Application of Support Vector Regression for Mid- and Long-term Runoff Forecasting in “Yellow River Headwater” Region

Capability assessment of conventional and data-driven models for prediction of suspended sediment load

Information about suspended sediment concentration (SSC) in the stream is vital for sustainability of water conservation and erosion control planning, designing and monitoring. In this research, prediction of SSC has been done using artificial neural network (ANN), support vector regression (SVR) and multi-linear regression (MLR) models. Performance evaluation of developed models has been carried out on the basis of root mean square error (RMSE), correlation coefficient (r), coefficient of efficiency (CE) and pooled average relative error (PARE). Cross-correlation function (CCF) validated that gamma test (GT) is an appropriate tool for the selection of most responsive input variables. On the basis of GT and CCF, GT-6 model was selected as the model with most effective input variables, with the values of gamma, standard error and V-ratio as 0.0643, 0.00583 and 0.2570, respectively. The ANN (6-3-1) model performed better than the other single and double hidden layered ANN models with the values of r, RMSE, CE and PARE as 0.939, 0.0063 g/l, 85.17 and 0.0160, respectively. The performance of the SVR model was found better with the values of r, RMSE, CE and PARE as 0.906, 0.018 g/l, 79.09 and 0001, respectively, but slightly poor than the selected ANN (6-3-1) model. The values of r, RMSE, CE and PARE were found as 0.899, 0.0312 g/l, 65.15 and - 0.0031, respectively, in the case of MLR model. The present study revealed that among the ANN, SVR and MLR models, the ANN model with a single hidden layer is most suitable for observed SSC. The present study offers the simple efficient model to estimate the suspended sediment concentration in the stream with minimum error using limited data set.

An Improved Method for Monitoring Multiscale Plant Species Diversity of Alpine Grassland Using UAV: A Case Study in the Source Region of the Yellow River, China

Plant species diversity (PSD) is essential in evaluating the function and developing the management and conservation strategies of grassland. However, over a large region, an efficient and high precision method to monitor multiscale PSD (α-, β-, and γ-diversity) is lacking. In this study, we proposed and improved an unmanned aerial vehicle (UAV)-based PSD monitoring method (UAVB) and tested the feasibility, and meanwhile, explored the potential relationship between multiscale PSD and precipitation on the alpine grassland of the source region of the Yellow River (SRYR), China. Our findings showed that: (1) UAVB was more representative (larger monitoring areas and more species identified with higher α- and γ-diversity) than the traditional ground-based monitoring method, though a few specific species (small in size) were difficult to identify; (2) UAVB is suitable for monitoring the multiscale PSD over a large region (the SRYR in this study), and the improvement by weighing the dominance of species improved the precision of α-diversity (higher R 2 and lower P values of the linear regressions); and (3) the species diversity indices (α- and β-diversity) increased first and then they tended to be stable with the increase of precipitation in SRYR. These findings conclude that UAVB is suitable for monitoring multiscale PSD of an alpine grassland community over a large region, which will be useful for revealing the relationship of diversity-function, and helpful for conservation and sustainable management of the alpine grassland.

Mid- to Long-Term Runoff Prediction Based on Deep Learning at Different Time Scales in the Upper Yangtze River Basin

Deep learning models are essential tools for mid- to long-term runoff prediction. However, the influence of the input time lag and output lead time on the prediction results in deep learning models has been less studied. Based on 290 schemas, this study specified different time lags by sliding windows and predicted the runoff process by RNN (Recurrent Neural Network), LSTM (Long–short-term Memory), and GRU (Gated Recurrent Unit) models at five hydrological stations in the upper Yangtze River during 1980–2018 at daily, ten-day, and monthly scales. Different models have different optimal time lags; therefore, multiple time lags were analyzed in this paper to find out the relationship between the time intervals and the accuracy of different river runoff predictions. The results show that the optimal time-lag settings for the RNN, LSTM, and GRU models in the daily, ten-day, and monthly scales were 7 days, 24 ten days, 27 ten days, 24 ten days, 24 months, 27 months, and 21 months, respectively. Furthermore, with the increase of time lags, the simulation accuracy would stabilize after a specific time lag at multiple time scales of runoff prediction. Increased lead time was linearly related to decreased NSE at daily and ten-day runoff prediction. However, there was no significant linear relationship between NSE and lead time at monthly runoff prediction. Choosing the smallest lead time could have the best prediction results at different time scales. Further, the RMSE of the three models revealed that RNN was inferior to LSTM and GRU in runoff prediction. In addition, RNN, LSTM, and GRU models could not accurately predict extreme runoff events at different time scales. This study highlights the influence of time-lag setting and lead-time selection in the mid- to long-term runoff prediction results for the upper Yangtze River basin. It is recommended that researchers should evaluate the effect of time lag before using deep learning models for runoff prediction, and to obtain the best prediction, the shortest lead-time length can be chosen as the best output for different time scales.

Runoff forecasting model based on variational mode decomposition and artificial neural networks

Accurate runoff forecasting plays a vital role in water resource management. Therefore, various forecasting models have been proposed in the literature. Among them, the decomposition-based models have proved their superiority in runoff series forecasting. However, most of the models simulate each decomposition sub-signals separately without considering the potential correlation information. A neoteric hybrid runoff forecasting model based on variational mode decomposition (VMD), convolution neural networks (CNN), and long short-term memory (LSTM) called VMD-CNN-LSTM, is proposed to improve the runoff forecasting performance further. The two-dimensional matrix containing both the time delay and correlation information among sub-signals decomposing by VMD is firstly applied to the CNN. The feature of the input matrix is then extracted by CNN and delivered to LSTM with more potential information. The experiment performed on monthly runoff data investigated from Huaxian and Xianyang hydrological stations at Wei River, China, demonstrates the VMD-superiority of CNN-LSTM to the baseline models, and robustness and stability of the forecasting of the VMD-CNN-LSTM for different leading times.

Spatio-Temporal Variation Characteristics of Aboveground Biomass in the Headwater of the Yellow River Based on Machine Learning

Accurate estimation of the aboveground biomass (AGB) of grassland is a key link in understanding the regional carbon cycle. We used 501 aboveground measurements, 29 environmental variables, and machine learning algorithms to construct and verify a custom model of grassland biomass in the Headwater of the Yellow River (HYR) and selected the random forest model to analyze the temporal and spatial distribution characteristics and dynamic trends of the biomass in the HYR from 2001 to 2020. The research results show that: (1) the random forest model is superior to the other three models (R2val = 0.56, RMSEval = 51.3 g/m2); (2) the aboveground biomass in the HYR decreases spatially from southeast to northwest, and the annual average value and total values are 176.8 g/m2 and 20.73 Tg, respectively; (3) 69.51% of the area has shown an increasing trend and 30.14% of the area showed a downward trend, mainly concentrated in the southeast of Hongyuan County, the northeast of Aba County, and the north of Qumalai County. The research results can provide accurate spatial data and scientific basis for the protection of grassland resources in the HYR.

The application of artificial neural networks and support vector regression for simultaneous spectrophotometric determination of commercial eye drop contents

In the present study, artificial neural networks (ANNs) and support vector regression (SVR) as intelligent methods coupled with UV spectroscopy for simultaneous quantitative determination of Dorzolamide (DOR) and Timolol (TIM) in eye drop. Several synthetic mixtures were analyzed for validating the proposed methods. At first, neural network time series, which one type of network from the artificial neural network was employed and its efficiency was evaluated. Afterwards, the radial basis network was applied as another neural network. Results showed that the performance of this method is suitable for predicting. Finally, support vector regression was proposed to construct the Zilomole prediction model. Also, root mean square error (RMSE) and mean recovery (%) were calculated for SVR method. Moreover, the proposed methods were compared to the high-performance liquid chromatography (HPLC) as a reference method. One way analysis of variance (ANOVA) test at the 95% confidence level applied to the comparison results of suggested and reference methods that there were no significant differences between them. Also, the effect of interferences was investigated in spike solutions.