THE USE OF MULTITEMPORAL NEAR_GROUND SPECTRAL REFLECTANCE TO DISCRIMINATE AMONG DEGRADED GRASSLANDS
利用多时相近地面反射波谱特征对不同退化等级草地的鉴别研究

From May to September 1998, we measured the seasonal near-ground spectral reflectance characteristics and biological parameters of non-degraded, moderately (or lightly) degraded and heavily degraded grassland communities of Leymus chinensis steppe and Stipa grandis steppe in Inner Mongolia, China. The results of Duncan multi-variance analysis indicated that the near-ground spectral reflectance varied with the degree of degradation, growing season and grassland type. For Leymus chinensis steppe grazing gradient, differences in the near-ground spectral reflectance among degraded communities were most distinct when measured at the end of July and were least at the end of May. For the Stipa grandis steppe grazing gradient, the spectral response was most distinct at the end of July and August and was least distinct at the end of May and June. No distinct differences in the Normalized Difference Vegetation Index (NDVI) among degraded community types were observed at the end of May, June and September. At the end of July and August, differences were observed but the degree of difference of the NDVI was less than that of the near-ground spectral reflectance. PCA and Pattern-Recognition Method, used to develop discriminate functions, indicated that the blue, red and near-infrared reflectance captured at the end of June had the greatest ability to discriminate among degrees of degradation along the L. chinensis grazing gradient ( mean probability of error was 0.7%) and had the poorest discriminatory power at the end of May (the mean probability of error was 12%). For S. grandis steppe grazing gradient, the discrimination ability was highest at the end of June and August when composed of blue, green and near-infrared reflectance (mean probability of error was lower than 4%) and lowest at the end of September (mean probability of error was about 10%).