A Hybrid Transformer-CNN Model for Interpolating Meteorological Data on the Tibetan Plateau

Atmosphere, Journal Year: 2025, Volume and Issue: 16(4), P. 431 - 431

Published: April 8, 2025

High-quality observational data play a crucial role in deepening the investigation of Tibetan Plateau’s influence on Asian climate. This study employs eight machine learning models (support vector regression (SVR), k-nearest neighbors (KNN), extreme gradient boosting (XGBoost), random forest (RF), long short-term memory (LSTM), gated recurrent unit (GRU), Transformer, and Transformer–convolutional neural network (Transformer-CNN)) to interpolate missing surface net radiation (Rn), soil temperature (Ts), water content (SWC), air (Ta), relative humidity (RH), wind speed (WS) from QOMS observation site. The covers period 1 January 2007 through 31 December 2016. A comparative evaluation these shows that Transformer-CNN model consistently outperforms other terms prediction accuracy. On test dataset, coefficients determination for interpolated results Ta, RH, WS, SWC, Ts, Rn were 0.97, 0.92, 0.79, 0.93, 0.98, respectively. Secondly, was then applied generate complete meteorological dataset full period. time series analysis this reveals statistically significant trends over past decade: (Ta) increased by 0.60 °C (p = 0.022) (Ts) 1.85 1.37 × 10−5). Meanwhile, (WS), (Rn) declined 0.42 m/s 1.18 10−12), 1.24% < 0.001), 9.21 W/m2 8.81 10−6),

Language: Английский

Climate change prediction in Saudi Arabia using a CNN GRU LSTM hybrid deep learning model in al Qassim region

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: May 10, 2025

Climate change, which causes long-term temperature and weather changes, threatens natural ecosystems cities. It has worldwide economic consequences. change trends up to 2050 are predicted using the hybrid model that consists of Convolutional Neural Network-Gated Recurrent Unit-Long Short-Term Memory (CNN-GRU-LSTM), a unique deep learning architecture. With focus on Al-Qassim Region, Saudi Arabia, assesses temperature, air dew point, visibility distance, atmospheric sea-level pressure. We used Synthetic Minority Over-sampling Technique for Regression with Gaussian Noise (SMOGN) reduce dataset imbalance. The CNN-GRU-LSTM was compared 5 classic regression models: DTR, RFR, ETR, BRR, K-Nearest Neighbors. Five main measures were evaluate performance: MSE, MAE, MedAE, RMSE, R². After Min-Max normalization, split into training (70%), validation (15%), testing (15%) sets. paper shows beats standard methods in all four climatic scenarios, R² values 99.62%, 99.15%, 99.71%, 99.60%. Deep predicts climate well can guide environmental policy urban development decisions.

Language: Английский