Successful use of stable isotopes (δ²H and δ¹⁸O) in ecohydrological studies relies on the accurate extraction of unfractionated water from different types of soil samples. Cryogenic vacuum distillation (CVD) is a common laboratory-based technique used for soil water extraction; however, the reliability of this technique in reflecting soil water δ²H and δ¹⁸O is still of concern. This study examines the reliability of a newly developed automatic cryogenic vacuum distillation (ACVD) system through a set of pure water extraction. The impacts of extraction parameters (i.e., extraction time, temperature, and vacuum) and soil properties on the recovery of soil water δ²H and δ¹⁸O were further assessed for the ACVD and traditional extraction systems (TCVD) systems. Finally, the potential influences of the CVD technique on the prediction of plant water uptake were investigated through a sensitivity analysis. We demonstrated that the ACVD system was reliable for recovering the isotopic composition of pure water, with negligible biases of − 0.1 ± 0.3‰ for δ²H and 0.04 ± 0.09‰ for δ¹⁸O. Both ACVD and TCVD similarly extracted water from the rewetted soils when the extraction time of the ACVD system reached 240 min, but none of the CVD systems successfully recovered the isotopic signatures of doped water from soil materials. Mean δ²H offsets of extracted soil water were − 2.6 ± 1.3‰ and − 2.4 ± 1.7‰ for ACVD and TCVD, respectively; while mean δ¹⁸O offsets were − 0.16 ± 0.14‰ and − 0.39 ± 0.37‰. The isotopic offsets of CVD systems were positively correlated with soil clay content, and negatively correlated with soil water content. The use of corrected soil data (with CVD offsets) could improve the prediction of plant water uptake based on its high correlation with environmental factors. This study identifies the isotopic offsets of CVD systems (i.e., ACVD and TCVD) and provides possible solutions for better-predicting plant water sources. Even so, the wide use of CVD techniques probably induces noticeable uncertainties in predicting plant water uptake depths. The dataset of soil water extraction in this study will have implications for the technological development of CVD techniques.

The isotopic data are collected from both the laboratory and the field. This study assessed the CVD approaches for stable isotope analysis and plant water source prediction through the following consecutive experiments. Experiment (Ⅰ): 56 pure water samples of known isotopic values were extracted using the newly designed ACVD system. Experiment (Ⅱ): 225 soil samples containing various rates of clay particles were oven-dried, doped with deionized water, and cryogenically extracted using ACVD and TCVD systems. Experiment (Ⅲ): a sensitivity analysis of plant water uptake under four types of input data was performed among seven sites in China.