| Active fire refers to a“hot spot”in which the brightness temperature(Tb)difference exceeds a given threshold compared with the brightness temperature of the surrounding land cover in thermal anomaly products monitored by satellites,also commonly referred to as active fire points.According to statistics,about 1%of the world's forests are disturbed by active fire every year.Active fires change the distribution of carbon“sources,sinks,and storage”by burning forest combustibles,thereby affecting the carbon balance and carbon cycle of terrestrial ecosystems.Active fires are frequently occurring in low-latitude and mid-latitude regions with global warming,which is focused widely.However,the current research work is insufficient upon the time scale of 10 to 20 years;meanwhile,there is a lack of reference and application of the emerging frontier spatiotemporal data mining methods,which leads to many challenges for active fire relational analysis and causal relational reasoning.Xishuangbanna is located on the northern edge of the tropics south of the Tropic of Cancer.It is the best-preserved,largest and most biologically diverse tropical rainforest ecosystem near the Tropic of Cancer in the world.In recent years,due to the influence of climate changes,the drought and high temperature in Xishuangbanna have been increased simultaneously.At the same time,land use changes driven by human activities have been intensified,resulting in an increasing risk of active fires.However,the current research on fires in Xishuangbanna mainly focuses on the effects of fires on ancient vegetation and paleoclimate in the geological history,and lacks research on the temporal and spatial dynamic patterns and spatial prediction of active fires.This research takes Xishuangbanna as the study area,integrates natural elements and human active elements,such as terrain,vegetation,rainfall,air temperature,wind speed,etc.using methods such as spatial correlation analysis,probability density estimation(PDE),and spatial association rule mining,aim to reveal the temporal and spatial dynamic characteristics of active fires in Xishuangbanna upon a 10-year scale(2011-2020),based on the spatio-temporal dynamics of MODIS's active fire;clarify the relationship and action mechanism of active fires with natural and human active elements;the key influencing factors of active fire and its spatial pattern,and the following conclusions have been obtained:(1)From 2011 to 2020,the distribution of active fires in Xishuangbanna has typical seasonality,but there is no strict periodicity.Majority of active fire occurs from winter to spring of the following year,especially from February to April in each year.At the same time,The Moran's I show that there is a significant spatial autocorrelation of active fires in Xishuangbanna,especially in the“high-high(H-H)”and“high-low(H-L)”patterns.(2)The Gaussian kernel density estimation method can better simulate the spatial distribution density of active fire.In addition,when the significance level is 0.01,the probability distribution of factors such as altitude,slope,distance to water area,winter and summer vegetation,rainfall,air temperature,wind speed,and humidity of the active fire location obeys the Gamma distribution,respectively.(3)The Apriori algorithm is used to mine spatial association rules,and different data discretization processing methods have a significant effects on the experimental results.For the data set studied in this paper,the equal interval length division method is more conducive to exploring the potential association rules in the data set.Compared with the equal quantile division method,the equal interval length division method is easier to obtain frequent and higher-order association rules,and its support(sup),confidence(conf),lift,Kulc and IR have a wider range of values.In addition,for the same data set,the same discretization processing method is adopted but the difference numbers of division have a significant impact on the mining results.However,there is no consistent law between the number of association rules generated and the change in the number of divisions.(4)The spatial prediction results show that the active hot spots in the study area presented spatial aggregation significantly,and the active fire points are concentrated on the east and west sides of Xishuangbanna with a confidence level of more than 90%.The western part is concentrated in Menghai County.The eastern part shows a trend of belt distribution along the Lancang River which is concentrated in the central part of Mengla County and the junction of Mengla County and Jinghong City.In addition,Jinghong City has the fewest active hot spots,with only a small number of occurrences at the north and south ends,but the confidence is high.Besides,among the areas with the highest distribution density of active fires from 2011 to 2020,the future active fires at the junctional area of Jinghong City and Menghai County will present a cooling trend,and the active fires in the junctional area of Jinghong City and Mengla County will present a trend toward to the west bank of the Lancang River;meanwhile the west of Menghai County may continue to maintain a high density of active fires.(5)From 2011 to 2020,the active fire spots in Xishuangbanna were concentrated in forest land and shrub land(accounting for 58.7%).Through association rule mining analysis,it can be seen that active fires are mainly distributed in areas 5-8 km away from residential areas and within 2.5 km away from roads.In addition,the relationship between various elements of human activities and active fire shows a strong direction,which means that human activities have significantly affected on the distribution of active fire in the study area,but active fires have not significantly affected on human activity in the past 10 years.This study integrates multidisciplinary theories and methods to carry out research on the spatiotemporal dynamic characteristics of active fire,which realizes the deep integration of data mining algorithms and geographic spatiotemporal data analysis methods.The research results have enriched the understanding of active fires in Xishuangbanna,expanded the ideas and cases of active fire research in tropical rainforest areas.It's helpful scientifically grasped the occurrence and development laws of active fires in tropical rainforest areas.It provides a reference and basis for the protection of tropical rainforest ecosystem and biodiversity. |
PDF Full Text Request