Evidano is an AI-powered qualitative data analysis platform that ingests transcripts, surveys, and classification tables into a single searchable corpus with linked metadata. Researchers and policy teams need fast, reproducible ways to turn satellite maps, surveys, and interviews into clear drivers and policy options. This post walks through a concise workflow for qualitative analysis of wetland conversion using the new PLOS study (published 2 July 2026) on Bure and Womberma Woredas, NW Ethiopia, and shows how to operationalize the same mixed-methods inputs in Evidano (Evidano) to produce thematic, frequency and cross-segment analyses in hours not weeks.
Key Takeaways
The PLOS ONE study (2 July 2026) reports 3, 355 ha (≈48.8%) of three riverine wetlands in Bure and Womberma were converted to farmland between 1985 and 2021, driven mainly by land allocation to youth and landless households and institutional priorities rather than only subsistence need.
- The core datasets were Landsat imagery (1985–2021), a household survey (n=396), an experts/head survey (n=75), and 12 key informant interviews, analyzed with supervised classification, descriptive statistics, ordered probit regressions, and thematic coding.
- Classification accuracy for 2021 was reported as overall accuracy 97.27% and Kappa 87.19%, supporting the LULC change findings.
- Triangated mixed-methods workflows that link annotated change maps to interview narratives and segment-level analyses (age, landholding, office affiliation) produce policy-ready decision briefs quickly.
Fast take + source
Fast take: A mixed-methods study published 2 July 2026 finds 3, 355 ha (≈48.8%) of three riverine wetlands in Bure and Womberma were converted to farmland between 1985 and 2021, driven largely by land allocation to young and landless households and by institutional priorities rather than simple subsistence need.
- Full study: PLOS ONE (Getnet Eneyew & Worie Assefa, PLOS ONE, July 2, 2026).
- Key datasets: Landsat imagery (1985–2021), household survey n=396, experts n=75, KIIs n=12; ordered probit regressions and thematic analysis.
Findings snapshot
| Metric | Value | Source / note |
|---|---|---|
| Publication date | 2 July 2026 | PLOS ONE article |
| Wetland area converted | 3, 355 ha (48.8% increase by 2021) | LULC change (1985–2021) |
| Key datasets | Landsat images; household survey (n=396); experts (n=75); KIIs (n=12) | Mixed-methods design |
| Landsat years used | 1985, 1995, 2000, 2010, 2021 | Classifications & accuracy checks |
| Classification accuracy (2021) | Overall accuracy 97.27%; Kappa 87.19% | Confusion matrix reported |
| Primary drivers identified | Land allocation to youth/landless; institutional job creation; stopped land redistribution | Ordered probit & KIIs |
How the study was done (plain English)
The study combined remote sensing, a structured household survey, an experts/head survey, and key informant interviews to measure LULC change and local perceptions.
- Remote sensing: Landsat TM/ETM+/OLI median composites with <10% cloud cover, supervised maximum-likelihood classification, validation via confusion matrices.
- Qualitative: audio-recorded KIIs, transcription, and thematic coding to capture local narratives on land allocation and institutional drivers.
- Quantitative: Likert items on ecosystem perceptions; ordered probit to model ordinal attitudes (1–5 scale).
So what for researchers and policy teams: three implications
1) Design: triangulate early and explicitly
Triangulate LULC change with perception data early in study design to distinguish demand-driven conversion from policy/institution-driven change.
The paper shows paired LULC change and perception data pinpoints whether change is demand-driven (subsistence) or policy/institution-driven; include targeted KIIs to validate remote-sensing epochs associated with policy events (for example, 1997 redistribution).
2) Segments matter: age, landholding, office affiliation
Compare demographic and institutional segments because segment-level propensity can differ from prevalence.
Ordered probit outputs show younger, landless households and experts from agriculture and land offices are more likely to support conversion; when you compare segments, report both prevalence (how many) and propensity (how much more likely).
3) Actionable outputs: go beyond p-values
Deliver maps linked to narratives and decision options rather than only statistical results to inform policy choices.
Policymakers need maps linked to narratives: annotated change maps, quote-backed decision logs, and scenario options (employment programs versus land allocation), delivered as a short decision pack.
Do more, faster with Evidano
Problem: scattered mixed-methods files → Solution
Evidano centralizes mixed-methods files so teams can search and link evidence across formats.
Ingest Landsat classification tables, survey spreadsheets, interview transcripts and KIIs into Evidano to produce a single searchable corpus with linked metadata (date, location, respondent segment).
Problem: slow thematic coding → Solution
Evidano accelerates thematic coding with AI-assistance while preserving human review.
Use Evidano to auto-suggest themes from transcripts and refine a codebook; export hierarchical code to subcode visualizations and quote extracts for each theme.
Problem: comparing segments (e.g., age, office) → Solution
Evidano supports cross-segment analyses that show where attitudes align or conflict.
Run cross-segment analyses to show theme prevalence by age, landholding, or office affiliation and generate tables and co-occurrence networks that surface where attitudes align or conflict.
Problem: multilingual field notes & PII → Solution
Evidano handles transcription, translation, and optional PII redaction to centralize sensitive KIIs safely.
Evidano does transcription and translation with custom dictionaries, plus optional PII redaction, so you can safely centralize sensitive KIIs and share sanitized extracts with stakeholders.
Security & compliance
Security and compliance: Data is encrypted and not used to train third-party models.
Data is encrypted and not used to train third-party models, important when working with government records or identifiable interviews.
7-step workflow to reproduce this study in Evidano
This 7-step workflow shows inputs and outputs you can hand to policy teams in two weeks for a pilot.
- 1) Ingest datasets: Landsat LULC tables, survey CSVs, interview audio and transcripts into an Evidano project.
- 2) Auto-transcribe and translate audio (Amharic to English) using a custom dictionary for local terms.
- 3) Predefine segments (age bands, landholding, office) and attach to each record.
- 4) Run automated thematic extraction on transcripts; review and lock a codebook with a human-in-the-loop.
- 5) Produce cross-segment frequency tables and co-occurrence maps that link themes to LULC change epochs.
- 6) Export stakeholder-ready artifacts: a one-page decision brief, annotated maps, quote bank, and reproducible tables for regression checks.
- 7) Iterate: launch targeted AI avatar interviews for follow-ups on policy options (for example, eco-friendly livelihoods).
FAQ: qualitative analysis of wetland conversion
Q: How do I compare attitudes across segments reliably?
You can compare attitudes across segments reliably by using standardized Likert items and attaching demographic metadata to each response.
Use standardized Likert items and attach demographic metadata; in Evidano, run cross-segment theme prevalence with confidence intervals and export contingency tables for ordered probit or other regressions.
Q: Can I link remote sensing epochs to interview themes?
Yes, you can link remote sensing epochs to interview themes by aligning LULC change dates with timestamped transcripts and coded references.
Align LULC change dates with timestamped transcripts and code co-occurrence (for example, references to '1997 redistribution') to support causal narratives.
Q: Is this ethical for sensitive respondents?
Yes, ethical practice requires informed consent and PII redaction when necessary.
Always collect informed consent and redact PII; Evidano supports PII redaction and encrypted storage, and analyses should be labeled non-diagnostic and research-focused.
Wrapping up & next steps
Wrapping up: Use a mixed-methods workflow that preserves quotes, quantifies theme prevalence, and links findings to maps to convert multi-source evidence into policy-ready insight for wetlands or any landscape study.
- The PLOS study (PLOS ONE) is a clear example: 48.8% conversion over 36 years driven by institutional allocation and youth land demands.
- Ready to try it? Try Evidano for free to ingest your transcripts and surveys, run thematic and cross-segment analyses, and generate stakeholder-ready briefs in days.