Real-time forecasting of suspended sediment concentrations in reservoirs by the optimal integration of multiple machine learning techniques

Huang, Cheng‐Han; Chang, Ming-Jui; Lin, Gwo‐Fong; Wu, Ming‐Chang; Wang, Po-Hsiang

doi:10.1016/j.ejrh.2021.100804

Cited by 13 publications

(7 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to match the operational forecasting schedule of NIER, Created a recurrent neural network (RNN) model that forecasts PM2.5 concentrations up to 2 days at 6-h intervals [6], [7], [8]. The RNN model is quick enough for real-time operational forecasting, and depending on the forecast lead time [9], [10], [11], the RNN-based prediction accuracy ranges from 74 to 80% (11% to 18% better than the CMAQ-based forecast). The CMAQ-based PM2.5 estimations are improved by the RNN model, but it is impossible to determine the exact steps that each input variable took to influence the prediction or the relative weight that each input variable had [12], [13].…”

Section: Levels In Recognition Of the Severely Harmful Impactsmentioning

confidence: 99%

The Use of TensorFlow in Analyzing Air Quality Artificial Intelligence Predictions PM2.5

Rahardja¹,

Aini²,

Sunarya³

et al. 2022

att

View full text Add to dashboard Cite

Artificial intelligence techniques to forecasts based on the Community Multiscale Air Quality (PM2.5) operational model can be known using TensorFlow. TensorFlow was used in this study to assess the scores of the Recurrent Neural Networks (RNN) input variables on the 6-hour forecast for July-October 2022. The relevance scores for the one- and two-day forecasts are represented by the sum of the relevance scores across the target prediction timeframe 2–5 and 4–7 previous time steps. The initial selection of input variables was based on their correlation coefficient with the measured PM2.5 concentration. Still, the order of contribution of the input variables measured by TensorFlow was different from the order of their correlation coefficients, which indicated an inconsistency between the linear and nonlinear variables of the method. It was found that the retraining of the RNN model using a subset of variables with a high relevance score resulted in a predictive ability similar to the initial set of input variables.

show abstract

Section: Levels In Recognition Of the Severely Harmful Impactsmentioning

confidence: 99%

The Use of TensorFlow in Analyzing Air Quality Artificial Intelligence Predictions PM2.5

Rahardja¹,

Aini²,

Sunarya³

et al. 2022

att

View full text Add to dashboard Cite

show abstract

“…In these works, complex patterns between the SSC and relevant environmental factors are extracted through either machine learning approaches (Cigizoglu, 2004), deep learning with more hidden layers (Hamshaw et al, 2018;Ying et al, 2020) or statistical modelling approaches (Kuhnert et al, 2012). Machine learning approaches mainly approximate a mapping for SSC pattern partitioning through a combination of nonlinear transformations or kernels, such as artificial neural networks (Kabiri-Samani et al, 2011;Khan et al, 2019;James et al, 2018;Teixeira et al, 2020), support vector machines (Kişi, 2012), and tree regression (Malik et al, 2017;Huang et al, 2021), and have achieved better prediction accuracy than physical models, but the physical interpretation is relatively limited. Statistical modelling approaches have become more popular for SSC forecasting in recent years (Wang et al, 2011;Wang and Tian, 2013;Liang et al, 2017;Wang et al, 2015), benefiting from their better interpretation of physical processes than machine learning approaches.…”

Section: Introductionmentioning

confidence: 99%

A physics-informed statistical learning framework for forecasting local suspended sediment concentrations in marine environment

Zhang

Wang

et al. 2022

Water Research

View full text Add to dashboard Cite

“…The consequences will be particularly severe in catchments with dammed reservoirs, trapping most of the sediment particles. Sediment entrapment has long been a major factor in reducing the capacity and the deterioration of water quality in many reservoirs around the world, endangering their durability, human health, and safety (Zarfl and Lucía 2018;Bilali et al 2020;Huang et al 2021). Multiple studies have shown today, as a result of the increase in sediment loads transported by rivers into reservoirs, that each year about 1% of their total capacity in the world is lost (Jain 2005;Rahmani et al 2018).…”

Section: Introductionmentioning

confidence: 99%

From the source to the reservoir and beyond—tracking sediment particles with modeling tools under climate change predictions (Carpathian Mts.)

et al. 2022

View full text Add to dashboard Cite

Purpose The study tracks spatial and temporal distribution of sediment particles from their source to the deposition area in a dammed reservoir. This is particularly important due to the predicted future climate changes, which will increase the severity of problems with sediment transport, especially in catchments prone to erosion. Methods Analyses were performed with a monthly step for two mineral and one mineral/organic sediment fractions delivered from the Carpathian Mts. catchment (Raba River) to the drinking water reservoir (Dobczyce) by combining SWAT (Soil and Water Assessment Tool), and AdH/PTM (Adaptive Hydraulics Model/Particle Tracking Model) modules on the digital platform—Macromodel DNS (Discharge Nutrient Sea). To take into account future changes in this catchment, a variant scenario analysis including RCP (representative concentration pathways) 4.5 and 8.5, and land use change forecasts, was performed. Results The differences between the two analyzed hydrological units (catchment and reservoir) have been highlighted and showed a large variability of the sediment load between months. The predicted climate changes will cause a significant increase of mineral fraction loads (silt and clay) during months with high flows. Due to the location and natural arrangement of the reservoir, silt particles will mainly affect faster loss of the first two reservoir zones capacities. Conclusions The increased mobility of finer particles (clay) in the reservoir may be more problematic in the future, mainly due to their binding pollutant properties, and the possible negative impact on drinking water abstraction from the last reservoir zone. Moreover, the study shows that the monthly approach to forecasting the impact of climate change on sediment loads in the reservoir is recommended, instead of a seasonal one.

show abstract

Real-time forecasting of suspended sediment concentrations in reservoirs by the optimal integration of multiple machine learning techniques

Cited by 13 publications

References 41 publications

The Use of TensorFlow in Analyzing Air Quality Artificial Intelligence Predictions PM2.5

The Use of TensorFlow in Analyzing Air Quality Artificial Intelligence Predictions PM2.5

A physics-informed statistical learning framework for forecasting local suspended sediment concentrations in marine environment

From the source to the reservoir and beyond—tracking sediment particles with modeling tools under climate change predictions (Carpathian Mts.)

Contact Info

Product

Resources

About