Ruzizi III: one project, two countries, and a sevenfold gap in physical displacement

The Ruzizi III Regional Hydropower Project sits on one river and affects three countries, but its resettlement footprint is not shared evenly. The World Bank's own project appraisal records 7 physically displaced households in Rwanda against 50 in DRC, and 841 economically affected households in Rwanda against 1,546 in DRC — more than 2,300 affected households in total, from a single 206 MW plant on a shared border.[1]

That asymmetry is the case study. A single engineering footprint — roughly 226 hectares, mostly reservoir inundation and access roads[1] — produces a sevenfold difference in physical displacement depending which bank of the river a household happens to sit on. Nothing about the dam changes at the border. What changes is population density, settlement pattern and land use either side of it.

The same project can be a modest acquisition in one country and a major resettlement programme in its neighbour, from the same act of construction.

The project, briefly

Ruzizi III is developed jointly by Burundi, DRC and Rwanda as a public-private partnership, with private sponsors including Industrial Promotion Services and a TotalEnergies-owned developer, through a special-purpose vehicle, Ruzizi III Energy Limited.[2] The plant sits on the Ruzizi River, which forms the natural border between the three countries as it runs from Lake Kivu to Lake Tanganyika.[1] The three national utilities — REGIDESO in Burundi, SNEL in DRC and EUCL in Rwanda — take equal shares of the output under power purchase agreements signed in 2019.[1]

Land acquisition and displacement occur only in Rwanda and DRC; the World Bank's disclosure is explicit that there is no land acquisition in Burundi for the hydropower component itself.[1] Burundi's exposure sits instead in the associated 220 kV transmission line to Bujumbura, a separately financed and separately governed component.[1]

Why the DRC and Rwanda figures diverge so sharply

The World Bank's disclosure gives a population-density figure for the combined project area: about 547 people per km² on average, ranging from 367/km² in Burundi to 753/km² in DRC and 874/km² in Rwanda.[1] On density alone, Rwanda's side is the more crowded — which makes the low physical displacement figure there, and the high one in DRC, counter-intuitive until the land-use pattern is considered rather than the population count alone.

A resettlement footprint tracks settlement pattern, not headcount. Where Rwandan households are concentrated in fewer, denser structures, a reservoir edge that skirts the settlement can take agricultural land from many households while displacing few. Where the equivalent strip on the DRC side crosses more dispersed structures, the same physical footprint removes more homes. The published totals do not explain which mechanism produced the divergence — only that it exists, and that it is large enough that a report describing "Ruzizi III's resettlement profile" without breaking out the two countries would materially mislead.

This is the operational lesson for any cross-border project: national aggregation is not optional detail, it is the fact that determines whether a resettlement budget and a livelihood-restoration programme are sized correctly on each side of the line. A project that plans one combined RAP with one blended average risks under-resourcing the country carrying most of the physical displacement.

The census that took three tries to see the Batwa

The project's general description states that "anthropological studies confirm the applicability of indigenous peoples' safeguards"[1] to the Historically Marginalized Communities in Rwanda and the Batwa ethnic group in DRC. Read only that line and the conclusion looks settled. The disclosure's own ESS7 screening section tells a more exact and more instructive story.

ESS7 — the World Bank's indigenous peoples standard, addressing free, prior and informed consent — was screened as relevant but its conditions determined not to apply: the Batwa in the project area were assessed as not depending on land for their livelihood and as living in mixed settlements rather than on traditionally-held land, so FPIC consultation was found not to be mandatory.[1] That is a specific, reasoned determination, not the blanket "safeguards apply" the summary line implies.

What actually happened to the Batwa population is the more interesting fact. Social assessments conducted in 2012 and 2021 recorded their presence but did not classify any of them as project-affected persons. A 2022 survey — the third attempt — identified roughly 20 Batwa families in Rwanda and 60 in DRC, and confirmed that some of them do qualify as PAPs: at least 11 households are affected by land acquisition.[1] Those households will be carried in the RAP as vulnerable, and a broader community development plan covering all Batwa beneficiaries — not only the confirmed PAPs — was to be prepared by Appraisal.[1]

That is the census failure this site's papers on vulnerability assessment and gender-differentiated outcomes describe in the abstract, caught in the act across a decade of the same project: a population correctly identified as present in 2012 was still not counted as affected nine years and a second survey later. What changed in 2022 was not the Batwa's circumstances — it was the census instrument's ability to see them. Naming a group in a project description is not the same achievement as counting them, and this disclosure is unusually honest about the gap between the two.

The cut-off date, and what the notice actually asks people to bring

Rwanda's cut-off date was set for 8 June 2026, following completion of the identification, inventory and valuation of land, houses, crops and trees across the affected sectors.[3] Compensation-agreement signing is being conducted sector by sector — Nzahaha first, Bugarama to follow — with each affected person notified at least three days before their scheduled signing date, and payment made only after all affected persons in a sector have signed.[3]

Two details in the public notice are worth flagging as good practice. First, legally married couples must both be present to sign the compensation forms — a direct, structural answer to the pattern this site's paper on gender-differentiated outcomes describes, where payment to a single household signatory can leave a spouse's interest unrecorded.[3] Second, the documentary requirements are explicit and varied: national ID, land title, marriage or widow certificate, and certificate of single status as applicable[3] — which means the process anticipates unmarried, widowed and single claimants as distinct categories with distinct evidence, rather than defaulting to one documentary standard for everyone.

What the register has to carry across a shared border

• Country as a first-class field on every entitlement, not an assumption embedded in which office administers the record — so a rate schedule, a cut-off date or a grievance route applicable to one country is never accidentally applied on the other bank.
• Physical and economic displacement reported separately per country, never blended into a single project total, given how far the two figures diverge here even within one engineering footprint — precisely the disaggregation Guidance Note 5 expects a census to preserve.[5]
• Named ethnic and marginalised-group status as its own field, carrying the indigenous-peoples safeguard screening result distinctly from the general vulnerability flag — including a screened- but-not-triggered outcome, which is a different fact from "not screened" and needs to be recorded as such. The private-sector component of the project is bound to the equivalent PS-5 requirement to make this determination explicit.[4]
• Joint signature evidenced for married claimants, and the specific document type relied on for each claimant's marital or civil status, recorded rather than inferred.
• A sector-by-sector rollout log — notification date, signing date, payment date — so that a household's position in the phased process is auditable independent of the project's aggregate completion percentage.

What this article does not establish

The World Bank source used here is a Concept-stage Environmental and Social Review Summary — an early appraisal disclosure, not a completed RAP or a completion audit — and its own text notes the draft RAP is still being updated to align with the Bank's template, ESS5 and national law.[1] Every figure in this article may change before Board approval and full disclosure of the final RAP. The Rwandan compensation-signing process described here was reported as underway in 2026; this article had no access to a completed payment record and cannot confirm how many households have in fact been compensated. Nothing here evidences the DRC-side process specifically, on which this article found no equivalent detailed reporting. This is not legal advice on Rwandan, Congolese or Burundian land law.