Accurate AQI forecasting in a high-altitude city using a simulated CVOCA-BiLSTM hybrid model: a case study of Lhasa, Tibet

Accurate AQI Forecasting in a High-Altitude City

A Case Study of Lhasa, Tibet

Forecasting the Air Quality Index (AQI) in high-altitude cities presents unique challenges due to nonlinear temporal patterns influenced by complex interactions between atmospheric physics and pollutant transport. To address this, a hybrid model combining a simulated Complex-Valued Optical Convolution Accelerator (CVOCA) and a Bidirectional Long Short-Term Memory (BiLSTM) network was developed to enhance both accuracy and computational speed.

Model Architecture and Function

The CVOCA module performs high-speed, complex-valued convolution, extracting amplitude and phase features from high-dimensional AQI sequences.
The BiLSTM network captures bidirectional temporal dependencies in the AQI data.

Evaluation and Performance

The model was tested on daily AQI data from Lhasa collected between 2014 and 2024, totaling 4018 samples. Metrics demonstrated superior performance compared to standalone LSTM and BiLSTM models:

Root Mean Square Error (RMSE): 1.60
Mean Absolute Error (MAE): 0.93
Mean Absolute Percentage Error (MAPE): 1.60%
Correlation coefficient (R): 0.997
Coefficient of determination (R²): 0.996

Significance and Application

This hybrid approach effectively captures the complex dynamics of AQI in plateau environments, offering a scalable solution for real-time air quality forecasting and supporting environmental decision-making.

In recent years, the escalating frequency and intensity of large-scale wildfires worldwide have become a critical driver of atmospheric pollution, severely undermining urban environmental sustainability and endangering public health.

Exceedance of hazardous Air Quality Index (AQI) thresholds constitutes a significant threat to ecosystem stability.

Author's summary: The study introduces a novel hybrid model combining CVOCA and BiLSTM that significantly improves AQI prediction accuracy in high-altitude cities, supporting real-time environmental management.

Nature — 2025-11-05