The section starts at Coștei

Ground from the Copernicus GLO-30 model, a surface model (stated vertical error: under 4 m at 90% of points). Water from surveyed heights (2–3 m of error between gauges), at the level of the months when the canal is held full. The section links the ends of the two canals, Timișoara with Șag and the Itebej lock with Jaša Tomić. Heights in metres above sea level. Heights are greatly exaggerated against distances.

Banat, western Romania · from Coștei to the Tisza

The Bega Canal enters Serbia five metres higher than the river Timiș

The Bega and the Timiș leave the hills of western Romania a few kilometres apart and cross the same plain. Where they enter Serbia, 17 km apart, Serbia’s hydrometeorological service reads the water level of each every day; the readings give the height of the water above sea level. From 2004 to 2025 there are 8,036 measured days. On 7,974 of them, 99%, the Bega’s water stood higher than that of the Timiș, by a median of 5.0 m. On the other 62 the Timiș was in flood and, on 56 of them, the canal was not being held full.

Where they leave the hills, at Coștei, it is the other way round: the Timiș is higher, by 5 to 8 m, and gives the Bega water through a canal. From there on the Timiș falls much faster, and from Topolovăț to the border the Bega is higher: by about 4 to 5.5 m (at Topolovăț, an estimate; at Timișoara, against the Timiș gauge at Șag; at the border, from the gauges). Part of the gap is held at the lock station near the border, at Srpski Itebej: while the canal is held full the gap is 5.5 m; when it is drawn down, usually from December to March, 3.2 m remain. The map shows the relief, measured from orbit; the section under it cuts the plain from one river to the other and moves as you scroll.

Scroll: the water sets out from Coștei ↓

Coștei · 48 km upstream of Timișoara

At Coștei the Timiș is higher and gives the Bega water

110.1 mthe crest of the Coștei weir, in the lowest of the published figures; it holds the Timiș at the level of the intake

Here the rivers are 9.4 km apart. The crest of the weir is given at 110.1 m, and the water of the Timiș stands at least that high. The Bega stands at 102.1 m at the Chizătău gauge, 3 km below the point where the canal meets it; at that point it is therefore 5 to 8 m below the Timiș (the section draws the lower end). Part of the Timiș’s water leaves for it along a canal of 10.1 km. On the map it is the dashed orange line; the arrow shows the way the water runs.

The Coștei works were built between 1758 and 1760 by Maximilian Fremaut, a hydraulic engineer from the Low Countries. (Banat Water Basin Administration; crest level: Panțu)

Topolovăț · 11.5 km further down

At Topolovăț the second canal runs the other way

6.4 kmthe mapped length of the Topolovăț–Hitiaș canal, in which water runs from the Bega to the Timiș (7.6 km in Panțu’s thesis)

From Coștei to Hitiaș, where this canal ends, the Timiș falls about 17 m (an estimate). The Bega, between the gauges at Chizătău and Topolovăț, falls 4.7 m. At Hitiaș the Timiș is about 4 m lower than the Bega at Topolovăț. The Bega’s level is that of the gauge at Topolovăț; that of the Timiș at Hitiaș is estimated between two gauges and may be wrong by 2–3 m. The canal itself shows the direction too.

So the second canal between them carries water in the opposite direction. In a flood, the water that would otherwise reach Timișoara leaves here for the Timiș. The Bega is given water when it has too little and relieved of it when it has too much.

Balinț, Chizătău, Remetea

Whose water runs under Timișoara’s bridges

7 → 29 m³/sthe long-term mean flow of the Bega at Balinț and at Chizătău, after the canal from Coștei joins it

The canal from Coștei brings 18 m³/s on average: more than twice what the Bega carries at Balinț, and about 60% of the flow at Chizătău. The gauges downstream give less: 17.5 m³/s at Topolovățu Mic and 17.3 at Remetea, where the river enters Timișoara. The table does not say which years each mean covers. (Timiș County spatial plan, table 35)

At low water the ratio between Remetea and Balinț is larger still: 4–6 times. The lowest flow of each year from 2015 to 2020 was between 1.6 and 2.7 m³/s at Balinț and between 7.0 and 12.9 m³/s at Remetea. The two minima need not fall on the same day. (recorded flows, in Șmuleac et al., 2023)

The plain between the rivers

In places the Bega runs higher than the plain beside it

11–55%of the land within 15 km of the Bega, from Topolovăț to the border, lies below its water: 39% as the central estimate. The ends show what a water level 3 m lower or higher would change; they are not a confidence interval

The blue on the map is land that, in the central estimate, lies lower than the Bega’s water at the nearest point of the canal. The count takes every point of the terrain model within 15 km of the canal whose nearest point on the Bega lies between Topolovăț and the border (2,570 km² in all) and compares it with the water level at that point of the canal. The result moves a great deal with the water level, which may be wrong by 2–3 m between gauges: 31–45% for 1 m lower or higher, 22–50% for 2 m, 11–55% for 3 m. This range shows only how far the result moves with the water level; the errors of the terrain model are not in it. The model also measures buildings and trees, so it shows the ground higher than it is: on open land, with no forest and no buildings (76% of the strip), the share is 45%; in forests, which also stand on higher ground, 7%; in built-up areas, 21%. What the check of the model shows ↓ “Lower than the water” does not mean “would be flooded”: dikes stand between the canal and that land.

A doctoral thesis on the canal says the same, without figures, of the embanked stretch above Timișoara: the Bega’s waters, in translation, “now run above the land between the rivers”. (Panțu, p. 151)

Work began in 1728, to drain the marshes of the plain and to float timber from Făget to Timișoara. (Beilicci and Beilicci, 2012) The dug bed carries the water of the hills past the low plain instead of letting it spill into it.

Timișoara

The Bega in Timișoara is about 5 metres above the Timiș at Șag

85.0 mthe normal operating level of the Bega in Timișoara

The Timiș passes south of the city. At Șag, 12 km from the centre, its gauge showed a median level of 80.2 m on the 148 days of 2023–2026 for which its reading has been kept; on eight days in ten it stood between 79.9 and 80.6 m. The Bega in the city is held at 85.0 m: almost 5 m higher. The two heights come from different reference systems, which can shift the gap by a few tens of centimetres. (Panțu, p. 201; readings of Romania’s hydrological institute on hidro.ro, from the Internet Archive)

From here their ways part. To the border the Bega covers 40.8 km on a course that measures 39.9 km as the crow flies. The Timiș, from Șag to Jaša Tomić, takes 42.2 km to cover 33.7: it is a lowland river, with meanders. The Bega is a dug channel.

Srpski Itebej and Jaša Tomić · Serbia

At the border: 5.5 m with the canal held full, 3.2 m with it drawn down

7,974of 8,036 days, 2004–2025: the days on which the Bega, entering Serbia, was higher than the Timiș

Between Timișoara and the border the Bega goes down in only two steps, at the locks of Sânmihaiu Român and Sânmartinu Maghiar: about 5 m in all. The Timiș, on its way from Șag to the border, falls about 6 m. How the gap changes on the way depends on the state of the canal. On the 132 days of 2023–2025 for which readings have been kept both at Șag and at the border, the gap at the border was, at the median, 0.8 m larger than at Timișoara while the canal was held full (81 days) and 1.4 m smaller when it was drawn down (40 days). For the Bega in Timișoara the calculation takes the operating level, not a reading of the day, and the thesis that gives it names no height system. If that system differs from the one of the Șag gauge, both figures shift by the same offset; the distance between them, 2.2 m, stays.

Above the lock at Srpski Itebej the gauge reads about 80.0 m while the canal is held full and about 78.0 m when it is drawn down. At Jaša Tomić, 17 km to the south, the Timiș usually stands at 74.5 m. Both heights are given above the Adriatic Sea. Each river is measured where it enters Serbia, and the two gauges do not face each other. On the other 62 days the Timiș was in flood; on 56 of them the canal was not being held full either. Every day, on one chart ↓

From the border to Klek

The water is held in steps

1.8 mthe step at Srpski Itebej while the canal is held full, from the station’s two gauges

An almost straight canal on an almost flat plain would be too shallow in summer if it flowed freely. Between 1901 and 1916 the canal was divided into steps by weir-and-lock stations; Olaru’s list has six: Sânmihaiu Român and Sânmartinu Maghiar in Romania; Itebej, Klek, Ečka and Titel in Serbia. Four are marked on the map with squares. (Olaru; Beilicci and Beilicci, 2012)

Freight stopped in 1958 and the passenger boats were withdrawn in 1967. (Timișoara City Hall, 2020) The lock and weir at Sânmihaiu Român and the stations at Srpski Itebej and Klek were repaired under a Romanian–Serbian project that closed in April 2022. (Timiș County Council, 2022) The border can be crossed by water only at temporary openings: for the summer of 2026 the County Council announced a single one for boats, from 24 to 26 July. (Timiș County Council, June 2026)

The freight is gone from the canal and the levels have stayed: while the canal is held full at Srpski Itebej the Bega stands 5.5 m above the Timiș at the border; when it is drawn down, 3.2 m.

Titel · 108 km from Timișoara

The Bega ends in the Tisza, the Timiș in the Danube

13 mfrom the Bega’s level in Timișoara to the usual level of the Tisza at Titel

Near Titel the Bega enters the Tisza, 108 km from the centre of Timișoara along the OpenStreetMap line. The Timiș and the Bega do not meet again: the Timiș joins the Danube at Pančevo, 46.5 km in a straight line from the Bega’s mouth.

Two rivers that were 9.4 km apart at Coștei and shared their water thus end in different waters: the Bega in the Tisza, the Timiș straight into the Danube.

The test: every measured day at the border, 2004–2025

Serbia’s hydrometeorological service prints, every year, the daily water level at each of its stations together with the height of each gauge’s zero. The 22 yearbooks for 2004–2025 give 8,036 days, with none missing, each with a reading for both the Bega (Srpski Itebej, above the lock) and the Timiș (Jaša Tomić). The chart shows all of them. The series starts in 2004 because the older volumes, 1991–2003, are published as scanned pages with no text layer.

Pick a day:Move a finger or the pointer over the chart to read a day.

Download all 8,036 days (CSV, 408 KB): date, the readings of the three gauges, the levels and the difference.

The Bega’s line has two steps a year. The state of the canal is read from the two gauges of the Srpski Itebej station, day by day, not from an operating log; the position of the weir is not observed. On 5,036 days (63%) there was a step of at least 1 m between the gauge above and the gauge below: the canal was held full. On 2,826 the step was under 25 cm. On those days the step vanishes because the water above falls, not because the water below rises: the gauge above reads about 78.0 m instead of 80.0 m, and the one below 78.0 m against 78.2 m. On only one of the 2,826 days did the water above reach the level it stays above on 95% of the days with the canal full. The other 174 days have a step in between: 55 spells, 44 of which lie between one with the canal full and one with it drawn down. The first day with the canal full fell, from year to year, between 7 March and 27 June; the last, usually in December. The Timiș’s line is that of a river without locks: it lies low and jumps in floods, with 7.3 m between the extreme daily readings of the period. On 56 of the 62 days when the Timiș reached or passed the Bega (on one they stood level) the canal was drawn down. With the zeros of the NVT2 system these days would number 49. The other six are two days of the flood of April 2005, one day in March 2007 and three days of the flood of June 2020. More than half of these days fall in 2005, 2006 and 2023. On all of these days the flow of the Timiș at Jaša Tomić was at least 180 m³/s, ten times the median daily flow (17.9 m³/s); only 3% of the period’s days reach that. But not every flood reaches the Bega’s level: the Timiș has passed 300 m³/s on days when it stayed below the Bega, because the canal was held full.

What a daily reading is. For the Bega’s gauge above the lock and for the Timiș gauge the yearbooks give, in all 22 years, the mean of the day from continuous recording: a chart recorder, then a digital sensor (from 2011 at Srpski Itebej, from 2006 at Jaša Tomić). So the gap between the rivers is worked out the same way over the whole period. The gauge below the lock, used only for the state of the canal, was until 2010 a staff read twice a day, with the mean of the two readings. Days with the canal full are 64% of the total in 2004–2010 and 62% in 2011–2025.

The 22 episodes when the Timiș reached or passed the Bega

Days	Highest daily reading of the Timiș	Highest daily flow
3	78.8 m	260 m³/s
6	79.3 m	362 m³/s
3	81.1 m	742 m³/s
2	78.7 m	248 m³/s
1	77.9 m	180 m³/s
4	79.4 m	382 m³/s
1	78.7 m	250 m³/s
2	78.5 m	246 m³/s
7	81.0 m	713 m³/s
3	78.7 m	258 m³/s
1	79.3 m	356 m³/s
3	79.1 m	285 m³/s
2	79.5 m	322 m³/s
2	79.8 m	352 m³/s
1	78.7 m	268 m³/s
2	79.5 m	380 m³/s
3	79.0 m	301 m³/s
3	80.8 m	651 m³/s
2	78.4 m	235 m³/s
1	78.2 m	218 m³/s
5	79.7 m	401 m³/s
5	78.4 m	230 m³/s

The median gap is 5.0 m and the mean 4.4 m; on 90% of days it was between 1.6 and 5.9 m. No episode lasted more than seven days.

Where the map is wrong

The relief is the Copernicus GLO-30 model, measured by radar from orbit in 2011–2015, with one point about every 30 m. Its producer states an absolute vertical error under 4 m at 90% of points and a relative one, between two points in the same one-degree square, under 2 m (also at 90%) on slopes of 20% or less. It is less accurate on water: on a canal one point wide the radar catches the banks and the trees, and on a wide river the value is flattened by the producer.

In Timișoara the model gives the Bega 89.0 m where the operating level is 85.0 m. At Jaša Tomić it puts the Timiș at 72.0 m: 2.5 m below the gauge’s usual level and 1.8 m below the lowest level measured in 22 years. At Șag it puts it at 78.5 m, 1.7 m below the gauge’s median level. Taken as it is, the model would show the Bega in Timișoara 10.5 m above the Timiș at Șag, where the gap is 4.9 m, and 7.3 m above it at the border, where the gauges give 5.0 m. The direction of the gap is right; its size is not.

So the water levels in the section and in the figures for each town come from surveyed heights: the Bega’s operating level in Timișoara, the two locks, the Serbian gauges and the readings of the Romanian gauges.

How good the ground is. Besides heights, Copernicus publishes quality layers for every point. In the strip counted in the chapter on the plain, 1.3% of cells have more than a tenth of their points edited, filled from other models or marked as water; without them the share of land below the Bega’s water stays at 39%. The height error the producer estimates has a median of 0.4 m per cell; it describes the noise of the radar measurement, not vegetation or buildings, and is not a confidence interval. A second public set of heights, the Tilezen mosaic, made from other measurements (SRTM and, in Europe, EU-DEM), shows the open land of the strip 2.3 m lower at the median (between 0.7 and 3.8 m in 90% of cells); with it the same count would give 53%. The two sets do not share a reference surface, and the comparison does not convert it. Which of them is closer to the ground cannot be settled from here.

That is why the share is given as a range, and why for a town or village the page says which is higher only from 6 m of difference up: 4 m, the stated error of the ground, plus 2.2 m, the largest error found in the water level; rounded, 6 m, and the sum is not a guaranteed bound. Until 13 October 2026 the threshold was 4 m. With 6 m there remain 75 places the page calls higher than the Bega’s water and none it calls lower; for 184 it shows the estimate only.

How good the drawn profile is. At Sečanj, on the Timiș, a gauge that had not been used in the calculation reads 74.3 m; the profile gave 72.4 m there, 1.9 m too low, and was corrected with it. When any one of the other surveyed points is left out, the profile misses it by 2.2 m at most. The largest error found in these tests is thus 2.2 m. It is not a guaranteed limit; the page works with 2–3 m throughout.

What remains uncertain. For the Romanian gauges there is no public daily series of the kind the Serbian yearbooks give. Romania’s hydrological institute (INHGA) does show each gauge’s reading of the day on hidro.ro, and the Internet Archive has kept that page on 123–149 days for each of the seven gauges used here, from January 2023 to October 2026. The drawn level is the median of those readings, taken only if dated within two days of the copy of the page, one per day: raw, unvalidated data, and a sample, not a full series. On the sample’s days in 2023–2025 the gauge at Jaša Tomić stood 18 cm below its 2004–2025 median, so the sample catches somewhat lower water than usual. Against its own years it is balanced: on its days in common with the border series each month has between 8 and 16 readings at Șag, the canal was held full on 61% of the sample’s days and on 61% of all days in 2023–2025, and 2023 is more thinly covered (24 readings, against 54 in 2024). Readings are given against each gauge’s zero, taken from DanubeHIS. That it is the same zero could be checked only at Lugoj, where the reading of 12 October 2026 (−26 cm) is the same in both places; at the other gauges it is an assumption. At Șag and Brod the water stands below the gauge’s zero almost all the time; at the median, by 1.3 and 0.9 m; at Chizătău and Lugoj, on most days, by 10–20 cm.

Correction, 13 October 2026: the first version of this page put a metre of water above the zero of every Romanian gauge, by analogy with the gauge at Jaša Tomić, and showed the Timiș at Șag almost 2 m too high. That is why the gap at Timișoara was then given as about 3 m.

The step at Sânmihaiu Român (2.2 m) is derived from the level in Timișoara, the published step of the Sânmartinu Maghiar lock (2.8 m) and the gauge at the border; a table of the canal’s weirs gives 2.4 m for it. The same table gives 2.5 m at Srpski Itebej, where the station’s two gauges show 1.8 m with the canal full; the table does not say which operating regime its values are for. At Klek the level (77.4 m) is the median of the gauge that worked there in 2004–2014, on the days when the Srpski Itebej weir was holding the water; the page takes it as the level above the Klek weir, and the step (2.4 m) is taken from the table; from Klek to the Tisza the line is interpolated. (Șmuleac et al., 2023, table 1)

For the Coștei weir Panțu’s thesis gives several heights: 110.10 m, the intake level at the crest of the spillway, from the hydrological data that predate the updated hydrological study; 110.50 m, the operating level to which the sill was brought after the damage of the high water of 2000; and 111.00 m on the upstream side, with 110.75 m downstream, for the crown of the rebuilt weir. The thesis does not explicitly reconcile them. The page uses the lowest; with the highest the Timiș would stand a further 0.9 m above the Bega at Coștei. A crest gives the lower bound of the water level, not the level itself: in high water the river runs over it. (Panțu, ch. 4.1 and 5.2)

The heights come from different reference systems: the Serbian gauges are given above the Adriatic, the Romanian ones above the Baltic (Lugoj: the Black Sea), the level in Timișoara is given in the thesis with no system named, and the terrain model is in EGM2008. The offsets between the two countries’ systems, probably a few tens of centimetres, are not applied: no published value against the Romanian system could be found for Serbia. They fall entirely on the derived step at Sânmihaiu Român and on the gap at Timișoara. The Lugoj gauge enters the profile of the Timiș above Coștei unconverted. At the border both gauges are in the same system; moved to NVT2, the gap would shift by 8 cm.

The terrain model measures the surface with everything on it: forests and built-up areas appear a few metres higher than the soil. Heights in the section are exaggerated several hundred times against distances; on the ground the plain looks flat. And the blue in the chapter on the plain means “lower than the canal’s water”, not “would be flooded”: dikes stand between the canal and that land.

Sources and method

Daily levels: Republic Hydrometeorological Service of Serbia, Hydrological Yearbook, 1. Surface Waters, editions 2004–2025: the stations Srpski Itebej (above and below the lock, zero at 76.00 m) on the navigable Bega (stations 44211 and 44212) and Jaša Tomić (zero at 73.46 m; station 42401) on the Timiș. Level = height of the gauge zero + the day’s reading. Days the yearbook marks with ice keep the printed reading. The zeros are the same in all 22 volumes and are given above the Adriatic; the editions from 2020 on also give them in the newer NVT2 system (75.71 and 73.09 m), in which the gap between the rivers would be 8 cm larger and the days when the Timiș reached or passed the Bega would be 49, not 62. Monthly means recomputed from the daily tables match the printed ones, to within 1 cm, in every volume. The whole series, CSV. The table holds only the daily values of the three gauges, transcribed from the yearbooks, and is not a product of the Serbian service.
Heights in Romania: the gauge zeros at Balinț, Chizătău, Topolovățu Mic, Remetea (Bega) and Lugoj, Brod, Șag (Timiș), from DanubeHIS (ICPDR), after Romania’s hydrological institute; the daily readings of the same gauges from hidro.ro (INHGA), in the copies kept by the Internet Archive (January 2023 – October 2026); the operating level in Timișoara (85.0 m) and the crest of the Coștei weir (110.1 m), from Panțu H., doctoral thesis, Politehnica University of Timișoara; the fall at Sânmartinu Maghiar, from Olaru M.. Between points the level follows the shape of the terrain model, corrected by its error at the surveyed points.
Relief: Copernicus DEM GLO-30 (produced using Copernicus WorldDEM-30 © DLR e.V. 2010–2014 and © Airbus Defence and Space GmbH 2014–2018 provided under COPERNICUS by the European Union and ESA; all rights reserved). The organisations in charge of the Copernicus programme by law or by delegation do not incur any liability for any use of the Copernicus WorldDEM-30. The quality layers (EDM, FLM, WBM, HEM) come from the same delivery. Land cover: ESA WorldCover 10 m 2021 v200 (© ESA WorldCover project, CC BY 4.0; contains modified Copernicus Sentinel data, 2021). The second set of heights: the Tilezen mosaic, which uses SRTM (NASA) and, in Europe, EU-DEM (produced using Copernicus data and information funded by the European Union).
Courses, canals, locks, places: © OpenStreetMap contributors, extract of 13 October 2026. Lengths are measured on these lines.
Flows: Timiș County Council, county spatial plan, vol. II, table 35; Șmuleac L. et al., Research Journal of Agricultural Science 55(2), 2023, tables 1 and 2.
History: Banat Water Basin Administration, “Istoric”; Beilicci E. and Beilicci R., 2012; Olaru M.; Timișoara City Hall, 2020. Sources give different lengths for the navigable canal (between 114 and about 119 km, of which 40–44 in Romania) and for the two link canals (9.6 and 7.6 km in Panțu, against 10.1 and 6.4 km on the map) and slightly different years for the Coștei works (1757–1761); the numbers on this page are measured on the map, from the centre of Timișoara.

More on Timișoara on this site: the fortress and the 948 m circle where building was banned · the old names of its streets · the electric street lights of 1884.