Research On Forest Dynamics Monitoring Based On Fusion Of Multi-source Remote Sensing Data

As the largest carbon reservoir of the terrestrial ecosystem,forest performs an increasingly important role in improving the ecological environment,maintaining the global carbon cycle,and mitigating climate change.The Three-North Shelterbelt Forest Program launched in 1978 was the largest and most representative ecological engineering project of China.Its effectiveness in ecologically fragile arid and semi-arid areas has always been a hot topic,which may due to the lack of more precise information on the specific distribution and spatial-temporal dynamics of forests.With the development of remote sensing data and science technology,the demand of our country for timely and accurate information on forest distribution and spatial and temporal dynamic has become more and more urgent.In response to the above questions,the main contents of our study were presented hereafter:（1）Our study used decision tree classification algorithm based on PALSAR/PALSAR-2 data and used NDVI_max index based on Landsat data to extract forest information respectively,and then used appropriate fusion algorithm to generate forest classification results maps of northern China ranging2007-2010 and 2015-2017.（2）Our study evaluated the accuracy of our forest classification result obtained by the fusion algorithm,compared with the four common medium and high-resolution land cover products qualitatively and quantitatively,and analyzed the uncertainties.The forest classification results obtained by the fusion algorithm have the highest overall accuracy（96%±1%）and the Kappa coefficient（0.66）,so it can capture more details of the forest distribution.（3）According to the forest classification result map obtained by the fusion algorithm,our research performed the spatial calculation at pixel level to monitor the forest dynamics of the forest distribution in northern China during 2007-2017,including forest loss,forest gain,unchanged forest and unchanged non-forest,and we carried out an accuracy assessment.（4）Our study realized the spatial combination of the forest dynamics and precipitation and analyzed the forest dynamics in different precipitation gradient zones.The forest loss rate in arid and semi-arid regions（35.01%）was significantly higher than that in humid and sub-humid regions（25.4%）.The linear least squares regression was performed between the annual mean precipitation rate and the annual mean precipitation.Forest loss rate was significantly negatively correlated with the annual mean precipitation.Our study generated accurate and effective forest distribution information through the multi-source remote sensing data fusion algorithm,which not only provided strong support at the remote sensing and computer level for the continuous implementation of the Three North Shelterbelt Project,but also can be applied to ecological assessment in other regions,which is beneficial to manages forests.