Study On Scale Transformation Of Forest Coverage And Land Type Division Technique

In the past50years, remote sensing technology exhibits rapid development and change and produces a large quantity of application fruits ranging from small-scale scientific research to large-scale actual production. With the rapid development of3S technologies, the application of remote sensing technology also showed a rapid development in forestry monitor industry, however, currently most studies are focusing on the experimental research of single-scale and single remote sensing data, while there are few application research on multi-source remote sensing data in a multi-measure spatial scale. Taking Fushun city as the study area and adopting multi-source remote sensing data combined with basic geographic data, this article studies scaling of multi-source remote sensing images and the forest land type division technique to provide technical reference for monitoring of forest resources.Content and conclusion of the research are as follows:(1) Selecting Fushun city as the research area, forest area as the research object, and MODIS and TM remote sensing images and aerial data to build a two-stage empirical model of forest vegetation coverage, where the model is tested with such a result:there is good correlation between TM NDVI average value and ground actually-measured forest area, and the two-stage model can relatively accurately extend the ground data interpreted by aerial photographs to the mesoscale range; the RMSE between MODIS predictive value of forest vegetation coverage and the actually measured value is0.175, deviation15.9%; when applying MODIS NDVI data and the two-stage model to estimate the forest area of Fushun, the difference between the estimated value and the Continuous Forest Inventory (CFI)/Forest Management Inventory (FMI) data is within15%, the overall estimated accuracy reaching85%.(2) Selecting the TM and Rapideye images, and Continuous Forest Inventory (CFI) data of year2009Qingyuan County of Liaoning Province, applying threshold value method to extract three vegetation types in TM and Rapideye (coniferous forest, broadleaf forest and mixed forest), and adopting scaling method based on mathematical statistics to create a linear spatial scale conversion model between the three types of forests and the result shows good correlation among the established spatial scale conversion models, with correlation coefficient above0.8. These models can be used to scale up the vegetation type conversion to provide reference to the accurate monitoring of large areas of different forest vegetation with the application of TM images.(3) Selecting Fushun city as the study area, the forest land type division technique of different scale (City-County-Township(management unit)) is formed based on multi-source remote sensing data. Taking Tukouzi county as an example, the types of land in the county were extracted by object-oriented approach, and then based on this method, adopting collaborative inversion TM approach to extract forest vegetation area information of Qingyuan county; then extracting forest vegetation area information by threshold value method with multi-temporal MODIS remote sensing data, and verifying the result which shows that RapidEy image classification accuracy reaches88.87%, Kappa coefficient0.61; classification accuracy of TM images reaches86.70%, Kappa coefficient0.63, the overall accuracy of MODIS images to extract the forest area reaches92.41%, Kappa coefficient0.73. The information of forest vegetation types extracted by the three images reaches fairly high accuracy, and the established of the forest land type division technique can provide reference to the monitoring of the area of forest vegetation types of different scales for the forestry department.