The challenge
Together with the international humanitarian NGO World Vision, we set out to investigate whether land degradation and conflict are correlated in countries in the IPC 4 and IPC 5 groups.
The difficulty lay in drawing meaningful insights from diverse data sources and establishing a clear relationship between land degradation and conflict variables. The project also called for a predictive model that could forecast conflict occurrence in the following six months from the state of land degradation or regeneration.
Our solution
Vegetation was assessed with the Normalised Difference Vegetation Index (NDVI), which ranges from -1 to +1. High values indicate healthy vegetation; low values indicate little vegetation or non-photosynthetic material. We extracted yearly median NDVI values for the IPC 4 and IPC 5 countries from 1981 onwards and analysed them alongside historical conflict records described by four variables: total deaths, total conflicts, total days of conflict and severity.
The analysis found a negative correlation between NDVI and all four conflict variables: as NDVI fell, indicating land degradation, the conflict variables rose over the following six months, and higher conflict values were followed by lower NDVI. Using feature engineering, we then built a classifier that predicts conflict occurrence in the next six months from land degradation and regeneration as measured by NDVI.
- Yearly median NDVI extracted for IPC 4 and IPC 5 countries since 1981
- Historical conflict records analysed across total deaths, total conflicts, total days of conflict and severity
- Correlation analysis between vegetation condition and conflict variables
- Feature-engineered classifier predicting conflict occurrence six months ahead, with 74% accuracy
The outcome
The research showed a clear link between land degradation and conflict prevalence in the countries studied, and the predictive model gives humanitarian and environmental organisations an additional input for early, preventive action.
World Vision has published the results on its website so that other organisations can use the findings.
