Aims The δ18O of soil evaporation (δE) is an important factor controlling the variations of atmospheric δ18O (δv), and it is also one of the key challenges of partitioning evapotranspiration into evaporation and transpiration components. δE is mostly simulated by the Craig-Gordon model, which is constrained by the δv of water vapor, the relative humidity (h), the equilibrium and kinetic factors and the δ18O of soil water (δs) at the evaporating front. Our objective is to investigate the diurnal variations of δE and factors affecting it. Methods We determined the δ18O of water vapor in a winter wheat-summer maize cropland based on the in-situ and continuous water vapor isotope ratio measurement system. We sampled soil water at different depths and analyzed it using the cryogenic vacuum distillation technique to acquire the δ18O of soil water at the evaporating front. Important findings During the growing period of winter wheat-summer maize, the diurnal variation of δE exhibited a bimodal pattern with peaks at 6:00 and 15:00. The h has a significant effect on the diurnal variation of δE in cropland ecosystems, and causes the Craig-Gordon model to be invalid under high humidity condition of h>95%. The in-situ and high resolution measurement of δv overcomes the uncertainty of using the local precipitation equilibrium method to evaluate δv, which improves the accuracy of δE. Different equilibrium factors have no significant influence on the accuracy of δE. Different kinetic factors, especially the canopy scale kinetic factor, influence the accuracy of δE significantly. The location of the evaporating front determines the h normalized to soil temperature and the δ18O of soil water directly and also influences the accuracy of δE significantly. Further research is needed to attain direct measurement of δE by combining isotope ratio infrared spectroscopy (IRIS) with the static chamber or dynamic chamber.