Water retention (WR) services of ecosystems can be evaluated using various methods based on different theories and data sources. However, there was still no quantitative comparison to comprehensively evaluate the variations and consistencies of results from different methods. To determine the influences of different methods, we used two methods, including ‘site data-comprehensive storage capacity method-meta-analysis’ and ‘spatial data-water balance method-InVEST model’, to compare forest WR services in Fujian Province, China. The results showed that the mean annual capacity of WR was 529.62 mm (653.68 mm), and the total amount of WR was 236.18×10⁸ m³ (291.07×10⁸ m³) from the InVEST model (meta-analysis). The WR capacities showed a large spatial variation, which were higher in the western and northern mountains than in the southeastern coastal plain. Meteorological, topographical and land cover factors were considered in the InVEST model, whereas specific stand conditions were applied in the meta-analysis. Therefore, the results of WR capacities from different methods in the same forest cover were different. In addition, we also conducted a site-scale analysis, and found the uncertainty sources of results were from multiple definitions of WR parameters, inconsistent measurement methods, measurement errors, and differences of coefficient localization. Our findings provide references for selecting appropriate evaluation methods in different regional studies and improving the theory and technology of evaluation in ecosystem WR services.