A Comparative Analysis of Data-Driven Models (SVR, ANFIS, and ANNs) for Daily Karst Spring Discharge Prediction

Transformer Versus LSTM: A Comparison of Deep Learning Models for Karst Spring Discharge Forecasting

“…On top of that, groundwater level forecasting has been explored in multiple studies (Daliakopoulos et al, 2005;Nayak et al, 2006;Wunsch et al, 2021). For the prediction of karst discharges, machine learning models have only recently been used (An et al, 2020;Cheng et al, 2021;Jeannin et al, 2021;Rahbar et al, 2022;Xu et al, 2022). For instance, Rahbar et al (2022) used several shallow and deep learning approaches, for example, an MLP combined with a hybrid gamma test-genetic algorithm approach, achieving high performances in the discharge forecasts for some of the investigated karst springs.…”

“…For the prediction of karst discharges, machine learning models have only recently been used (An et al, 2020;Cheng et al, 2021;Jeannin et al, 2021;Rahbar et al, 2022;Xu et al, 2022). For instance, Rahbar et al (2022) used several shallow and deep learning approaches, for example, an MLP combined with a hybrid gamma test-genetic algorithm approach, achieving high performances in the discharge forecasts for some of the investigated karst springs. Generally, deep learning approaches (e.g., MLP and Long Short-Term Memory (LSTM)) showed high performances for the prediction of karst spring discharges (Cheng et al, 2021;Rahbar et al, 2022).…”

“…Pipe flow models are based on a conceptualization of different geological units of the aquifer such as the epikarst, conduits, and the permeable matrix (Çallı et al., 2022; Mazzilli et al., 2017). Data‐driven model approaches include machine learning models, which are statistical models that learn certain tasks through training with data (Rahbar et al., 2022).…”

“…(2022). The state‐of‐the‐art deep learning approaches applied in hydrology so far include the multi‐layer perceptron (MLP), convolutional neural networks (CNN), and recurrent neural networks (RNN), such as LSTMs (Rahbar et al., 2022; Wunsch et al., 2021, 2022). LSTMs are neural networks that were developed to overcome shortcomings of previous RNNs, which have difficulties learning long‐term dependencies.…”

“…On top of that, groundwater level forecasting has been explored in multiple studies (Daliakopoulos et al., 2005; Nayak et al., 2006; Wunsch et al., 2021). For the prediction of karst discharges, machine learning models have only recently been used (An et al., 2020; Cheng et al., 2021; Jeannin et al., 2021; Rahbar et al., 2022; Xu et al., 2022). For instance, Rahbar et al.…”