The Spatio-temporal Pattern Of Wildfires In Inner Mongolia

Wildfires are widespread in the world.The distribution of wildfires presents a complex and diverse pattern.when the fire occurs uncontrollably or abnormally frequently,it will threaten the ecology and the human social life.At the same time,the study of the spatiotemporal pattern of wildfires is the basic work to find out the driving factors,and provides the basis for the study of fire ecology,fire risk assessment and accurate fire prediction.The purpose of this paper is to detect whether the distribution of Inner Mongolia wildfires from 2001 to 2018 is independent or interactive in time and space,and to explore the clustering features from a cross spatio-temporal perspective,to further explore the meteorological driving conditions of the high-density gathering of wildfires and wildfires season.In this paper,the Ripley’s K function,kernel density estimation and other methods are used to explore whether the fire points adjacent in space are similarly adjacent in time and how the interaction behaves in different spatio-temporal scales.The hot spots of wildfire are characterized by the continuous density surface in space and time.The location and duration of clusters are accurately detected by means of spatio-temporal permutation scanning statistics method.The correlation degree of fires among clusters in time is analyzed by simple correlation analysis method.The corresponding meteorological conditions in the spatio-temporal range of clustering are evaluated,and the seasonal range of wildfires is given according to the clustering results.The results are as follows:(1)The occurrence of wildfire has periodicity.Every 4-5 years,there is a peak.From 2001 and 2018,wildfires were unusually frequent in 2003 and 2008,and the number of fires in2014,2015 and 2018 was also significantly higher than average.In different ecological regions,the number of wildfires in the humid region was extremely high in 2003,the number of wildfires in the semi-humid region was high in 2008,and the number of wildfires in the semiarid and arid region was high in 2014 and 2017,respectively.The occurrence of wildfires also has interlunar variation.March and September are the months with high occurrence of wildfires in the whole study area.In different ecological zones,the concentration of wildfires in humid and sub-humid areas was much higher in March than in September.While the frequency of wildfires in March in semi-arid areas was close to that in September,and more in arid areas in April and September.The spatial distribution of wildfires mainly concentrated in the semihumid region,followed by the humid region,the semi-arid region is widespread,and the arid region has a small local distribution.(2)The spatio-temporal distribution of wildfires in each ecological region shows an obvious aggregation pattern,that is,ignition points similar the time is also very close in space.The degree of aggregation in the semi-humid region is the highest.At different scales of time and space,the degree of ignition aggregation increases with the increase of spatial range,and then decreases with the increase of temporal range.In the humid and semi-arid regions,the concentration of wildfires was the most obvious in one year,while in the semi-humid region,the concentration was the highest in two years.(3)Most of the detected high-density areas were located at the junction of humid,semihumid and semi-arid regions.They were clustered in different locations at different time.The precise spatio-temporal location of the inter-annual wildfires cluster shows that the movement path of the wildfires in the region from 2001 to 2018,which is from the northern part of Greater Khingan to the southern part of Greater Khingan and back again.Based on the correlation among these clusters,it is found that they form two types of high-risk areas in the eastern and western foothills of Greater Khingan,and the temporal variation of the occurrence of wildfires in the same area is consistent.The detected monthly wildfires cluster revealed that wildfires frequently occurred on the eastern side of the northern part of Greater Khingan in March,wildfires frequently occurred on the northernmost part of Inner Mongolia in May and June,and wildfires frequently occurred on the southern part of Greater Khingan in September and October,the only summer wildfire cluster was detected on the west side of Greater Khingan in August.(4)Based on the effective detection of the cluster location and the spatio-temporal range by the spatio-temporal permutation scanning statistics method,the identification of the wildfire seasons in each ecological region shows that the humid region’s wildfires season is from March1 to June 2,and from August 29 to September 26;the semi-humid wildfire season is from March 1 to March 31,from April 25 to May 31 and from August 27 to October 15;and the semi-arid wildfire season is from February 28 to April 13,from May 3 to June 1 and from June20 to September 12.(5)The analysis of meteorological conditions of the frequent occurrence of wildfires shows that the clustering of wildfires is characterized by abnormally high temperature and abnormally low relative humidity.The temperature anomaly is as high as 10 °C and the maximum relative humidity anomaly is 15% below the multi-year average.In summary,it can be concluded that the spatio-temporal distribution of wildfire presents a significant aggregation pattern,and there are obvious space-time interactions at different scales.The wildfires clustering in different time appeared in different places,but most of them were distributed in the junction of humid area,semi-humid area and semi-arid area.Most of the clusters accurately detected by the spatio-temporal permutation scan statistics last for one year or two years,with time shifting from northern Greater Khingan to southern Greater Khingan and gradually back to northern Greater Khingan between 2001 and 2018.This paper reveals the hot spots of local wildfires driven by high temperature and low relative humidity,and identifies the seasons of wildfires in each ecological region base on clustering results.