A Comparison of Ensemble and Deep Learning Algorithms to Model Groundwater Levels in a Data-Scarce Aquifer of Southern Africa

Abstract

Machine learning and deep learning have demonstrated usefulness in modelling various groundwater phenomena. However, these techniques require large amounts of data to develop reliable models. In the Southern African Development Community, groundwater datasets are generally poorly developed. Hence, the question arises as to whether machine learning can be a reliable tool to support groundwater management in the data-scarce environments of Southern Africa. This study tests two machine learning algorithms, a gradient-boosted decision tree (GBDT) and a long short-term memory neural network (LSTM-NN), to model groundwater level (GWL) changes in the Shire Valley Alluvial Aquifer. Using data from two boreholes, Ngabu (sample size = 96) and Nsanje (sample size = 45), we model two predictive scenarios: (I) predicting the change in the current month’s groundwater level, and (II) predicting the change in the following month’s groundwater level. For the Ngabu borehole, GBDT achieved R2 scores of 0.19 and 0.14, while LSTM achieved R2 scores of 0.30 and 0.30, in experiments I and II, respectively. For the Nsanje borehole, GBDT achieved R2 of −0.04 and −0.21, while LSTM achieved R2 scores of 0.03 and −0.15, in experiments I and II, respectively. The results illustrate that LSTM performs better than the GBDT model, especially regarding slightly greater time series and extreme GWL changes. However, closer inspection reveals that where datasets are relatively small (e.g., Nsanje), the GBDT model may be more efficient, considering the cost required to tune, train, and test the LSTM model. Assessing the full spectrum of results, we concluded that these small sample sizes might not be sufficient to develop generalised and reliable machine learning models.

A Review of Geological and Climatic Variables in Groundwater Availability Prediction in Africa: Machine Learning Approaches

“…Gradient Boosting Decision Trees (GBDT), Extreme Gradient Boosting (XGBoost), Random Forest Regression (RFR) and Decision Trees are considered part of the tree decision family. Regression, classification, and ranking tasks respond well to these traditional machine learning methods (Gaffoor et al, 2022). GBDT is a machine learning algorithm that combines multiple decision trees to make predictions.…”

“…Gaffoor et al, (2022) employed two machine learning algorithms, gradient-boosted decision tree (GBDT) and long short-term memory neural network (LSTM-NN), to model groundwater level changes in the Shire Valley Alluvial Aquifer (Southern South-Africa). The algorithms were trained using hydro-climatic inputs and groundwater level changes from two boreholes (namely Ngabu and Nsanje).…”

A Systematic Review of Machine Learning Algorithms in Groundwater Level Simulations and Forecasting

“…The DEBi-LSTM model outperformed existing models like LSTM, bagging ensemble, and an ensemble model. Gaffoor et al (2022) analyzed groundwater level variations in the Shire Valley Alluvial Aquifer using two ML techniques: gradient-boosted decision trees (GBDT) and LSTM-NN. Input variables included soil moisture, runoff, evapotranspiration, groundwater storage anomaly, precipitation, and surface temperature.…”

Groundwater Level Prediction Using Deep Recurrent Neural Networks and Uncertainty Assessment

“…In addition, in recent decades, climate change has greatly affected all human lives (McMichael et al, Seyed Mousavi et al, 2022). Observed the climate changes such as temperature increase and changes in the trend of rainfall, relative humidity, soil moisture, as well as evaporation and transpiration have led to significant changes in groundwater sources, especially in arid and semi-arid areas (Afrifa et al, 2022;Gaffoor et al, 2022;Wunsch et al, 2022;Yifru et al, 2021). Hence, appropriate and effective models can be necessary tools to make informed decisions about the status of groundwater sources and their environmental effects.…”