| Eucalyptus in Guangxi accounts for 7/1000 of China’s forestland area,but it meets a quarter of China’s timber demand,which plays an important role in maintaining China’s timber safety and protecting natural forest felling.In China,eucalyptus plantations are mainly operated in short rotation period,while the existing forest resource inventory data cycle is relatively long,and the data update frequency is difficult to meet the data needs of short-rotation eucalyptus plantation management.Therefore,based on Landsat,Sentinel-2 and MODIS data in Guangxi Zhuang Autonomous Region,this study compared the spatial distribution of eucalyptus plantations extracted from different data sources and different classification algorithms.Based on the optimal model,the spatial distribution of eucalyptus plantation was extracted year by year from 2006 to 2020 to realize the dynamic monitoring of the spatial distribution of eucalyptus plantation.Based on the extraction results of time series spatial distribution,Land Trendr algorithm was used to monitor the disturbance change of eucalyptus plantation area.At the same time,the annual age of eucalyptus plantation was extracted by long-term spatial distribution extraction results,Land Trendr algorithm and CCDC algorithm respectively.Based on the age inversion results,Landsat and MODIS data were fused by data fusion algorithm to generate monthly image data of eucalyptus plantations during the felling year.According to logging and planting of eucalyptus in NBR index falling to their lowest level throughout the year when the change law of construction of eucalyptus plantation in coupled inversion model in order to realize accurate inversion of eucalyptus plantation in plantation age,in order to short rotation of eucalyptus plantation management to provide data support,and eucalyptus resources to realize the sustainable development,study the main conclusion is as follows:(1)Different data sources and algorithms have different effects on spatial distribution extraction results of eucalyptus plantations.Among the four algorithms of random forest,gradient lifting decision tree,support vector machine and decision tree,the random forest algorithm has the highest extraction accuracy with an overall accuracy of 91.07% and a Kappa coefficient of 0.86,followed by the gradient lifting decision tree with an overall accuracy of 86.93% and a Kappa coefficient of 0.80,and the support vector machine has the worst results.The overall accuracy is only 73.23%,and the Kappa coefficient is 0.70.Among the three kinds of landsat-8,Sentinel-2 and MODIS image data,the random forest algorithm based on Landsat-8 image has the highest extraction accuracy with the overall accuracy of91.07%,Kappa coefficient of 0.86,and Sentinel-2 image with the overall accuracy of 86.62%.The Kappa coefficient was 0.85,and the results based on MODIS data were the worst,with the overall accuracy of 77.24% and the Kappa coefficient of 0.76.The results showed that the random forest algorithm based on Landsat-8 images could accurately extract the spatial distribution of eucalyptus plantation.It can provide a method support for monitoring the dynamic change of spatial distribution of eucalyptus plantation based on time series image.(2)The monitoring results of the dynamic change of spatial distribution of eucalyptus plantation showed that the area of eucalyptus plantation increased firstly and then decreased to a stable trend from 2006 to 2020.Such as: In 2006,the area of eucalyptus was 6,584km~2.In2008,the area of eucalyptus decreased by 480km~2 compared with 2007 due to the severe snow disaster in south China.After that,the area of eucalyptus increased year by year,and reached the local maximum of 22881km~2 in 2015.Among them,2018 is relatively special,with a sudden increase in area,reaching 26,304 km~2.At the same time,it was found that the extraction results of eucalyptus plantation area based on time series satellite images were roughly the same as those of other studies and yearbook statistics.The results indicated that the spatial distribution results of eucalyptus plantation based on time series could provide basic data for follow-up disturbance change monitoring and annual stand age extraction.(3)Land Trendr algorithm can monitor the time and area of disturbance in eucalyptus plantation,and the overall monitoring accuracy is 87%.The Land Trendr algorithm was used to monitor the disturbance of eucalyptus plantation in Guangxi from 2007 to 2020.The disturbance of eucalyptus plantation in Guangxi was mainly mild disturbance,and the area with mild disturbance was 32,098 km~2,and the area with severe disturbance was 13,804 km~2.Among the disturbance events with the largest variation range of NBR in the eucalyptus planting area,the disturbance area was 466.842 km~2 in 2011 and 1,546,354 km~2 in 2018,and0.71,441 km~2 in the area with both mild disturbance and severe disturbance.(4)Based on the change of vegetation index in the time series before and after eucalyptus fusing,precise inversion of the annual/monthly age of eucalyptus plantations can be achieved.Among them,the accuracy of CCDC algorithm is the highest,accounting for87.33%.In 2020,the annual age of eucalyptus plantations in the whole region mainly focuses on 1 to 8 years,accounting for 71.34% of the total area of eucalyptus in the whole region.Eucalyptus plantations less than 2 years old accounted for 38.2% of the total area,and eucalyptus plantations more than 15 years old accounted for only 1.4%.Based on the inversion results of annual stand age,data fusion algorithm was used to fuse Landsat and MODIS data to obtain the accurate monthly stand age of eucalyptus plantation,in which the monthly stand age was mainly concentrated in 3 months to 8 months,and the cutting and planting time was mainly distributed in May to October,which was consistent with the actual survey results.The results showed that Landsat image data based on time series could accurately extract the ages of eucalyptus plantations at the annual and monthly scales,which could provide timely data support for the operation of eucalyptus plantations in short rotation period. |
PDF Full Text Request