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

One to twelve-month-ahead forecasting of MODIS-derived Qinghai Lake area, using neuro-fuzzy system hybridized by firefly optimization

Lakes, as the main sources of surface water, are of great environmental and ecological importance and largely affect the climatic conditions of the surrounding areas. Lake area fluctuations are very effective on plant and animal biodiversity in the areas covered. Hence, accurate and reliable forecasts of ​lake area might provide the awareness of water and climate resources and the survival of various species dependent on area fluctuations. Using machine learning methods, the current study numerically predicted area fluctuations of ​China's largest lake, Qinghai, over 1 to 12 months ahead of lead time. To this end, Moderate Resolution Imaging Spectroradiometer (MODIS) sensor images were used to monitor the monthly changes in the area of ​the lake from 2000 to 2021. Predictive inputs included the MODIS-derived lake area time latency specified by the autocorrelation function. The data was divided into two periods of the train (initial 75%) and test (final 25%), and the input combinations were arranged so that the model in the test period could be used to predict 12 scenarios, including forecast horizons for the next 1 to 12 months. The adaptive neuro-fuzzy inference system (ANFIS) was utilized as a predictive model. The firefly algorithm (FA) was also used to optimize ANFIS and improve its accuracy, as a hybrid model ANFIS-FA. Based on evaluation criteria such as root mean square error (RMSE) (477-594 km2) and R2 (88-92%), the results confirmed the acceptable accuracy of the models in all forecast horizons, even long-term horizons (10 months, 11 months, and 12 months). Based on the normalized RMSE criterion (0.095-0.125), the models' performance was reported to be appropriate. Furthermore, the firefly algorithm improved the prediction accuracy of the ANFIS model by an average of 16.9%. In the inter-month survey, the models had fewer forecast errors in the dry months (February-March) than in the wet months (October-November). Using the current method can provide remarkable information about the future state of lakes, which is very important for managers and planners of water resources, environment, and natural ecosystems. According to the results, the current approach is satisfactory in predicting MODIS-derived fluctuations of Qinghai Lake area and has research value for other lakes.

Coupling ANFIS with ant colony optimization (ACO) algorithm for 1-, 2-, and 3-days ahead forecasting of daily streamflow, a case study in Poland

Finding an efficient and reliable streamflow forecasting model has always been an important challenge for managers and planners of freshwater resources. The current study, based on an adaptive neuro-fuzzy inference system (ANFIS) model, was designed to predict the Warta river (Poland) streamflow for 1 day, 2 days, and 3 days ahead for a data set from the period of 1993-2013. The ANFIS was additionally combined with the ant colony optimization (ACO) algorithm and employed as a meta-heuristic ANFIS-ACO model, which is a novelty in streamflow prediction studies. The investigations showed that on a daily scale, precipitation had a very weak and insignificant effect on the river's flow variation, so it was not considered as a predictor input. The predictor inputs were selected by the autocorrelation function from among the daily streamflow time lags for all stations. The predictions were evaluated with the actual streamflow data, using such criteria as root mean square error (RMSE), normalized RMSE (NRMSE), and R². According to the NRMSE values, which ranged between 0.016-0.006, 0.030-0.013, and 0.038-0.020 for the 1-day, 2-day, and 3-day lead times, respectively, all predictions were classified as excellent in terms of accuracy (prediction quality). The best RMSE value was 1.551 m³/s and the highest R² value was equal to 0.998, forecast for 1-day lead time. The combination of ANFIS with the ACO algorithm enabled to significantly improve streamflow prediction. The use of this coupling can averagely increase the prediction accuracies of ANFIS by 12.1%, 12.91%, and 13.66%, for 1-day, 2-day, and 3-day lead times, respectively. The current satisfactory results suggest that the employed hybrid approach could be successfully applied for daily streamflow prediction in other catchment areas.

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 (T_w) using only air temperature (T_a). The proposed models relied on a nonlinear relationship between the T_w and T_a 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 T_w 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 T_a 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 T_w 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.

Comparing linear and non-linear data-driven approaches in monthly river flow prediction, based on the models SARIMA, LSSVM, ANFIS, and GMDH

River flow variations directly affect the hydro-climatological, environmental, and ecological characteristics of a region. Therefore, an accurate prediction of river flow can critically be important for water managers and planners. The present study aims to compare different data-driven models in predicting monthly flow. Two river catchments located in the Guilan province in Iran, where rivers play an essential role in agricultural productions (mainly rice), are studied. The monthly river flow dataset was provided by Guilan Regional Water Authority during 1986-2015. The models are derived from two different numerical types of stochastic and machine learning (ML) models. The stochastic model is seasonal autoregressive integrated moving average (SARIMA), and the MLs are least square support vector machine (LSSVM), adaptive neuro-fuzzy inference system (ANFIS), and group method of data handling (GMDH). The inputs were selected by autocorrelation and partial autocorrelation functions (ACF and PACF) from the flow rates of the previous months. The data was divided into 75% of training and 25% of testing phases, and then the mentioned models were implemented. Predictions were evaluated by the criteria of root mean square error (RMSE), normalized RMSE (NRMSE), and Nash Sutcliff (NS) coefficient. According to the calculated values of different criteria during the test phase, RMSE = 1.138 cms, NRMSE = 0.109, and NS = 0.826, it can be concluded that the SARIMA model was superior to its ML competitors. Among the ML models, GMDH had the best performance (by RMSE = 1.290 cms, NRMSE = 0.124, and NS = 0.777) because it has more optimization parameters and sample space for network make-up. The models were also evaluated in hydrological drought conditions of both rivers. It was resulted that the rivers' flow can be well predicted in drought conditions by using these models, especially the SARIMA stochastic model. According to the NRMSE values (ranged between 0.1 and 0.2), the accuracy of predictions is evaluated in the appropriate range, and the present study shows promising results of the current approaches. Consequently, a comparison between the performance of linear stochastic models and complex black-box MLs, reveals that linear stochastic models are more suitable for the current region's monthly river flow prediction.