CDOM fluorescence as a proxy of DOC concentration in natural waters: a comparison of four contrasting tropical systems

Chromophoric dissolved organic matter (CDOM) fluorescence or absorption is often proposed as a rapid alternative to chemical methods for the estimation of bulk dissolved organic carbon (DOC) concentration in natural waters. However, the robustness of this method across a wide range of systems remains to be shown. We measured CDOM fluorescence and DOC concentration in four tropical freshwater and coastal environments (estuary and coastal, tropical shallow lakes, water from the freshwater lens of two small islands, and soil leachates). We found that although this method can provide an estimation of DOC concentration in sites with low variability in DOC and CDOM sources in systems where the variability of DOC and CDOM sources are high, this method should not be used as it will lead to errors in the estimation of the bulk DOC concentration.     

Why do satellite imageries show exceptionally high chlorophyll in the Gulf of Mannar and the Palk Bay during the Norteast Monsoon?

The Gulf of Mannar (GoM) and the Palk Bay (PB) are two least studied marine environments located between India and Sri Lanka. Exceptionally high chlorophyll a concentration in the GoM and the PB during the Northeast Monsoon (November–February) is a consistent feature in satellite imageries, which has been attributed to the intrusion of the Bay of Bengal (BoB) waters. The analyses of the Moderate Resolution Imaging Spectroradiometer (MODIS) and field chlorophyll data collected from 30 locations in the Indian sector of the GoM and the PB in January 2011 showed significant overestimations in the satellite data. This error was much higher in the PB (60–80 %) as compared to the GoM (18–28 %). The multivariate analyses evidenced that the exceptionally high satellite chlorophyll in the PB is contributed largely by turbidity, colored dissolved organic matter (CDOM), and bottom reflectance. The paper cautions that though MODIS is superior in estimating chlorophyll a in optically complex waters, there are still chances of overestimations in regions like the PB.     

Estimation of chlorophyll a content in inland turbidity waters using WorldView-2 imagery: a case study of the Guanting Reservoir,Beijing, China

Complex optical properties, such as non-pigment suspension and colored dissolved organic matter (CDOM), make it difficult to achieve accurate estimations of remotely sensed chlorophyll a (Chla) content of inland turbidity. Recent attempts have been made to estimate Chla based on red and near-infrared regions where non-pigment suspension and CDOM have little effect on water reflectance. The objective of this study is to validate the applicability of WV-2 imagery with existing effective estimation methods from MERIS when estimating Chla content in inland turbidity waters. The correlation analysis of measured Chla content and WV-2 imagery bands shows that the Chla sensitive bands of WV-2 are red edge, NIR 1, and NIR 2. The coastal band is designed for seawater Chla detection. However, the high correlation with turbidity data and low correlation with Chla made coastal band unsuitable for estimating Chla in inland waters. The high-resolution water body images were extracted by combining the spectral products (NDWI) with the spatial morphological products (sobel edge detection). The estimation results show that the accuracy of the single band and NDCI is not as good as the two-band method, three-band method, stepwise regression algorithm (SRA) and support vector machines (SVM). The SVM estimation accuracy was the highest with an R², RMSE, and URMSE of 0.8387, 0.4714, and 19.11%, respectively. This study demonstrates that the two-band and three-band methods are effective for estimating Chla in inland water for WV-2 imagery. As a high-precision estimation method, SVM has great potential for inland turbidity water Chla estimation.     

紫外光诱导荧光分析仪结合多元线性回归算法在城市河流常规污染指标监测中的应用

为建立一种针对城市河流水体常规污染指标的快速原位监测方法,首次运用紫外光诱导荧光分析仪对扬州市60条城市河流进行水体三维荧光光谱(EEM)测量,形成了具有多样性的水质样本集合。利用峰值拾取法、相关性分析和主成分分析3种方式从三维荧光光谱中提取溶解性有机物(DOM)污染信息,结合多元线性回归算法(MLR),建立与化学需氧量(COD_Cr)、高锰酸盐指数(I_Mn)、氨氮(NH₃-N)和总磷(TP)4项常规水质污染指标相关的预测模型。研究结果表明,峰值拾取法结合相关性分析可以有效地反映水体EEM中的污染特征和状况,由此建立的4项水质指标预测模型训练集决定系数均>0.82,预测结果与国家及行业标准方法分析值之间具有较低的均方根误差,说明该预测方法具有较好的准确度和精密度,为城市广域水体的高效、原位监测提供了一种有效的解决方案。     

Estimation of chlorophyll-a concentration using field spectral data: a case study in inland Case-II waters, North China

In the remote sensing of chlorophyll-a (Chla) in inland Case-II waters, the assumption that the optical parameter of Chla specific absorption coefficient a*ph remains constant usually restrains application of many models. In this paper, we presented a newly developed model [Rrs(-1)(lambda1) - Rrs(-1)(lambda2)] x Rrs(lambda3) x a*ph(-1)(lambda1) which was improved on a previous three-band model to isolate interferences from a*ph. In terms of the importance of water optical properties in the model development, spectral and absorption characteristics were analyzed for Shitoukoumen Reservoir and Songhua Lake in Northeast China, as typical examples of inland Case-II waters. Both waters showed overwhelming absorption sum of tripton and chromophoric dissolved organic matter (CDOM) owing to their relatively low Chla contents (1.53 to 19.35 microgl(-1)). According to the optical characteristics of waters studied, optimal positions for lambda (1), lambda (2) and lambda (3) were spectrally tuned to be at 664, 684 and 705 nm, respectively. The model allowed accurate Chla estimation with a determination coefficient (R (2)) close to 0.98 and a root mean square error (RMSE) of 0.87 microgl(-1). Comparison of different models further showed the stability of the improved model, implying its potential use in water color remote sensing. Although the findings underline the rationale behind the improved model, an extensive database containing data in different water conditions and water types is required to generalize its application.     

基于BP神经网络的贵阳市空气质量指数预报模型

采用贵阳市2013年1月1日—2015年12月31日的空气质量指数(AQI)日均值,常规的地面和高空观测资料,基于不同季节,调整BP神经网络的隐藏层个数和隐藏层节点数,建立不同的BP神经网络预报模型,进行参数检验,最终选取预报效果最好的模型带入实况进行检验。结果表明,夏季的预报效果最好,采用的模型TS评分为81.6%,平均绝对误差为9.1,正确率为97.4%,用该模型检验预报效果,实况和预报的相关系数为0.71,平均误差为9;而冬季的预报效果明显低于其他季节,采用的模型TS评分为65.7%,平均绝对误差为19.5,正确率为72.9%,用该模型检验预报效果,实况和预报的相关系数为0.79,平均误差为19。而且BP神经网络模型的预报效果同隐藏层个数与隐藏层节点数没有显著关系。     

基于小波分析优化PM 2.5浓度预测模型

采用多元线性回归方法(MLR)和BP神经网络方法(BPNN),按1 h、3 h、6 h、12 h、24 h、48 h预测时长对贵港市2015—2018年PM_2.5浓度建模并检验对比模型准确率。结果表明,基于MLR与BPNN都能对PM_2.5浓度作预测,预测效果随着预测时长的增加而下降,MLR、BPNN模型预测结果平均绝对误差(MAE)分别为4.01μg/m³~15.48μg/m³、3.89μg/m³~15.63μg/m³。采用小波分析方法对污染物数据优化并再次建模,结果表明,小波-多元线性回归(W-MLR)模型与小波-神经网络(W-BPNN)模型均得到优化,3 h~24 h预测时长优化效果尤为显著,W-MLR、W-BPNN模型预测结果分别使MAE降低1.6%~13.5%、0.8%~9.8%,且后者预测效果优于前者。     

Monitoring diel dissolved oxygen dynamics through integrating wavelet denoising and temporal neural networks

Diel dissolved oxygen (DO) time series measured continuously using proximal sensors in situ for a temperate lake were denoised using discrete wavelet transform (DWT) with the orthogonal wavelet families of coiflet, daubechies, and symmlet with order of 10. Diel DO time series denoised were modeled using nine temporal artificial neural networks (ANNs) as a function of water level, water temperature, electrical conductivity, pH, day of year, and hour. Our results showed that time-lag recurrent network (TLRN) using denoised data emulated diel DO dynamics better than the best-performing TLRN using the original data, time-delay neural network (TDNN), and recurrent network (RNN). Daubechies basis dealt with diel DO data slightly better than the other bases given its coefficient of determination (r ²?=?87.1 %), while symmlet performed slightly better than the other bases in terms of root mean square error (RMSE?=?1.2 ppm) and mean absolute error (MAE?=?0.9 ppm).     

Optical Resilience of the Paraíba do Sul River (Brazil) during a Toxic Spill of a Wood-Pulping Factory: The Cataguazes Accident

This study investigates the inherent optical properties (IOP) of a Brazilian river during a non-natural, anthropogenically mediated, toxic spill of a wood-pulping factory (the ‘Cataguazes accident’). The results indicated an outstanding transformation in the river water chromophoric dissolved organic matter (CDOM) pools. For instance, increases in CDOM absorption coefficients, a _CDOM (λ), which were averaged at specific spectral intervals, , ranged from 58-fold at the UV-B and UV-A ranges to 95-fold at the PAR range. As a result, the water color expressed as CDOM absorption at 440 nm, a _CDOM (440), varied from 4.16 to 365.03 m^-1. For S-coefficient, the variations ranged from ∼1.1 to 5.6-fold, respectively, at the 300–650 nm and UV-B range. The variability of S as a proxy of dissolved chromophores was thus clearly influenced by the spectral range used. Optical proportions were also investigated through the use of and S ratios at the UV-B, UV-A, and PAR ranges and, in the case of , also at the NIR range. This approach also showed clear variations between the water samples, likely reflecting changes in the composition of optically active substances in the river system. As a whole, the findings obtained here indicated that both the quantity and quality of the chromophoric material dissolved in the river water were greatly altered by the toxic spill. The changes in the optical properties of the river water, although extreme and likely with no parallel in the literature, were quite rapid as indicated by the optical resilience of the system. Overall, this study indicates that IOP might be thought, and possibly used, as a metric tool for monitoring the state of waters and aquatic ecosystems.     

Forecasting of cyanobacterial density in Torrão reservoir using artificial neural networks

The ability of general regression neural networks (GRNN) to forecast the density of cyanobacteria in the Torr?o reservoir (Tamega river, Portugal), in a period of 15 days, based on three years of collected physical and chemical data, was assessed. Several models were developed and 176 were selected based on their correlation values for the verification series. A time lag of 11 was used, equivalent to one sample (periods of 15 days in the summer and 30 days in the winter). Several combinations of the series were used. Input and output data collected from three depths of the reservoir were applied (surface, euphotic zone limit and bottom). The model that presented a higher average correlation value presented the correlations 0.991; 0.843; 0.978 for training, verification and test series. This model had the three series independent in time: first test series, then verification series and, finally, training series. Only six input variables were considered significant to the performance of this model: ammonia, phosphates, dissolved oxygen, water temperature, pH and water evaporation, physical and chemical parameters referring to the three depths of the reservoir. These variables are common to the next four best models produced and, although these included other input variables, their performance was not better than the selected best model.     

Artificial neural network modeling of dissolved oxygen in reservoir

The water quality of reservoirs is one of the key factors in the operation and water quality management of reservoirs. Dissolved oxygen (DO) in water column is essential for microorganisms and a significant indicator of the state of aquatic ecosystems. In this study, two artificial neural network (ANN) models including back propagation neural network (BPNN) and adaptive neural-based fuzzy inference system (ANFIS) approaches and multilinear regression (MLR) model were developed to estimate the DO concentration in the Feitsui Reservoir of northern Taiwan. The input variables of the neural network are determined as water temperature, pH, conductivity, turbidity, suspended solids, total hardness, total alkalinity, and ammonium nitrogen. The performance of the ANN models and MLR model was assessed through the mean absolute error, root mean square error, and correlation coefficient computed from the measured and model-simulated DO values. The results reveal that ANN estimation performances were superior to those of MLR. Comparing to the BPNN and ANFIS models through the performance criteria, the ANFIS model is better than the BPNN model for predicting the DO values. Study results show that the neural network particularly using ANFIS model is able to predict the DO concentrations with reasonable accuracy, suggesting that the neural network is a valuable tool for reservoir management in Taiwan.     

Water Quality Prediction of Water Sources Based on Meteorological Factors using the CA-NARX Approach

With the increasingly serious problem of surface water environmental safety, it is of great significance to study the changing trend of reservoir water quality, and it is necessary to establish a water quality prediction and early warning system for the management and maintenance of water resources. Aiming at the problem of water quality prediction in reservoirs, a CA-NARX algorithm is designed, which combines the improved dynamic clustering algorithm with the idea of machine learning and the forward dynamic regression neural network. The improved dynamic clustering algorithm is used to classify the eutrophication degree of waterbodies according to the total phosphorus and total nitrogen content. Considering four meteorological factors, air temperature, water temperature, water surface evaporation, and rainfall, synthetically for each water quality condition, the total phosphorus and total nitrogen in the waterbody are forecasted by an improved forward NARX dynamic regression neural network. Based on this, the CA-NARX prediction algorithm can realize short period water quality prediction. Compared with the traditional support vector regression machine model, improved GA-BP neural network, and exponential smoothing method, the CA-NARX model has the least prediction error.

     

Assessment of temporal and spatial water quality in international Gomishan Lagoon,Iran, using multivariate analysis

Coastal lagoon ecosystems are vulnerable to eutrophication, which leads to the accumulation of nutrients from the surrounding watershed over the long term. However, there is a lack of information about methods that could accurate quantify this problem in rapidly developed countries. Therefore, various statistical methods such as cluster analysis (CA), principal component analysis (PCA), partial least square (PLS), principal component regression (PCR), and ordinary least squares regression (OLS) were used in this study to estimate total organic matter content in sediments (TOM) using other parameters such as temperature, dissolved oxygen (DO), pH, electrical conductivity (EC), nitrite (NO₂), nitrate (NO₃), biological oxygen demand (BOD), phosphate (PO₄), total phosphorus (TP), salinity, and water depth along a 3-km transect in the Gomishan Lagoon (Iran). Results indicated that nutrient concentration and the dissolved oxygen gradient were the most significant parameters in the lagoon water quality heterogeneity. Additionally, anoxia at the bottom of the lagoon in sediments and re-suspension of the sediments were the main factors affecting internal nutrient loading. To validate the models, R², RMSECV, and RPDCV were used. The PLS model was stronger than the other models. Also, classification analysis of the Gomishan Lagoon identified two hydrological zones: (i) a North Zone characterized by higher water exchange, higher dissolved oxygen and lower salinity and nutrients, and (ii) a Central and South Zone with high residence time, higher nutrient concentrations, lower dissolved oxygen, and higher salinity. A recommendation for the management of coastal lagoons, specifically the Gomishan Lagoon, to decrease or eliminate nutrient loadings is discussed and should be transferred to policy makers, the scientific community, and local inhabitants.     

The distribution of organic matter in songkhla lake basin (SLB)

The concentrations and the distribution of organic matter in SLB (e.g. the outer part of Thale Sap Songkhla area) were examined during the period of September 1988 to September 1989. Organic matter in water samples were analyzed by gravimetry. Total organic matter and dissolved organic matter concentrations ranged from 74 to 7908 mg/L and from 23 to 7813 mg/L, respectively. Except for a set of observations which showed low concentrations during the flood season in November 1988, organic matter was regularly distributed throughout SLB. Domestic effluent and aquaculture are hypothesized as major sources of organic matter contamination. The levels of organic matter concentrations in SLB are significantly higher than would be expected for an area considered to be uncontaminated. The data also indicated a linear relationship between concentrations of suspended organic matter and suspended solids. This may be due to the organic matter associated with suspended solids.     

Hybrid Neural Networks and Boosted Regression Tree Models for Predicting Roadside Particulate Matter

This paper examines the application of artificial neural network (ANN) and boosted regression tree (BRT) methods in air quality modelling. The methods were applied to developing air quality models for predicting roadside particle mass concentration (PM₁₀, PM_2.5) and particle number counts (PNC) based on air pollution, traffic and meteorological data from Marylebone Road in London. Elastic net, Lasso and principal components analysis were used as feature selection methods for the ANN models to reduce the number of predictor variables and improve their generalisation. The performance of the ANN with feature selection (ANN hybrid) and the BRT models was evaluated and compared using statistical performance metrics. The performance parameters include root mean square error (RMSE), fraction of prediction within a factor of two of the observation (FAC2), mean bias (MB), mean gross error (MGE), the coefficient of correlation (R) and coefficient of efficiency (CoE) values. The input variables selected by the elastic net produced the best performing ANN models. The ANN hybrid produced models performed only slightly better than the BRT models. The R values of the ANN elastic net and BRT models were 0.96 and 0.95 for PM₁₀, 0.96 and 0.96 for PM_2.5 and 0.89 and 0.87 for PNC, respectively. Their corresponding CoE values were 0.72 and 0.70 for PM₁₀, 0.74 and 0.76 for PM_2.5 and 0.81 and 0.71 for PNC respectively. About 80–99% of all the model predictions are within a factor of two of the observed particle concentrations. The BRT models offer more advantages regarding model interpretation and permit feature selection. Therefore, the study recommends the use of BRT over ANN where the model interpretation is a priority.     

Chemical speciation and partitioning of trace metals (Cd, Co, Cu, Ni, Pb) in the lower Athabasca river and its tributaries (Alberta, Canada)

Concentrations of Cd, Co, Cu, Ni and Pb were measured in particulate and dissolved phases at 11 sites located upstream and near Athabasca oil sands development. The in situ discrimination between non-labile and labile dissolved metals was done using diffusive gradients in thin-films (DGT) devices. The DGT-labile fraction of Co and Ni was 30% lower near development sites whereas Cu, Cd and Pb showed minor changes spatially. It was found that an 8-fold increase in dissolved organic matter (DOM) near development induced a rapid decrease in DGT-labile metals. Dissolved metal concentrations were used along with DOM, major ions, nutrients, pH and conductivity to calculate the distribution of dissolved metal species using the speciation model WHAM. Labile-DGT metal concentrations agreed well with WHAM-predicted concentrations. It was also found that a significant amount of metals were associated with the non-DGT labile fraction (i.e. colloidal DOM) and colloid abundance was more important than suspended particulate matter abundance in influencing metal mobility near Athabasca oil soils development. Since changes in colloidal DOM levels are likely to be the result of surface mining activities, this confirms the serious effects of oil sands activities on metal biogeochemical cycles in the lower Athabasca River.     

Artificial neural network analysis for reliability prediction of regional runoff utilization

Many factors in the reliability analysis of planning the regional rainwater utilization tank capacity need to be considered. Based on the historical daily rainfall data, the following four analyzing procedures will be conducted: the regional daily rainfall frequency, the amount of runoff, the water continuity, and the reliability. Thereafter, the suggested designed storage capacity can be obtained according to the conditions with the demand and supply reliability. By using the output data, two different types of artificial neural network models are used to build up small area rainfall–runoff supply systems for the simulation of reliability and the prediction model. They are also used for the testing of stability and learning speed assessment. Based on the result of this research, the radial basis function neural network (RBFNN) model, using the Gaussian function that has a similar trend as the nature as basic function, has better stability than using the back-propagation neural network (BPNN) model. Despite the fact that RBFNN was more reliable than BPNN, it still made a conservative estimate for the actual monitoring data. The error rate of RBFNN was still higher than the correction of BPNN 4-3-1-1. This should have significant benefit in the future application of the instantaneous prediction or the development of related intelligent instantaneous control equipment.     

基于随机森林与神经网络的高光谱土壤重金属Zn含量反演

在河北省保定市白洋淀区域采集115个土壤样品进行重金属含量分析和室内光谱测量,分别将BP神经网络、随机森林、决策树、多元线性回归、K近邻回归、AdaBoost回归和偏最小二乘回归法应用于全部原始波谱数据和基于双层随机森林选择后的波段数据。结果表明,基于原始波谱数据的土壤重金属Zn元素含量的反演模型精度较低,而通过双层随机森林选择出光谱数据中与土壤重金属Zn信息相关的波段,减轻了网络模型的过拟合问题,提高了模型预测精度;与其他模型比较,结合双层随机森林和BP神经网络构建的反演模型对研究区土壤重金属Zn含量预测效果最佳。     

NARX神经网络模型在昆明市环境空气质量预测中的应用

神经网络模型作为一种重要的手段被广泛应用于数学计算、物理建模、水文模拟、环境预测、人工智能等研究领域。为验证神经网络模型在高原山地城市环境空气质量预测中的作用,以昆明市环境空气自动监测站气象因子和污染物浓度数据为基础,构建NARX神经网络模型,对污染物浓度进行预测。结果表明,基于NARX神经网络建立的预测模型具有很强的非线性动态描述能力,能够对环境空气6参数做出较为准确的预测,其预测浓度相对误差显著低于CMAQ、NAQPMS空气质量数值模式以及LSTM统计模型预测结果。优化后的NARX神经网络对污染物浓度变化趋势的预测较其他几个模式更为敏感。