闽江河口湿地土壤全磷高光谱遥感估算

磷是湿地生态系统必需和限制性元素,利用高光谱遥感数据对其进行估算对实现湿地土壤磷素快速和准确定量具有重要意义。选取闽江河口湿地作为研究区,于2013年5月,采集16个土壤剖面80个样本作为估算与验证模型样本;基于光谱指数建立土壤全磷(TP)含量估算模型,其中光谱指数包括原始光谱反射率(R)、比值土壤指数(RSI)、归一化土壤指数(NDSI)和有机质诊断指数(OII)。此外进一步分析反射光谱与不同形态磷,TP与有机质之间关系,以期初步揭示河口湿地土壤TP估算的机理。研究结果表明,闽江河口湿地土壤TP含量与R相关系数较高的区域分布在360-560 nm,并在406 nm处达到最大值-0.816;光谱指数RSI(R_(430),R_(830))、RSI(R_(460),R_(810))、RSI(R_(560),R_(580))、NDSI(R_(430),R_(830))、NDSI(R_(460),R_(830))、NDSI(R_(560),R_(580))和OII(R_(446))与土壤TP含量均有较高的相关系数,能较好的用于TP含量的估算;各估算模型决定系数(r~2)和均方根误差(RMSE)分别在0.657-0.805和0.052-0.067之间;验证模型r~2和RMSE分别在0.606-0.893和0.037-0.044之间。分潮滩建立TP含量估算模型是可行的,并且能提高部分光谱指数的估算精度。土壤TP含量的估算精度与磷素的组成有关,其中与铁吸附态磷关系较为密切,钙吸附态和铝吸附态磷关系较弱。土壤TP与有机质和氧化还原环境的存在密切关系可能是湿地土壤TP含量估算的重要机理。

Estimating the soil total phosphorus content based on hyper-spectral remote sensing data in the Min River estuarine wetland

Phosphorus (P) is an essential and limiting nutrient in wetland ecosystems and it plays a vital role in eutrophication. Remote sensing (RS) offers an up-to-date and relatively accurate means to measure the soil P content. Recently, some studies have shown that it was feasible to estimate the total P (TP) content of terrestrial ecosystem soil based on hyper-spectral RS data. However, little information is available on TP content estimation by RS technology on wetland soil. The aim of this study was to estimate the TP content of wetland soil using hyper-spectral RS data. Min river estuarine wetland, located in the subtropical zone, is one of the most typical and important estuarine wetlands in southeast China. Soil samples, from Shanyutan tidal marsh in the Min River Estuary, were collected in sixteen profiles at five depths (0-10 cm, 10-20 cm, 20-30 cm, 30-40 cm, and 40-50 cm) along an elevation gradient, in May of 2013. Estimation and validation models were constructed by spectrum parameters, including original spectral reflectance (R), simple ratio soil index (RSI), normalized difference soil index (NDSI), and organic matter diagnosis index (OII) calculated by optimal bands. The results indicated that the spectral reflectance of the soil increased with depth at 350-580 nm, while an opposite trend was observed at 580-2500 nm. Soil TP content showed a negative correlation with R at 350-600 nm, whereas a positive correlation was observed at 600-2500 nm. The highest correlation coefficient value was -0.816 and occurred at 406 nm. The correlation coefficient between soil TP content and OII exhibited a bimodal distribution, with peaks at 446 nm (r = -0.843) and 634 nm (r = 0.798). NDSI and RSI were each calculated by bands in three zones, (420-440 nm and 440-590 nm, 460-470 nm and 590-1000 nm, and 550-590 nm and 550-590 nm, respectively), which had higher correlation coefficients with TP content than those in other zones. The determination coefficient (r²) and root means square error (RMSE) of estimation models ranged from 0.657-0.805 and 0.052-0.067, respectively, and those of the validation models ranged from 0.606-0.893 and 0.037-0.044, respectively. These results indicate that TP content of the Min River estuarine soil could be estimated by most of the selected parameters. The evaluation parameters of the estimation models supported that estimating the TP content of high and middle tidal flats soil individually could improve the estimation accuracy of some parameters such as RSI(R₄₃₀, R₈₃₀), RSI(R₄₆₀, R₈₁₀), and NDSI(R₄₃₀, R₈₃₀). Additionally, the estimation accuracy of soil TP content also depended on the P fractions. Iron bound phosphorus (Fe-P), occluded phosphorus (O-P), and organic phosphorus (Org P) had higher correlation coefficients with R than did aluminum bound phosphorus (Al-P) and calcium bound phosphorus (Ca-P). The corresponding changes in the contents of TP within organic matter and a redox environment in wetland soil could be used as important mechanisms for estimating soil TP content. In conclusion, it was feasible to estimate TP content of subtropical estuarine wetland soils based on hyper-spectral RS data.