Extraction And Management Of Forestry Resources Information And Its Application In Forest Fire Prevention

There are much rich forestry resources with large carbon storage in China.These resources played a significant important role on water conservation,soil and water conservation,wind and sand fixation,climate regulation,disaster prevention and mitigation.The traditional method of field measurement of forestry resources has a huge disadvantage with extremely low efficiency and high cost.In contrast,remote sensing technology has more great advantage of large-scale,multi-category,multitemporal,multi-band,etc.,which make it possible to conduct forestry resources survey and extract key information.Rapid extraction,efficient management and sharing of forestry resource information will help forest fire prevention and provide a scientific basis for decision-making on fire prevention for forestry administration.This thesis selected Yanyuan county of Sichuan Province as the study area.Firstly,multi-category forestry resource information was extracted using Landsat and MODIS remotely sensed data.Secondly,the thesis designed forestry resource information management and application service system and updated a new scheme for the integrated management of forestry resource information.Finally,the extracted forestry resource information,weather,terrain and vegetation variables were regarded as fire drivers and used to model forest fire risk.The main work and achievements can be summarized as follows:1)Multi-source remote sensing reflectance data were collected and used to obtain vegetations indices.The thesis analyzed the relationship between these indices and canopy density,mean diameter at breast height,mean tree height,forest volume and stand density.On this basis,the multiple linear regression model for extracting forestry resource information was established based on vegetation indices and principal component analysis.The validation results show that RMSE is slow and R of the model for these parameters was greater than 0.3,except for stand density.It indicated that the established model can be applied in large-scale extraction of forestry resource information.2)This thesis made an in-depth investigation and analysis on forestry resource information management and sharing mode.It found that the timely downloading and processing of multi-source remote sensing data has become the biggest obstacle to the rapid extraction of forestry resource information.At present,the acquisition methods of various data sources are not unified,so it is difficult to realize the collaborative processing of multi-source data;The existing dynamic evaluation and monitoring modes of forestry resource information are analyzed,which mainly include time series feature extraction and change monitoring.On this basis,the forestry resource information management and application service system,which is feasible and expandable,were designed from the perspective of framework and functional requirements.3)The forest fire risk model based on forest resource information was established using logistic regression.And forest fire risk was calculated to characterize the degree of forest fire danger quantitatively.The historic forest fire dataset of Yanyuan county from 2003 to 2008 was constructed,where 70% of them was used to train forest fire risk model and the rest was used to validate the established model.The quantitative metrics: AUC value of ROC curve,mean absolute error（MAE）,root mean square error（RMSE）and logarithmic score（LS）,were used to evaluate the model performance.Higher AUC value and other better quantitative metrics were observed on the forest fire risk model added forestry resource information（AUC=0.81）,while AUC is only 0.74 when forestry resource information was not considered.This showed that the forest fire risk model with forest resource information has better performance and proved that forest resource information extracted in this thesis plays an important role in forest fire prevention.