JOURNAL OF NATURAL RESOURCES ›› 2014, Vol. 29 ›› Issue (3): 507-515.doi: 10.11849/zrzyxb.2014.03.014

• Resources Research Methods • Previous Articles     Next Articles

Extracting Forest Vegetation Types from Multi-temporal MODIS Imagery in Sichuan Province

YANG Cun-jian1, ZHOU Qi-lin2, REN Xiao-lan1, CHENG Wu-xue1, WANG Qin1   

  1. 1. Research Center of RS & GIS Applications, Key Laboratory of Land Resources Evaluation and Monitoring in Southwest (Sichuan Normal University), Ministry of Education, Chengdu 610068, China;
    2. The Forestry Bureau of Suining City, Suining 629000, China
  • Received:2013-06-17 Revised:2013-09-16 Online:2014-03-20 Published:2014-03-20

Abstract:

Information of forest vegetation types is very important for ecological planning, protection and construction. In this paper, we discussed a method to extract vegetation types from multi-temporal MODIS imageries in order to overcome the limitation of singletemporal imagery in identifying vegetation types. The forest vegetation was classified into five types: the evergreen and deciduous mixed forest, the evergreen broadleaf forest, the evergreen coniferous forest, the deciduous broadleaf forest and the deciduous coniferous forest in Sichuan Province. The multi-temporal MODIS feature data were selected based on analyzing the growth difference of the vegetation types through a year. The multi-temporal Normalization Different Vegetation Index (NDVI) was calculated using the red band and near infrared band of MODIS images acquired on January 9, February 26, April 22, July 19 and October 23, which were respectively presented as NDVI(1-9), NDVI(2-26), NDVI(4-22),NDVI(7-19) and NDVI(10-23).
The knowledge "NDVI(1-9) > T1 and NDVI(10-23) > T2" for evergreen forest was discovered by multi-temporal image analysis, which was used to formulate model of extracting the evergreen feature of forest. The knowledge"NDVI(7-19) > T3, NDVI(2-26)< T4 and NDVI(4-22)> T5"for deciduous forest was discovered, which was used to formulate model of extracting the deciduous feature of forest. The knowledge"NDVI(1-9) > T6 and B2 < T7"for coniferous forest was discovered, which was used to create model of extracting coniferous forest. B2 is near infrared band of MODIS images acquired on January 9. The evergreen feature, deciduous feature and coniferous forest were obtained by using the models, multi-temporal NDVI and B2. The five vegetation types were obtained by judging and combining evergreen feature, deciduous feature and coniferous forest. The overall accuracy was about 84%. The lowest accuracy of vegetation type was 76%. It was shown that the method proposed here is labor-saving and cost-effective, which is of high application value in investigating and monitoring vegetation types in large extent. It was also shown that forest vegetation covered 28.43% of the total area of Sichuan Province in 2005. According to their percentage, vegetation types arranged in descending order were the deciduous broadleaf forest, the evergreen coniferous forest, the evergreen broadleaf forest, the deciduous coniferous forest and the evergreen and deciduous mixed forest, and the percentages were respectively 32%, 29%, 18%, 14% and 7%. The vegetation type data are of important value in protecting and utilizing the forest vegetation in Sichuan Province.

Key words: vegetation extraction, NDVI, spectrum feature, MODIS data

CLC Number: 

  • S771.8