Kim et al. (2025) Monthly Temperature Prediction in the Han River Basin, South Korea, Using Long Short-Term Memory (LSTM) and Multiple Linear Regression (MLR) Models

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Water
• Year: 2025
• Date: 2025-12-31
• Authors: Chul‐Gyum Kim, Jeongwoo Lee, Jeong-Eun Lee, H. Y. Kim
• DOI: 10.3390/w18010098

Research Groups

The provided text does not explicitly list specific research groups, labs, or departments involved in the study.

Short Summary

This study compares Multiple Linear Regression (MLR) and Long Short-Term Memory (LSTM) models for monthly mean temperature prediction in the Han River Basin, South Korea, finding both highly accurate but complementary, with LSTM excelling in non-linear dynamics and MLR offering greater stability and interpretability.

Objective

• To compare and evaluate the performance of a statistical model (Multiple Linear Regression) and a deep learning model (Long Short-Term Memory) for predicting monthly mean temperature in the Han River Basin, South Korea.

Study Configuration

• Spatial Scale: Han River Basin, South Korea.
• Temporal Scale: Monthly mean temperature prediction; predictor selection based on 40 years of historical data; lag times for predictors ranging from 1 to 18 months; LSTM model utilized an 18-month input sequence.

Methodology and Data

• Models used: Multiple Linear Regression (MLR), Long Short-Term Memory (LSTM).
• Data sources: Historical climate indices (dynamically selected as predictor variables) and monthly mean temperature observations.

Main Results

• Both MLR and LSTM models accurately reproduced the seasonal variability of monthly temperature, achieving high performance metrics (Nash–Sutcliffe efficiency > 0.97, Pearson’s correlation coefficient > 0.98).
• The LSTM model demonstrated slightly higher predictive skill in several periods but also exhibited larger prediction variance, reflecting its sensitivity to non-linear predictor–response relationships.
• The MLR model showed more stable predictive behavior with narrower uncertainty bounds, particularly under low signal-to-noise conditions, attributed to its structural simplicity.
• The study concludes that the two approaches are complementary: LSTM is better at capturing non-linear temporal dynamics, while MLR provides superior interpretability and robustness.

Contributions

• Provides a direct comparative evaluation of a traditional statistical model (MLR) and a deep learning model (LSTM) for medium-range temperature prediction in a specific hydrological basin.
• Highlights the complementary strengths of both model types, demonstrating that while deep learning can capture complex non-linearities, simpler statistical models offer superior stability and interpretability, especially under challenging conditions.
• Informs future research directions by suggesting advanced hybrid architectures (e.g., CNN–LSTM, Transformer-based models) and multi-model ensemble methods for enhancing prediction accuracy and reliability.

Funding

The provided paper text does not contain information about funding sources.

Citation

@article{Kim2025Monthly,
  author = {Kim, Chul‐Gyum and Lee, Jeongwoo and Lee, Jeong-Eun and Kim, H. Y.},
  title = {Monthly Temperature Prediction in the Han River Basin, South Korea, Using Long Short-Term Memory (LSTM) and Multiple Linear Regression (MLR) Models},
  journal = {Water},
  year = {2025},
  doi = {10.3390/w18010098},
  url = {https://doi.org/10.3390/w18010098}
}