Grassland Biomass On North-western Plateau Of Sichuan And Vegetation Indexes Relation Using Landsat TM Image

Grassland ecosystem is the most important and most widely distributed type of ecosystems.But it is more fragile ecosystems,easily affected by natural factors and man-made factors.And grassland biomass is the main criteria to reflect and assess o grassland ecosystem.At present,the use of remote sensing technology for monitoring natural grassland biomass has become a scientific research at the forefront of international grass topics,but results are very different.It's not really predictable which kind of vegetation index and model of relationship between vegetation index and grassland biomass be more suitable for monitoring regional grassland biomass at this point.It is in urgent need of solving the problem of contemporary grassland and remote sensing science.Using thematic mapper(TM) images for North-western Plateau of Sichuan area in 2005 and extracting vegetation indexes(NDVI,RVI,DVI,SAVI,MASVI,PVI and GVI) from the image of the research region,the monadic linear regression models and the non-linear regression models were established,respectively,to express the relations between grassland biomass and the vegetation indexes.Results showed that the correlations between sampled biomass and the seven vegetation indexes were highly positive significant,with RVI being highest(0.884),NDVI and SAVI 0.850,again GVI, MASVI,DVI being respectively 0.837,0.822,0.813,PVI lowest,at 0.760.Generally speaking,compared to the monadic linear regression(R~2 = 0.781),there was an increase in fitting accuracy of curve models,such as cubic polynomial equation(R~2 = 0.816), Second-degree polynomial(R~2 = 0.788) and Power(R~2 = 0.782).So the curve models were better to reflect the relationship between vegetation index and measured biomass.For VI-biomass regression model,the cubic polynomial model was better than the monadic linear regression models and the other non-linear regression models(Power, Exponential,Logarithm,Second-degree polynomial,S curve,Growth).And multiple correlation coefficients(R~2 = 0.816) of the cubic polynomial model based on RVI was higher than the others,and the error of the model verified by the observation value was very small,with an average error of 13%,fitting accuracy of 87%,not only indicating that it was better suited to monitor the grassland growing of North-western Plateau of Sichuan,and that it met regional grassland monitoring needs,but also the RVI cubic polynomial model to monitor regional grassland biomass was a simple and effective and practical method.