Development of the Simplified Meta-Statistical Extreme Value (SMEV) Model for Analyzing Extreme Wind Events and Particle Transport Hazards in Eastern Lake Urmia

Introduction

The ongoing desiccation of Lake Urmia in northwestern Iran has transformed its former lakebed into a significant source of airborne dust and salt particles, posing escalating environmental and public health risks. These storms pose serious environmental and health risks by elevating particulate matter concentrations (PM₁₀ and PM₂.₅), degrading air quality, and impairing agricultural productivity. Wind events exceeding 5 m/s can start wind storm mobilization and atmospheric dust generation in arid and semi-arid environments. Wind direction is important for the transport of dust to cities. The east of the Urmia Lake is more affected by wind because of the dominant wind direction in the Urmia Lake basin. This part is more important for risk assessment studies. Classical models, such as the Generalized Extreme Value (GEV) distribution, often fall short of capturing the full complexity of wind extremes under nonstationary conditions. To overcome these limitations, the Simplified Meta-Statistical Extreme Value (SMEV) model is developed and used for the first time, in this study, as a method that integrates both ordinary and extreme wind data into a unified distribution framework. This study aims to estimate return period wind speeds with SMEV and benchmarked against GEV, and evaluate wind direction probabilities for storm prediction. Results will inform regional dust storm risk management and advance extreme value modeling in the Lake Urmia basin.

Materials and Methods

Using three-hourly wind speed and direction data from 2005 to 2024 across four synoptic stations (Tabriz, Maragheh, Bonab, and Shabestar) in the eastern Lake Urmia Basin, SMEV was employed to estimate return period wind speeds and assess directional probabilities. In this research, the CEEMDAN method has been used as a method to remove noise and trends from wind speed data. At the stations, wind events were divided into extreme and ordinary events, based on the wind speed threshold, using the peak-over-threshold (POT) approach by applying the 90th percentile. 5, 10, 20, 50, 100, and 200 periods were chosen for the return period. The model combines a two-parameter Weibull distribution for ordinary winds with annual extreme wind counts to generate composite cumulative distribution functions (CDFs) per dominant direction sector. The bootstrap method was used for SMEV model performance evaluation. The GEV model was used as a benchmark and employed to estimate return period wind speeds, and both models were evaluated using AIC, BIC, FSE, WFSE, and leave-one-out cross-validation (LOO). Additionally, a random forest algorithm was trained to predict the likelihood of wind directions associated with dust transport.

Results and Discussion

SMEV predicted critical wind speeds exceeding 7 m/s with high confidence. In all 4 stations, wind speeds predicted more than 7 m/s, and wind direction analysis revealed over 70% probability of wind-driven dust transport from the southwest and south to the east, toward residential areas. The random forest method has predicted the corresponding wind directions for selected stations east of Lake Urmia. The dominant directions for extreme storm events are southwest and south for short to medium return periods. In longer time periods, the dominance of south and west continues at Shabestar and Maragheh stations, and for Tabriz and Maragheh stations, the dominant direction changes to east. GEV extreme values predicted more than 12 m/s for wind speeds. It shows the GEV overestimated. For Urmia Lake Basin, wind speeds of more than 12 happen rarely and are not common. The SMEV model outperformed the GEV model, providing more stable and realistic estimates of return-level wind speeds, particularly for long recurrence intervals. Error metrics confirmed the superiority of SMEV (FSE = 0.014; WFSE = 20.7) compared to GEV (FSE = 0.081; WFSE = 196), highlighting its improved performance in estimating environmental hazards. The advantage of this method over other classical methods is in distinguishing between extreme and normal events, as well as distinguishing extreme events with the corresponding dominant directions of extreme wind speeds. In addition, the use of a wind speed threshold limit, unlike other statistical methods such as GEV, which only focus on maximum wind speeds in the analysis of extreme events, can provide reliable accuracy for this method in estimating extreme events.

Conclusion

This study focused on the analysis of extreme wind speeds in the eastern part of the Urmia Lake watershed, and using a simplified metastatistical limit value model, was able to provide reliable estimates of strong winds in different return periods. The results showed that speeds exceeding 7 m/s occur with high probability in this area, and this amount is sufficient to initiate the transport of suspended particles and the formation of dust storms in the study area. In conclusion, SMEV demonstrates significant potential for use in regional wind hazard assessments, early warning systems, and dust storm risk mitigation in the Urmia Lake Basin. This model relies solely on wind speed and direction and does not consider other environmental drivers such as soil moisture, land cover, vegetation, or surface roughness that can significantly affect the potential for dust emission. This approach can also help universities, along with other tools, to identify high-risk areas susceptible to dust transport from the dry bed of Lake Urmia. Overall, this model can be used as an effective tool in analyzing climate risks associated with wind and dust storms in the region. In addition, the use of the 90th percentile threshold and 24-hour separation criteria raises statistical assumptions that more extreme events may have been identified, which has increased the accuracy of the model and, on the other hand, has made the model more sensitive to extreme phenomena. However, it is suggested that in future studies, the integration of environmental variables such as relative humidity and precipitation should be considered to improve the SMEV model. Also, combining this model with wind datasets based on satellite images can also improve the spatial representation of wind patterns.