世界银行-探索基于地理空间的方法_在亚美尼亚制定人口普查前的国家抽样框架（英）-2025.1_37页_6mb

Background and Problem:

• Lack of accessible and accurate national sampling frames hinders representative surveys in Armenia due to population displacement and conflicts.
• Existing frames (census settlements, electoral precincts, grids) are outdated, large-scale, or unavailable.

Methodology:

• Data Sources: Gridded population (WorldPop 2020), OpenStreetMap visible features, administrative and settlement boundaries.
• Tool: Semi-automatic preEA tool within QGIS for generating enumeration areas (EAs) with constraints (urban/rural split, maximum population/area).
• Validation: Cross-compared with census data, identified non-residential areas, adjusted population estimates.

Results:

• Created 7,413 pre-EAs (3,813 urban; 3,600 rural) with homogenous populations (100–1,000), avoiding outliers.
• Population estimates matched census data at marz/urban/rural levels (correlation ≥0.99).
• Boundaries respected administrative and natural features, reducing geometric errors.

Application:

• Used for “Listening to Armenia” survey (L2Arm) baseline survey, enabling cost-effective stratified sampling.
• Targeted 4,000 households across strata (regions and urban/rural), saving time and resources compared to manual methods.

Conclusion:

• New national sampling frame addresses traditional constraints and supports national surveys.
• Methodology innovates by leveraging geospatial data for automatic EAs, applicable to other conflict-affected nations.
• Challenges include population estimate inaccuracies and boundary management, but overall efficient and practical.