关键词: 城市化进程, 生态系统服务, InVEST模型, CLUE-S模型, 驱动因子

Abstract: Nitrogen (N) is an essential biogenic element in wetland ecosystems. It is very important to estimate total N (TN) concentration in wetland soil by the hyperspectral remote sensing data with nondestructive, quick and accurate quantification. In this study, Minjiang River estuarine wetland was chosen as the study area, and 80 samples of 16 soil profiles were collected along a hydrological gradient (from high tidal to middle tidal flat) in May, 2013. Soil spectral reflectance and TN concentration were determined in the laboratory. Estimation and validation models were constructed by original spectral reflectance (R) and spectrum parameters including ratio index (RI), normalized difference index (NDI) and deference index (DI). Moreover, the correlations of spectral reflectance with NH₄⁺-N, NO₃^--N, SOM and EC were analyzed in order to reveal the mechanism of estimating soil TN concentration based on hyperspectral remote sensing data. The results showed that the spectral reflectance of middle tidal soil was higher than that of high tidal soil at 350-600 nm, while the spectral reflectance of high tidal soil was higher than that of middle tidal soil at 600-2500 nm. Soil TN concentration showed a significant correlation with R at near 500 nm; the highest correlation coefficient value was -0.508, occurring at 490 nm. The spectrum parameters of RI, NDI and DI were calculated by bands at 600-1000 nm respectively, which greatly improved correlation coefficients with TN concentration, especially RI (590, 640), RI (610, 940), NDI (940, 590), NDI (940, 610), DI (640, 920) and DI (640, 940). The models built could well realize the inversion of wetland soil TN concentration in the study area, in which the determination coefficients (r²) and the root means square errors (RMSE) were all larger than 0.610 and less than 0.208, respectively. The best estimate parameter was RI (610, 940), and the r² values of its estimation and validation models were 0.832 and 0.631, while RMSE values were 0.178 and 0.202, respectively. The close relationship of soil TN concentration with SOM concentration is an important mechanism for estimating soil TN concentration, while NH₄⁺-N, NO₃^--N and EC had little impact on estimation accuracy of TN concentration.

Key words: ecosystem services, urbanization, InVEST model, driving factor, CLUE-S model