Quantitative evaluation of different soil–water characteristic curve models on bare soil evaporation simulation

• FX-integrated model shows superior accuracy in predicting evaporation for BCS and NS. • FX-integrated predicts cumulative evaporation with variances of −6.34%, 10.01%, and 11.25%. • VG-integrated and LN-integrated models accurately capture NS evaporation rates. • FX-integrated model excels in simulating bare soil evaporation due to precise air-entry value estimation. The bare soil evaporation process is the link between the atmosphere and soil surface in the hydrologic cycle and, therefore, is a key issue in many fields of hydrological sciences. The selection of different soil–water characteristic curve (SWCC) models is essential to simulate bare soil evaporation processes. Remarkably, there is a notable lack of understanding regarding the quantitative characterization of the impact of some SWCC models on the simulation accuracy of bare soil evaporation, including the van Genuchten model (VG model), Brooks-Corey model (BC model), Fredlund and Xing model (FX model), Log-Normal Distribution model (LN model), modified van Genuchten model (mVG model), modified Brooks-Corey model (mBC model) and modified Log-Normal Distribution model (mLN model). In our study, to evaluate different SWCC methods for the estimate of evaporation frombare soils, we collected three distinct sets of evaporation data from column tests representing different lithologies. Utilizing the numerical simulation, we integrated VG, BC, LN, FX, mVG, mBC and mLN models to Richards equation to construct the simulation models (abbreviated as VG-integrated, BC-integrated, LN-integrated and FX-integrated models, respectively) of unsaturated water flow, comparing the evaporation rates and cumulative evaporation obtained from those integrated models. The FX-integrated model exhibited superior accuracy in predicting evaporation dynamics for Beaver Creek sand (BCS) and Natural silt (NS), with slightly diminished performance for Coarse sand (CS). The FX-integrated model predicts cumulative evaporation pretty well for BCS, NS, and CS, with variances of −6.34%, 10.01%, and 11.25%, respectively. The VG-integrated and LN-integrated models captured the experimentally measured evaporation rates of NS well, with the values of R 2 equal to 0.9390 and 0.9467, respectively. The BC-integrated model excelled in simulating CS with the values of R 2 equal to 0.9409. The modified integrated model group (mVG-integrated, mBC-integrated, and mLN-integrated model) exhibits systematic improvements—particularly the mBC-integrated model achieves enhanced CS evaporation rate predictions ( R 2 = 0.9897 vs. BC-integrated’s 0.9409) with 69% lower root mean square error ( RMSE ; 0.25 vs. 0.83 mm/d), but their performance in simulating BCS and NS evaporation remains inferior to the FX-integrated model. Further analysis underscores the FX-integrated model’s superiority in simulating bare soil evaporation due to the FX model’s ability to estimate air-entry values and fitting SWCC dry-end data more accurately than the VG, BC, and LN models. Consequently, our findings suggest that the FX-integrated model is the preferred choice for simulating bare soil evaporation. The research findings provide practical guidance, especially in accurately assessing evaporation under sustained evaporation conditions in arid areas.