Why rivers make poor borders — and why we keep using them

In 1864, the Rio Grande flooded. When the water receded, 630 acres of land that had been on the Mexican side of the river were now, physically, on the American side. The Treaty of Guadalupe Hidalgo, signed in 1848, had drawn the international boundary along the river. The river had moved. Mexico said the border was where the treaty had put it. The United States said the border was the river, and the river was here now. Neither position was unreasonable. Both could not be right.

The dispute over this strip of desert — called the Chamizal — ran for 99 years.

Not because the land was particularly valuable. Not because either government was intractable on principle. It ran for 99 years because the premise was broken. When your border is a river, and the river moves, you don’t have a border. You have a negotiation that never ends, dressed up as geography.

The map problem

Rivers define roughly 23% of international borders globally and approach 50% of South American borders, according to the Global Subnational River-Borders database compiled by geographers Laurence Smith and Sarah Popelka in 2020. The appeal is easy to understand from a distance. Rivers are visible, continuous, and appear on maps without anyone needing to visit the place. When the British were drawing South Asia’s borders, the French partitioning West Africa, and the Soviets carving Central Asian republics, rivers were the one feature identifiable from London, Paris, or Moscow. Mountains are visible from a distance too, but they clump and disperse, leaving ambiguous gaps. A river runs for hundreds of miles in a single, followable line. A border drawn along one can be explained in a conference room ten thousand miles from the water.

The logic seems airtight until you examine the premise: that rivers divide.

They don’t. The history of human settlement runs the opposite direction. Major civilizations built themselves at rivers because rivers drew people together. Cairo grew on both banks of the Nile, because the Nile was the reason for Cairo. Baghdad straddles the Tigris. Paris has been on the Seine since the Romans built Lutetia on an island in the middle of it. Rivers were where boats docked, where fish were caught, where goods changed hands — and none of those activities cared which bank you’d been born on. Both banks developed together, exploiting the same resource from two angles.

A line drawn down a river’s center doesn’t find a natural separation. It cuts through a natural integration.

The thalweg

International law has known for over two centuries that "the river" isn't a single point. The thalweg principle — from the German for "valley way" — holds that the boundary in a navigable river follows the deepest navigable channel, not the geometric center line. First adopted for the Rhine in the 1801 Peace of Lunéville, it has become the default rule in international boundary law. The reasoning is practical: the deepest channel is where navigation happens, and navigation rights are what states care most about dividing. The catch is that the deepest channel is not fixed. Sedimentation, flooding, and erosion shift it continuously, without informing either government. International law's solution to the imprecision of river borders introduced a new imprecision of its own.

The river moves

Even if you accept that rivers look like borders, you immediately run into the fact that they refuse to stay put. The thalweg principle acknowledges this with serene indirectness. The Chamizal dispute made it concrete.

The 1848 Treaty of Guadalupe Hidalgo made the Rio Grande the US-Mexico boundary without specifying what happened when rivers shifted. After the 1864 flood moved the channel south, the two countries negotiated the 1884 Convention: accretion — gradual, imperceptible change — shifts the border with the river; avulsion — sudden change, like a flood — does not. Under avulsion, the border stays in the old channel. Sensible on paper. In practice, the distinction collapsed because rivers are indifferent to legal categories, and “gradual” does a lot of work that neither government could agree on.

The Chamizal Convention of 1963 finally settled the matter by splitting the contested acreage and authorizing the Rio Grande Rectification Project: 155 miles of wandering river straightened and locked into 88 miles of concrete channel, 167 feet wide, 15 feet deep. More than 5,600 El Paso residents were displaced from five South El Paso neighborhoods to build it. The border didn’t follow the river. The river was engineered to follow the border.

Both the 1884 Convention and the Rectification Project are attempts to force a river to stop being a river. The Convention failed; the concrete channel succeeded, at the cost of displacing over five thousand people and flattening five neighborhoods. The real price of a river border is whatever it costs to suppress the river.

The Sir Creek dispute between India and Pakistan — a tidal creek in the Rann of Kutch whose channel has shifted since the 1914 map used to define the boundary — remains unresolved for the same reason: the legal definition of the border references a natural feature that won’t hold still. India argues for the thalweg; Pakistan argues for the eastern bank. Twelve rounds of talks, no breakthrough.

Avulsion, accretion, and why it matters which one you call it

The 1884 Convention's accretion/avulsion distinction has been applied inconsistently across jurisdictions for over a century. Courts and commissions have reached different conclusions about the same type of river movement depending on context, because the question of whether a shift was "gradual" or "sudden" is often impossible to answer definitively. Rivers change continuously; the line between steady erosion and a discrete flood event is rarely clean. The same physical event can produce different legal outcomes depending on which side's arbitration bodies you're in. What was designed as a precise rule turns out to depend, in practice, on whoever has the political leverage to define the categories.

The river doesn’t divide the water

Engineering can stabilize a moving border. Concrete holds. What it cannot do is change the fact that everything an upstream country does to the water still determines what the downstream country receives. A border is not a dam.

The Amu Darya, 2,540 kilometers from the Pamir mountains to the Aral Sea, runs along the border between Afghanistan and the former Soviet republics of Uzbekistan and Tajikistan. Afghanistan contributes roughly 30% of the river’s total flow and currently uses approximately 2% of it. Uzbekistan holds a 48.2% allocation of the river’s total managed flow — a figure that traces directly to Soviet-era planning, when these weren’t international borders at all but internal administrative lines drawn during the 1924 National Territorial Delimitation. They became international borders when the Soviet Union collapsed in 1991, and the Amu Darya suddenly ran between sovereign states with competing interests and no binding agreement on how to share it.

What happened next to the Aral Sea was not caused by borders. Soviet planners diverted both the Amu Darya and the Syr Darya beginning in the 1960s to irrigate cotton fields — accomplished entirely under internal administrative authority. By 1988 Uzbekistan was the world’s largest cotton exporter. The Aral Sea, once 68,000 square kilometers and one of the world’s four largest lakes, shrank to less than 10% of its mid-century size. A salt-crusted desert replaced it. The fishing towns of Muynak and Aralsk found themselves stranded miles from the retreating waterline. Forty to sixty thousand fishermen lost their livelihoods.

All of this happened before the borders arrived.

What the post-Soviet border transformation did was convert what might have been a reversible internal management failure into a structural governance gap. A bad Soviet planning decision could in principle be reversed by the same central authority that made it. Once the Amu Darya became an international boundary between sovereign states, correcting the overdiversion required negotiation among governments with competing interests and no enforceable treaty framework. The sovereign border structure made that kind of correction categorically harder.

Afghanistan’s Qosh Tepa Canal is the same governance gap in its pre-disaster phase. The 285-kilometer canal, planned operational by 2028, would divert between roughly 17 percent of the river’s annual flow, by geospatial estimate, and close to a third of it, by Carnegie Endowment’s projection for full operational capacity. Uzbekistan has protested diplomatically. Afghanistan calls it a domestic issue. No comprehensive water-sharing treaty covers all riparian states. Hydrology does not recognize the distinction between domestic and diplomatic.

Six countries share the Mekong — China, Myanmar, Laos, Thailand, Cambodia, Vietnam — and the Laos-Thailand border follows roughly 1,500 kilometers of it. The Mekong River Commission, established in 1995, has no enforcement powers and does not include China as a full member — China is a “dialogue partner,” exempt from the consultation requirements and data-sharing obligations that membership entails. China operates 11 hydropower dams on the upper Mekong (which China calls the Lancang), plus 95 on tributaries. A 2020 investigation by the Stimson Center, using satellite data, found that during the 2019 drought — when downstream countries were experiencing acute water shortages — China’s dams were withholding more water than usual while receiving normal precipitation. Seventeen Vietnamese provinces declared drought emergencies. The Mekong River Commission has documented a 40% decline in fish stocks over the preceding decade. The river’s seasonal flood pulse, which recharges Cambodia’s Tonle Sap lake and supports the fisheries and agriculture of tens of millions of people across the lower basin, is disrupted when upstream dams regulate flow for hydropower. China’s position: dam operations are a matter of domestic sovereignty.

The river doesn’t end at the border.

Any river boundary creates a situation in which one sovereign’s water management decision is another sovereign’s flood, drought, or fish collapse. The border divides governance; the hydrology stays connected. This isn’t a fixable flaw. It’s inherent to what river borders are.

The Mekong River Commission and the limits of voluntary cooperation

The 1995 Mekong Agreement that established the Mekong River Commission was a genuine diplomatic achievement — the first formal framework for lower-basin consultation. But its limitations were built into the design. The Commission has no enforcement mechanism; it can issue reports and recommendations, not compel behavior. China and Myanmar joined as dialogue partners rather than full members, meaning the countries operating the most consequential infrastructure have the fewest binding obligations. The Commission's mandate covers the lower basin only, putting the upper-basin dams outside its formal scope. Twenty-five years of operation have produced detailed hydrological data and no binding constraints on upstream operations. It answers the question "can't they just make agreements?" with a precise illustration of what those agreements actually look like.

The river runs through it

Rivers don’t run between communities. They run through them.

Every city that built itself on a river built on both banks, because the river’s value — water, fish, transport, commerce — was accessible from both sides. When a border arrives and declares the river’s center to be an international line, it doesn’t discover a division that was already there. It imposes one onto a community that had organized itself around a shared resource.

The Oder-Neisse line makes this visible because the process of selection left a record.

At the Potsdam Conference in July and August 1945, the Allied powers drew Poland’s new western boundary along the Oder and Neisse rivers, transferring approximately 25% of Germany’s pre-1937 territory to Polish administration. Germans had lived east of the Oder for centuries. The border didn’t recognize an existing division; it created one. Between 1945 and 1950, approximately 12 million ethnic Germans were displaced from territories east of the line — expelled in conditions ranging from organized transport to lethal chaos. Mortality estimates range from 500,000 to 2 million; the records were too disrupted for a more precise figure.

Historian Włodzimierz Suleja has argued that Stalin partly engineered this border to entrench a territorial grievance between Poles and Germans, ensuring Polish dependency on Soviet protection against any future German revision claim. Whether or not that specific reading is correct, the broader point isn’t in dispute: the Oder and Neisse rivers provided natural-looking visual authority for a political decision made in Moscow and Washington and London. The rivers looked like a border. They provided cover for the decision that made them one.

Recognition of the Oder-Neisse line unfolded across 45 years, and the distinctions matter. From 1945 to 1970, West Germany refused any formal recognition — the line was provisional, expellee organizations maintained revision claims. In December 1970, under Chancellor Willy Brandt’s Ostpolitik, West Germany signed the Treaty of Warsaw, accepting the Oder-Neisse as Poland’s western frontier and affirming its inviolability. This was real de facto recognition — but it carried a formal legal reservation: only a reunified German parliament could give the line its final binding force. That reservation kept the border politically live throughout the 1970s and 1980s, with conservative parties and expellee organizations contesting its permanence even as the diplomatic reality was accepted. Not until the Two Plus Four Treaty of September 12, 1990, as a condition of reunification, was the reservation dropped unconditionally. The German-Polish Border Treaty of November 14, 1990, ratified in January 1992, completed the confirmation.

The river border survived — but only after 45 years of oscillation between non-recognition, qualified recognition, and unconditional acceptance, while millions of Poles built lives west of the line and millions of expelled Germans nursed claims to land east of it. The Oder and Neisse simply flowed on.

Alsace-Lorraine, on the Rhine’s western bank, changed national identity four times between 1871 and 1944 — annexed by the German Empire, returned to France by Versailles, annexed again by Nazi Germany, returned to France by liberation. The Rhine ran through all of it. The population didn’t move; the political meaning of the border did. A river border doesn’t settle a territorial dispute. It marks the site where the dispute will recur.

Rivers run through communities, not between them. The cost of treating them otherwise is paid by whoever lives on both banks. In the Oder-Neisse case: 12 million people.

Why cartographers keep drawing the line

The naive answer is that cartographers don’t know better. They do. The real answer is that river borders serve functions for the people who draw them, regardless of what they do to the people who live near them.

Start with the simplest one: rivers can be seen from very far away. When the British were drawing the subcontinent, the French partitioning West Africa, the Soviets designing Central Asian republics, the people doing the drawing were operating from maps in colonial capitals — not from knowledge of terrain, communities, or how people organized their lives around the water. Rivers were identifiable in London, Paris, and Moscow without anyone needing to go and look. A border that follows a river can be explained and defended in a conference room ten thousand miles from the water. The appeal of river borders is inseparable from the distance from which they are drawn.

But visibility is the lesser mechanism. The more durable case for rivers is that they look natural, and a natural-looking border converts a political decision into an apparent act of geography. A straight line through the Sahara looks like what it is — a ruler applied to an empty map. A river border looks like something that was already there. The Oder-Neisse line, on any map, looks like an obvious western boundary for Poland. It wasn’t obvious; it was chosen among several options at Potsdam, by three men who were simultaneously deciding the futures of Germany, Japan, Korea, and Eastern Europe. The rivers naturalized the politics. That naturalization has value: contesting the line requires arguing not just against a political choice but against a river, and rivers look unchallengeable.

Then there is the argument that is genuinely hard to dismiss: the alternatives are worse. Without geographic features, borders require explicit political decisions about where one people ends and another begins — either arbitrary (straight lines through Africa and the Middle East) or resting on contested criteria like ethnicity and language. The post-WWI attempt to draw borders along ethnic lines produced the Sudetenland crisis, the Danzig corridor, and eventually part of the logic of the next war. Geographic borders, for all their failures, at least produce a visible, continuous line that both sides can point to.

This argument for river borders isn’t dishonest. It’s the real reason thoughtful administrators reach for rivers even knowing the problems. But it rests on a false comparison. The choice isn’t between river borders and perfect ethnic borders — it’s between river borders and borders drawn with actual knowledge of how the communities near a specific river live, which communities the river connects, what the water does for people on both banks. That harder work is what river borders excuse the administrator from doing. The cost is paid by the people near the water, not the people in the conference room.

The Berlin Conference and the systematic river border

The 1884–85 Berlin Conference formalized what colonial powers had already been doing ad hoc: using geographic features to partition Africa. The Congo, the Niger, the Zambezi became international boundaries not because they divided African communities but because European negotiators could identify them from conference rooms in Berlin. The Congo River was so central to the partition that territory on each bank was assigned to different colonial powers — Belgium on the left, France on the right — with a precision that had nothing to do with who actually lived there. Fourteen European governments and the United States attended. No African government was present.

The confession in concrete

The Rio Grande, in El Paso, now runs through 88 miles of concrete channel. Fixed geometry where a wandering river used to be. Dug out, lined, and locked in place so that the border could stop being a negotiation and start being a line.

That is the solution. It is also the confession: the border required the river to stop being a river.

Every river border is, in some sense, a continuous project of the same kind — not poured concrete in most cases, but treaties, commissions, legal doctrines, diplomatic standoffs, and engineering interventions, all aimed at making a fluid thing hold still long enough to be a line. The Amu Darya is ungoverned. The Mekong is governed badly. The Oder-Neisse spent 45 years oscillating between non-recognition, qualified recognition, and unconditional acceptance — at the cost of 12 million people — and now requires ongoing political maintenance even though it has ceased to be a physical question. The Sir Creek negotiators have met twelve times and counted nothing as settled.

River borders are made by people who need a line they can point to on a map — people usually far from the water and not the ones who’ll pay the consequences when the line doesn’t hold. They are maintained by the difficulty of acknowledging, once a border is drawn, that the premise was wrong. And they are paid for, over and over, by the communities that built themselves on both banks long before anyone arrived with a treaty and told half of them they were on the wrong side.

The river keeps moving. The concrete keeps it from showing.

Gen AI Disclaimer

Some contents of this page were generated and/or edited with the help of a Generative AI.

Media

Jere Carbon – Pexels

Key Sources and References

Smith, L. and Popelka, S. (2020). Rivers as political borders: a new subnational geospatial dataset. Water Policy, 22(3), 293–312. DOI 10.2166/wp.2020.032.

Treaty of Guadalupe Hidalgo (1848). Established the Rio Grande as the US-Mexico international boundary.

Convention of 1884 (US-Mexico). Established the accretion vs. avulsion doctrine for Rio Grande boundary shifts.

Chamizal Convention Act (1963). Resolved the Chamizal dispute by splitting the contested acreage and authorizing the Rio Grande Rectification Project. Ratified 1964; the Chamizal Zone was formally incorporated into Mexico on October 28, 1967.

Rio Grande Rectification Project. Straightened 155 miles of river into 88 miles of concrete channel, 167 feet wide, 15 feet deep. More than 5,600 El Paso residents were displaced from five South El Paso neighborhoods (Rio Linda, Cotton Mill, Cordova Gardens, El Jardín, southern Segundo Barrio). Texas State Historical Association / IBWC historical documentation / Historical Marker Database (HMDB.org).

Thalweg principle. First adopted for the Rhine in the 1801 Peace of Lunéville. Shah, S.A. (2010). River Boundary Delimitation and the Resolution of the Sir Creek Dispute between Pakistan and India. Vermont Law Review, 34(2).

Sir Creek dispute (India-Pakistan). A tidal creek in the Rann of Kutch; channel shifted since the 1914 Bombay Government Resolution map used to define the boundary. India argues for thalweg; Pakistan for the eastern bank. Twelve rounds of talks, dispute confirmed unresolved as of 2025. Shah, S.A. (2010) op. cit. The Diplomat (2025).

Uzbekistan water allocation (48.2% of Amu Darya managed flow). ICWC–Aral Basin Water Organization; FAO data. Reflects Soviet-era allocation framework.

Afghanistan’s contribution to Amu Darya flow (~30%) and current consumption (~2%). FAO and ICWC-Aral Basin Water Organization data.

Afghanistan’s Qosh Tepa Canal. Length: 285 km. Target operational date: 2028. Diversion estimates: Crosslin, Enerel L. Monitoring the Qosh Tepa Canal Project: A Geospatial Timeline of Taliban Water Diversion. Honors Thesis, University of South Dakota, Spring 2025 . Ibragimova, Galiya. Afghanistan’s Qosh Tepa Canal Could Trigger a Central Asian Water Crisis. Carnegie Endowment for International Peace / Politika, September 3, 2025 . The Diplomat (2025); Euronews (May 2025); Jamestown Foundation; Eurasianet.

NASA Earth Observatory; published academic record.

Soviet cotton production and Aral catastrophe. By 1988, Uzbekistan was the world’s largest cotton exporter. Soviet 1924 National Territorial Delimitation created the Central Asian administrative borders. 40,000–60,000 fishermen lost livelihoods. Caspian Policy Center.

Mekong River Commission (MRC). Established 1995 under the Agreement on the Cooperation for the Sustainable Development of the Mekong River Basin. China and Myanmar are dialogue partners, not full members; neither is subject to the consultation requirements or data-sharing obligations of full membership.

China’s Mekong dams. 11 hydropower dams on the upper Mekong (Lancang), confirmed in Stimson Center reporting. The figure of 95 dams on tributaries appears in broader Lancang basin dam-inventory literature.

Stimson Center / Eyes on Earth (2020). Eyler, B. and Weatherby, C. New Evidence: How China Turned Off the Tap on the Mekong River. Stimson Center, April 13, 2020. Satellite-data investigation finding that China’s Lancang dams withheld water during the 2019 drought while receiving normal precipitation upstream.

2019 Mekong drought. Seventeen Vietnamese provinces declared drought emergencies. Stimson Center; Al Jazeera; Foreign Policy.

Mekong fish stocks. 40% decline documented over the preceding decade. Mekong River Commission fisheries monitoring; NBC News; Mongabay (2024).

Potsdam Conference (July–August 1945). Allied powers established Poland’s new western boundary along the Oder and Neisse rivers. Approximately 25% of Germany’s pre-1937 territory transferred to Polish administration.

Oder-Neisse expulsions. Approximately 12 million ethnic Germans displaced from territories east of the Oder-Neisse line, 1945–1950. La Lumia, Cristiano, and Iris Nachum. Displacement and Compensation in Germany after the First and Second World Wars. Central European History, Vol. 58, No. 3 (2025), pp. 293–313. DOI: 10.1017/S0008938925000020. Mortality estimates 500,000–2 million: German Federal Archive (Bundesarchiv). Vertreibung und Vertreibungsverbrechen 1945–1948: Bericht des Bundesarchivs vom 28. Mai 1974. Published Bonn, 1989. Documents over 600,000 confirmed deaths; broader range reflects genuine historical uncertainty in disrupted wartime records.

Suleja, Włodzimierz, Stanisław Jankowiak, and Czesław Osękowski. Polski Dziki Zachód [Polish Wild West]. Roundtable discussion conducted by Barbara Polak. Biuletyn Instytutu Pamięci Narodowej. See also: Raack, R.C. Stalin Fixes the Oder-Neisse Line. Journal of Contemporary History, Vol. 25, Issue 4 (1990). DOI: 10.1177/002200949002500405.

Treaty of Warsaw (1970). West Germany accepted the Oder-Neisse line as Poland’s western frontier and affirmed its inviolability. Signed December 7, 1970; ratified May 17, 1972. American Journal of International Law 86, no. 1 (1992): 163–173. DOI: 10.2307/2203147.

Two Plus Four Treaty (September 12, 1990). As a condition of German reunification, Germany dropped the legal reservation regarding the Oder-Neisse line unconditionally.

German-Polish Border Treaty (November 14, 1990, ratified January 1992). Final formal confirmation of the Oder-Neisse line as Poland’s permanent western boundary.

Alsace-Lorraine border changes: 1871 (annexed by German Empire following Franco-Prussian War); 1919 (returned to France, Treaty of Versailles); 1940 (re-annexed by Nazi Germany); 1944 (returned to France by liberation). Standard historical record.

Berlin Conference (1884–85). Formalized the use of geographic features — rivers, meridians, parallels — to partition Africa among European colonial powers. No African government was represented. Standard historical record.