The last improvement of levelling records in Riga city dated at 1975-77. The first triangulation network was developed during 1882-85. 288 levelling benchmarks were fixed at 1880-82. Currently it is available un Rigas GeoMetrs SIA archive. 200km was levelled with an accuracy 2,7-3.7 mm/km. 1902-1906 is the next epoch of the extension of Riga levelling network with an accuracy 1.29-2.18 mm/km. Properly made and calibrated rod was used, additionally new benchmarks were installed. The number of benchmarks achieved 743 together with 222 still existing from the first network. The national 1st order levelling network covering whole territory of Latvia was developed in 1929-1939. The levelling results were published at 1941 and the catalog is available now. The improved Riga 2nd order levelling network was develoed in period 1933-1940. High quality 171 benchmarks were fixed in solid walls of stable houses and bridges. Levelling length of 196.9 km in 13 loops were carried out with an accuracy of 1.38 mm/km. The network was tied to the national levelling network. The levelling time span appears 7 years long. The discrepancies appeared within the range of 0 to -28 mm. It was assumed that the vertical movement occurred. Levelling RMS were different in range of 0.86 mm/km till 4.72 mm/km. However, the Riga levelling average RMS is 1.8 mm/km. In order to tie the Latvian national levelling network to Soviet one at 1947/48 the Lomonosovo-Riga and Riga Baranovitch 1st order levelling were performed by Soviet geodesists. Commencing at 1967 the II order levelling by Soviets were done in Latvia. The last levelling network adjustment were performed at 1977 and the resulted system is named Baltic 1977 normal height system. The last 2nd order levelling campaign in Riga were carried out at 1975-1977 by Leningrad geodetic enterprise the vertical movement of Riga levelling network discovered recently by using RTK DGNSS technique in Riga the Sea level rise is about 26 cm in 100 years. Mainly DGNSS and SLR technologies should be used for EUVN development (Fig.9). J.Šimek tests (Fig. 10) approve the instability of the ground level. It seems that several years long levelling campaigns are capable to represent the random epochal elevations between closest benchmarks.