In Network RTK (Real-Time Kinematic) positioning, the multiple corrections from the reference stations, which constitute a network, are interpolated for the user location through appropriate interpolation models. There exist various methods to model spatial decorrelation errors from the tropospheric and ionospheric delay, which are the main contributors of the multiple corrections. Since tropospheric delay is largely affected by height differences, the heights of the multiple reference stations should be considered when selecting the appropriate interpolation methods. This work provides a comparative analysis of the different levels of performance of each height-related multiple correction interpolation method. In addition, this study proposes to add constraints to the conventional height-related interpolation methods that are derived from the characteristics of the tropospheric zenith delay variation over height. The actual Global Positioning System (GPS) observations are collected from selected reference station networks located in the USA for performance evaluation. As a result, the proposed solution yields improved vertical positioning accuracy by approximately 10% compared to the conventional interpolation methods for the selected networks.