Accurate modeling of zenith tropospheric delay (ZTD) is beneficial for high-precision navigation and positioning. Many models with good performance have been developed for calibrating ZTD, such as the GPT3 model, which is recognized as an excellent global model and is widely used. However, certain limitations still remain in current models, such as the adoption of only single gridded data for modeling, and the model parameters need to be further optimized. In our previous research, a new approach based on the sliding window algorithm was proposed and applied to develop the GZTD-H model to address some of these limitations. However, this model is only suitable for the vertical adjustment of ZTD, not for estimating ZTD directly. In this study, an improved global grid ZTD model considering height scale factor (GGZTD-H) is derived from the initial GZTD-H model for estimating ZTD. The RMSs of the GGZTD-H model are 4.11 cm and 3.29 cm as validated by radiosonde data and IGS data, respectively. Compared with the UNB3m model and the canonical GPT3 model, the new model exhibits better performance. Moreover, three resolutions of the GGZTD-H model have been developed to reduce the quantity of gridded data delivered to users and optimize the ZTD computation process. Compared with the GPT3 model, the GGZTD-H model shows better performance with lower resolution and requires fewer model parameters for ZTD estimation, greatly optimizing ZTD computation. Users may select the best model that meets their needs in terms of the balance between resolution and accuracy. The high-precision GGZTD-H model could be used as a ZTD vertical stratification model for the vertical adjustment of atmospheric data and as an empirical model for ZTD estimation, which has potential applications in GNSS precise positioning, such as for the establishment and maintenance of the global terrestrial reference frame.