AI Qualitative Analysis of Wetland Conversion

Evidano is an AI-powered qualitative data analysis platform that ingests transcripts, spreadsheets, and remote-sensing tables to accelerate mixed-methods synthesis.

This post demonstrates a reproducible 7-step workflow to synthesize Landsat change maps with n=396 household surveys, 75 experts, and 12 KIIs to surface institutional drivers of wetland conversion.

• The study reports a 48.8% increase in cultivated area across three wetlands from 1985–2021, equal to 3, 355 hectares converted.
• Remote sensing LULC maps achieved high validation, with overall accuracy 97.27% and Kappa 87.19% for 2021.
• The mixed-methods corpus included 396 households, 75 sector experts, and 12 key-informant interviews, enabling cross-segment thematic and statistical analysis.
• Evidano shortens the time from multisource inputs to policy-ready outputs by automating transcription, hierarchical coding, and cross-segment comparisons.

This section explains what the study did in plain English: the mixed-methods study combined supervised Landsat classification (1985, 1995, 2010, 2021) with structured household and expert surveys and 12 key-informant interviews.

The study used thematic analysis on KIIs, descriptive statistics on surveys (SPSS), and an ordered probit regression (STATA) to link attitudes to household and institutional attributes. Training samples (50–60 per class) and Google Earth Engine compositing produced high-quality LULC maps; household fieldwork ran March–May 2022.

• Remote sensing: Landsat 5/7/8 images composited via Google Earth Engine, supervised maximum likelihood classification, accuracy assessed by confusion matrix.
• Surveys: n=396 households (17 villages) plus 75 sector experts/department heads, using Likert scales for perceptions and attitudes.
• Qualitative: 12 KIIs transcribed and thematically coded to contextualize policy and local decision-making.

This section lists a 7-step workflow to reproduce the study in a two-week pilot.

Step 1: Collect and standardize inputs, export Landsat LULC tables, household survey CSVs (n=396), expert CSV (n=75), and KII audio files.

Step 2: Upload to Evidano, enable transcription with a custom Amharic dictionary and PII redaction.

Step 3: Auto-code and review, run AI-assisted thematic coding, review suggested codebook, and lock hierarchical codes.

Step 4: Run cross-segment analysis, compare attitudes by age, landholding, and office affiliation (agriculture versus environment).

Step 5: Merge remote-sensing metrics, import LULC change tables as a spreadsheet and link rows to villages/kebeles for integrated analysis.

Step 6: Visualize and annotate, generate co-occurrence network and export top quotes per theme for policymaker slides.

Step 7: Deliverables, produce a reproducible report (PDF plus CSV codebook) and a short executive brief linking the 48.8% conversion to institutional drivers.

This section summarizes next steps: institutional decisions are central to conversion and AI-enabled qualitative workflows speed synthesis from data to decision.

The PLOS ONE study (PLOS ONE) makes a clear case that institutional decisions, not ignorance of wetland value, are central to conversion.

• Try a 2-week pilot to reproduce key tables and thematic maps from the study's data.
• Start by uploading one survey CSV and two KII transcripts to test transcription, code suggestions, and cross-segment outputs.

Ready to turn multisource wetland evidence into policy-ready insight? Try Evidano for free.