Water Level Prediction through Hybrid SARIMA and ANN Models Based on Time Series Analysis: Red Hills Reservoir Case Study

Abstract

Reservoir water level (RWL) prediction has become a challenging task due to spatio-temporal changes in climatic conditions and complicated physical process. The Red Hills Reservoir (RHR) is an important source of drinking and irrigation water supply in Thiruvallur district, Tamil Nadu, India, also expected to be converted into the other productive services in the future. However, climate change in the region is expected to have consequences over the RHR’s future prospects. As a result, accurate and reliable prediction of the RWL is crucial to develop an appropriate water release mechanism of RHR to satisfy the population’s water demand. In the current study, time series modelling technique was adopted for the RWL prediction in RHR using Box–Jenkins autoregressive seasonal autoregressive integrated moving average (SARIMA) and artificial neural network (ANN) hybrid models. In this research, the SARIMA model was obtained as SARIMA (0, 0, 1) (0, 3, 2)12 but the residual of the SARIMA model could not meet the autocorrelation requirement of the modelling approach. In order to overcome this weakness of the SARIMA model, a new SARIMA–ANN hybrid time series model was developed and demonstrated in this study. The average monthly RWL data from January 2004 to November 2020 was used for developing and testing the models. Several model assessment criteria were used to evaluate the performance of each model. The findings showed that the SARIMA–ANN hybrid model outperformed the remaining models considering all performance criteria for reservoir RWL prediction. Thus, this study conclusively proves that the SARIMA–ANN hybrid model could be a viable option for the accurate prediction of reservoir water level.

Time-Series Analysis for the Number of Foot and Mouth Disease Outbreak Episodes in Cattle Farms in Thailand Using Data from 2010–2020

“…The auto.arima function utilized the Hyndman–Khandarkar algorithm for automatic autoregressive integrated moving average (ARMA) modeling. As previously described, this algorithm performed various steps in the model selection procedure [ 29 , 43 ]. Based on such procedures, the final model was defined as the model with the lowest AIC.…”

Comparing three types of data-driven models for monthly evapotranspiration prediction under heterogeneous climatic conditions

“…Also, evaluating the agricultural drought status, which is done by famous indicators such as standardized precipitation-evapotranspiration index (SPEI) and Palmer drought severity index (PDSI), directly requires the monthly scale ET0 rate of the region. Data-driven models like stochastic and artificial intelligence methods are efficient approaches that have shown good performance in modeling and predicting hydrometeorological variables in recent years (Essam et al 6 ; Dehghanisanij et al 7 ; Elbeltagi et al 8 ; Azad et al 9 ; Zhang et al 10 ; Zarei et al 11 ; Graf and Aghelpour 12 ; Chen et al 13 ). In ET0 cases, Karbasi 14 have used AIs for ET0 forecasting in 1, 2, 3, 7, 10, 14, 18, 24, and 30 days lead times.…”

Viable forecasting monthly weather data using time series methods

“…The validity of the model may be checked through residuals. It was concluded that the ARIMA model has more applications in predicting floods, harvesting and crop management Azad et al (2022). revealed that the SARIMA-ANN hybrid model outperformed the remaining models seeing all concert criteria for reservoir RWL forecast, furthermore it was verifies that the SARIMA-ANN hybrid model can be a viable choice for the correct forecast of reservoir water level.…”