Running Dry, Running North

Displacement in the Red Sea and Horn of Africa Corridor, Special Commentary. August, 2026

Executive Summary

The corridor running from the Ethiopian highlands and the Somali interior through Djibouti and Bosaso, across the Gulf of Aden into Yemen, and onward to the Saudi frontier is now one of the busiest irregular migration routes in the world. It is also one of the least understood, because it is persistently described in the wrong terms. Governments and donors treat it as a smuggling problem, a border problem, or a security problem. It is at least equally a financing problem, and the financing part is the part they control.

The land runs dry and then the people run north, and the interval between those two events is where policy either works or does not. That interval is measured in seasons, and what fills it is money: the finance that lets a household rebuild a herd, repair a water point, or survive a failed harvest without liquidating the asset base that makes the next harvest possible. Adaptation finance and migration are not separate ledgers. They are the same ledger, read at different points in the sequence.

• Displacement inside Somalia is now overwhelmingly climatic wherever it is being measured. In the five monitored districts worst affected by the current drought, drought accounts for three of every four new displacements. IOM has cautioned that the national total is likely to exceed 300,000 people but that its data covers those five districts only.

• Movement along the Eastern Route is rising steeply, not falling, despite a decade of enforcement investment on both sides of the water. Recorded arrivals in Yemen from the Horn rose 77 per cent across 2025, from 60,900 to 107,900.

• The spending pattern is inverted. Of roughly USD 17 billion in development assistance to Somalia between 2012 and 2021, about USD 1.5 billion went to climate resilience. Somalia’s humanitarian appeals for 2021 to 2023 alone sought around USD 5.8 billion, close to four times a decade of resilience spending.

• Global adaptation finance is going backwards. International public flows fell from USD 28 billion in 2022 to USD 26 billion in 2023 against projected needs of USD 310 to 365 billion a year by 2035. Fragile and conflict-affected states receive around USD 5 per person annually, against USD 7 in non-fragile developing countries and about USD 2 where conflict is intense.

• Deterrence has changed the shape of the route without changing its volume. Through the first half of 2025, transit fell by a quarter through Bosaso while more than doubling through Obock, and total arrivals in Yemen rose.

• The humanitarian fallback that has absorbed the adaptation shortfall is itself now failing. By April 2026, the Somali response had received 14 per cent of its requested funding.

• Delivery is the binding constraint, not intent. Armed actors tax water, tax assistance, and attack the people who build infrastructure. Any credible adaptation programme has to be designed against that reality rather than around it, and there is now enough operational evidence to do so.

This is a claim about which lever is available, not a claim that climate is the only driver. Conflict runs through this corridor at every point, and section 1 sets out how the two interact.

The recommendations are organised around three shifts: repricing risk so that fragility stops being a disqualification, financing institutions rather than only projects, and treating mobility as a legitimate adaptation strategy rather than as evidence that adaptation has failed. Section 3.1 sets out the costing and its limits, section 5 the delivery problem in contested territory, and Annex A the implementation design for the two recommendations that turn on execution rather than principle.

1. The Wrong Question

Almost all policy attention on this corridor is organised around a single question: how do we stop the movement. That question has produced patrol boats, biometric registration, return agreements, awareness campaigns in villages of origin, and a substantial architecture of externalised border management. It has not produced a reduction in movement.

A more productive question is why people stay where they are for as long as they do, and what it costs to make staying possible. Most people in drought-affected districts do not move. They sell livestock, reduce meals, borrow from kin, send one household member to a town, and wait for the next season. Movement across a border is typically the last option, taken after the domestic coping repertoire has been exhausted. The policy-relevant fact is not the departure but the sequence of failures that precedes it. This reframing changes what counts as a migration intervention: a functioning borehole, a livestock insurance scheme that pays out before the animals die, and a cash transfer triggered by a rainfall forecast rather than a famine declaration are all migration policy instruments.

What this frame does not claim

The obvious objection is that this corridor is shaped by conflict, not by climate. It is. Al-Shabaab territorial control and taxation determine which land can be farmed and which water points can be reached. Ethiopia has generated large displacement flows for reasons unrelated to rainfall. The war in Yemen defines what happens to people once they have crossed. The two drivers also interact rather than compete: insecurity is frequently what converts a survivable drought into a displacing one, by blocking movement to pasture, severing market access, and preventing assistance from arriving. Attribution between them is genuinely contested, and any figure presented as a clean split should be treated with caution, including the figures here.

The claim is therefore narrower than a claim about causation. Of the forces operating on this route, adaptation finance is the one that donors, multilateral funds, and destination governments directly control, and the one they are getting wrong in ways that are documented and correctable within a budget cycle. A finance frame that ignores conflict will fail, which is why section 5 treats delivery under armed actors as a design problem rather than an excuse. A conflict frame that ignores finance has now been tried for a decade.

2.The Corridor Today

The Eastern Route carries people from Ethiopia and Somalia through Djibouti or the Somali coast, across the Gulf of Aden into Yemen, and northward toward Saudi Arabia, where the Ethiopian and Somali labour markets have been embedded for decades. The route is dangerous, well established, and expanding.

Recorded arrivals in Yemen from the Horn reached 107,900 across 2025, up 77 per cent on the 60,900 recorded in 2024. In the final quarter alone there were 57,340 arrivals, 28 per cent above the same period a year earlier. Transit through the Djiboutian port of Obock rose 58 per cent over the year, from 76,600 to 121,300, with the sharpest proportional increases among women. These figures should be read as floors. Flow monitoring in Yemen was suspended between May and July 2025 for funding reasons, and monitoring at Bosaso was interrupted in January of the same year. The 2025 totals were recorded despite those gaps, which is the strongest reason to treat them as conservative, and it means the corridor is now less consistently observed than it was five years ago.

On the origin side the picture is unambiguous. Somalia declared a national drought emergency in November 2025 after four consecutive failed rainy seasons. By early 2026, around 6.5 million people faced high levels of acute food insecurity, with at least two million in the emergency phase and more than 1.8 million children under five facing acute malnutrition, including 483,000 in the severe category. The World Food Programme estimated cereal production at 83 per cent below the long-term average. Nearly 62,000 people were displaced by drought in the first months of 2026 across five monitored districts, with drought accounting for three of every four new displacements there, a rise of 22 per cent on the previous year. IOM has been explicit that the national figure is likely to exceed 300,000 and that its data covers five districts only, so the monitored share should not be read as a national attribution. A further 125,000 drought displacements were projected for the second quarter of 2026 even under normal or above-normal Gu rainfall.

At the far end of the route, enforcement operates at industrial scale: Saudi authorities have reported arrests for irregular entry running close to a million a year, and Ethiopia has run repeated large-scale repatriation rounds under bilateral arrangements since 2022. The route absorbs these returns and continues to fill.

Table 1. Corridor and finance indicators at a glance

Indicator	Latest reading	Direction
Arrivals in Yemen from the Horn	107,900 in 2025	Up 77 per cent on 2024 (60,900)
Arrivals, fourth quarter 2025	57,340	Up 28 per cent year on year
Transit through Obock, Djibouti	121,300 in 2025	Up 58 per cent on 2024 (76,600)
Transit through Bosaso, Somalia	20,800, January to June 2025	Down 25 per cent on same period 2024
Somalis in acute food insecurity	Around 6.5 million, early 2026	Rising after four failed seasons
Drought share of new displacement	Three in four, five monitored districts	Up 22 per cent on 2025
International public adaptation finance	USD 26 billion in 2023	Down from USD 28 billion in 2022
Adaptation finance per capita, fragile states	About USD 5 per person per year	Against USD 7 in non-fragile states; USD 2 in high-intensity conflict

Sources: IOM Displacement Tracking Matrix and Eastern Route quarterly reporting; Mixed Migration Centre, fourth quarter 2025; IPC and OCHA assessments; UNEP Adaptation Gap Report 2025; World Bank (2024).

3. Why Adaptation Finance Is the Right Frame

Adaptation finance is the money that allows a household, a district, or a country to absorb a shock without losing its productive base. It is distinct from mitigation finance, which reduces emissions, and from humanitarian finance, which sustains life once the productive base has already been lost. The Horn needs all three. It receives mostly the third.

At the global level the arithmetic is stark. Developing countries are projected to need USD 310 to 365 billion annually for adaptation by 2035, against international public flows of USD 26 billion in 2023, down from USD 28 billion the year before. The shortfall is between twelve and fourteen times current provision, the Glasgow doubling commitment is on course to be missed, and the successor goal is not calibrated to close the gap. Composition compounds scale: a clear majority of adaptation finance now arrives as debt, much of it non-concessional. A country borrowing to build drought resilience is being asked to service a loan out of an economy the drought has already suppressed. For states with limited fiscal capacity and existing debt distress, this is a structurally unsound proposition dressed as support.

The distributional pattern is the sharpest problem and the one most directly relevant here. World Bank analysis puts per capita adaptation finance in fragile and conflict-affected states at around USD 5 a year, against roughly USD 7 in non-fragile developing countries and about USD 2 in high-intensity conflict settings. Those gaps look modest as ratios and are severe as allocations, because the same countries carry the highest exposure and the least capacity to self-finance. A country facing several times the climate risk and receiving less than the average allocation is not being treated equally; it is being treated regressively.

Somalia illustrates the pattern at its extreme. It ranks 71st out of 79 low and lower-middle income countries by funding approved from multilateral climate funds, and has been almost entirely dependent on bilateral channels as a result. The Somali government has put the annual climate requirement at around USD 5 billion; SPARC has assessed the costed adaptation requirement to 2030 at USD 48.5 billion, more than six times 2022 gross domestic product. The largest single climate fund approval to date is the USD 95 million project agreed with the Green Climate Fund and FAO in October 2024. This is the inversion at the centre of the problem: the instrument designed to reduce climate-driven vulnerability is allocated in inverse proportion to vulnerability, and the countries generating the largest climate-driven outflows are those least able to access the finance that would reduce them.

3.1 What the two approaches cost

The claim that prevention is cheaper than response is made constantly in this policy area and almost never costed. What follows is an attempt at an order-of-magnitude comparison. It is not a like-for-like costing, the limitations are set out below, and readers should treat the ratios rather than the absolute figures as the finding.

Line	Approximate figure	Period and basis
Development assistance to Somalia, all purposes	USD 17 billion	2012 to 2021, OECD disbursements
Of which climate resilience	USD 1.5 billion, about 9 per cent	Same period and basis
Somali humanitarian appeals	USD 1.09bn, 2.1bn, 2.6bn	2021, 2022, 2023; requested, not received
Somali assessed climate requirement	Around USD 5 billion a year	Government estimate, 2024
Somali costed adaptation need to 2030	USD 48.5 billion	SPARC; over six times 2022 GDP
Largest single climate fund approval	USD 95 million, about 1.2m people	GCF and FAO, October 2024
EU Trust Fund, Horn of Africa window	About EUR 1.8 billion	2015 to 2021, nine countries, mixed purposes
Of which regular migration schemes, fund-	EUR 56 million, under 1.5 per cent	Oxfam analysis of the full EUTF

Line	Approximate figure	Period and basis
wide
Regional Migrant Response Plan appeal	USD 81 million, severely underfunded	2025, corridor-specific
CERF anticipatory action, Somalia	USD 20 million, about 903,000 people	2021 activation, roughly USD 22 per person
Modelled return on resilience investment	USD 2.80 per USD 1	USAID, Somalia, 15-year model
Modelled savings from resilience	USD 794 million, about USD 53m a year	USAID, 15 years, avoided losses included

Sources: OECD disbursement data as compiled by International Crisis Group; OCHA appeal documents; SPARC; GCF and FAO; European Council and Oxfam; IOM; CERF allocation records; USAID Economics of Resilience to Drought. See references.

Three comparisons

The first is internal to the aid budget and is the most defensible of the three. Somalia received roughly USD 17 billion in development assistance between 2012 and 2021, of which about USD 1.5 billion, some nine per cent, went to climate resilience. Over an overlapping and much shorter window, three humanitarian appeals covering 2021 to 2023 sought around USD 5.8 billion. The system requested nearly four times as much for three years of emergency response as it spent on a decade of resilience in the same country. Both figures come from the same donors and are recorded in the same accounts, which is why this comparison survives scrutiny in a way the others do not.

The second is the unit cost of acting early. The 2021 anticipatory action activation released USD 20 million through the Central Emergency Response Fund across seven agencies and five sectors, reaching about 903,000 people before the drought peaked, at roughly USD 22 a head. The humanitarian appeal for the following year sought USD 2.1 billion. USAID modelling for Somalia puts the return on early response and resilience at about USD 2.80 for every dollar invested, with total savings of USD 794 million over fifteen years once avoided income and livestock losses are included, and a 30 per cent reduction in humanitarian outlay from a proactive rather than a late response. The USAID figures are modelled rather than observed and rest on assumptions about household income effects that should be read as indicative.

The third comparison is the weakest and the most suggestive. The European Union committed roughly EUR 1.8 billion to the Horn of Africa window of its Emergency Trust Fund between 2015 and 2021, an instrument created explicitly to address irregular migration. Across the fund as a whole, Oxfam found that under 1.5 per cent, about EUR 56 million, went to expanding regular migration channels. Meanwhile the corridor-specific Regional Migrant Response Plan, which covers protection, reintegration, and the route monitoring on which this analysis depends, appealed for USD 81 million in 2025 and remained severely underfunded.

What these figures cannot show

Four limitations should be stated plainly, because the comparison is worth less than it appears if they are not.

Category contamination. Development and trust fund totals bundle genuine development spending with migration management, and the boundary is drawn differently by different reporting systems. The EUTF in particular funded employment programmes, refugee services, and road rehabilitation alongside border capacity building; treating the whole envelope as enforcement spending would be dishonest, and no published disaggregation is reliable enough to separate them.

Requests are not receipts. Humanitarian appeal figures record what was asked for. Somalia’s 2023 appeal was 46 per cent funded; the 2026 response stood at 14 per cent funded in April. Comparing appeals against disbursements overstates the humanitarian side, and comparing disbursements against disbursements would narrow the ratio without reversing it.

The largest line is invisible. Enforcement expenditure by Gulf destination states, including detention and deportation costs associated with arrests running close to a million a year, is not published. This is probably the single largest item in the whole system and its absence is the main reason a proper costing does not exist.

No counterfactual. None of these figures establishes what displacement would have occurred under a different spending pattern. The USAID model is the closest available and it addresses humanitarian caseload rather than cross-border movement.

The study that would settle it

The specification is straightforward. A corridor-level cost model comparing, on a per-averted-displacement basis, a sustained adaptation package in origin districts against the combined annual outlay on humanitarian response, border management, detention, and return across origin, transit, and destination states. It requires four inputs: disaggregated donor reporting on migration-related and adaptation-related spending in Somalia, Ethiopia, and Djibouti; published or estimated enforcement and return costs for Saudi Arabia and the EU external action instruments; district-level displacement data from the Displacement Tracking Matrix, which already exists; and a matched comparison design across districts receiving and not receiving adaptation investment.

Two of the four inputs already exist in usable form. The binding gap is destination-state enforcement expenditure, which no origin-state institution can compel and which a coalition of donor governments could publish tomorrow. Commissioning the study would cost a small fraction of one quarter of any of the lines in Table 2, and until it is done the central claim in this policy area will continue to be asserted rather than demonstrated, including by us.

4. Five Structural Failures

4.1 The fragility discount

Climate funds are risk-averse institutions staffed by people whose professional exposure runs one way. A project that fails in a fragile state is career-damaging; a project not attempted there is invisible. Fiduciary standards, co-financing expectations, and accreditation requirements are applied uniformly to states with radically different administrative capacity, which functions as a screening device against the weakest applicants. The average interval from concept to disbursement in Somalia has been measured at over five years. A five-year approval cycle is not a delay in a drought that arrives on an eighteen-month rhythm. It is a refusal expressed procedurally.

4.2 Humanitarian substitution, and its collapse

When adaptation finance does not arrive, humanitarian finance has historically arrived in its place, at higher cost and lower durability. Somalia’s humanitarian appeal more than doubled between 2021 and 2022. Emergency money keeps people alive, which is not a small thing, but it cannot build a water point that lasts a decade or capitalise an insurance scheme; its planning horizon is a season. The result has been a system that pays repeatedly for the consequences of an investment it declines to make once, and a rural political economy in which the relief pipeline becomes the principal source of external revenue.

That substitution is now failing on its own terms, which changes the urgency of everything else here. By April 2026, UN agencies and partners had received 14 per cent of the funding requested for the Somali response, and delivery reached under a quarter of assessed need across the 21 priority districts. Arguments for adaptation finance have usually been arguments about efficiency, on the reasoning that prevention is cheaper than relief. In the current funding environment the argument is no longer about efficiency. There is no longer a relief backstop of sufficient size to be the expensive alternative.

4.3 Leakage and the corruption interaction

The interaction between climate stress and governance failure is multiplicative rather than additive, and it operates through the delivery layer. Climate stress erodes livelihoods; corruption weakens the projects meant to restore them, distorts relief allocation, and restricts lawful mobility by turning passports, exit permits, and labour contracts into rationed goods sold at a premium. When both forces are present, displacement becomes structural and stops responding to enforcement at the border, because the border is not where it is generated. The mechanism is set out at greater length elsewhere (OHirsi 2026).

The independent evidence is substantial and does not rest on any single source. Transparency International’s 2024 index placed Somalia at the bottom of the global range and warned that climate funds are exposed to theft precisely where climate vulnerability and weak controls coincide. A 2026 U4 Anti-Corruption Resource Centre review documents the mechanism at settlement level, where gatekeepers extract fees and resell in-kind assistance, a practice a leaked UN report found widespread in the 2022 drought response; a confidential UN report in September 2023 documented extortion of cash assistance across 55 displacement camps; and a 2025 re-registration in Baidoa found households genuinely in need some 45 per cent below the previous register. The same review notes a 2025 media allegation, which remains unverified, that officials at the Ministry of Environment and Climate Change embezzled more than USD 5.7 million from a climate resilience project.

Two qualifications matter for credibility. First, the U4 review found no substantiated cases of corruption channelling assistance directly to Al-Shabaab; the group’s extraction is coercive taxation rather than corruption in the conventional sense, and conflating the two overstates the case. Second, the sums involved in documented diversion are small relative to the sums not disbursed at all. The fragility discount described above costs Somalia more than leakage does.

The policy implication is uncomfortable for both sides. Donors citing fiduciary risk as grounds for withholding are describing a real problem, but withholding does not reduce the corruption. It removes the resource base on which resilience would be built while leaving the governance deficit untouched, and it guarantees the displacement the same donors then spend far more to police. The answer is to fund with tracking, audit, and expenditure transparency built into the instrument from the start.

4.4 Mobility is absent from adaptation planning

Most national adaptation plans in the region treat migration as an outcome to be prevented rather than a strategy to be supported. This is analytically backwards. Seasonal and circular labour migration has been a core Horn of Africa adaptation mechanism for centuries, and remittances are among the most reliable counter-cyclical transfers available to a Somali or Ethiopian household. A framework that codes all movement as failure will under-invest in the one adaptation channel that already works, and will push regulated, temporary, reversible mobility into the smuggling economy.

4.5 What adaptation finance can and cannot be expected to do

One objection deserves a direct answer. If mobility is itself an adaptation strategy, and if poverty constrains movement as much as it drives it, then better-resourced households may migrate more rather than less. The mobility transition literature broadly supports this: emigration tends to rise with income before it falls, and the very poorest are often too poor to leave. The objection is well founded and it narrows the claim rather than defeating it. Adaptation finance should not be sold as a mechanism for reducing departure numbers, because on the available evidence it will not reliably do that within a political time horizon, and a report that promises it will be found out within two reporting cycles.

What adaptation finance changes is the character of the movement: whether it is chosen or forced, seasonal or permanent, documented or smuggled, undertaken with an asset base intact or after the herd has died and the household has borrowed against a journey it cannot repay. A pastoralist who moves after an insurance payout and one who moves after total loss are both recorded as migrants. They are not in the same position, they do not enter the same labour markets, and they do not generate the same costs downstream for anyone. This is the more defensible case and it is the one consistent with Group C below. The policy objective is not stillness. It is that movement should be a decision rather than a collapse.

5. Delivering Under Armed Actors

Everything above concerns whether the money is allocated. This section concerns whether it can be spent. In roughly the areas where need is highest, adaptation delivery has to pass through or around an insurgency that treats water as revenue and infrastructure as a target. A financing argument that does not engage this is an argument for money that will sit undisbursed, and undisbursed money is how the fragility discount reproduces itself.

5.1 Three modes of extraction

The first is direct taxation of the resource. In Jubaland, Al-Shabaab charges herders around USD 0.50 each time a camel drinks at a trough it controls. Farmers in Jilib must pay at least USD 20 before sowing, regardless of acreage, and the group obstructs water to the fields of anyone who refuses. In Adan Yabaal, residents were required to maintain diesel pumps the group managed and then to pay for their own livestock’s water. The implication for programme design is immediate and frequently missed: a borehole built in territory the group controls does not simply serve a community. It becomes a toll gate, and the donor has capitalised the collector.

The second is taxation of the delivery itself. The group has demanded levies of up to 40 per cent on assistance or cash routed through local intermediaries, alongside registration fees for agencies seeking access, a practice dating to the 2011 famine. Western, Turkish, and Gulf-based agencies typically refuse, and counter-terrorism financing rules in donor jurisdictions make refusal close to mandatory. The consequence is that coverage stops at the territorial line rather than at the line of need. In 2022, roughly 660,000 people in areas under the group’s control were largely unreachable, of whom around 375,770 were assessed as in urgent need.

The third is violent interdiction. Crisis Group identified at least 65 attacks on food convoys and water points between October 2020 and December 2022, nearly double the number in the preceding two years, with the rate rising during the government offensive that began in August 2022. In September 2022, twelve Somali men drilling a well at Geriley in Gedo were killed and their equipment burned. This is not incidental risk to be priced into a contingency line. It is a deliberate strategy of denying the state and its partners the capacity to deliver, and it intensifies precisely when the state is advancing.

Alongside the insurgency sits the distinct domestic diversion problem described in section 4.3. The two operate on different logics, coercion in one case and rent-seeking in the other, and they require different countermeasures. Treating them as a single corruption problem produces controls that address neither.

5.2 What the operational record shows is possible

The record is not one of uniform failure, and the exceptions are instructive. An EU-funded irrigation scheme at Marka and Janale, costing around USD 4 million, ran from 2021 to completion in March 2023 despite sustained interference. FAO and IOM convened water management committees that brought feuding clans together and included women and young people. Al-Shabaab threatened staff, planted improvised explosive devices along the canal, and held people connected to the project for three months. The scheme was finished, and it now supplies roughly 3,000 families, including households in areas the group controls.

Smaller precedents point the same way. A Somali organisation built a sand dam in Gedo in 2021 after residents themselves persuaded the group that the structure mattered. In Afmadow in 2020, the group first refused permission for a community rainwater cistern and then allowed it. Flood risk warnings now reach people in areas under the group’s control by automatic mobile alert, which is delivery requiring no physical access at all. And the group has itself built a reservoir, a canal, and a water tower at Buulo Fulaay in Bay region, confirmed by satellite imagery. Water infrastructure is not categorically opposed. Control of it is contested.

One further datum is worth holding alongside the others. During the 2020 to 2023 drought the group’s conduct, including blockades of Wajid, Hudur, Qansadheere, and Dinsor and the destruction of water points, generated sufficient local resentment to help produce the clan uprising that enabled the 2022 offensive. Denial of water is not costless for the group either, which is what makes community-level negotiation over access a real lever rather than a wishful one.

5.3 Six design rules

1. Assume extraction and price it explicitly. A delivery model whose viability depends on zero leakage will not operate in the districts with the highest need. The choice is not between clean delivery and compromised delivery. It is between bounded, visible leakage and no delivery at all, and that trade-off should be stated in programme documents rather than discovered in an audit.
2. Prefer modalities that are hard to tax and hard to capture. Mobile money transfers to verified individuals, mobile early warning, and index insurance settlements move value without moving a convoy. Large contracted infrastructure requires exactly the physical presence the insurgency attacks and the supply chain it taxes.
3. Do not build assets that convert into rents. Fixed water points under contested control transfer as revenue to whoever holds the ground. Payouts, transfers, and portable inputs do not. Where fixed infrastructure is genuinely necessary, cross-clan community management committees are the mechanism with the best operational record, and their composition should be a condition of disbursement rather than an implementation detail.
4. Sequence fixed infrastructure to durable territorial control. The Jowhar and Balad rehabilitation, estimated at around USD 140 million with roughly half pledged, is viable only where the offensive has held. Portable and community-managed measures carry no such precondition and should be the default in contested districts.
5. Separate the countermeasures for coercion and for rent-seeking. Insurgent taxation is addressed through modality choice, community negotiation, and access diplomacy. Gatekeeper extraction is addressed through verified direct transfer, independent re-registration, and published beneficiary counts. Applying either toolkit to the other problem wastes the control and produces the false comfort of having done something.
6. Stop criminalising the negotiations that make access possible. Elders in Somalia can face arrest for meeting or agreeing terms with Al-Shabaab, yet those negotiations are frequently the only route by which a community secures water in the districts concerned. Crisis Group’s recommendation, that Mogadishu should at minimum refrain from interfering with community-level dialogue over water and relief access, is uncomfortable and correct. Donors should say so rather than leaving Somali intermediaries to carry a risk their own counter-terrorism financing rules created.

None of this makes delivery in contested territory straightforward, and no design rule survives contact with a group that has killed well-drillers. What the record establishes is that the binding constraint is design rather than possibility, and that the sector has enough operational experience to design against the constraint instead of citing it as a reason to withhold.

6. What Deterrence Has Purchased

The record of recent years supports a reasonably confident assessment. Enforcement changes the route. It does not change the volume. Intensified operations in Djibouti and northern Somalia during 2025 produced precisely the pattern the reframing predicts: transit through Bosaso fell 25 per cent in the first half of the year, from 27,800 to 20,800, as local authorities ordered returns and patrols tightened, while transit through Obock more than doubled, from 32,000 to 65,000, and recorded arrivals in Yemen more than tripled against the equivalent 2024 window. Smugglers adopted new landing points in Ta’iz to evade coastguard patrols. Enforcement at one node moved traffic to another; it did not reduce the traffic. It also raises the price charged by smugglers, lengthens and endangers the journey, and increases the leverage of trafficking networks over people carrying more debt. It does not alter the calculation of a pastoralist who has lost his herd.

Mass return without reintegration finance is the same policy in a different register. Returnees arrive in origin communities that are, by definition, the communities that could not sustain them the first time, and that are now under greater climate stress than when they left. Survey evidence from Ethiopian returnees consistently shows re-migration intention surviving detention, deportation, and documented abuse. A return programme unaccompanied by livelihood investment is a recycling mechanism charged to the humanitarian budget.

7. Recommendations

Group A concerns how adaptation finance is allocated, Group B what it is spent on, and Group C the treatment of movement itself. Each names the decision point at which it would have to be taken. Recommendations 4 and 5 turn on implementation detail rather than principle; their sequencing and pilot design are set out in Annex A.

Group A: Reprice fragility

1. Establish a fragility-weighted allocation floor in the major multilateral climate funds, so that a minimum share of adaptation resources is reserved for fragile and conflict-affected states and cannot be reallocated to lower-risk portfolios at the margin. Decision point: the Green Climate Fund and Adaptation Fund Boards, at portfolio review.
2. Adopt a grant-first rule for least developed countries in conflict or post-conflict conditions, with an explicit ceiling on the proportion of adaptation support delivered as non-concessional debt. Financing resilience through instruments that deepen debt distress is self-cancelling. Decision point: bilateral donors at the point of programming, and the climate funds at replenishment.
3. Create a simplified direct-access window with a hard decision clock, ideally twelve months from concept to approval, accepting smaller project sizes and higher supervision intensity as the price of speed. Against a Somali average of over five years, this is the change with the largest effect per unit of effort. Adaptation finance arriving after the displacement has been converted into humanitarian finance at a loss. Decision point: the Green Climate Fund Board and its accreditation panel.

Group B: Fund institutions, not only projects

4. Capitalise national designated authorities as permanent institutions rather than project management units, on the four-phase sequence in Annex A. Somalia’s access constraint is administrative before it is political. The precedent is documented: World Bank support to Somali public financial management preceded roughly USD 700 million in on-budget donor support and USD 90 million from the Green Climate Fund, while South Sudan, without the equivalent, received neither. Decision point: bilateral donors, in a pooled multi-year facility.

5. Extend the existing anticipatory action framework rather than building a parallel one, adding an asset-protection window alongside the humanitarian one and piloting it across three districts including one contested. Design specification in Annex A. Money released on a forecast is worth several times the same money released on a famine classification, and Somalia already hosts the longest-running framework of its kind. Decision point: the Emergency Relief Coordinator and CERF, with co-funding from the climate funds.

6. Make expenditure tracking and independent audit a condition of scaling rather than a barrier to entry. Small tranches with tight verification, escalating on demonstrated performance, will move more money into fragile settings than large tranches gated behind fiduciary standards those settings cannot yet meet.

7. Invest specifically in the rural water, rangeland, and livestock asset base, applying the design rules in section 5.3 to anything sited in contested territory. Livestock is the productive capital of the Somali interior, and its loss is the most reliable single predictor of onward movement.

Group C: Treat mobility as adaptation

8. Negotiate a regional labour mobility framework covering skills recognition, portable social protection, transparent recruitment, and enforceable contract standards. Substantial migration between these regions is going to occur regardless; the only variable in policy control is whether it occurs through a regulated channel or a criminal one. Decision point: the IGAD Council of Ministers and Gulf Cooperation Council labour ministries, building on existing bilateral labour agreements rather than replacing them.

9. Attach reintegration finance to every return and readmission agreement as a binding component rather than an aspiration, benchmarked against livelihood restoration in the receiving district rather than arrival numbers. Decision point: the parties to each bilateral readmission arrangement, at renewal.

10. Fund route monitoring on a multi-year basis rather than annually, so coverage cannot lapse for budgetary reasons as it did between May and July 2025, and establish a standing corridor observatory with shared reporting across origin, transit, and destination states. Commission alongside it the corridor cost model specified in section 3.1. This is the cheapest recommendation here and the precondition for the rest. Decision point: the donors to the Regional Migrant Response Plan.

8. What to Watch

Five indicators will show within twelve to eighteen months whether any of this is changing.

• Performance of the coming Deyr and Gu seasons, and whether the current sequence of failed rains extends further.
• Monthly transit volumes through Obock and Bosaso, read together rather than separately, since enforcement at one displaces traffic to the other. The 2025 divergence is the clearest available evidence of substitution.
• Recorded arrivals in Yemen, read alongside monitoring coverage. A fall in recorded arrivals during a coverage gap means nothing, and coverage lapsed twice in 2025.
• The grant share of new adaptation approvals to Somalia, Ethiopia, and Djibouti, which is the clearest single test of whether the debt problem is being addressed.
• Time from concept note to first disbursement for climate fund applications from the region. The benchmark to beat is five years.

Conclusion

The Red Sea corridor is usually described as a crisis of movement. It is better understood as a crisis of what happens before the movement. Land runs dry across the Somali interior and the Ethiopian lowlands on a rhythm that is now well documented and increasingly well forecast. Whether people then move, and in what condition, is not fixed by the rainfall. It is decided in the interval that follows, by whether adaptation finance arrives in time, in a usable form, and through channels that reach the household rather than stopping several layers above it.

The point is not that finance stops people from moving. It may not, and it should not be sold on that promise. The point is that finance determines whether they move with something left or with nothing, into a labour contract or into a smuggler’s debt. At present it does not arrive at all in the places where that distinction is being made, and the countries where it fails to arrive are precisely the countries generating the flows that destination states then spend heavily to contain.

There is a version of this policy debate in which the enforcement budget and the adaptation budget are recognised as two entries in the same account. That recognition would not resolve the corruption problem, the conflict problem, or the rainfall problem. It would, however, stop the international system from paying twice for the failure to pay once.

Annex A. Sequencing and Pilot Design

Recommendations 4 and 5 fail or succeed on implementation detail. Both have been attempted before in forms that did not hold. What follows specifies the sequence, the decision gates, and the tests that determine whether each phase proceeds. It is written for programme designers, fund secretariats, and donor desk officers, and can be read independently of the argument above.

A.1 Institutional capitalisation: a four-phase sequence

The binding constraint on Somali access to climate finance is accreditation, not appetite. The USD 2 million the Ministry of Environment and Climate Change obtained from the Green Climate Fund in 2021 came indirectly, through UNDP and the Global Water Partnership. The USD 95 million project agreed in October 2024 is executed by FAO. Somalia is pursuing accredited entity status with UN support, and until it holds that status every dollar arrives with an intermediary’s overhead and an intermediary’s timetable attached.

Table 3. Institutional capitalisation, phased with gates

Phase	Action	Indicative scale	Gate to next phase
0 (months 0 to 12)	Fund the accreditation process itself as a standalone line, not as an activity inside a project	USD 3 to 5 million	Accreditation application filed and under assessment
1 (months 12 to 30)	Core multi-year funding for the national designated authority: salaries, systems, audit function, retained rather than seconded staff	USD 10 to 15 million over three years	Published expenditure tracking; external audit with no material findings
2 (months 30 to 48)	First directly accessed tranche, deliberately small, with independent audit and public reporting from day one	USD 20 to 40 million	Disbursement within stated clock; audit clean; results verified independently
3 (month 48 onward)	Escalate tranche size on demonstrated performance, with each increment gated the same way	Doubling per verified cycle	Continuous; failure freezes rather than cancels

Indicative scales illustrate the order of magnitude required. They are not costed estimates, and they are set deliberately below typical multilateral project size, which is the point.

Three features of the sequence matter more than the numbers. The first is that failing a gate freezes escalation rather than cancelling the programme. Cancellation returns the country to the humanitarian channel, which is the outcome the whole exercise exists to avoid, and everyone involved knows it, which is why cancellation threats are not credible and are therefore not deterrents. A freeze is credible because it is survivable.

The second is that phase 1 funds retained staff rather than seconded consultants. The failure mode of the past decade has been externally staffed programme units that dissolve at project close, leaving no residual capacity and requiring the next programme to rebuild from zero. The World Bank public financial management precedent worked because it built a function that outlasted the intervention.

The third is that phase 0 is the cheapest phase and the one most likely to be skipped, because accreditation support is administrative and unglamorous and does not photograph well. It is also the gate on everything downstream.

A.2 Anticipatory action: extending what already exists

Somalia hosts the longest-running collective anticipatory action framework in the humanitarian system. It was developed under the Humanitarian Coordinator with the federal government, OCHA, and the World Bank, and first triggered in June 2020 with USD 15 million from the Central Emergency Response Fund. It was triggered again in early 2021, releasing USD 20 million across seven agencies and five sectors and reaching about 903,000 people before the drought peaked. The infrastructure of triggers, pre-agreed actions, and pre-arranged finance exists. It does not need to be invented, and inventing a parallel one would be the most reliable way to waste the next three years.

The independent evaluation identified a specific weakness worth designing around: the framework was built for drought and then activated against several different shocks, which degrades trigger precision. One shock, one trigger.

Pilot specification

1. Scope. Add an asset-protection window alongside the existing humanitarian window, so that a triggered payout funds commercial destocking, fodder provision, water trucking to fixed points, and index-based livestock insurance settlement, not consumption support alone. Commercial destocking has been costed at roughly USD 4.50 per person in comparable pastoral settings.
2. Geography. Three districts: two under government control and one contested. Testing the model only where access is straightforward answers the easy question and leaves the binding constraint untested. The contested district applies the design rules in section 5.3.
3. Trigger discipline. Publish the trigger threshold before the season, pre-commit the finance, and publish the payout when it fires. A trigger renegotiated at the moment of activation is not a trigger; it is a slower version of discretionary appeal funding.
4. Financing. Blend CERF pre-arranged finance with a climate fund contribution. This is the mechanism by which anticipatory action stops being humanitarian expenditure recorded as such and starts drawing on the adaptation envelope where it belongs, which matters for the allocation argument as much as for the operation.
5. Evaluation. Measure against a displacement counterfactual, not a caseload target. Compare district-level displacement recorded by the Displacement Tracking Matrix in pilot districts against matched non-pilot districts over the following two seasons. Caseload reached measures effort; displacement averted measures the thing the money is for.
6. Horizon. Three seasons minimum before any judgement. A single failed activation in a single season is not evidence, and treating it as evidence is how promising instruments get abandoned in this sector.

A.3 Where the two tracks meet

The two designs are complementary and should not be run by the same unit. Anticipatory action operates on a seasonal clock through existing humanitarian machinery and can begin within one budget cycle. Institutional capitalisation operates on a four-year clock and produces nothing visible for the first eighteen months. Bundling them means the slower track absorbs the faster one’s reporting requirements and the faster one absorbs the slower one’s delays.

They meet at one point, and it is the point worth protecting: the national designated authority, once capitalised, is the entity that should eventually hold the anticipatory action trigger and disburse against it. That is the exit condition for the whole arrangement, and it should be written into the design at the start rather than discovered as an aspiration at the end.

References

Crisis Group, International. 2024. Fighting Climate Change in Somalia’s Conflict Zones. Africa Report No. 316. Nairobi and Brussels, 10 December.

European Council. 2021. EU Emergency Trust Fund for Africa: allocations by window. Brussels.
Food and Agriculture Organization and Green Climate Fund. 2024. GCF approves funding for FAO-led projects in Somalia and Iraq. Press release, 23 October.

Integrated Food Security Phase Classification and OCHA. 2026. Somalia acute food insecurity analysis. Mogadishu and Nairobi.

International Organization for Migration. 2025. Regional Migrant Response Plan for the Horn of Africa to Yemen and Southern Africa, quarterly report, April to June 2025. Nairobi.

International Organization for Migration. 2026. Eastern Route quarterly report, fourth quarter 2025. Regional Office for East, Horn and Southern Africa, Nairobi.

International Organization for Migration. 2026. Drought now drives 75 per cent of displacement in Somalia’s most affected districts. Press release, 24 April. Geneva and Mogadishu.

Mixed Migration Centre. 2026. Quarterly Mixed Migration Update, Eastern and Southern Africa, fourth quarter 2025. Copenhagen.

OHirsi, A.I. 2024. Reducing vulnerability of Somali women to the global climate crisis: a call to action for gender-sensitive adaptation and mitigation strategies. American Journal of Climate Change 13: 779 to 792.

OHirsi, A.I. 2026. The Wheelbarrow Motion Theory: climate change and corruption as interdependent drivers of irregular south-to-north migration, practical solutions and policy responses. The Journal of Population and Sustainability 10 (2): 61 to 91.

Oxfam. 2020. The EU Trust Fund for Africa: Trapped between Aid Policy and Migration Politics. Oxford.

SPARC. 2023. Why we urgently need to fund climate adaptation in fragile and conflict-affected countries like Somalia.

Stimson Center. 2025. Closing the climate finance gap in conflict-affected states. Washington DC.

Transparency International. 2025. 2024 Corruption Perceptions Index: climate funds at risk of theft. Berlin.

U4 Anti-Corruption Resource Centre. 2026. Aid Diversion and Corruption in Somalia. Helpdesk Answer 2026:3. Bergen.

United Nations Central Emergency Response Fund. 2021. Somalia anticipatory action allocation: summary of learning. New York.

United Nations Environment Programme. 2025. Adaptation Gap Report 2025: Running on Empty. Nairobi. doi:10.59117/20.500.11822/48798.

United Nations Office for the Coordination of Humanitarian Affairs. 2021. Anticipatory Action in Somalia: Learning from 2020. New York.

United Nations Office for the Coordination of Humanitarian Affairs. 2021. Beneficiary assessment for the independent evaluation of the anticipatory action pilot in Somalia. New York.

United Nations Office for the Coordination of Humanitarian Affairs. 2026. Somalia 2025 to 2026 drought emergency, situation report no. 5, 6 April. Mogadishu.

United States Agency for International Development. 2018. Economics of Resilience to Drought: Somalia Analysis. Washington DC.

World Bank. 2024. Closing the Gap: Trends in Adaptation Finance in Fragile and Conflict-affected Settings. Washington DC.

Suggested citation: Foresight for Practical Solutions. 2026. Running Dry, Running North: Displacement in the Red Sea and Horn of Africa Corridor. Mogadishu: FPS.

Foresight for Practical Solutions is a policy research and action think tank based in Mogadishu, working on governance, migration, climate change, and Horn of Africa affairs.