Short-Term PM<sub>10</sub> Retrieval during Dust Storms Using GNSS ZNHD, Meteorological Factors, and PM<sub>2.5 </sub>

Dust storms pose serious environmental and climatic challenges, requiring timely and accurate monitoring. Retrieving dust-related indicators using Global Navigation Satellite System (GNSS) signals is an emerging remote sensing method with high temporal resolution. While most existing studies emphasize the correlation between PM?? and GNSS-derived tropospheric delay or precipitable water vapor (PWV), effective PM?? retrieval models remain scarce. This study proposes a short-term PM?? retrieval model in response to the March 2021 dust storm in northern China. The model integrates GNSS zenith tropospheric non-hydrostatic delay (ZNHD), meteorological factors, and PM2.5 concentration to explore their influence on retrieval outcomes and dust storm characteristics. ZNHD was calculated for two stations—BJFS (Beijing) and CHAN (Changchun)—and decomposed using Ensemble Empirical Mode Decomposition (EEMD) to extract relevant signal components. Correlations between PM?? and both original and reconstructed ZNHD series were analyzed. Using hourly data, a nonlinear Least Squares Support Vector Machine (LS-SVM) rolling model was developed for short-term PM?? retrieval. The study confirms that integrating GNSS-derived ZNHD with environmental variables enhances short-term PM?? retrieval accuracy, offering a promising approach for dust storm monitoring and early warning.