The determination of geodetic heights with high accuracy is one of the main tasks in geodesy. In 2015, a resolution for the definition of the International Height Reference System (IHRS) and its realization by the International Height Reference Frame (IHRF) was released by the International Association of Geodesy (IAG). In the past years, intensive efforts have been devoted to developing the IHRF and its temporal changes. The main goal of the proposed study is to estimate temporal variations of orthometric/normal heights (ΔH/ΔH*) at the locations of proposed IHRF sites using data from space geodetic techniques (GRACE, GNSS, SLR, VLBI, and DORIS) and hydrological models. Moreover, it is also aimed at investigating the possibilities of estimating ΔH/ΔH* in (near) real-time. Monthly GRACE/GRACE-FO-based and GLDAS-based Global Geopotential Models were utilized to estimate ΔH/ΔH* at all proposed IHRF sites. Moreover, ellipsoidal height changes from GNSS, SLR, VLBI, and DORIS stations co-located with the proposed IHRF sites were utilized to estimate those ΔH/ΔH*. Machine learning with a long short-term memory networks approach and an empirical approach based on a seasonal decomposition method used historical data to estimate (or extrapolate) ΔH/ΔH* for the current month at selected IHRF sites distributed worldwide and characterized by different mass change signals. The results obtained were analyzed and discussed considering the contemporary requirements for precise height measurements as well as the modernization and unification of international/national levelling networks. Keywords: orthometric/normal heights changes, reference system/frame, IHRF, space geodetic techniques, hydrological models