Daily River Water Temperature Research Articles

Bat algorithm optimised extreme learning machine (Bat‐ELM): A novel approach for daily river water temperature modelling

AbstractHere, the capability of the Bat algorithm optimised extreme learning machines ELM (Bat‐ELM) is demonstrated for river water temperature (Tw) modelling in the Orda River, Poland. Results using the multilayer perceptron neural network (MLPNN), the classification and regression Tree (CART) and the multiple linear regression (MLR) models were presented for comparison. The models were developed according to two scenarios: (1) using air temperature (Ta) as input for predicting Tw, and (2) using Ta and the periodicity (i.e., day, month and year number). River Tw calibration and validation results derived from air temperature and the periodicity show its potential application. The Bat‐ELM accurately predicts the Tw and surpassed all other models with coefficient of correlation (R) values ranging within the limits of 0.973 to 0.981, and the Nash‐Sutcliffe efficiency (NSE) values will fall within the interval of 0.947 to 0.963. Findings from this research also highlight the robustness of the Bat‐ELM using the periodicity by enhancing its ability to estimate river Tw.

Least square support vector machine-based variational mode decomposition: a new hybrid model for daily river water temperature modeling.

Machines learning models have recently been proposed for predicting rivers water temperature (Tw) using only air temperature (Ta). The proposed models relied on a nonlinear relationship between the Tw and Ta and they have proven to be robust modelling tools. The main motivation for this study was to evaluate how the variational mode decomposition (VMD) contributed to the improvement of machines learning performances for river Tw modelling. Measured data collected at five stations located in Poland from 1987 to 2014 were acquired and used for the analysis. Six machines learning models were used and compared namely, K-nearest neighbor's regression (KNNR), least square support vector machine (LSSVM), generalized regression neural network (GRNN), cascade correlation artificial neural networks (CCNN), relevance vector machine (RVM), and locally weighted polynomials regression (LWPR). The six models were developed according to three scenarios. First, the models were calibrated using only the Ta as input and obtained results show that the models were able to predict consistently water temperature, showing a high determination coefficient (R2) and Nash-Sutcliffe efficiency (NSE) with values near or above 0.910 and 0.915, respectively, and in overall the six models worked equally without clear superiority of one above another. Second, the air temperature was combined with the periodicity (i.e., day, month and year number) as input variable and a significant improvement was achieved. Both models show their ability to accurately predict river Tw with an overall accuracy of 0.956 for R2 and 0.955 for NSE values, but the LSSVM2 have some advantages such as a small errors metrics, and high fitting capabilities and it slightly surpasses the others models. Thirdly, air temperature was decomposed into several intrinsic mode functions (IMF) using the VMD method and the performances of the models were evaluated. The VMD parameters appeared to cause much influence on the prediction accuracy, exhibiting an improvement of about 40.50% and 39.12% in terms of RMSE and MAE between the first and the third scenarios, however, some models, i.e., GRNN and KNNR have not benefited from the VMD. This research has demonstrated the high capability of the VMD algorithm as a preprocessing approach in improving the accuracies of the machine learning models for river water temperature prediction.

River water temperatures are higher during lower flows after accounting for meteorological variability

AbstractRiver water temperature is known to be important for water quality and ecosystem processes. We quantified the degree to which lower river flows are associated with warmer river water, after accounting for seasonality and meteorological variability. We applied a systematic methodology to analyse observed mean daily river water temperature and mean daily river flow from 47 sites draining mountain, hill, and lowland catchments across the Canterbury region of Aotearoa New Zealand. We fitted regression models to remove seasonal patterns from all variables, then removed correlations between water temperature and each of three meteorological variables (solar radiation, air temperature, and earth temperature) before quantifying water temperature‐river flow relationships. Strong seasonal patterns were present in water temperature and each meteorological variable across all sites. Many sites also showed strong seasonal patterns in river flows. We demonstrated that seasonal patterns must be accounted for before day‐to‐day associations between water temperature and meteorological variables or river flow can be characterised. Higher water temperatures were associated with lower flows for 46 of 47 sites, even after having accounted for seasonality and associations with each meteorological variable. Increases in water temperature associated with a hypothetical reduction in river flow from the median to the fifth percentile varied with the site but were 0.5°C on average. This finding has important implications for river flow management because it indicates that increased river water temperatures would accompany reduced river flows regardless of site or catchment characteristics.

Daily River Water Temperature Prediction: A Comparison between Neural Network and Stochastic Techniques

The temperature of river water (TRW) is an important factor in river ecosystem predictions. This study aims to compare two different types of numerical model for predicting daily TRW in the Warta River basin in Poland. The implemented models were of the stochastic type—Autoregressive (AR), Moving Average (MA), Autoregressive Moving Average (ARMA) and Autoregressive Integrated Moving Average (ARIMA)—and the artificial intelligence (AI) type—Adaptive Neuro Fuzzy Inference System (ANFIS), Radial Basis Function (RBF) and Group Method of Data Handling (GMDH). The ANFIS and RBF models had the most fitted outputs and the AR, ARMA and ARIMA patterns were the most accurate ones. The results showed that both of the model types can significantly present suitable predictions. The stochastic models have somewhat less error with respect to both the highest and lowest TRW deciles than the AIs and were found to be better for prediction studies, with the GMDH complex model in some cases reaching Root Mean Square Error (RMSE) = 0.619 °C and Nash-Sutcliff coefficient (NS) = 0.992, while the AR(2) simple linear model with just two inputs was partially able to achieve better results (RMSE = 0.606 °C and NS = 0.994). Due to these promising outcomes, it is suggested that this work be extended to other catchment areas to extend and generalize the results.

Investigation on River Thermal Regime under Dam Influence by Integrating Remote Sensing and Water Temperature Model

River water temperature (RWT), a primary parameter for hydrological and ecological processes, is influenced by both climate change and anthropogenic intervention. Studies on such influences have been severely restricted due to the scarcity of river temperature data. This paper proposed a three-stage method to obtain long-term daily water temperature for rivers and river-type reservoirs by integrating remote sensing technique and river water temperature modelling. The proposed three-stage method was applied to the Three Gorges Reservoir (TGR) and validated against in situ measured RWTs in the two study sites, Cuntan and Huanglingmiao. The result showed improvements in the method: the quadrate window selection and RWT correction jointly reduce RMSE from 1.8 to 0.9 °C in Cuntan and from 2.1 to 1.2 °C in Huanglingmiao. As a whole, the estimated daily RWT has a consistent RMSE of 1.1–1.9 °C. Meanwhile, by analysing the Landsat-derived daily RWT, we demonstrated that the TGR had a significant impact on the outflow’s thermal regime. At the downstream reach of TGR, an apparent increase in RWT in the cold season and interannual thermal regime delay compared to inflow were found with the increasing water level after the dam construction. All the results and analyses indicate that the proposed three-stage method could be applied to obtain long time series of daily RWT and provide a promising approach to qualitatively analyse RWT variation in the poorly gauged catchment for river water quality monitoring and management.

Derivation of consistent, continuous daily river temperature data series by combining remote sensing and water temperature models

Scarcity of water temperature data in rivers may limit a diversity of studies considering this property, which regulates many physical, chemical, and biological processes. We present a robust method to generate a consistent, continuous daily river water temperature (RWT) data series for medium and large rivers using the combined techniques of remote sensing and water temperature modelling. In order to validate our approach, we divided this study into two parts: (i) we evaluated methods to derive RWT from Landsat 7 ETM+ and Landsat 8 TIRS imagery; and (ii) we evaluated the calibration and validation of river temperature models, using these data, to generate the continuous RWT data series. A 1.2 km section of the White River located near Hazleton, IN, USA, was selected to assess this method mainly due to river width and data availability. We tested three methods to retrieve RWT from Landsat 7 and four from Landsat 8, and we also applied a simple thermal sharpening technique. For Landsat 7, the methods showed bias and RMSE of 0.01–0.46 °C and 1.32–1.84 °C, while for Landsat 8, the methods showed bias and RMSE of 0.08–1.27 °C and 1.74–2.17 °C, and in both cases, the best results were found applying the radiative transfer equation with NASA's Atmospheric Correction Parameter Calculator. For the second part of the validation process, we compared a stochastic model and a hybrid model, air2stream, using as input two datasets: the RWT data derived from Landsat 7 only, and a combined dataset of both Landsat 7 and 8 derived RWT. The air2stream model outperformed the stochastic model when calibrated with Landsat 7 data only, with RMSE of 1.83 °C, but both models showed similar results when calibrated with the combined Landsat data, when air2stream showed RMSE of 1.58 °C. Due to its physical basis, better calibration procedure, and higher consistency, air2stream was considered the best model for deriving the continuous RWT data series. When compared to the measured daily mean RWT data, there was no observed tendency in under or overestimating the RWT in low or high temperature conditions by the modelled series. While further tests are needed in order to evaluate if our approach can be applied to analyse past behaviour and present trends, and the impacts of climate change on the temperature of rivers, the consistent results indicate that this approach has the potential to be applied in rivers with no measured temperature data, for example, in the spatial modelling of longitudinal profiles of rivers and the modelling of tributary river temperatures.

Assessing the performance of a suite of machine learning models for daily river water temperature prediction.

In this study, different versions of feedforward neural network (FFNN), Gaussian process regression (GPR), and decision tree (DT) models were developed to estimate daily river water temperature using air temperature (Ta), flow discharge (Q), and the day of year (DOY) as predictors. The proposed models were assessed using observed data from eight river stations, and modelling results were compared with the air2stream model. Model performances were evaluated using four indicators in this study: the coefficient of correlation (R), the Willmott index of agreement (d), the root mean squared error (RMSE), and the mean absolute error (MAE). Results indicated that the three machine learning models had similar performance when only Ta was used as the predictor. When the day of year was included as model input, the performances of the three machine learning models dramatically improved. Including flow discharge instead of day of year, as an additional predictor, provided a lower gain in model accuracy, thereby showing the relatively minor role of flow discharge in river water temperature prediction. However, an increase in the relative importance of flow discharge was noticed for stations with high altitude catchments (Rhône, Dischmabach and Cedar) which are influenced by cold water releases from hydropower or snow melting, suggesting the dependence of the role of flow discharge on the hydrological characteristics of such rivers. The air2stream model outperformed the three machine learning models for most of the studied rivers except for the cases where including flow discharge as a predictor provided the highest benefits. The DT model outperformed the FFNN and GPR models in the calibration phase, however in the validation phase, its performance slightly decreased. In general, the FFNN model performed slightly better than GPR model. In summary, the overall modelling results showed that the three machine learning models performed well for river water temperature modelling.

Modeling daily water temperature for rivers: comparison between adaptive neuro-fuzzy inference systems and artificial neural networks models.

River water temperature is a key control of many physical and bio-chemical processes in river systems, which theoretically depends on multiple factors. Here, four different machine learning models, including multilayer perceptron neural network models (MLPNN), adaptive neuro-fuzzy inference systems (ANFIS) with fuzzy c-mean clustering algorithm (ANFIS_FC), ANFIS with grid partition method (ANFIS_GP), and ANFIS with subtractive clustering method (ANFIS_SC), were implemented to simulate daily river water temperature, using air temperature (Ta), river flow discharge (Q), and the components of the Gregorian calendar (CGC) as predictors. The proposed models were tested in various river systems characterized by different hydrological conditions. Results showed that including the three inputs as predictors (Ta, Q, and the CGC) yielded the best accuracy among all the developed models. In particular, model performance improved considerably compared to the case where only Ta is used as predictor, which is the typical approach of most of previous machine learning applications. Additionally, it was found that Q played a relevant role mainly in snow-fed and regulated rivers with higher-altitude hydropower reservoirs, while it improved to a lower extent model performance in lowland rivers. In the validation phase, the MLPNN model was generally the one providing the highest performances, although in some river stations ANFIS_FC and ANFIS_GP were slightly more accurate. Overall, the results indicated that the machine learning models developed in this study can be effectively used for river water temperature simulation.

Climatic and dam-induced impacts on river water temperature: Assessment and management implications

In a changing climate with a warming trend in air temperature, river water temperature increases as a result of heat exchange with the atmosphere. Moreover, of the different types of anthropogenic activity impacting rivers, the construction of dams appears to have multi-dimensional effects on the river environment, and it especially affects the thermal condition of rivers. The aim of the study is to identify and assess the impact of these two distinct sources of water temperature distortion in relation to the natural thermal conditions of rivers. In the study, linear trend analysis and a complex wavelet transform are used. The analysis focuses on daily river water temperature data for time periods before and after the construction of selected reservoirs in the Polish Carpathians, and on daily air temperature data for neighboring meteorological stations. Three rivers were selected for analysis: (1) Dunajec River, 22km downstream of the Czorsztyn–Sromowce Wyżne reservoir complex, (2) Ropa River, 16km downstream of the Klimkówka Reservoir, and (3) Wisłok River, 33km downstream of the Besko Reservoir. Research has shown that the significant increasing trends identified for water temperature are weaker than analogous trends in air temperature. The Czorsztyn–Sromowce Wyżne and Klimkówka reservoirs appear to exert considerable influence on natural air–water temperature synchronization, because the phase difference increases 5-fold in comparison with conditions prior to the construction of the reservoirs. The weakening of the natural air–water temperature synchrony implies diminished impact of air temperature on stream water temperature. However, this creates an opportunity for preparing appropriate management practices mitigating an increasing temperature trend in order to shape more favorable (natural) thermal conditions for native aquatic biota in impounded rivers.

A regional neural network ensemble for predicting mean daily river water temperature

Summary Water temperature is a fundamental property of river habitat and often a key aspect of river resource management, but measurements to characterize thermal regimes are not available for most streams and rivers. As such, we developed an artificial neural network (ANN) ensemble model to predict mean daily water temperature in 197,402 individual stream reaches during the warm season (May–October) throughout the native range of brook trout Salvelinus fontinalis in the eastern U.S. We compared four models with different groups of predictors to determine how well water temperature could be predicted by climatic, landform, and land cover attributes, and used the median prediction from an ensemble of 100 ANNs as our final prediction for each model. The final model included air temperature, landform attributes and forested land cover and predicted mean daily water temperatures with moderate accuracy as determined by root mean squared error (RMSE) at 886 training sites with data from 1980 to 2009 (RMSE = 1.91 °C). Based on validation at 96 sites (RMSE = 1.82) and separately for data from 2010 (RMSE = 1.93), a year with relatively warmer conditions, the model was able to generalize to new stream reaches and years. The most important predictors were mean daily air temperature, prior 7 day mean air temperature, and network catchment area according to sensitivity analyses. Forest land cover at both riparian and catchment extents had relatively weak but clear negative effects. Predicted daily water temperature averaged for the month of July matched expected spatial trends with cooler temperatures in headwaters and at higher elevations and latitudes. Our ANN ensemble is unique in predicting daily temperatures throughout a large region, while other regional efforts have predicted at relatively coarse time steps. The model may prove a useful tool for predicting water temperatures in sampled and unsampled rivers under current conditions and future projections of climate and land use changes, thereby providing information that is valuable to management of river ecosystems and biota such as brook trout.

Implementation of Artificial Neural Networks in Modeling the Water-Air Temperature Relationship of the River Drava

Water temperature directly affects the physical, biological and chemical characteristics of the river and determines the fitness and life of all aquatic organisms. It has direct and indirect effects on nearly all aspects of stream ecology. Accurately estimating water temperature is a complex problem. The purpose of this article is to analyze the relationship between the air and water temperature of the River Drava by constructing an artificial neural network (ANN) model and choosing appropriate network architectures for the River Drava’s daily river water temperature as well as demonstrating its application in improving the interpretation of the results. A linear regression model, as well as a stochastic model are also constructed and compared to ANN models consisting of a multilayer perceptron neural network and a radial basis function network. The results indicate that the ANN models are much better models and that ANNs are powerful tools that can be used for the estimation of daily mean river temperature.

Daily river water temperature forecast model with a k‐nearest neighbour approach

AbstractWater temperature is a key abiotic variable that modulates both water chemistry and aquatic life in rivers and streams. For this reason, numerous water temperature models have been developed in recent years. In this paper, a k‐nearest neighbour model (KNN) is proposed and validated to simulate and eventually produce a one‐day forecast of mean water temperature on the Moisie River, a watercourse with an important salmon population in eastern Canada. Numerous KNN model configurations were compared by selecting different attributes and testing different weight combinations for neighbours. It was found that the best model uses attributes that include water temperature from the two previous days and an indicator of seasonality (day of the year) to select nearest neighbours. Three neighbours were used to calculate the estimated temperature, and the weighting combination that yielded the best results was an equal weight on all three nearest neighbours. This nonparametric model provided lower Root Mean Square Errors (RMSE = 1·57 °C), Higher Nash coefficient (NTD = 0·93) and lower Relative Bias (RB = − 1·5%) than a nonlinear regression model (RMSE = 2·45 °C, NTD = 0·83, RB = − 3%). The k‐nearest neighbour model appears to be a promising tool to simulate of forecast water temperature where long time series are available. Copyright © 2011 John Wiley & Sons, Ltd.