Hybrid artificial intelligence-time series models for monthly streamflow modeling

Modelling daily soil temperature at different depths via the classical and hybrid models

“…Likewise, Tabari et al (2015) reported better performance of an AI-based model including the MLP using the antecedent ST data for modelling the ST at deeper layers than the surface layers. Furthermore, the results of the current study verify the outcomes of previous studies such as Mehdizadeh (2018b), Mehdizadeh and Kozekalani Sales (2018), Mehdizadeh et al (2017cMehdizadeh et al ( , 2018bMehdizadeh et al ( , 2019, and Fathian et al (2019). The authors developed different types of the hybrid models through hybridization of the different time-series-and AI-based models for improving the modelling efficiency of classical models in modelling hydrological variables.…”

A Comparision Using Statistical and Machine Learning Methods for Streamflow Time Series

“…Thus assisting hydrologists, water resource planners, public institutions, hydroelectric companies, and policy-makers in the effective management of water and the preservation of this natural resource. Therefore, in recent decades the development of new approaches, together with the improvement of available ones, has received much attention from hydrologists around the world to model and accurately estimate flow processes to address the problems mentioned above (MOHAMMADI et al, 2006;MEHDIZADEH et al, 2019;WU et al, 2010;NIU et al,2020).…”

“…However, the main disadvantage of these physics-based models is the presence of numerous region-dependent parameters that require calibration and validation, making it difficult to optimize them (FARFÁN et al, 2020). Besides, there can be enormous uncertainty in the models predictions, due to uncertainty in the input data, because inaccessibility to the various sources of information, limits the use of these models (MEHDIZADEH et al, 2019;WAGENA et al, 2020). In this regard, many researchers have studied the adoption of empirical models to predict future flow based on a long collection of historical flow records (WU et al, 2010;NIU et al,2019;TIKHAMARINE et al, 2020).…”

“…In this regard, many researchers have studied the adoption of empirical models to predict future flow based on a long collection of historical flow records (WU et al, 2010;NIU et al,2019;TIKHAMARINE et al, 2020). These empirical models are widely used and easier to apply to flow modeling as they do not require complex data sets or physical requirements (MEHDIZADEH et al, 2019). Empirical models include: statistics models, the most recognized for hydrological simulations, exponential smoothing methods (CHATFIELD & YAR, 1988) e Box & Jenkins (BOX et al, 2008) as ARMA (autoregressive moving average) and ARIMA (integrated autoregressive moving average).…”

“…Empirical models include: statistics models, the most recognized for hydrological simulations, exponential smoothing methods (CHATFIELD & YAR, 1988) e Box & Jenkins (BOX et al, 2008) as ARMA (autoregressive moving average) and ARIMA (integrated autoregressive moving average). And the datadriven models (machine learning algorithms), among the most used by hydrological scientists are the artificial neural network (ANN), for its learning ability, generalizability and satisfactory results in accurate flow forecast estimates (MONTANARI et al, 2000;SHAHWAN & ODENING, 2007;MEHDIZADEH et al, 2019).…”