Prediction of soil organic carbon and total nitrogen in mattic layer of typical distribution area of alpine meadow on Qinghai-Tibet plateau
Description
The Mattic layer (ML) is the surface layer of alpine meadow soil, which is characterized by a highly developed and dense complex of living and dead Kobreia roots, humus, and minerals. This layer is rich in organic matter and roots and contributes significantly to nutrient cycling, water retention, and ecological stability in alpine meadow ecosystems. Some studies have reported increasing degradation of the mattic layer due to global warming exacerbated by global warming, overgrazing, and human activities, which reduces the carbon and nitrogen fixation capacity of the soil and may accelerate the release of these elements into the atmosphere, creating feedback loops that exacerbate global warming. We selected two typical distribution areas of alpine meadows on the Qinghai-Tibet Plateau, set up sample sites, collected soil in the mattic layer, and measured the organic carbon and total nitrogen. Combined with high-resolution terrain and vegetation data obtained by UAV, we accurately predicted the distribution of soil organic carbon and total nitrogen in the sample area. The study found that the UAV had excellent utility in predicting soil organic carbon and total nitrogen in the mattic layer. At the same time, the uncertainty of soil organic carbon and total nitrogen distribution was deeply analyzed, and the factors affecting the distribution of soil organic carbon and total nitrogen were revealed. The distribution of SOC and TN is highly uncertain in areas with large topography, and topography is the most important factor controlling the distribution of SOC and TN. These maps can play a role in grazing and soil management.
Files
Steps to reproduce
We selected two typical distribution areas of alpine meadow in the Qinghai-Tibet Plateau, and set up 62 sampling points in the sample areas by Latin hypercube sampling. At each sampling point, the mattic layer profile was dug down along the surface, the thickness of the mattic layer was determined and recorded, and a 1500 g soil sample was collected from the mattic layer. In addition, a 1 m 2 plot was established near each soil sampling site, and a core of 5 cm in diameter and 10 cm in height was used to collect root soil samples. Soil organic carbon and total nitrogen were measured in the soil samples of the substrate layer, and the volume ratio of root to soil and the mass ratio of root to soil were measured in the root to soil ratio samples. High resolution (1×1m) digital elevation model (DEM) and multispectral images of the two sample areas were collected by UAV. Digital elevation model (DEM), multispectral image, and root-soil ratio data were processed to obtain topographic, multispectral, and bioenvironmental covariates, respectively. The environmental covariates and soil organic carbon and total nitrogen measured values were used to construct the original database. The regression model of soil organic carbon and total nitrogen was constructed by random forest, gradient boosting machine, and geographic weighted regression in R, and the regression model was corrected by residual Kriging. At the same time, in R, we used bootstrap to quantify the uncertainty of soil organic carbon and total nitrogen in the sample area, and evaluated the variable importance of the model based on SHAP. Finally, the content distribution map and uncertainty distribution map of soil organic carbon and total nitrogen in the sample area were drawn respectively.
Institutions
- Shenyang Agricultural University