Spatial assessment of migration impacts

Spatial analysis helps practitioners connect movement and settlement with the places where benefits and adverse effects may occur. Its purpose is to inform a decision: where to verify conditions, change an access route, expand reliable service capacity, protect a resource, or revise an accommodation plan.

A map of likely arrival destinations is one input. It cannot, by itself, establish impact severity. The assessment also needs evidence about people, environmental conditions, service reliability, and the ability of institutions to respond.

Separate three spatial questions

IFC's in-migration handbook distinguishes the identification of likely destinations from assessment of the impacts that may develop there. It calls for place-specific analysis informed by local and regional conditions (IFC, 2009, pp. 62–68). The map portfolio proposed here develops that distinction through three linked views.

Settlement pressure: where might additional residents, temporary occupants, or service users concentrate? Examine employment access, transport, rents, available accommodation, and social connections. A concentration estimate is not a classification of the people living there.

Exposure and vulnerability: what may be affected? Examine tenants' affordability, flood exposure, hazardous traffic, customary resource access, barriers to care, and sensitive habitats. These conditions can vary within a settlement.

Capacity and response: what can absorb or reduce additional demand? Examine functioning water systems, waste collection, staffed health facilities, safe housing, public budgets, and feasible upgrades. Record reliability and accessibility rather than nominal capacity alone.

Keep these maps separate before combining findings. A composite index can conceal the mechanism that a particular intervention needs to address.

Assemble a fit-for-purpose data stack

Begin with project records, national statistics, and local administrative data. Use global datasets to fill clearly described gaps and test broad patterns. The following resources support different questions; they are not interchangeable measures of migration.

WorldPop's foundational research describes spatial population modelling; Stevens and colleagues show how census totals can be redistributed using remotely sensed and ancillary data (Tatem, 2017; Stevens et al., 2015). This distinction is essential: a fine grid describes a modelled allocation of people, not necessarily a fine-resolution observation of them.

Prepare data before overlaying it

Create a source register that records the custodian, title, version, observation dates, release date, coordinate reference system, units, resolution, method, licence, restrictions, and validation status. Preserve the original data and script the transformations where possible.

For distance and area analysis, select an appropriate projected coordinate system. Geographic longitude and latitude are useful for interchange, but degrees are not metres. Check geometry validity, duplicate identifiers, missing attributes, and boundary alignment before joining records.

Population rasters require particular care. Determine whether a pixel contains a count or density. Sum counts over mutually exclusive areas; integrate densities using pixel area. Reprojection or resampling needs to preserve totals. Nearest-neighbour treatment is appropriate for categorical classes, while population redistribution requires a mass-preserving method. Record edge-cell treatment and test whether totals remain consistent with the source.

Do not treat absence from a map as absence on the ground. An unmapped settlement, road, or clinic is a verification priority. Likewise, a zero, a missing value, and a suppressed value need distinct codes.

Map pathways with networks and local knowledge

Road and transport networks can help explain travel between recruitment areas, markets, accommodation, and project facilities. A straight-line buffer is suitable for an initial search area; it is often a poor representation of access across rivers, hills, gated land, or seasonal roads.

Build alternative travel scenarios: walking and vehicle access, wet and dry seasons, bridge availability, operating hours, and affordability. Field-check decisive links. A five-minute route that depends on an unavailable vehicle may not be a meaningful option for the affected household.

WHO's AccessMod distinguishes physical accessibility from the availability of sufficient service capacity and supports travel-time, referral, and service-expansion analysis (WHO, AccessMod). In project assessments, this allows two different questions: can people reach the facility, and can it provide the needed care when they arrive?

Social networks also shape access. Interview evidence can identify places where newcomers stay with relatives, obtain introductions, or find rental housing. Represent such findings as qualitative annotations or aggregated flows. Do not publish household networks or sensitive locations simply because they can be geocoded.

Estimate pressure on services

For a settlement or service catchment i, a planning calculation can express demand as:

P⁰ is the population expected without the project, M is additional project-linked demand expressed as resident equivalents, q is the relevant demand per person, and Dᵒᵗʰᵉʳ represents non-household uses. s identifies the scenario. In practice, disaggregate groups with different demand patterns instead of relying on a single resident-equivalent conversion.

A potential deficit is:

C is reliable delivered capacity in the same units and period. For water, distinguish source yield, treatment, storage, distribution, quality, losses, and actual household access. For health services, staff time, service type, equipment, and opening hours may be the relevant constraints. A generic population-per-facility ratio does not reveal them.

These are proposed planning calculations. The worked example demonstrates them with fictional numbers, including the difference between an absolute deficit and a proportional shortfall.

Do not allocate the same spare capacity to several overlapping catchments. Use an explicit allocation rule, document existing users, and test whether the proposed intervention transfers shortages elsewhere. Regional totals can appear adequate while particular neighbourhoods remain underserved.

Detect environmental and settlement change

Use consistent seasonal windows, cloud screening, comparable processing, and stable analysis boundaries. Inspect candidate changes and validate a sample in the field. Building expansion is evidence of physical change; occupancy, tenure, and migration require additional investigation.

ESA explicitly warns that differences between WorldCover 2020 and 2021 combine actual change with changes in the classification algorithm (ESA WorldCover, data documentation). For change detection, prefer a consistent processing method across dates and quantify classification error where feasible. A subtraction of categorical land-cover maps is not a verified conversion inventory.

Overlay verified change with flood-prone land, drainage, customary use, water-source protection areas, or habitats only after checking the meaning and scale of each layer. An apparent overlap can indicate a question for assessment without establishing legality, impact magnitude, or attribution to the project.

Use indices cautiously

If a screening index is useful, publish its variables, units, scaling, weights, missing-data rules, and rationale. Show the original values beside the score. Test alternative weights and geographic units. Label it a screening index, rather than a predicted probability of migration or harm.

Severity overrides can be useful for credible, high-consequence concerns. A contaminated drinking-water source should remain visible even if employment access or settlement growth receives a low score. Uncertainty should not automatically become a reassuring zero.

Aggregation also changes interpretation. A district average may conceal a rental cluster; a household survey cannot automatically support a precise raster surface. Compare results at more than one reasonable spatial unit and state where evidence is insufficient for finer interpretation.

Make maps useful and safe

Every map needs a defined question, reference dates, units, sources, method, and limitations. Use colour together with labels or symbols, provide a readable legend, and include a table or text description of the main findings. Maintain consistent classes across comparable dates and scenarios. Avoid unexplained “high-risk” labels that attach danger to residents rather than conditions.

The OCHA data-responsibility guidance offers a useful model for examining risks in collecting and sharing operational data; its institutional requirements apply to OCHA, rather than automatically to investment projects (OCHA, 2025). For this application, define public, restricted analytical, and protected case-management outputs. Small-area aggregation may still identify people where a group is rare or a settlement is small.

A productive map review ends with a decision: verify a missing settlement, revise camp transport, protect an access route, repair a water system, or seek specialist advice. Preserve the evidence and rationale so the decision can be revisited through monitoring.