Soil chemical properties improve model performance in estimating ecosystem carbon exchange of cropland in China
摘要
Accurate assessment of carbon (C) budget in cropland provides a better understanding of the potential of agroecosystems to mitigate global greenhouse gas emissions. Cropland net ecosystem CO2 exchange (NEE) is one of the most crucial parts of terrestrial C fluxes because they are greatly influenced by human activities (e.g., field management), which is defined as ecosystem respiration minus gross primary productivity. However, the effects of soil physical and chemical properties associated with field management on cropland NEE have often been neglected in previous model studies. In this study, we simulated cropland NEE using random forest model combined with 49 observations from 18 site-level experiments in China from 2000 to 2019. The random forest model included 23 predictor variables, covering climate, vegetation, soil physical properties, soil chemical properties, farming practice, and location. Our results show that mean annual NEE was –338.30 g Cm−2 yr−1 in cropland of China over the past 20 years. Cropland NEE showed a decreasing trend in China from 2000 to 2019. The outputs of the random forest model showed that soil chemical properties were the most important variables for predicting cropland NEE, followed by soil physical properties, with relative importance values of 36.1% and 29.8%, respectively. Our results highlight the necessity to incorporate soil physical and chemical properties into cropland C flux simulations to improve the accuracy of C budget estimates. This study provided an updated data map of regional NEE in cropland for improving the assessment of terrestrial C cycle.