%0 Journal Article
%T Extraction of Forest Resources in Thailand based on MODIS Data<br>基于MODIS数据的泰国林地资源提取
%A LV Ting-ting
%A SUN Xiao-yu
%A YU Bo-hua
%A <br>吕婷婷
%A 孙晓宇
%A 于伯华
%J 资源科学
%D 2008
%I 
%X Forest monitoring using satellite data has become an important tool for investigating and managing the exploitation of forest land, especially in tropical forest regions where large areas of forest have disappeared. Thailand was once a forest-rich country, particularly tropical forest, but over the last fifty years it has suffered serious forest losses. According to official statistics from the Thailand Royal Forest Department, in the early 20th century the rate of forest cover was 75%, but in 2004 it was only 32.66%. Thailand is a tropical country with extensive cloud cover from May to October, which presents a significant limitation to the availability of remote sensing data for operational monitoring of forest areas. In order to improve classification accuracy and effectively reduce cloud noise, a set of normalized indices including NDVI (Normalized Difference Vegetation Index), NDSI (Normalized Difference Soil Index) and NNDVI-LST (Normalized NDVI and Land Surface Temperature Index (LST)) were introduced by expanding the idea of NDVI. The NDVI has been demonstrated to be a robust and sensitive vegetation measure and has been used widely for continental and global-scale land cover classification. But because it is very sensitive to cloud noise, the classification accuracy is poorer in tropical areas than in other areas of the world. Compared with NDVI, NDSI uses near-infrared bands and short-wave infrared bands which can greatly reduce the influence of cloud noise. LST is also an important parameter to describe the characteristics of land surface cover, which is highly sensitive to different land cover types on a large scale. The results show that NNDVI-LST can more effectively reflect vegetation information than LST alone. By combining these three indices, Thailand's forest boundaries can be clearly extracted. MODIS Land Products such as 8-day MOD09Q1, 8-day MOD09A1 and 8-day MOD11A2 were used to construct NDVI, NDSI and NNDVI-LST for this study. The resulting MODIS-derived forest map was compared to national forest statistical data and the 2000 forest map provided by the Thailand Royal Forest Department. The forest boundaries extracted by MODIS are very similar to those in the forest map of 2000. Forest area estimation using MODIS data was highly correlated with the statistical data, with a correlation coefficient of 0.9264. Classification using Landsat ETM was also used to validate the MODIS-derived forest result. With a confusion matrix, the overall accuracy was 86.78%.
%K Thailand
%K Forest
%K MODIS
%K Multiple Indices
%K NDVI
%K NDSI<br>泰国
%K 林地
%K MODIS
%K 多参数
%K MODIS
%K 数据
%K 泰国
%K 林地
%K 资源提取
%K Data
%K based
%K Thailand
%K Forest
%K Resources
%K 决定系数
%K 线性关系
%K 存在
%K 发现
%K 回归分析
%K 精度验证
%K 分类方法
%K 参数综合
%K 利用
%K 结果
%K 影像
%U http://www.alljournals.cn/get_abstract_url.aspx?pcid=B5EDD921F3D863E289B22F36E70174A7007B5F5E43D63598017D41BB67247657&cid=B47B31F6349F979B&jid=9DEEAF23637E6E9539AD99BE6ABAB2B3&aid=7FD1B159FACA4FD0F62582E2C1A7DF4B&yid=67289AFF6305E306&vid=340AC2BF8E7AB4FD&iid=DF92D298D3FF1E6E&sid=BB5084A31068995F&eid=74253B3A525E9002&journal_id=1007-7588&journal_name=资源科学&referenced_num=1&reference_num=11