Analysis of food system drivers of deforestation highlights foreign direct investments and urbanization as threats to tropical forests.

Approximately 90% of global forest cover changes between 2000 and 2018 were attributable to agricultural expansion, making food production the leading direct driver of deforestation. While previous studies have focused on the interaction between human and environmental systems, limited research has explored deforestation from a food system perspective. This study analyzes the drivers of deforestation in 40 tropical and subtropical countries (2004-2021) through the lenses of consumption/demand, production/supply and trade/distribution using Extreme Gradient Boosting (XGBoost) models. Our models explained a substantial portion of deforestation variability globally (R2 = 0.74) and in Asia (R2 = 0.81) and Latin America (R2 = 0.73). The results indicate that trade- and demand-side dynamics, specifically foreign direct investments and urban population growth, play key roles in influencing deforestation trends at these scales, suggesting that food system-based interventions could be effective in mitigating deforestation. Conversely, the model for Africa showed weaker explanatory power (R2 = 0.30), suggesting that factors beyond the food system may play a larger role in this region. Our findings highlight the importance of targeting trade- and demand-side dynamics to reduce deforestation and how interventions within the food system could synergistically contribute to achieving sustainable development goals, such as climate action, life on land and zero hunger.

Prioridades de atención en zonas rurales en escenarios posconflicto: un análisis de minería de texto / Attention priorities in rural areas in post-conflict scenarios: a text mining analysis

RESUMEN Las herramientas automatizadas de análisis de texto resumen grandes volúmenes de información y permiten generar, de forma eficiente, conocimiento a partir de datos desestructurados, como las opiniones. El objetivo de este trabajo fue identificar prioridades en comunidades afectadas por el conflicto armado, utilizando ejercicios participativos de 13 municipios de Antioquia, Colombia. Se analizaron 15.534 opiniones, de 9.765 personas; tras una limpieza de texto, se describió el uso, la asociación, la diferenciación y la importancia de los términos, según los enfoques temáticos y tipos de opinión expresados, utilizando minería de texto en R. Se encontró que las prioridades giraban en torno a la disponibilidad de infraestructuras, dotación e insumos, ya que eran las problemáticas más mencionadas por las comunidades y correspondía con la realidad territorial; por otra parte, las oportunidades estaban representadas, principalmente, por los recursos naturales y humanos. El análisis de minería de texto del ejercicio participativo permitió identificar las prioridades de las comunidades, a nivel socioeconómico, de forma satisfactoria; sin embargo, la preparación de la información requiere mucho trabajo y los resultados se deben revisar cuidadosamente, para asegurar su coherencia. Que la información pueda ser analizada por agentes externos a la colección de datos, representa otra ventaja de esta herramienta.

ABSTRACT Automated text analysis tools summarize large volumes of information and allow efficient generation of insights from unstructured data such as opinions. The objective of this work was to identify priorities in communities affected by the armed conflict using participatory exercises in 13 municipalities of Antioquia, Colombia. 15,534 opinions from 9,765 people were analyzed. After a text cleaning, the use, association, differentiation and importance of the terms were described according to the thematic approaches and types of opinion expressed using text mining in R. It was found that the priorities revolved around the availability of infrastructures, endowment and inputs, since they were the most mentioned problems by the communities, and that this corresponded to the territorial reality. On the other hand, the opportunities were mainly represented by natural and human resources. The text mining analysis of the participatory exercise allowed to identify the priorities of the communities at the socio-economic level in a satisfactory way. However, the preparation of the information is labor intensive and the results must be carefully reviewed to ensure consistency. Another advantage of this tool is that the information can be analyzed by external agents to the data collection.

Semiparametric Modeling of Daily Ammonia Levels in Naturally Ventilated Caged-Egg Facilities.

Ammonia concentration (AMC) in poultry facilities varies depending on different environmental conditions and management; however, this is a relatively unexplored subject in Colombia (South America). The objective of this study was to model daily AMC variations in a naturally ventilated caged-egg facility using generalized additive models. Four sensor nodes were used to record AMC, temperature, relative humidity and wind speed on a daily basis, with 10 minute intervals for 12 weeks. The following variables were included in the model: Heat index, Wind, Hour, Location, Height of the sensor to the ground level, and Period of manure accumulation. All effects included in the model were highly significant (p<0.001). The AMC was higher during the night and early morning when the wind was not blowing (0.0 m/s) and the heat index was extreme. The average and maximum AMC were 5.94±3.83 and 31.70 ppm, respectively. Temperatures above 25°C and humidity greater than 80% increased AMC levels. In naturally ventilated caged-egg facilities the daily variations observed in AMC primarily depend on cyclic variations of the environmental conditions and are also affected by litter handling (i.e., removal of the bedding material).