Gravity-based models have been extensively utilized in urban studies for measuring geographic disparities in access to urban parks over the past several decades. However, despite methodological advancements incorporating various aspects of accessibility, there has been limited focus on the impact of variable selection (e.g., attractiveness factors) and transport modes on accessibility evaluations. This study investigates the differences in gravity-based models for assessing park accessibility based on varying assumptions about attractiveness factors and travel impedance. Semi-structured interviews with local residents were conducted to identify the reasons for park visits in Shanghai. Our bivariate correlation analyses reveal that factors such as park openness and access to public transport were crucial, in addition to conventional factors identified in the literature (i.e., park size and driving accessibility). This insight led to the development of localized accessibility measurements that incorporate park inclusiveness (i.e., entrance fees and opening hours) and multimodal travel options (based on multinomial logistic mode choice models). The results indicate that the refined model produces lower and more varied accessibility levels, which can better capture accessibility gaps across different geographic contexts. This accurate and practical identification of accessibility gaps can assist local planners and decision-makers in formulating effective policies and strategies to promote equitable access to urban public parks.