When ice blocked the Dnieper: a lake over Gomel and two ways out to the Desna
Could the Sozh once have flowed into the Desna instead of the Dnieper? On today's relief it cannot. If glacier ice blocked the Dnieper below Gomel, it could: the Sozh backs up into a lake, and the lake spills over a low pass into the Desna basin. A model scenario in 48 seconds. Nothing in it is dated.
What you are seeing
Today the Sozh joins the Dnieper at Loyew. East of Gomel two wide, nearly empty valleys cross the divide towards the Desna: Zamglai near Chernihiv and Khoroput–Tsata near Dobrush (dotted in the film). The Zamglai through-valley, the largest in the Gomel and Chernihiv Polesie, connects the Dnieper and Desna valleys. Another through-valley, Khoroput and Tsata, connects the valleys of the Iput and the Snov.
The ice is a model. The Dnieper glaciation covered the whole territory of Belarus. The film starts with the model ice at its maximum, over the whole area; the margin then retreats north and leaves a tongue of dead ice in the Dnieper valley. Matoshko (2004) interprets the Dnieper ice stream as stagnating and breaking into blocks of dead ice, with runoff carried mainly by the Desna and Zamglai streams within the former glacier bed. (His study area lies north of Kyiv; he describes no dammed lake or spillway.)
While the tongue blocks the Dnieper, the Sozh backs up into a lake. In the film run the lake stands at about 135 m and spills through the Khoroput–Tsata trough near Dobrush, down the Snov and the Desna, for about 4,700 model years; it covers up to about 3,400 km². When the tongue narrows, the way to Zamglai opens: the outflow cuts the Zamglai pass in about 750 model years at 1,000 m³/s and the lake falls from about 134 to 126 m (on the simulation's coarser grid, which lowers pass heights slightly). When the dead ice is gone, the lake drains and the Sozh flows to the Dnieper again.
Which pass opens depends on how wide the ice was. The split screen shows two ice geometries from a phase diagram of 693 simulated runs. On today's relief (306 runs), 282 form a dammed lake; all 246 of them whose lake stays inside the model area drain to the Desna through one of the two valleys — Zamglai when the ice in the Dnieper valley is a narrow tongue (up to ±10 km) or a detached plug of dead ice below Loyew, Khoroput–Tsata when it is a broad lobe (±20 km or more). The other 36 reach the edge of the model area, so their real outlet cannot be seen; they are not counted either way.
Water in the channels
The water you see moving is computed, not drawn: a two-dimensional flood model (the local-inertial shallow-water scheme of Bates, Horritt & Fewtrell 2010) on the same terrain, with the ice as a wall. At the model's scenario discharge (1,000 m³/s of meltwater and rivers) the lake near Gomel fills up to the Dobrush pass in about 18 model days (a level-pool estimate: stored volume divided by the discharge), then stands about 1.1 m above the pass while it spills. The overflow runs down the Snov and reaches the Desna at Chernihiv within 60 model hours. None of the overflow has left the model area by the end of that run: all 0.2 km³ that came in is still filling low ground in the Desna lowland. When the dead ice is gone, the Sozh reaches the Dnieper below Loyew within 126 model hours. The rivers of today (Dnieper, Sozh, Besed, Iput, Snov, Desna, Pripyat) carry their long-term mean discharges, 1,289 m³/s in all. Hours and days are model time, slowed down; they are not dates.
Why the Sozh cannot reach the Desna on its own
The lowest direct pass from the Sozh into the Desna basin is Zamglai, 127 m on today's terrain, about 18 m above the Dnieper at Loyew, which the Sozh reaches first. The divide near Dobrush is higher still, 135 m. Without an ice dam, water from Gomel always flows to the Dnieper (tested on the terrain with outlets at Loyew and on the Desna; walling off Loyew in a test sends it to the Desna, so the test can tell the two apart). In a separate test covering the last glacial cycle (project note 2026-09-25-sozh-tilt-x3), tilting the land by the glacial isostasy of the ICE-6G model, even amplified many times over, did not change that either. Free-flowing river capture is ruled out; only the ice dam works.
How it was made
- Terrain: FABDEM (bare-earth elevation) at 3 arc-seconds. Pass heights are measured as the lowest connecting level between two points (minimax).
- Ice: our own model, because ICE-6G covers only the last glacial cycle and has no Dnieper ice. A plastic ice profile; the margin retreats to the published Novozybkov-stage line and leaves dead ice in the Dnieper valley that melts over 7,000 model years. Gursky & Levitskaya (1989) divide the retreat of the Dnieper ice sheet in Belarus into three stages; the second, the Mozyr (Novozybkov) stage, is marked by ice-margin features near Mozyr and Novozybkov.
- Lake: flat water behind the ice, filled by priority-flood to its lowest outlet; a lake that touches the edge of the model area is flagged as unobservable.
- Pass incision: sand transport (Engelund–Hansen) through the active, not-drowned pass, starting from the valley filled to its brim (an upper bound).
- The film shows the run without isostasy (see the limits). Vertical exaggeration: terrain ×30, ice thickness ×10. Rivers in the "today" frames are the flood model at mean discharges.
Limits of the model
- Nothing is dated. The age of the Dnieper glaciation itself is disputed: The Belarusian scheme correlates the Dnieper glaciation with marine isotope stage 8 and cites a thermoluminescence age of 216±18 thousand years for Dnieper late-glacial deposits. Shik (2010) finds no reliable evidence that the moraine of the Dnieper glacial tongue belongs to isotope stage 8 and assigns it to the end of the Middle Neopleistocene, stage 6; he adds that TL dates older than 150 thousand years are often much too young. We found no numerical date for the valley fills of Zamglai or Khoroput–Tsata in the sources we read.
- Model years are not calendar years; the timing of retreat and melt is a scenario.
- Isostasy: with the land fully depressed by the ice (elastic plate over a relaxing mantle), the model's first lake would leave a shoreline at about 210 m near Gomel in today's terms, and even the weaker variants we modelled put it at 175–185 m; in the sources we read we found no continuous lake deposits up to such heights (deposits this old may also have been eroded or buried). So only runs with no depression, or one weaker than any variant we modelled, fit; the film shows the run without it. In every isostatic run we followed in detail, whenever the lake spills south through Dobrush or Zamglai it also touches the northern edge of the model area, so a northern outlet along the ice margin cannot be excluded there.
- The Dobrush (Khoroput–Tsata) trough is deep and saddle-shaped; the lake in this model could not cut it from scratch, so the film run assumes it already existed.
- Wide abandoned valleys in the Dnieper basin are known elsewhere, and those with a known age are young: Near Orsha and Shklov the Dnieper has abandoned 'dead valleys' up to 1 km wide, which Kalicki & San'ko read as channels used when the river carried meltwater from the last (Valdai) ice sheet, after 17,000 BP; they also describe an ice-dammed Lake Orsha at that ice margin. Where wide Dnieper palaeochannels have been dated (the upper Dnieper near Smolensk), they were active in the Early Holocene, 12-8.5 thousand years ago, and by their age cannot be related to glacial meltwater of the Last Glacial Maximum. These do not date our two valleys.
- An older reading exists: Citing Goretsky (1970), Mander writes that Dnieper water overflowed along the Zamglai valley into the Desna valley in the early and middle Anthropogene (Quaternary).
- Context: Chernihiv Polissia is cut by modern river valleys and by old through-valleys of glacial meltwater runoff; Zamglai is classed as a valley mire (8,334 ha), and loess in the Zamglai valley is only 1-1.5 m thick against 5-9 m on the right bank of the Snov. (the source does not name Zamglai as a through-valley).
- The flood model runs on the simulation's cells (about 170 × 280 m), so channels are wider and shallower than real ones; it carries no sediment during the flood; the dammed lake is held at a constant level while it spills (its volume is huge against days of flow); today's rivers are the model with mean discharges, not observed water.
- What would test it: optically stimulated luminescence dates of the sands under the peat at both passes, and lake deposits at the heights the model predicts.
The interactive viewer is in private preview.
Video licence: CC BY-NC-SA 4.0, because it is derived from FABDEM. Share it with a link to this page; no commercial use.
Sources
- Terrain: FABDEM V1-2 — Hawker et al. 2022, Environmental Research Letters, doi:10.1088/1748-9326/ac4d4f
- Kalicki T. & San'ko A.F. 1992, Geographia Polonica 60
- Mander E.P. (ed. Goretsky G.I.) 1973, Anthropogene deposits and the development of the relief of Belarus (in Russian), p. 103
- Matoshko A.V. 2004, Quaternaire 15, doi:10.3406/quate.2004.1759
- Matoshko A.V., Gozhik P.F. & Ivchenko A.S. 2002, Netherlands Journal of Geosciences 81, doi:10.1017/S0016774600022642
- Panin A., Adamiec G. & Filippov V. 2015, Quaternaire 26, doi:10.4000/quaternaire.7141
- Lindner L. et al. 2004, Geological Quarterly 48; Shik S.M. 2010 (in Russian)
- Gursky B.N. & Levitskaya R.I. 1989 (in Russian); Karopa G.N. 2009 (in Russian); Elovicheva Ya.K. (attributed), Belarusian State University teaching chapter (in Russian)
- Polyanska K.V. et al. 2017, Podesennya protection-zone study (in Ukrainian)
- Flood model: Bates, Horritt & Fewtrell 2010, Journal of Hydrology 387, doi:10.1016/j.jhydrol.2010.03.027
- Ice sheet flexure: Le Meur & Huybrechts 1996, Annals of Glaciology 23, doi:10.3189/S0260305500013586
Mean river discharges:
- EUWI+ 2018. Dnieper River Basin Management Plan in Belarus. Annex 1: Attached tables. Minsk, October 2018 (authors Bulak I., Korneev V., Pakhomau A.), link
- Клименко В.Г. 2010. Гідрологія України: навчальний посібник для студентів-географів. Харків: ХНУ імені В.Н. Каразіна, 124 с., link
- Велика українська енциклопедія (е-ВУЕ), article «Снов». ДНУ «Енциклопедичне видавництво», link
- EUWI+ 2020. Pripyat River Basin Management Plan in Belarus (Draft Pripyat RBMP - Belarus), report, 20.03.2020, link