Abstract: Chlorophyll content is an important indicator of plant growth. The chlorophyll content of Stellera chamaejasme can provide a basis for both monitoring the growth and controlling the hazard of S. chamaejasme. A typical degraded meadow, which was dominated by S. chamaejasme in Xinghai County, Qinghai Province, was chosen for the experiment. Five methods were adopted to predict, contrast and analyze the SPAD values so as to construct the optimal prediction model of the chlorophyll content of S. chamaejasme in Qinghai Province, which included partial least squares (PLS) in the whole wavelength region of 400-1000 nm, multiple linear regression (MLR) and PLS based on successive projections algorithm (SPA), the red edge parameters and vegetation index. Results indicated that the optimal prediction performance was achieved by SPA-PLS model that was established by 9 characteristic wavelengths with SPA algorithm, and the correlation coefficient was predicted as 0.778, while the root mean square error was 1.895. Compared with the PLS model built on the full spectrum, the SPA-PLS model significantly reduced the computational complexity and improved the modeling efficiency. Compared with the SPA-MLR model, SPA-PLS model effectively solved the collinear problem among variables and also improved the forecasting accuracy, thus, it was the best model for predicting chlorophyll content of S. chamaejasme. Among predicting models built on the red edge parameters and vegetation index, a model constructed by MCARI index possessed the highest predicting accuracy with a correlation coefficient of 0.808 and a root mean square error of 1.969. Consequently, it could be the optimal vegetation index for inversing chlorophyll content of S. chamaejasme.

Key words: broadleaf plant, specific leaf area., leaf dry mass, leaf width, leaf area, leaf length