Case 1

Coastal Bangladesh

175.7Mpeople in the delta (UNFPA, 2025)

The Ganges–Brahmaputra–Meghna delta, coastal Bangladesh. A stylized vector map: parchment land over the teal Bay of Bengal, with the three major rivers — the Ganges (Padma), Brahmaputra (Jamuna), and Meghna — converging through the delta to the coast. Physical geography only; no district boundaries or study-scope points. Not to scale.
Stylized schematic of the Ganges–Brahmaputra–Meghna delta — physical geography only, not to scale and not an official study scope.

Bangladesh's coastal zone faces one of the most studied combinations of climate stress and population pressure in the world. For 3MIP's first benchmark case, participating teams work from a single curated dataset for coastal Bangladesh — each team choosing its own research question, model architecture, key variables, and outputs.

Why Bangladesh

Why Bangladesh is the first case

Three factors converge on Bangladesh. The empirical literature on internal migration there is among the densest in the climate-migration field, providing a real benchmark against which model outputs can be sanity-checked. The country's population of 175.7 million (UNFPA SWOP 2025) and its low-lying delta geography make the policy stakes consequential. And the data infrastructure — population grids, modeled migration flows, salinity layers, flood histories — is already openly licensed and usable.

The coastal districts face coastal erosion, sea-level rise, seasonal monsoons, tropical cyclones, and salinity encroachment — and their populations are highly mobile. The case was chosen for data availability and the depth of the existing literature, not because the dynamics are simple.

Bell et al. (2021) projected that migration toward Bangladesh's coast, not away from it, will continue through 2100 across the sea-level rise scenarios they studied. That finding sits in productive tension with the popular narrative of mass climate exodus and provides the empirical motivation for revisiting the problem with multiple model architectures.

Migration toward Bangladesh's coast — not away from it — is projected to continue through 2100. — Bell et al. (2021), Environmental Research Letters
Bell, A. R., Wrathall, D. J., Mueller, V., Chen, J., Oppenheimer, M., et al. (2021). Migration towards Bangladesh coastlines projected to increase with sea-level rise through 2100. Environmental Research Letters , 16(2), 024045 https://doi.org/10.1088/1748-9326/abdc5b

The benchmark

The benchmark design

Teams choose how to define and model migration within their own architecture. The intercomparison does not require harmonized output variables; it requires only harmonized inputs. Where outputs diverge — and they will — the synthesis documents the divergence and the methodological choices that drive it.

Model architectures

The six model architectures

The signature figure on the home page renders the same curated dataset through six families of migration model. Each encodes how people move differently — which is exactly why their outputs diverge.

Agent-based

Simulates individual households deciding whether, when, and where to move.

Gravity

Predicts flows between places from their populations and the distance between them.

Radiation

Derives flows from the population lying between an origin and a destination — parameter-free.

Integrated assessment

Embeds migration within a coupled model of climate, economy, and policy.

Machine learning

Learns migration patterns statistically from observed data.

Cellular automaton

Evolves a grid of places by local transition rules at each time step.

Geographic scope

The case centers on the Ganges-Brahmaputra-Meghna delta and Bangladesh's low-lying coastal zone along the Bay of Bengal (mapped in the hero above). The curated dataset covers the country at the resolutions documented on the Data page; 3MIP does not prescribe a fixed list of study districts, so the spatial scope of any analysis is the team's own choice.

Curated input dataset

3MIP curates 16 datasets for the case — population and socioeconomic, environmental, and mobility data — documented with citations and access links on the Data page. Teams use this curation as their common input. The data list may be expanded, but any new dataset must be made known and available to all teams; contact project coordination to request the inclusion of an additional dataset.

  • Population & socioeconomic 7
  • Environmental 5
  • Mobility 4

See the full data documentation .

Expected outputs and timeline

  1. 3MIP launches; registration opens.
  2. First participant webinar.
  3. Regular participant webinars; teams run their models and share preliminary findings.
  4. First synthesis at iEMSs 2026, University College Dublin (Session C7 + Workshop WSC7).
  5. Regular webinars continue.
  6. Submissions to Climatic Change topical collection, received through 2027.
  7. Synthesis paper drafted from the collected outputs.

Background literature

Background literature

The papers below frame the Bangladesh case. Bell et al. (2021) provides the direct empirical motivation; the others document the migration dynamics the participating models must contend with.

Bell, A. R., Wrathall, D. J., Mueller, V., Chen, J., Oppenheimer, M., et al. (2021). Migration towards Bangladesh coastlines projected to increase with sea-level rise through 2100. Environmental Research Letters , 16(2), 024045 https://doi.org/10.1088/1748-9326/abdc5b
Carrico, A. R., & Donato, K. (2019). Extreme weather and migration: evidence from Bangladesh. Population and Environment , 41(1), 1–31 https://doi.org/10.1007/s11111-019-00322-9
Chen, J., & Mueller, V. (2018). Coastal climate change, soil salinity and human migration in Bangladesh. Nature Climate Change , 8, 981–985 https://doi.org/10.1038/s41558-018-0313-8
Hassani-Mahmooei, B., & Parris, B. W. (2012). Climate change and internal migration patterns in Bangladesh: an agent-based model. Environment and Development Economics , 17(6), 763–780 https://doi.org/10.1017/S1355770X12000290
Mallick, B., Best, K., Carrico, A. R., Ghosh, T., Priodarshini, R., Sultana, Z., & Samanta, G. (2023). How do migration decisions and drivers differ against extreme environmental events?. Environmental Hazards , 22(5), 475–497 https://doi.org/10.1080/17477891.2023.2195152

Participation guidelines

3MIP runs on a small set of shared commitments rather than a formal written protocol:

  • Register as a modeler or a domain expert.
  • The data inputs are fixed. Every team works from the same curated dataset for coastal Bangladesh.
  • The research question is open. Model design, key variables, dynamics, and outputs are each team's own choice.
  • The dataset list can grow — but only when a new dataset is made known and available to all teams.
  • Regular meetings are held with times rotated across time zones, so the inconvenience of out-of-hours meetings is shared.
  • Code and findings are shared with the network.

Register to take part .