Hybrid the long short-term memory with whale optimization algorithm and variational mode decomposition for monthly...

Area of Science:

• Environmental Science
• Hydrology
• Artificial Intelligence

Background:

• The sustainability of desert vegetation relies on the water balance, making accurate evapotranspiration (ET) estimation vital.
• Determining appropriate artificial vegetation types and distribution in sandy areas requires precise ET data.

Purpose of the Study:

• To develop and validate a novel hybrid model for accurate monthly ET estimation in desert environments.
• To assess the performance of the proposed model against existing methods for ET prediction.

Main Methods:

• A hybrid model combining Variational Mode Decomposition (VMD), Whale Optimization Algorithm (WOA), and Long Short-Term Memory (LSTM) networks (VMD-WOA-LSTM) was developed.
• VMD decomposed the ET time series, WOA optimized LSTM hyperparameters, and LSTM extracted nonlinear features for estimation.
• The model was applied to estimate monthly ET in the southeast Tengger Desert.

Main Results:

• The VMD-WOA-LSTM model demonstrated superior accuracy and reliability in estimating monthly ET compared to standalone LSTM and Support Vector Machine (SVM) models.
• Performance metrics confirmed the enhanced predictive capability of the VMD-WOA-LSTM approach.

Conclusions:

• The VMD-WOA-LSTM model is a highly effective tool for estimating evapotranspiration in desert regions.
• This method can serve as an essential auxiliary tool for managing artificial sand-binding vegetation and water resources in arid environments.