Differences in rainfall partition into throughfall among different primeval forests distributed along an altitude gradient are inadequately investigated and understood. Through continuous and automatic monitoring of natural rainfall and throughfall along an elevation gradient on Mountain Gongga, we examined the response of throughfall to various rainfall patterns in the broadleaved forest (BF), broadleaved-coniferous mixed forest (MF), and coniferous forest (CF) across individual rain events from May to October in 2019. A series of linear models that estimate throughfall amount were obtained and compared. Results showed that throughfall was jointly controlled by rainfall characteristics (including amount, duration, average, and peak intensity) and leaf area index (LAI). Rainfall amount was the primary control for throughfall amount. The models with all rainfall parameters and LAI as variables did not markedly outperform (R-2 enhancement by 0-0.02) the simple linear models with rainfall amount as the only variable; therefore, the latter are recommended due to simplicity and easiness of use. Although the correlation of throughfall with LAI was less prominent compared to rainfall parameters, LAI showed a significant positive linear correlation (p < 0.05) with the estimated rainfall amount threshold (the rainfall required to saturate the canopy) by the single-variable linear models at the monthly scale. Over the study period, penetration proportions of rainfall in BF, MF, and CF were 83%, 75%, and 80%, respectively. The rainfall amount threshold in CF (0.70 mm) was less than those in BF (0.80 mm) and MF (0.92 mm). Rain events of higher intensity exhibited a higher mean penetration proportion than lower intensity rain events. The use of single-variable linear models developed here, despite some overestimations of throughfall amount, could lead to an overall satisfactory estimation of rainfall redistribution in mountainous areas.