Soil hydraulic parameters are influenced by various inherent soil properties, such as pore structure and organic matter content, which can vary with changes in the ecosystem. However, identifying the temporal variations of soil hydraulic parameters in a co-evolving soil-vegetation system remains a challenge. This study focused on a tropical forest with significant seasonal variations in vegetation attributes, evaporation, and carbon fluxes over a five-year monitoring period. The particle batch smoother algorithm was integrated with an unsaturated flow model to identify the seasonally varied soil hydraulic parameters through assimilation of in-situ measured soil moisture. As a benchmark, the Generalized Likelihood Uncertainty Estimation method was applied to optimize soil hydraulic parameters without considering temporal variation. The results indicated that the temporally varying soil hydraulic parameters exhibited regular seasonal patterns and outperformed the unvaried soil hydraulic parameters in terms of reducing the errors in modeling of soil moisture and evaporation. Moreover, the seasonal variations in soil hydraulic parameters were closely linked to changes in the litterfall and terrestrial carbon fluxes over time. Specifically, due to the hysteresis of the transformation from litterfall to soil organic matter, the accumulated litterfall in Hot-dry season can replenish the soil organic matter, resulting in an increase in field capacity and saturated hydraulic conductivity in the Hot-rainy season. However, the intense decomposition of soil organic matter under high temperature in Hot-dry season led to a decrease in field capacity and saturated hydraulic conductivity. This study emphasizes the value of the particle batch smoother algorithm in detecting temporal variations in soil hydraulic parameters within a coevolving soil-vegetation system, thereby contributing to a more comprehensive understanding of the intricate dynamics within the ecohydrological system under a changing environment.