期刊
GEODERMA
卷 323, 期 -, 页码 31-40出版社
ELSEVIER
DOI: 10.1016/j.geoderma.2018.02.031
关键词
Heterotrophic soil respiration; Environmental soil classes; PLSR; Random Forest
类别
资金
- Deutsche Forschungsgemeinschaft (DFG) [SFB/TR32]
Spatial patterns of soil respiration (SR) and its sensitivity to temperature (Q10) are one of the key uncertainties in climate change research but since their assessment is very time-consuming, large data sets can still not be provided. Here, we investigated the potential of mid-infrared spectroscopy (MIRS) to predict SR and Q10 values for 124 soil samples of diverse land use types taken from a 2868 km(2) catchment (Rur catchment, Germany/Belgium/Netherlands). Soil respiration at standardized temperature (25 degrees C) and soil moisture (45% of maximum water holding capacity, WHC) was successfully predicted by MIRS coupled with partial least square regression (PLSR, R-2 = 0.83). Also the Q10 value was predictable by MIRS-PLSR for a grassland submodel (R-2 = 0.75) and a cropland submodel (R-2 = 0.72) but not for forested sites (R-2 = 0.03). In order to provide soil respiration estimates for arbitrary conditions of temperature and soil moisture, more flexible models are required that can handle nonlinear and interacting relations. Therefore, we applied a Random Forest model, which includes the MIRS spectra, temperature, soil moisture, and land use as predictor variables. We could show that SR can be simultaneously predicted for any temperature (5-25 degrees C) and soil moisture level (30-75% of WHC), indicated by a high R-2 of 0.73. We conclude that the combination of MIRS with sophisticated statistical prediction tools allows for a novel, rapid acquisition of SR and Q10 values across landscapes and thus to fill an important data gap in the validation of large scale carbon modeling.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据