In this study, multi-source water samples from the soil-plant-atmosphere continuum (SPAC) were collected for hydrogen and oxygen stable isotopes measurement in an evergreen coniferous forest across the source area of Xin'an River during both the growing and non-growing seasons of 2021. By analyzing the differences and their variation characteristics of water isotopic compositions (δ¹⁸O and δ²H), we evaluated the influence of different sampling schemes (plant part and soil depth) on evapotranspiration partitioning in different seasons. Finally, the sample collection scheme based on isotope technique was optimized to partition evapotranspiration in forest ecosystems in humid areas of southern China. The results showed that the δ¹⁸O and δ²H of water from different sources were gradually enriched in the soil-plant water transport process, and more enriched in the non-growing season. The intensity of dynamic fractionation of water in different parts of plants increased gradually with the enrichment of isotopes. The distribution of isotopic compositions of stream water and spring water was similar, and the atmospheric water vapor was the most depleted compared with the water from other sources. The vertical distribution of isotopic compositions of soil water is different, which mainly presents three different forms:decreased with increased depth, increased first and then decreased, or decreased first and then increased. The variation of soil water isotopic compositions in shallow layer was larger than that in deep layer. The inflection points of δ¹⁸O and δ²H were ranged from 50 to 90 cm vertically. According to the distribution characteristics and differences of water isotopic compositions of soil and different parts of plant, phloem was the best sampling part that met the isotope steady-state assumption in Cunninghamia lanceolata forest. By comparing the T/ET (transpiration to evapotranspiration ratio) based on the isotopic compositions of water vapor in soil evaporation (δ_E) of different soil depths, it was found that the range of T/ET in the growing season was 13.46%, lower than that in the non-growing season (21.42%). It means that the variation of soil sampling depth has a greater influence on T/ET under relatively dry and cold condition, further it can be inferred that the suitable soil sampling depth of Cunninghamia lanceolata is about 20-30 cm. The research results can provide a scientific basis in sampling scheme design and model establishing of evapotranspiration estimation, and lay an effective foundation for the separation of evapotranspiration components in ecosystems and the traceability of plant transpiration water across other humid and semi-humid regions.