The Modeling Analysis Of The Fire Risk In The Wildland-urban Interface At The Provincial Scale

At present,the fires are increasingly frequent and cause serious damage to people and buildings in the Wildland-urban interface(WUI).In view of the potential fire risk brought about by the vigorous development of the wildland-urban interface(WUI)in China,there is still a phenomenon that the relevant research is not comprehensive enough or even lags behind.At the same time,from the perspective of time and space,it is less likely to analyze the occurrence of fire risk in the WUI.Therefore,with the help of relevant mathematical models,this paper has conducted detailed studies on the temporal and spatial characteristics and related impacts of the WUI fires.The influencing factors of fires in the WUI are discovered and the relationship between them can also provide theoretical support for improving the scientificity of fire management in the WUI.The main work is as follows:1.By using the random forest algorithm to fit the WUI fires at the provincial scale,the spatial distribution characteristics of fires in the Wildland-urban interface(WUI)are explained.Firstly,the results of resampling for feature selection show that,the independent variables involved in the WUI fire risk modeling are:the road density,the railway density,the monthly average minimum temperature,the normalized difference vegetation index,population density and elevation.The independent variables involved in forest fire risk modeling are:the monthly average temperature,the normalized difference vegetation index,the road density,the railway density,population density and elevation.Afterwards,according to the random forest algorithm,the occurrence of fire was fittted.It was found that the simple correlation coefficient between the predicted value and the true value of the training set and the test set was higher than 0.90.It can be considered that the RF algorithm can effectively fit the fire occurrence in the WUI and Forest.In summary,by using the random forest algorithm to model the WUI fire risk and forest fire risk,and six variables with high contribution to the WUI fire risk and forest fire risk were selected.2.Through the use of global linear regression model(LM),geographically weighted regression model(GWR)and geographically and temporally weighted regression(GTWR),the three models were studied in terms of the fitting and interpreting capabilities of fire risk in the WUI and Forest.We analyze and establish the Spatio-temporal model for the best performing fire risk distribution.In this process,we first used the method of spatial cross-validation(SCV)to perform independent variable screening,and then included the seven significant independent variables that were selected into the modeling process of the GWR and GTWR models.The results show that the GTWR has the best ability to match fire risk and its spatial structure.Finally,when the model was independently tested,it was found that the GTWR had stronger interpreting power than the other two models.Therefore,the GTWR model has better fitting and interpreting capabilities than the GWR and LM models,which shows that the GTWR model can reveal scientifically the difference of influence of independent variables for the fire occurrence in different time cross-sections and different spatial positions,and it is a meaningful attempt to use it to fit and interpret the WUI fires risk and Forest fires risk.3.We efficiently and meaningfully fit the fire risk by using the GTWR model.The output of the GTWR model can deeply analyze the spatio-temporal heterogeneity of the independent variables that affect the fire occurrence in the WUI and compare with that of the Forest fires.This also shows the advantages of the GTWR model in evaluating the laws of spatio-temporal dynamic and uncertainty of the WUI fires risk.Firstly,results of linear regression training using spatial cross-validation methods show that NDVI and Road are the most important variables in the modeling of the WUI fires risk,and LST and NDVI are the most important variables in the modeling of the Forest fires risk.After that,according to the spatial and temporal distribution of the independent variables,it can be found that the normalized difference vegetation index(NDVI)has a positive correlation with the WUI fire risk and the Forest fire risk,but NDVI has a greater impact on the WUI fires occurrence than that of the Forest fires occurrence.It shows that there is a close correlation between the WUI fires occurrence and the Forest fires occurrence,and the two kinds of fires have a certain relevance;Road is positively correlated with the WUI fires occurrence and changes significantly over time.This variable represents the frequency of personnel activity,indicating that with the rapid development of infrastructure and urbanization in China,the occurrence of the WUI fires has become more frequent.