[Identifying Driving Factors and Their Interacting Effects on Sources of Heavy Metal in Farmland Soils with Geodetector and Multi-source Data].

The identification of heavy metal sources in farmland soils is essential for the rational health condition management and sustainable development of soil. Using source resolution results(source component spectrum and source contribution)of a positive matrix factorization(PMF)model, historical survey data, and time-series remote sensing data, integrating a geodetector(GD), an optimal parameters-based geographical detector(OPGD), a spatial association detector(SPADE), and an interactive detector for spatial associations(IDSA)model, this study explored the modifiable areal unit problem(MAUP) of spatial heterogeneity of soil heavy metal sources and identified the driving factors and their interacting effects on the spatial heterogeneity of soil heavy metal sources in categorical and continuous variables, respectively. The results showed that the spatial heterogeneity of soil heavy metal sources at small and medium scales was affected by the spatial scale, and the optional spatial unit was 0.08 km2 for detecting spatial heterogeneity of soil heavy metal sources in the study region. Considering spatial correlation and discretization level, the combination of the quantile method and discretization parameters with an interruption number of 10 could be implied to reduce the partitioning effects on continuous variables in the detection of spatial heterogeneity of soil heavy metal sources. Within categorical variables, strata(PD 0.12-0.48) controlled the spatial heterogeneity of soil heavy metal sources, the interaction between strata and watersheds explained 27.28%-60.61% of each source, and the high-risk areas of each source were distributed in the lower sinian system, upper cretaceous in strata, mining land in land use, and haplic acrisols in soil types. Within continuous variables, population (PSD 0.40-0.82) controlled the spatial variation in soil heavy metal sources, and the explanatory power of spatial combinations of continuous variables for each source ranged from 61.77% to 78.46%. The high-risk areas of each source were distributed in evapotranspiration (41.2-43 kg·m-2), distance from the river (315-398 m), enhanced vegetation index (0.796-0.995), and distance from the river (499-605 m). The results of this study provide a reference for the research of the drivers of heavy metal sources and their interactions in arable soils and provide an important scientific basis for the management of arable soil and its sustainable development in karst areas.

[Influence of sensor spectral parameters on the simulation of hypespectral data based on the spectral reconstruction approach].

Data simulation has been widely used in various applications of remote sensing, especially in design of new type sensors, test of new developed algorithms and other associated applications. However, change of parameters of sensor and its system can affect the accuracy of data simulations. Based on spectral reconstruction, the present study employs convolution of spectral response function (SRF) of four bands-blue, green, red, and infrared red within the wide field of view multispectral imager to analyze the impact central wavelength and bandwidth have on the accuracy of spectral reconstruction. The results show that root mean square error (RMSE) caused by central wavelength displacement is less than 0.025, while RMSE caused by bandwidth shift is less than 0.012, indicating the good accuracy of data simulations. Apparently, central wavelength and bandwidth have impact on accuracy of spectral reconstruction to some extent. Therefore, hyperspectral reconstruction depending on central wavelength and bandwidth is conducive for users to further understand hyperspectral imaging system, to find the main factor(s) influencing system performance, to better simulate hyperspectral data and to broaden the application range of remotely sensed data.