Mapping soil total nitrogen in an estuarine area with high landscape fragmentation using a multiple-scale approach

The mapping of soil total nitrogen (STN) in estuarine areas is important for agricultural management and ecological conservation, but it is complicated due to high landscape fragmentation and complex influencing factors. In this study, a multiple-scale approach was proposed by adopting various predictors covering different aspects of spectral values, ecological indices, geographical position, land cover composition, and landscape fragmentation that represent comprehensive land surface characteristics. The former three aspects were independent of scales, whereas the latter two aspects were scale-dependent predictors. Multiple scales of 100 m, 200 m, 400 m, and 800 m were adopted for STN mapping using different algorithms to achieve the best simulation effect and search for the most suitable scale. Chongming Island, a typical and important estuarine area in China, was selected to demonstrate the approach. A partial least square regression method using a leave-one-out cross validation approach achieved the best simulation effect among all the algorithms. The results at the 100 m scale possessed the highest accuracy with a root mean squared error and a mean absolute error of 0.2541 and 0.2119 g/kg, respectively. In contrast, the results at the 200 m scale had the second highest accuracy and the lowest uncertainty. The STN possessed a mean value of 1.22 g/kg for the entirety of Chongming Island and it exhibited distinct spatial heterogeneity that was driven by complex influencing factors. Among all of the predictors, the vegetation condition and soil salinity contributed the most to the STN spatial variance. Human activity was the fundamental driving factor of the STN change, and it generally increased the STN because of long-term and island-wide agricultural development. Our approach is highly applicable in estuarine areas, and the scales of 100 m and 200 m can both meet the spatial heterogeneity demand, and thus, these scales are suitable for STN mapping.