years from 2006 to 2020. The long-term homogeneity of the generated time series is analyzed. The harmonized dataset will be applied to generate a continuous time series on fog and low stratus (FLS) occurrence for a climatological time scale of 30 years.

Impact of R&D Investment and Network Penetration on Human Development: Evidence from China.

This paper measures the human development indices of 31 inland provinces (municipalities) in China in a continuous time series during 2000-2017 according to the 2010 HDI compilation method. It uses a geographically and temporally weighted regression model for conducting an empirical study on the effects of R&D investment and network penetration on human development in each province (municipality) of China. There is significant spatial and temporal heterogeneity in the impact of R&D investment and network penetration on human development across provinces (municipalities) in China due to differences in resource endowments and economic and social development. For R&D investment, eastern provinces (municipalities) have mostly positive effects on human development, and central regions have mostly weak positive or negative effects. In contrast, western provinces (municipalities) show different development paths, with weak positive effects in the early stage and significant positive effects after 2010. Most provinces (municipalities) show a continuous and increasing positive effect for network penetration. The marginal contributions of this paper are mainly in improving the shortcomings in research perspectives, empirical methods, and research data in the study of human development influencing factors in China relative to the study of HDI itself in terms of measurement or application dimensions. This paper constructs a human development index for China, analyzes its spatial and temporal distribution, and explores the impact of R&D investment and network penetration on human development in China, with the hope of providing lessons for China and developing countries to promote the level of human development and cope with the pandemic.

Quantification and Diversity Analyses of Major Glucosinolates in Conserved Chinese Cabbage (Brassica rapa L. ssp. pekinensis) Germplasms

The genebank at the National Agrobiodiversity Center (RDA-Genebank, Jeonju, Republic of Korea), conserves approximately 8000 germplasms of Brassica spp., of which Chinese cabbage (Brassica rapa L. ssp. pekinensis) is one of the major crops actively used as food in Northeast Asia, including Korea, as the main ingredient for kimchi. Glucosinolates are a major class of compounds in Chinese cabbage that are responsible for their distinctive flavor, and RDA-Genebank is constantly building a related database (DB) to select suitable germplasms required by consumers and provide resources for breeding programs. In this study, ten glucosinolates were analyzed in sixty Chinese cabbage germplasms. Six aliphatic glucosinolates were the major components, accounting for 85.00% to 91.98% of total glucosinolates in each germplasm. Among them, gluconapin (333.26 to 23,501.58 μmol∙kg-1 DW) was highly represented, followed by glucobrassicanapin (545.60 to 10,344.70 μmol∙kg-1 DW) and progoitrin (155.28 to 8536.51 μmol∙kg-1 DW). In addition, we selected germplasms with a high content of each studied glucosinolate. To analyze the diversity and distribution of glucosinolates among the studied germplasms, Pearson's correlation was performed, and the related results were interpreted through their biosynthetic pathways. The k-means clustering indicated four optimal clusters, which were confirmed through principal component analysis. Orthogonal projection to latent structure discriminant analysis (OPLS-DA) was also performed on the status (landrace and cultivar) and origin (Korea, China, Taiwan, and Japan) passport data of the germplasms, followed by the calculation of variable importance in the projection (VIP) values. These results are part of a continuous series of studies to analyze the glucosinolates of Brassica germplasms that are being conserved at RDA-Genebank. We aim to provide related results through a public platform accessible to everyone and thereby improve the distribution of Brassica germplasms.

The vertical velocity field of the Tibetan Plateau and its surrounding areas derived from GPS and surface mass loading models

Obtaining a precise velocity field of the Earth's surface is critical for probing Earth's tectonic movements. In this contribution we construct a precise vertical velocity field with high-spatial resolution for the Tibetan Plateau (TP) and its surrounding areas using both 113 continuous and 969 campaign GPS time series over the period 2008 to 2019. Specifically, we compared the statistical distribution of residual continuous GPS time series to find the optimal regional surface mass loading (SML) models. The optimal modeled nontectonic displacements were used to correct these campaign GPS time series to reduce their scatter and accordingly improve the precision of the uplift rate estimates. The results showed that we could reduce the root mean square (RMS) on the vertical coordinate time series by more than 75% of the campaign GPS stations. We computed and removed vertical deformation due to horizontal strain rates obtained from GPS time series, and discussed dynamic sources (mantle dynamic topography, isostatic adjustment, and flexural loading) that drive the residual vertical tectonic motion. The study presents the most recent surface kinematic constraints for comprehending the dynamic sources of tectonic uplift or subsidence between and within tectonic units.Plain Language Summary: Movements of the Tibetan Plateau (TP) have a profound impact on adjacent areas. We collected all available GPS data to determine a precise vertical velocity field with high-spatial resolution. We make use of optimal surface mass loading (SML) models to improve estimation accuracy of campaign GPS stations, including an estimate of the vertical movement caused by the horizontal tectonic motions. We combine geological and geophysical data with our vertical velocity field to understand the dynamic sources of uplift or subsidence between and within tectonic units of the TP.

A new species of naked-toed gecko Cyrtopodion sensu lato (Squamata: Sauria: Gekkonidae) from Gujarat, India

We describe a new species of naked-toed gecko of the genus Cyrtopodion Fitzinger, 1843, from the Dahod and Panchmahal districts of Gujarat state, India. The new species is diagnosable by the following suite of characters: a medium-sized Cyrtopodion (adult, snout to vent length up to 50 mm); dorsal scalation on trunk granular, intermixed with enlarged, regularly arranged transverse rows of 15 trihedral tubercles; 6 transverse rows of tubercles on the second segment of the tail; midbody scale rows across belly 20-22; midventral scales 89-97; males with a continuous series of 29-33 precloacal-femoral pores. The new species, Cyrtopodion vindhya sp. nov. is the fifth endemic species of reptile described in the last 15 years from the state of Gujarat that highlights the rich and unique diversity of this understudied region.

Development and validation of a pavement temperature profile prediction model in a mechanistic-empirical design framework

An accurate temperature prediction tool is an important part of any mechanistic-empirical (M-E) pavement design and performance prediction method. In this paper, a one-dimensional finite control volume (FCV) model is introduced that predicts the temperature within a pavement structure as a function of time and depth. The main input data required for the model are continuous time series of air temperature for conductive heat transfer, solar radiation for radiative heat transfer, and wind speed for convective heat transfer. The heat balance equation for each control volume of the FCV model is solved using an implicit scheme. To validate the numerical model, comparisons were made to measured temperature data from four test sections in Sweden located in regions with different climatic conditions. A good agreement was obtained between the calculated and measured temperature values within the asphalt layer, and temperature in the granular layers with the values of the coefficient of determination R2 ranging from 0.866 to 0.979. The model is therefore suitable to be implemented as a pavement temperature prediction tool in M-E design.

Challenges and perspectives in recurrence analyses of event time series

The analysis of event time series is in general challenging. Most time series analysis tools are limited for the analysis of this kind of data. Recurrence analysis, a powerful concept from nonlinear time series analysis, provides several opportunities to work with event data and even for the most challenging task of comparing event time series with continuous time series. Here, the basic concept is introduced, the challenges are discussed, and the future perspectives are summarized.

Intelligent Fault Prognosis Method Based on Stacked Autoencoder and Continuous Deep Belief Network

Mechanical fault prediction is one of the main problems in condition-based maintenance, and its purpose is to predict the future working status of the machine based on the collected status information of the machine. However, on one hand, the model health indices based on the information collected by the sensors will directly affect the evaluation results of the system. On the other hand, because the model health index is a continuous time series, the effect of feature learning on continuous data also affects the results of fault prognosis. This paper makes full use of the autonomous information fusion capability of the stacked autoencoder and the strong feature learning capability of continuous deep belief networks for continuous data, and proposes a novel fault prognosis method. Firstly, a stacked autoencoder is used to construct the model health index through the feature learning and information fusion of the vibration signals collected by the sensors. To solve the local fluctuations in the health indices, the exponentially weighted moving average method is used to smooth the index data to reduce the impact of noise. Then, a continuous deep belief network is used to perform feature learning on the constructed health index to predict future performance changes in the model. Finally, a fault prognosis experiment based on bearing data was performed. The experimental results show that the method combines the advantages of stacked autoencoders and continuous deep belief networks, and has a lower prediction error than traditional intelligent fault prognosis methods.

Ankle strategies for step-aside movement during quiet standing.

The mediolateral ankle strategy plays a crucial role in providing ankle stability in ground obstacle-avoidance behavior. This is achieved by changing basic walking patterns according to the characteristics of the obstacle. In daily life, it is more common to use step-aside movement (i.e., dodging step) for collision avoidance rather than side-stepping (i.e., widening standing base) when encountering an approaching pedestrian or bicycle. While studies have been conducted on the mediolateral ankle strategy contribution in obstacle avoidance using side-stepping, knowledge of step-aside movement is still inadequate. Therefore, we conducted an electromyography (EMG) analysis on the tibialis anterior (TA), peroneus longus (PL), and soleus (SOL) muscles, as well as measured center of pressure (CoP) displacement, and vertical ground reaction force (vGRF) of the standing leg, in order to understand the role of ankle muscles in step-aside movement during quiet standing. Fifteen healthy young men repeated twelve step-aside movements in both left and right directions. A Bayesian one-sample t-test was used to determine the sufficient step and participant counts. Multiple linear regression analysis was used to investigate the correlation between the muscle activity and CoP displacement or vGRF. The regression coefficients (β) of the left push phase and the right loading phase were tested against zero using a Bayesian one-sample t-test to identify the correlation between independent and dependent variables. We used the one-dimensional statistical parametric mapping (SPM1d) method to analyze the differences between and within the groups of EMG data based on the continuous time series. The results showed that the PL displayed a substantial contribution to the mediolateral ankle strategy during the push phase of step-aside movement, and also contributed to maintaining ankle stability during the loading phase. This suggested that screening for PL weakness and providing appropriate interventions and/or training approaches is especially critical for populations with walking stability problems.

How subsurface and double-core anticyclones intensify the winter mixed-layer deepening in the Mediterranean Sea

Abstract. The mixed layer is the uppermost layer of the ocean, connecting the atmosphere to the subsurface ocean through atmospheric fluxes. It is subject to pronounced seasonal variations: it deepens in winter due to buoyancy loss and shallows in spring while heat flux increases and restratifies the water column. A mixed-layer depth (MLD) modulation over this seasonal cycle has been observed within mesoscale eddies. Taking advantage of the numerous Argo floats deployed and trapped within large Mediterranean anticyclones over the last decades, we reveal for the first time this modulation at a 10 d temporal scale, free of the smoothing effect of composite approaches. The analysis of 16 continuous MLD time series inside 13 long-lived anticyclones at a fine temporal scale brings to light the importance of the eddy pre-existing vertical structure in setting the MLD modulation by mesoscale eddies. Extreme MLD anomalies of up to 330 m are observed when the winter mixed layer connects with a pre-existing subsurface anticyclonic core, greatly accelerating mixed-layer deepening. The winter MLD sometimes does not achieve such connection but homogenizes another subsurface layer, then forming a multi-core anticyclone with spring restratification. An MLD restratification delay is always observed, reaching more than 2 months in 3 out the 16 MLD time series. The water column starts to restratify outside anticyclones, while the mixed layer keeps deepening and cooling at the eddy core for a longer time. These new elements provide new keys for understanding anticyclone vertical-structure formation and evolution.

Sample based water quality monitoring of coastal seas: How significant is the information loss in patchy time series compared to continuous ones?

The high temporal and spatial variability of tidal dominated coastal areas poses a challenge for characterising water quality. Water quality monitoring relies often on information collected by water sampling from a vessel or by satellites, and covers limited time periods and therefore limited tidal and meteorological conditions. To assess the loss of information from discrete sampling, continuous time series of one year (suspended particulate matter (SPM) concentration, SPM flux and Chlorophyll a (Chl) concentration) were used. Eight different schemes of sampling into these time series were applied that are typical for many monitoring programs. They differ in the time between sampling events (synodic or half-synodic) and the duration of the sampling (tidal cycle, half a tidal cycle, one or more samples). The information loss was quantified by applying a bootstrap method to calculate the mean and standard deviation over the considered period. These were then compared with the true mean calculated from the continuous series. The probability to match the true mean within a certain margin depends on the sampling period and the season, but it is always low, especially if the allowed uncertainty is stringent (e.g., ±2.5 % about the true mean). For the SPM concentration this probability is lower than 10 % and for Chl concentration lower than 20 %. Similarly, conclusions arise for the detection of trends in a 20 year time series of SPM concentration with an artificial yearly increase of 0.5 %. None of the sampling schemes was able to assess statistical significant interannual trends with probabilities above 60 %. Further, the significant trends overestimated the increase by a factor 2 to 8. Here, present modus operandi is thus inadequate for basic trend detection, but may be acceptable for the more marine, lower turbid areas where higher probabilities were obtained in this study.

Understanding the states and dynamics of mangrove forests in land cover transitions of The Gambia using a Fourier transformation of Landsat and MODIS time series in Google Earth Engine

IntroductionMangroves are resilient forests of transitional zones between land, ocean and freshwater for they are tolerant to salinity. In The Gambia, mangrove forests are found along the coast of Atlantic Ocean and River Gambia where they support the livelihoods of millions through multiple ecosystems services. Despite their importance in The Gambia, consistent country-wide information on their coverage, dynamics and change hotspots are lacking. Thus far, it remains unclear whether the coverage of mangroves has decreased or increased over the last few decades. Often, the existing estimates vary strongly across sources and the methodologies in the available literature are not always reproducible. This study attempts to fill these gaps by providing up-to-date spatial information on mangrove forests in The Gambia.MethodsTo provide a reproducible workflow and a comprehensive assessment, we used continuous time series of freely available data to study the extent and dynamics of mangrove forests in The Gambia. To construct gap-free image time series, we used statistical models to describe land surface phenology based on monthly composites derived from Landsat and MODIS land surface reflectance acquired between 2000 and 2020. We used 1212 Landsat and 407 MODIS tiles spread across multiple spectral indices along with a calibrated set of training and validation data to train and validate a random forest classifier for accurate land cover classifications.Results and discussionThe overall accuracy and Kappa statistics of the classifications range between 0.96 and 0.98. Our findings suggest a net increase of 4,000 ha in mangrove forests over the last 2 decades, corresponding to an annual rate of 200 ha. The net increase is largely due to strong policy making which results in participative forest resource management through the national forest action plan. The net increase in mangrove forests should not mask the substantial degradation in some places across the country. We estimated a total loss of nearly 5,670 ha, most of which appears to have taken place during the last decade in favor of wetlands. The Gambian mangroves are amongst the most promising green business in Gulf of Guinea, deserving an integrated governance - policy framework that address the key requirements for ecological sustainability.

Comparison of Statistical Production Models for a Solar and a Wind Power Plant

Mathematical models to characterize and forecast the power production of photovoltaic and eolian plants are justified by the benefits of these sustainable energies, the increased usage in recent years, and the necessity to be integrated into the general energy system. In this paper, starting from two collections of data representing the power production hourly measured at a solar plant and a wind farm, adequate time series methods have been used to draw appropriate statistical models for their productions. The data are smoothed in both cases using moving average and continuous time series have been obtained leading to some models in good agreement with experimental data. For the solar power plant, the developed models can predict the specific power of the next day, next week, and next month, with the most accurate being the monthly model, while for wind power only a monthly model could be validated. Using the CUSUM (cumulative sum control chart) method, the analyzed data formed stationary time series with seasonality. The similar methods used for both sets of data (from the solar plant and wind farm) were analyzed and compared. When compare with other studies which propose production models starting from different measurements involving meteorological data and/or machinery characteristics, an innovative element of this paper consists in the data set on which it is based, this being the production itself. The novelty and the importance of this research reside in the simplicity and the possibility to be reproduced for other related conditions even though every new set of data (provided from other power plants) requires further investigation.

Feature Extraction and Prediction of Water Quality Based on Candlestick Theory and Deep Learning Methods

In environmental hydrodynamics, a research topic that has gained popularity is the transmission and diffusion of water pollutants. Various types of change processes in hydrological and water quality are directly related to meteorological changes. If these changing characteristics are classified effectively, this will be conducive to the application of deep learning theory in water pollution simulation. When periodically monitoring water quality, data were represented with a candlestick chart, and different classification features were displayed. The water quality data from the research area from 2012 to 2019 generated 24 classification results in line with the physics laws. Therefore, a deep learning water pollution prediction method was proposed to classify the changing process of pollution to improve the prediction accuracy of water quality, based on candlestick theory, visual geometry group, and gate recurrent unit (CT-VGG-GRU). In this method, after the periodic changes of water quality were represented by candlestick graphically, the features were extracted by the VGG network based on its advantages in graphic feature extraction. Then, this feature and other scenario parameters were fused as the input of the time series network model, and the pollutant concentration sequence at the predicted station constituted the output of the model. Finally, a hybrid model combining graphical and time series features was formed, and this model used continuous time series data from multiple stations on the Lijiang River watershed to train and validate the model. Experimental results indicated that, compared with other comparison models, such as the back propagation neural network (BPNN), support vector regression (SVR), GRU, and VGG-GRU, the proposed model had the highest prediction accuracy, especially for the prediction of extreme values. Additionally, the change trend of water pollution was closer to the real situation, which indicated that the process change information of water pollution could be fully extracted by the CT-VGG-GRU model based on candlestick theory. For the water quality indicators DO, CODMn, and NH3-N, the mean absolute errors (MAE) were 0.284, 0.113, and 0.014, the root mean square errors (RMSE) were 0.315, 0.122, and 0.016, and the symmetric mean absolute percentage errors (SMAPE) were 0.022, 0.108, and 0.127, respectively. The established CT-VGG-GRU model achieved superior computational performance. Using the proposed model, the classification information of the river pollution process could be obtained effectively and the time series information could also be retained, which made the application of the deep learning model to the transmission and diffusion process of river water pollution more explanatory. The proposed model can provide a new method for water quality prediction.

Non-spherical microparticle shape in Antarctica during the last glacial period affects dust volume-related metrics

Abstract. Knowledge of microparticle geometry is essential for accurate calculation of ice core volume-related dust metrics (mass, flux, and particle size distributions) and subsequent paleoclimate interpretations, yet particle shape data remain sparse in Antarctica. Here we present 41 discrete particle shape measurements, volume calculations, and calibrated continuous particle time series spanning 50–16 ka from the South Pole Ice Core (SPC14) to assess particle shape characteristics and variability. We used FlowCAM, a dynamic particle imaging instrument, to measure aspect ratios (width divided by length) of microparticles. We then compared those results to Coulter counter measurements on the same set of samples as well as high-resolution laser-based (Abakus) data collected from SPC14 during continuous flow analysis. The 41 discrete samples were collected during three periods of millennial-scale climate variability: Heinrich Stadial 1 (18–16 ka, n=6; ∼250 years per sample), the Last Glacial Maximum (LGM) (27–18 ka, n=19; ∼460 years per sample), and during both Heinrich Stadial 4 (42–36 ka, n=8; ∼620 years per sample) and Heinrich Stadial 5 (50–46 ka, n=8; ∼440 years per sample). Using FlowCAM measurements, we calculated different particle size distributions (PSDs) for spherical and ellipsoidal volume estimates. Our calculated volumes were then compared to published Abakus calibration techniques. We found that Abakus-derived PSDs calculated assuming ellipsoidal, rather than spherical, particle shapes provide a more accurate representation of PSDs measured by Coulter counter, reducing Abakus to Coulter counter flux and mass ratios from 1.82 (spherical assumption) to 0.79 and 1.20 (ellipsoidal assumptions; 1 being a perfect match). Coarser particles (>5.0 µm diameter) show greater variation in measured aspect ratios than finer particles (<5.0 µm). While fine particle volumes can be accurately estimated using the spherical assumption, applying the same assumption to coarse particles has a large effect on inferred particle volumes. Temporally, coarse and fine particle aspect ratios do not significantly change within or among the three time periods (p value >0.05), suggesting that long-range transport of dust is likely dominated by clay minerals and other elongated minerals.

Comparison of Artificial Neural Network and Regression Models for Filling Temporal Gaps of Meteorological Variables Time Series

Continuous meteorological variable time series are highly demanded for various climate related studies. Five statistical models were tested for application of temporal gaps filling in time series of surface air pressure, air temperature, relative air humidity, incoming solar radiation, net radiation, and soil temperature. A bilayer artificial neural network, linear regression, linear regression with interactions, and the Gaussian process regression models with exponential and rational quadratic kernel were used to fill the gaps. Models were driven by continuous time series of meteorological variables from the ECMWF (European Centre for Medium-range Weather Forecasts) ERA5-Land reanalysis. Raw ECMWF ERA5-Land reanalysis data are not applicable for characterization of specific local weather conditions. The linear correlation coefficients (CC) between ERA5-Land data and in situ observations vary from 0.61 (for wind direction) to 0.99 (for atmospheric pressure). The mean difference is high and estimated at 3.2 °C for air temperature and 3.5 hPa for atmospheric pressure. The normalized root-mean-square error (NRMSE) is 5–13%, except for wind direction (NRMSE = 49%). The linear bias correction of ERA5-Land data improves matching between the local and reanalysis data for all meteorological variables. The Gaussian process regression model with an exponential kernel based or bilayered artificial neural network trained on ERA5-Land data significantly shifts raw ERA5-Land data toward the observed values. The NRMSE values reduce to 2–11% for all variables, except wind direction (NRMSE = 22%). CC for the model is above 0.87, except for wind characteristics. The suggested model calibrated against in situ observations can be applied for gap-filling of time series of meteorological variables.