Soil color mapping based on Munsell system in the northeast of China

Soil color is the most readily available soil property, soil scientists often use the Munsell system to describe soil color and to quickly understand the physical, chemical, and biological properties of soil through soil color variation. Currently, soil color mapping studies often rely on indirect acquisition of Munsell color data from soil spectral information. As a result, there may be some discrepancies between the mapped colors and the colors perceived by humans. In this study, we collected 880 dry and 827 moist Munsell samples from the Northeast China to test the method developed for soil color mapping. We selected a set of covariates (climate, topography, evapotranspiration, regolith thickness, and remote sensing images) for digital soil color mapping, these covariates were related to soil formation or organic matter accumulation processes. Finally, the dry and moist topsoil color mapping with 30-m resolution was obtained for cropland in Northeast China. These covariates all played positive roles in the soil classification process, among which annual temperature mean (MAT), mean annual precipitation (MAP), land surface temperature (LST), and mean annual potential evapotranspiration (MAPET) had the highest contribution. We summarized the difficulties that usually occur in soil color mapping, and put forward a 'stepwise classification' method. According to the three-dimensional attributes of the Munsell system, the measured color data were divided into seven color groups (four dry and three moist color groups). Among all color groups, the highest accuracy of mapping validation was 0.69 and the lowest was 0.58. The kappa coefficient, usually affected by classification bias, was small (0.3). The color differences between measured and predicted colors were found to be 1.21 (dry color) and 1.36 (moist color). Compared to direct classification, the stepwise classification method improved classification accuracy and reduced mapping color differences. Our soil color mapping for two cases of dry and moist corresponded well with spectral color maps. The correlation between soil organic matter (SOM) content and soil color (dry and moist) was analyzed in 807 samples. The darker the soil color was, the higher the SOM content was. In summary, soil color holds the potential to serve as a surrogate for assessing cropland quality, particularly in evaluating SOM.