Airborne pollen is a significant risk factor for allergic diseases. However, the specific health risks posed by different pollen types and across seasons remain incompletely understood. Utilizing data from Yulin, a representative city in the semi-arid region of China (January 2023 to December 2024), we conducted a time-series analysis with a Generalized Additive Model (GAM). The model integrated daily data on specific pollen deposition counts (Cupressaceae, Artemisia, Chenopodiaceae, etc.), meteorological conditions, air pollutants, and number of pediatric outpatients for allergic diseases (allergic rhinitis, asthma, and wheeze). The pollen deposition counts exhibits pronounced seasonal characteristics, with Artemisia pollen average deposition counts and relative risks are higher during the pollen season (June-October). A significant nonlinear positive exposureresponse relationship was observed. Across the different study periods, the relative risk is higher in the summer-autumn pollen peak (August to October). The optimal lag day for relative risks varies by disease: immediate effects (lag0) were identified for allergic rhinitis, whereas asthma exhibited a delayed response (lag3). Our findings demonstrate that in semi-arid environments, the risk of pollen-induced allergic diseases in children shows temporal heterogeneity, with critical disparities driven by pollen types and seasons. This evidence underscores the necessity of implementing pollen-specific and seasonally-targeted strategies for public health interventions and urban greening plans in comparable regions worldwide.