Chen Songxi's team published a study on dust tracking detection and traceability analysis in northern China

A few days ago, the research paper "Detecting and Evaluating Dust-Events in North China with Ground Air Quality Data" by Professor Chen Songxi's team from Guanghua School of Management, School of Mathematical Sciences and Statistical Science Center on the tracking detection and traceability analysis of sand and dust in northern China was published in the United States .
Published online by the Geophysical Society Earth and Space Science
.

In the spring of 2021, northern China has experienced several large-scale and high-intensity sand and dust weather attacks
.
How to automatically identify the area of ​​origin of sand and dust, track the movement path of sand and dust, and scientifically and reasonably evaluate the pollution contribution of sand and dust weather to northern China plays an important role in the prevention and control of sand and dust

This paper selects 688 air quality monitoring stations and 258 meteorological stations in northern China during the spring of 2015 to 2020, with a total of 120 million ground monitoring data, combined with the reanalysis of the European Centre for Medium-Range Weather Forecasts (ECMWF).
Based on wind field data, a set of statistical algorithms to automatically identify and track sand and dust processes is proposed
.
Traditional dust identification mainly relies on satellite remote sensing data, which has a trade-off of spatial and temporal resolution.

The study found that the dust in northern China mainly comes from the Taklimakan Desert (42.
1%), the Alxa Desert (23.
4%) and the Horqin Sandy Land (16.
6%)
.
Since the influence of the Taklimakan Desert is mainly limited to Xinjiang Province, if the local influence is excluded, the Alxa Desert contributes 59.

Dust Identification Algorithms and Data

This paper establishes a two-stage detection model with low model assumptions and high computational efficiency, extracting information from the temporal and spatial dimensions, respectively
.
The first-stage detection is a multivariate time series change point identification that comprehensively considers the correlation between PM10 concentration and pollutants

Since the high occurrence of sand dust occurs in the spring period in northern China, the article selects the monitoring data of the state-controlled pollutant monitoring stations in 15 provinces and cities north of 32° north latitude in the spring of the six years from 2015 to 2020, including six pollutants such as PM2.
5.
concentration data
.
Each pollutant monitoring station is matched with adjacent meteorological stations to provide surface meteorological data and reanalysis wind field data of 500hPa isobaric surface

Figure 1 Evolution of a large-scale dust process identified by data and algorithms from May 2 to 6, 2017
.
The figure shows nine snapshots at 12-hour intervals, and the red area marks the range of influence of the dust process at each moment

Transport network analysis

The identification algorithm proposed in this paper gives the overall "distribution" of each dust process, from which various statistical indicators can be extracted
.
As shown in Figure 2, with the help of the concept in the directed graph model, the article takes each province as an image node, eliminates the edge that points to itself within the node, builds a sand and dust transmission network, and calculates the outdegree of each node.

The left panel of Fig.
2 shows the dust transport network after removing the self-circulation
.
Each node represents a province, the edge with the same color as the node represents the sand and dust transmission process from this node to other nodes, and the thickness of the edge refers to the total number of accumulated hours of transmission

Structural changes related to pollutant concentration during sand-dust period

If the time when the sand-dust event was detected at each station in the spring of six years is taken as time t0, and the time series of pollutant concentration and the correlation coefficient sequence are aligned, the evolution law of each major pollutant in the sand-dust event can be analyzed
.
As shown in Figure 3, the red dotted line marks the moment when the algorithm detects the sand-dust event (marked as time 0).

Fig.
3 The evolution process of the average chemical composition in a sand-dust event, the red dotted line marks the beginning time of the sand-dust event identified by the algorithm
.
Left panel: PM10 concentration and its covariance with other pollutants
.
Right: Normalized concentration curves for six pollutants
.
From top to bottom, the average results for northern China, the Alxa desert, and the Beijing-Tianjin-Hebei region

The first author of this article is Tong Peifeng (a second-year doctoral student at Guanghua School of Management), and Chen Songxi is the corresponding author of this article
.
Tang Tengyong, professor of statistics at Temple University, is the other author
.