Database: The conflict–land use–nutrition nexus reveals pathways for integrated climate mitigation and sustainable development strategies

The 2030 Agenda for Sustainable Development emphasizes that progress towards the Sustainable Development Goals (SDGs) requires coordinated, cross-sectoral action, recognizing that contemporary societal challenges are interconnected rather than isolated (Liu et al., 2015; Nilsson et al., 2016). However, armed conflict, land-use change and associated environmental degradation, and food and nutrition insecurity are often addressed independently in both research and policy, even as evidence increasingly shows that they co-evolve through dynamic and mutually reinforcing pathways (Morales-Muñoz et al., 2020; Segovia, 2017). A dataset was constructed to capture the spatial variation in conflict dynamics, land use-driven environmental degradation, and food and nutritional insecurity (FNI) in Colombia between 2000 and 2024. This dataset allowed the analysis done in the report here: https://hdl.handle.net/10568/180621
Methodology:Each municipality formed the unit of analysis in this national-scale study. The official municipal boundaries of the National Administrative Department of Statistics (DANE) were used, ensuring spatial consistency with official statistics.
The conflict dimension was characterized by indicators that captured both direct and indirect forms of violence. Our intent was to analyze institutional fragility as an indicator of the instability of the territory. Variables of the environmental degradation dimension reflect changes in the land-use. The food and nutrition insecurity dimension integrates socio-economic, health, accessibility and agricultural variables to reflect multiple pathways affecting food availability, access and utilization in the country.
We included as armed attacks those attributed to guerrillas, paramilitary groups, dissident factions, as well as unidentified actors or incidents where the perpetrators are disputed. Collectively, these categories encompass episodes of armed violence, regardless of the possible involvement of state actors, that are typically linked to dynamics of territorial control.
To estimate the rate of forest loss, we defined forest as all-natural vegetation cover, including natural forest, mangroves, and floodplains, in accordance with the MapBiomas classification. The category “woody vegetation on sand” was treated as non-forested natural vegetation, in accordance with IDEAM. We defined land-use change as the transition from forest to non-forest.
To approximate the livestock load in each municipality, we used the data available in the MapBiomas reports on total agricultural areas, which includes both cropland and pasture. We then divided the number of cattle by the number of hectares dedicated to agricultural activities in the corresponding year.
Market access was measured as the average Euclidean distance from the municipal centroid to the three nearest formal food markets, using georeferenced data from OpenStreetMap (OSM) and the amenity=marketplace tag to identify only food markets. Meanwhile, the variable for access to urban centers was defined as the average travel time by land from each municipality to the nearest city, based on the “Accessibility to Cities 2015” raster and calculations of zonal statistics.
We normalized variables such as number of murders and number individual displacement and number of people affiliated to a public health insurance to be expressed as rates aiming to account for differences in population size across municipalities and years. Specifically, we divided the absolute number of events by the total municipal population in the corresponding year.
Additionally, we calculated a forest loss rate by dividing the number of hectares of tree cover loss (per year) by the total forest area (in hectares) in the baseline year 2000. We also calculated a ratio for land-use change by the number of hectares transitioned in the periods 2000-2005, 2006-2011, 2012-2017, 2018-2023, and 2022-2023, and dividing them by the hectares of forest existing in 2000, 2006, 2012 and 2018, respectively, to obtain a percentage change in forest cover. Cattle density was calculated dividing the number of cattle by the area of the municipality.
Due to heterogeneous data availability across municipalities and years, some variables showed incomplete temporal coverage, particularly in the early years of the study period. To address this limitation and ensure comparability across municipalities, we summarized each variable using the median of available annual observations over the period 2000–2024 for each municipality. We preferred the median over the mean to reduce sensitivity to outliers and irregular patterns. For the variable related to the number of armed attacks, and given the high proportion of missing values, the information was incorporated dichotomously, indicating the presence (1) or absence (0) of proven or documented records at the municipal level. This decision responds to limitations in the comparability of counts between municipalities, associated with heterogeneities in record coverage and possible underreporting.
Finally data was normalized.