Evidano is an AI-powered qualitative data analysis platform that centralizes transcripts, survey data, and metadata for reproducible mixed-methods workflows. On 2 July 2026 a PLOS ONE study showed that three wetlands in Bure and Womberma woredas (NW Ethiopia) lost 3, 355 ha, a 48.8% decline over 36 years, mostly converted to farmland. This post shows how to run a rigorous qualitative analysis of wetland conversion and translate findings into policy or program action using integrated tools and a compact workflow you can run on similar datasets.
Key Takeaways
The PLOS ONE study (published 2 July 2026) documents that 48.8% of three study wetlands were converted to farmland between 1985 and 2021, and that institutional drivers rather than only subsistence motives explain much of the change.
Reproducing the study requires combining Landsat-derived LULC maps with household surveys, expert surveys, and transcribed key informant interviews in a single, queryable workspace.
- The study reports a loss of 3, 355 hectares (48.8%) across three wetlands from 1985 to 2021, using Landsat supervised classification and accuracy assessment.
- Household surveys (n=396), expert surveys (n=75) and 12 KIIs supported thematic validation and ordered probit analysis of attitudes.
- Institutional events such as 1997 land redistribution and allocations to landless youth were identified as primary drivers, shifting policy responses toward employment diversification and agro-ecological appraisal.
Fast take + source
The PLOS ONE study (published 2 July 2026) documents that 48.8% of three study wetlands were converted to farmland between 1985 and 2021; drivers include land redistribution events (1997), allocation to landless youth and institutional priorities favoring agriculture.
- Why this matters: mixed remote-sensing and qualitative evidence expose policy and institutional drivers, not just poverty-driven subsistence.
- Quick payoff for researchers: reproduce the authors' land-use-to-attitudes linkage by combining transcripts, survey sheets and Landsat metadata into a single AI-assisted analysis pipeline with reproducible codebooks.
Source: PLOS ONE.
Findings snapshot
| Metric | Value | Source / note |
|---|---|---|
| Publication date | 2 July 2026 | PLOS ONE |
| Wetland loss (1985–2021) | 3, 355 ha (48.8%) | Landsat supervised classification, study |
| Survey sample | Households n=396; Experts n=75; KIIs n=12 | Face-to-face and self-administered (Mar–May 2022) |
| Landsat epochs | 1985, 1995, 2010, 2021 | TM/ETM+/OLI collections; <10% cloud cover |
| Classification accuracy (2021) | 97.27% overall; Kappa ~87% | Confusion matrix reported in study |
What the study did (plain English)
The authors combined multi-date Landsat classifications with household and expert Likert surveys and 12 oral key informant interviews to link mapped change with stakeholder narratives.
- Mixed inputs included raster LULC maps, structured survey spreadsheets, and audio-recorded KIIs (Amharic translated to English).
- Key methods were supervised maximum-likelihood classification in ArcGIS, accuracy assessment via confusion matrices, thematic analysis of KIIs, descriptive statistics in SPSS for perceptions, and ordered probit regression in Stata for attitudes' drivers.
So what for researchers & policy teams
For qualitative researchers
Qualitative researchers should link spatial change to stakeholder narratives because remote sensing shows where change occurred and interviews explain why it happened.
Guardrails include preserving transcripts and survey metadata such as timestamps, enumerator IDs and geotags to enable cross-segment analysis and reproducibility.
For UX / field teams collecting data
UX and field teams should design questionnaires and KIIs so answers can be triangulated with mapped change by asking respondents to reference parcel locations or seasons.
Capture language variants and create a custom dictionary of local terms (for example Amharic terms for wetland services) to support accurate transcription and translation.
For policy & environmental analysts
Policy and environmental analysts should prioritize institutional drivers because the dominant driver in this study was institutional, not just subsistence.
Analysts should combine LULC area change with household dependency metrics to estimate ecosystem service loss and target policies where conversion risk is highest.
Do more, faster with Evidano
Ingest heterogeneous inputs
Evidano centralizes Landsat-derived reports, GIS metadata, survey spreadsheets and audio files into one secure workspace so you can query across modalities.
Evidano keeps codebooks, geotags and enumerator notes linked so researchers can run cross-segment queries and preserve provenance.
Transcribe, translate, normalize
Evidano transcribes audio with custom dictionaries for local terms and supports PII redaction to maintain privacy during analysis.
Evidano supports translation with a user glossary to preserve local terms such as cheffee during processing.
Automated thematic + cross-segment analysis
Evidano runs thematic extraction on KIIs, frequency analysis on open-ended survey items, and cross-segment comparisons (age, landholding, office) in minutes for rapid iteration.
Evidano can replicate the ordered-probit angle with AI-assisted coding and exportable codebooks for transparency.
Visualize & share evidence
Evidano generates co-occurrence networks, hierarchical code to subcode trees, and downloadable stakeholder-ready visuals that link quotes to mapped changes.
Security: data are encrypted and never used to train third-party models, see Evidano.
7-step checklist: reproduce this study faster
Follow these seven steps to recreate a mixed-methods analysis of wetland conversion and scale the workflow across sites.
- 1) Collect and centralize inputs: Landsat epochs, survey CSVs, audio KIIs with geotags and consent metadata.
- 2) Transcribe and translate with a custom dictionary of local wetland terms, and redact PII.
- 3) Auto-generate themes and frequency counts from KIIs and open-ended survey items.
- 4) Link LULC polygons to household and respondent IDs to enable parcel-level segment analysis.
- 5) Run cross-segment contrasts (age, landholding, livelihood) and export an annotated codebook.
- 6) Validate with selected quotes and Kappa or accuracy checks, and iterate the codebook where disagreement exceeds 10%.
- 7) Produce a one-page stakeholder brief combining map snapshots, top themes, and three policy recommendations.
FAQ: qualitative analysis of wetland conversion
What did the PLOS ONE study find about wetland conversion?
The PLOS ONE study found that three wetlands lost 3, 355 hectares, a 48.8% decline between 1985 and 2021, and that institutional drivers were dominant.
The study linked land redistribution events, allocations to landless youth, and institutional priorities favoring agriculture to most conversions, using Landsat classifications validated by surveys and interviews.
What data and methods did the authors use to link maps and narratives?
The authors used multi-date Landsat supervised maximum-likelihood classification, household and expert Likert surveys, and 12 key informant interviews with thematic coding.
The methods included accuracy assessment with confusion matrices, descriptive statistics in SPSS, and ordered probit regression in Stata for attitudes analysis.
How can a research team reproduce the mixed-methods analysis faster?
A research team can reproduce the analysis faster by centralizing raster LULC maps, survey spreadsheets and transcribed KIIs in one workspace and using automated thematic extraction and linked metadata.
Follow the seven-step checklist to collect inputs, transcribe and translate with custom dictionaries, auto-generate themes, link polygons to respondent IDs, run contrasts, validate coding, and produce a stakeholder brief.
What policy implications arise from institutional drivers of conversion?
Institutional drivers shift policy responses toward employment diversification and mandatory agro-ecological appraisal before land allocations.
Policy analysts should quantify impacts by combining LULC area change with household dependency metrics to target interventions where conversion risk is highest.
Wrapping up: next moves
If you analyze land-use change with interviews and surveys, treat spatial and textual evidence as a single dataset to make findings actionable.
The PLOS ONE study (2 July 2026) is a clear example where institutional drivers dominated conversion and mixed qualitative methods explained why allocations happened despite acknowledged ecosystem value, see PLOS ONE.
- Start small by piloting the seven-step checklist on one wetland polygon and 20 transcripts.
- Ready to try an integrated pipeline? Try Evidano for free to convert mixed evidence into stakeholder-ready insights.