Geo-AI
Knowledge discovery of Middle East dust sources using Apriori spatial data mining algorithm
Project highlights
Abstract
Identifying the areas susceptible to dust storm formation is one effective way of dealing with this destructive environmental phenomenon. This study is the first attempt to employ the Apriori spatial data mining algorithm to dust source susceptibility mapping (DSSM). The research process was based on extracting association rules between spatial-temporal patterns of dust drivers (including soil, vegetation, and climate parameters) in the Middle East’s hotspots dust sources (HDSs). For this purpose, HDSs were identified using visual interpretation of sub-daily MODIS-Terra/Aqua RGB images from 2000 to 2021. The Middle East’s HDSs mainly correspond to desert areas with poor vegetation cover and ephemeral/dried-up water bodies. A total of three million rules were extracted by running the Apriori algorithm. Accordingly, bare and non-vegetated lands, high soil thickness, low soil moisture, very high wind speed, and high temperature were identified as the most common features of HDSs. Using three measures including support, confidence, and lift, 54 frequent, reliable, and logical rules were selected, and the related maps were generated. Then, the susceptible dust sources (SDSs) map of the Middle East was produced in five classes of extreme (13% of the areas), high (14%), moderate (16%), low (17%), and no (40%) susceptibility through the weighted linear combination of the rule maps. The accuracy of the identified SDSs was estimated at 83.7% using the verification points. A sensitivity analysis was performed using the leave-one-out method to determine the isolated effect of the selected rules on the produced SDSs map. The model uncertainty varied between 15.7% and 16.8% for different rules. The variation range of uncertainty was 1.1%, demonstrating that a single rule does not significantly affect the model’s performance; however, some rules have a more influential role. Our results revealed that Apriori’s ability to provide generalizable association rules is a robust algorithm for DSSM.
Introduction
The dust storm is an integral component of the Earth system (Shao et al., 2011) with wide-spreading impacts on weather and climate (Huang et al., 2006), biogeochemical cycles (Jickells et al., 2005), and human health (Goudie, 2014). Obtaining information about dust emission sources coupled with local environmental knowledge can effectively deepen our understanding of their characteristics and help to control the ensuing environmental problems. In this regard, remotely sensed (RS) satellite data, given their expansive and unique spatial-temporal coverage (Ashpole and Washington, 2013) have played a key role in identifying dust sources (Tsolmon et al., 2008). Accordingly, global dust sources have been identified and mapped based on atmospheric aerosol content (by analysis of aerosol indices (AIs)) using polar orbit satellites data such as Total Ozone Mapping Spectrometer (TOMS)/Nimbus-7 (Prospero et al., 2002; Washington et al., 2003) and Moderate Resolution Imaging Spectroradiometer (MODIS)/Terra and Aqua (Ginoux et al., 2012). The largest global dust sources are identified in the northern hemisphere within a dust belt stretching from the western coast of North Africa, the Middle East, and Central and South Asia to East Asia and China. There is considerably less dust activity outside this zone (Goudie, 2014).
In addition to the significance of the highly dynamic nature of dust plumes and their transportability by wind, a significant issue is that dust events occurring between two daily acquisitions by polar orbit satellites remain undetected. Polar orbit satellites (e.g., Nimbus-7, Terra, and Aqua) can acquire data only at the specific local passing time given their orbital trajectories. Therefore, a spatial offset may exist between the location of the modeled dust sources based on AIs and the ground reality (Ashpole and Washington, 2013). Various approaches have been employed in recent years to address this discrepancy in dust source mapping. The first and most accurate solution involves using ground-based data to understand the characteristics of the surface features, such as land cover, geological formations, and physicochemical compositions of the underlying soil (Parajuli et al., 2014). However, such data is not always easily accessible. Therefore, this approach often faces certain constraints. Various environmental factors play a role in the formation of dust sources (henceforth dust drivers), including vegetation, wind speed, precipitation, soil moisture, and temperature (Shepherd et al., 2016). Considering these drivers based on RS data coupled with AIs within the spatial-temporal modeling framework makes it possible to overcome the constraints of the first approach and effectively identify dust sources (Papi et al., 2022). The second approach is called “dust sources susceptibility mapping (DSSM)”, which has been studied extensively in the past half-decade. Some of these studies include the identification of susceptible dust sources (SDSs) using RS data and various models such as multi-criteria decision-making (MCDM) (Darvishi Boloorani et al., 2020), machine learning (Darvishi Boloorani et al., 2022a; Papi et al., 2021), and data mining (Gholami et al., 2020a).
Data mining is an intelligent data processing tool to data-driven knowledge discovery, i.e., understanding the structures, patterns, and associations between large and complex datasets (Hand and Adams, 2014). Numerous drivers make it difficult to model and analyze their mechanisms and complex associations with dust sources’ formation in knowledge extraction and, ultimately, decision-making. Accordingly, taking advantage of intelligent models that can manage large and complex datasets can yield more satisfactory results. Many studies in the past two decades have benefited from the capabilities of data mining techniques in various geo-environmental fields, such as air pollutant concentration prediction (Siwek and Osowski, 2016), aerosol optical depth (AOD) prediction (Soleimani et al., 2021), and land subsidence susceptibility mapping (Mohammady et al., 2019). However, taking advantage of this approach and its techniques has not been given much attention in dust studies, especially in dust source identification.
The Apriori method, a potent rule-based data mining tool, was used in the current work for the first time to identify SDSs based on discovery of association rules by analyzing the complex spatial-temporal patterns of different dust drivers in hotspot dust sources (HDSs). This algorithm has successfully been used for knowledge discovery and prediction applications such as wind speed (Guo et al., 2014), landslides (Wu et al., 2016), dust-driven diseases (Hunter and Cervone, 2017), road accidents (Hong et al., 2020a, Hong et al., 2020b), and mouth cancer (Tang et al., 2013).
Studies in the field of dust storms (e.g., Boroughani et al., 2020; Darvishi Boloorani et al., 2022a; Ebrahimi-Khusfi et al., 2021; Gholami et al., 2020a; Gholami et al., 2020b; Gholami et al., 2020c; Gholami et al., 2020d; Rahmati et al., 2020a; Rahmati et al., 2020b) generally use different machine learning algorithms such as artificial neural network (ANN), support vector machine (SVM), and random forest (RF) and data mining techniques such as BCART and boosted generalized additive model (BGAM)— to identify SDSs based on modeling the statistical relations between dust drivers. Despite their proven capabilities, these approaches cannot provide association rules concerning dust occurrence and the contributing drivers (Wu et al., 2016). Therefore, these methods are not generalizable, and for applying in other regions, complex models coupled with all parameters must be assigned and executed again. The main advantage of the Apriori algorithm is that based on association rules obtained by this technique for a region, it is possible to generalize these rules to similar regions and identify SDSs without re-running the algorithm. Accordingly, practical steps can be taken to improve the generalizability of DSSM approaches by reducing processing costs, minimizing ground data requirements, and avoiding related problems.