Monitoring And Prediction Of Tomicus Spp. Disaster With Remote Sensing Images At Multi-scale In Yunnan Pine Forests

Tomicus yunnanensis Kirkendall and Faccoli and Tomicus minor Hartig are the major forest pests in southwest China,which have caused devastating damage to Yunnan pine forests.However,one of the main reasons for the disaster widespread is the monitor technology is low timeliness that cannot accurately monitor and predict the occurrence of the shoot beetles.Remote sensing technologies has been demonstrated that can obtain the information of the damage level and insect infested area accurately,realize the dynamic monitoring and prediction of insect diseases,and provide a strong basis for effective disaster prevention.This paper takes Xiangyun county,Yunnan province as the study area,chose Tomicus spp.as the research objects.We combined multi-remote sensing data,ground survey data and historical data to explores the outbreak and dynamic spread pattern of shoot beetles,and build monitoring and prediction models at multi-scales.The main results of this paper are as follows:（1）The damage characteristics of the shoot beetles in the“shoot transferring to trunk”period were determined.Based on single wood analysis,Yunnan pines with high shoot damage ratios were preferred by Tomicus spp.during the“shoot transferring to trunk”periods have been verified.The pine resin plays an important role to resist invasion of Tomicus spp.The water content of Yunnan pine phloem and xylem significantly decreased after the successful invasion of Tomicus spp.which accelerate the death of Yunnan pines.The host selection habit of“shoot transferring to trunk”period provides a key monitoring object for remote sensing identification at single wood scale-Yunnan pines with high shoot damage ratios.（2）Aggregation distribution pattern of Tomicus spp.during the shoot feeding period were determined.The sequence of the distance of spatial dependences in three levels of damaged forest:slightly infested forest>moderately infested forest>severely infested forest,and the sequence of the intensities of local spatial continuity:slightly infested forest>moderately infested forest>severely infested forest.The aggregated damage characteristics would weaken Yunnan pine that conducive for adults to bore trunk and to reproduce.The spatial aggregation damage characteristics of Tomicus spp.provide spatial information for the accurate monitoring at plot scale.（3）At the needle scale,the change rules of the hyperspectrum,color,chlorophyll content and water content of needles after Tomicus spp.damaged were clarified.The reflectance of red bands and long-wave near-infrared bands（1400-1700nm）increased,while reflectance of short-wave near-infrared bands（780-850nm,950-1100nm）decreased with the change of the needle color,the red edge of plant shifts from originally red to shorter blue.The relative content of chlorophyll and water content of needles decreased significantly.Based on correlation analysis and non-parametric Anova-Kruskal-Wallis test,it is clear that chlorophyll and water are the main factors affecting the spectrum of needles.The selected sensitive bands can be used to establish spectral identification system to monitor the damage of Tomicus spp.（4）At the individual tree scale,the WV-3 satellite data were suitable for classifying different levels of damaged trees（healthy,slightly,moderately and severely）.The normalized band index could improve the classification accuracy was proved based on the difference analysis.The stepwise linear discriminant classifier（SDA）was the best classifier,an overall accuracy of 78.33%was achieved.The best mapping results of different damaged trees was predicted based on the best classification model.The results provide target positioning for accurate prevention and control of forestry departments.（5）At the plot scale,Landsat time-series data were demonstrated that can be used to monitor the dynamic change of Yunnan pine forest damaged by Tomicus spp.,including the outbreak time,damage levels and infested area.The water stress index（MSI）was selected by the change detection algorithm-temporal anomaly,an overall accuracy of 86.38%was achieved.The spread directions of Tomicus spp.was determined based on the geometrical centroid analysis of Arc GIS.Combined with the wind direction,we speculated wind to be a vector for long-distance dispersal for Tomicus spp.The dynamic monitoring of Tomicus spp.disaster provides a foundation for the study of spreading mode.（6）At the landscape scale,a prediction model of shoot damage ratios was constructed in Yunnan pine forest after the outbreak of Tomicus spp.The inversion model of shoot damage ratio（Y）was constructed based on the water stress index anomaly（MSI’）,the model is:Y=1.954 X+0.153,which has the fitting accuracy of 0.837.The distance to previous year’s severely infested forests（DS）,the previous year’s shoot damage ratio（SDR’）,the conversion value of aspect(T_Aspect),crown density（FC）and the distance to road（DR）were selected as the major factors to build the shoot damage ratio（Y）predicting model,the model is:Y=30.335-0.007*DS+0.551*SDR’+5.175*T_Aspect-0.131*FC-0.012*DR,which has the fitting accuracy of 0.679.The model could meet the application requirement that predict the spread of Tomicus spp.in the outbreak year.（7）At the county scale,a damage aera prediction model was constructed based on meteorological factors.The significantly relationship between rainfall,temperature,sunshine duration and shoot beetle outbreak were demonstrated based on correlation analysis.The average annual total precipitation of the previous 3 years（P）,the average temperature of the previous 3 years（T）,the average annual total sunshine time of the previous 3 years（S）were selected as the major factors to build the shoot damage ratio（Y）predicting model by stepwise regression,the model is:Y=29.555*P+15.910*S+3463.185,which has the fitting accuracy of 0.827 and standard error of 2707.17 hectares.The model provides the basis for monitoring the shoot beetles at large scale and taking measures to prevent the beetle damage.（8）At spatiotemporal scale,a prediction model of shoot damage ratio was constructed based on six independent variables-the distance to previous year’s severely infested forests（DS）,the previous year’s shoot damage ratio（SDR’）,the conversion value of aspect(T_Aspect),the distance to road（DR）,the average annual total precipitation of the previous 3 years（P）,the average temperature of the previous 3 years（T）,the average annual total sunshine time of the previous 3 years（S）,the model is:Y=67.68+0.652*SDR’-0.0009*DS-0.026*P+0.77*T-0.202*S-0.0003*DR,which has the fitting accuracy of 0.555 and standard error of 0.1288,which can provide data support for the monitoring and prediction of shoot beetle damage at large area,especially in the years with severely damage.By integrating ground survey and multi remote sensing data sources,the damage characteristics,the spatial distribution of shoot beetles during shoot damage periods and the spatial distribution of different levels of damaged Yunnan pines were definite which were conducive to the spread pattern study of shoot beetles in the woods.Supported by stand factors,beetle pressure factors,topographical factors,landscape factors and meteorological factors,a complete set of monitoring and prediction model were constructed at different scales which can provide practical monitoring and early warning technology system,which were useful for forestry department to take reasonable measures to reduce the harm to Yunnan pine forest,and has a good application prospect.