U radu je analizirano ponašanje dnevnih razina podzemne vode mjerenih u tri duboka piezometra locirana u slivu Vranskog jezera na otoku Cresu. Obuhvaćeno je razdoblje od 10. rujna 1997. do 31. prosinca 2000., dakle ukupno 1209 dana. Ponašanje razina u sva tri piezometra bilo je gotovo identično, jer su se vrijednosti kvadrata koeficijenata linearne korelacije između parova piezometara kretale između R2=0,955 i R2=0,998. Izračunati su i dnevni dotoci, kao i gubitci iz jezera. Maksimalni srednji dnevni dotok u jezero u analiziranom razdoblju iznosio je Q=8,9 m3/s, dok je maksimalni srednji dnevni gubitak iz jezera iznosio Q=-1,9 m3/s. Analizirane su razlike između istovremeno mjerenih razina podzemne vode u svakom pojedinom piezometru i vodostaja jezera. Utvrđeno je da je na piezometru B1, udaljenom 335 m od obale jezera, razina podzemne vode samo 56 dana (4,6 % analiziranog vremena) bila viša ili jednaka vodostaju jezera. Na piezometru B2, udaljenom 540 m od obale jezera, razina podzemne vode bila je 262 dana (21,7 % analiziranog vremena) viša ili jednaka vodostaju jezera, dok je na najudaljenijem piezometru B3, izbušenom 790 m od obale jezera, razina podzemne vode bila viša ili jednaka vodostaju jezera 89,1 % analiziranog vremena. Ovakav odnos je neočekivan i teško ga je objasniti ili korigirati bez dodatnih mjerenja, a što je nužno s obzirom da postoji sumnja u realnost izmjerenih razina podzemnih voda. Činjenica da Vransko jezero na Cresu predstavlja u svijetu jedinstvenu vodnu masu formiranu na krškom otoku zahtijeva da se procesi koji se u i oko njega zbivaju pouzdano rasvijetle, što predstavlja osnovni cilj ovog rada..; The paper analyses the behaviour of groundwater levels recorded in three deep piezometers located in the Like Vrana basin on the Cres Island in the period from 10 September 1997 to 31 December 2000, i.e. a total of 1209 days. The behaviour of groundwater levels in all three piezometers was nearly identical, with the values of square coefficients of linear correlation among the piezometer pairs ranging from R2=0.955 and R2=0.998. The daily inflows and lake water losses were calculated as well. The maximum mean daily inflow into the lake in the analysed period equalled Q=8,9 m3/s, whereas the maximum mean daily loss from the lake was Q=-1,9 m3/s. The differences between simultaneously measured groundwater levels in each individual piezometer and the lake water level were analysed. It was determined that in the piezometer B1, 335 m away from the lake shore, the groundwater level was higher or equal to the lake water level for a period of only 56 dana (4.6 % of the analysed time period). In the piezometer B2, 540 m away from the lake shore, the groundwater level was higher or equal to the lake water level for a period of 262 dana (21.7 % of the analysed time period),whereas in the farthest piezometer B3, drilled 790 m away from the lake shore, the groundwater level was higher or equal to the lake water level 89.1 % of the analysed time period. This relation was unexpected and difficult to explain or correct without further measurements, which is necessary due to doubts as to the veracity of the measured groundwater levels. Lake Vrana on the Cress island, a unique water formed on a karst island mass in the world, deserves that the processes occurring in the lake itself and its environment should be reliably clarified, which is the primary objective of this paper.