Abstract:

Aims Our objective was to estimate grassland vegetation coverage (VC) in Inner Mongolia prairie, China, using a statistical model and a sub-pixel model and determine which model was more applicable in this area.

Methods Field experiments were conducted around three experimental stations in the Inner Mongolia prairie using a digital camera and a LAI-2000 plant canopy analyzer. A spectrum information model was used to extract the measured VC value from the photos. Statistical models were built between those VC values and six vegetation indexes. Then the VC map was made by using the model with the highest R². The measured leaf area index (LAI) values were used in the sub-pixel model to make another VC map.

Important findings The simple ratio vegetation index extracted from the Landsat-5 TM (thematic mapper) image had the higher correlation with VC values calculated from the photos taken in the field (R²= 0.761). The VC spatial distribution maps generated by the two models were generally similar, but the VC average value gained from the statistical model was 0.09 lower than that of VC values retrieved by the sub-pixel model. These two methods had similar results in the areas with higher and lower VC values, but the results from the sub-pixel model were higher than the statistical model with mid VC values.

Key words: Baiyanxile grassland in Inner Mongolia, grassland vegetation coverage, Landsat-5 TM image, statistical model, sub-pixel model