EGU2020-6314, updated on 12 Jun 2020
EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

How biotic and abiotic factors affect stemflow production? Insights from both local and global scales

Yafeng Zhang, Xinping Wang, Yanxia Pan, and Rui Hu
Yafeng Zhang et al.
  • Shapotou Desert Research and Experiment Station, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou, China (

Stemflow production has been reported to be influenced by a suite of biotic and abiotic factors, and those factors would be quite different considering local and global scales. Although the number of published stemflow studies showed a steady increasing trend in recent years, the relative contributions of biotic and abiotic factors to stemflow production were still largely unclear due to the large number of influencing factors and the complex interactions among those factors. Here we present stemflow results conducted from both from local scale and global scale: (1) stemflow of nine xerophytic shrubs of Caragana korshinskii were measured in nearly nine growing seasons from 2010 to 2018 within a desert area of northern China, accompanying with observing on six biotic variables (shrub morphological attributes) and ten abiotic variables (meteorological conditions); (2) a global synthesis of stemflow production results (stemflow percentage was reported) derived from Web of Science for more than 200 peer-reviewed papers published in the last 50 years (1970-2019), and ten most reported biotic factors (vegetation life form, phenology, leaf form, bark form, community density, community age, vegetation height, diameter at breast height, leaf area index, stemflow measuring scale) and four abiotic factors (climate types, mean annual precipitation, elevation, mean annual temperature) were considered. We performed a machine learning method (boosted regression trees) to evaluate the relative contribution of each biotic and abiotic factor to stemflow percentage, and partial dependence plots were presented to visualize the effects of individual explanatory variables on stemflow percentage, respectively.

How to cite: Zhang, Y., Wang, X., Pan, Y., and Hu, R.: How biotic and abiotic factors affect stemflow production? Insights from both local and global scales, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-6314,, 2020

Display materials

Display file

Comments on the display material

AC: Author Comment | CC: Community Comment | Report abuse

Display material version 1 – uploaded on 02 May 2020
  • CC1: Comment on EGU2020-6314, Anke Hildebrandt, 06 May 2020

    This is really impressive data collection.

    You state that the contribution of biotic and abiotic factors (combined) was below 20%, but the model fit was rather good. Was the remainder attributed to differences between shrub individuals? Or individual events?

    • AC1: Reply to CC1, Yafeng Zhang, 06 May 2020

      Thank you Anke for your interesting.

      Actually, the default total contribution of all biotic and abiotic variables are 100%. What you pointed out I assume may the mean relative contribution of each variable in two categories in the bottom right table of slide 14, respectively. Percent deviance explained (81% for SFv, 66% for SFp, and 49 % for FR) generated by BRT in the top right table of slide 14 is a statistical measure of fit analogous to the R-square value.

      For detailed information, please refer to our recently published paper in AGRFORMET: 10.1016/j.jhydrol.2015.05.060

      The best,



      • CC2: Reply to AC1, Anke Hildebrandt, 06 May 2020

        Thank you for the response! I understand now. I misinterpreted the meaning "mean relative contribution". Thanks for the clarification.

    • AC2: Reply to CC1, Yafeng Zhang, 06 May 2020

      For the former reply, I may not explain clearly. In BRT, the relative influence of each variable is scaled so that all variables (in our case we have 14 variables) included in the model sum to 100, with higher values indicting stronger influence on the response. Percent deviance explained shows how good the model is.