Deforestation in the Peruvian Amazon: Modeling of Drivers

Deforestation in the Peruvian Amazon, a serious ecological threat, has resulted in the loss of 2,284,889 hectares of forest between 2001 and 2018, with Loreto, Ucayali, San Martín, Huánuco, Madre de Dios, and Junín being the most affected departments. Despite monitoring efforts by the National Forest Conservation Program of the Ministry of the Environment (MINAM), gaps remain in understanding the direct and indirect drivers of deforestation. This study focuses on developing predictive models using artificial intelligence techniques to identify the underlying drivers of deforestation in Peru, thus addressing a critical gap in existing research. The dataset used in this study is part of the "Promotores de la deforestación en el Perú" project. It encompasses a wide range of data variables including climate, land cover and use, population, population density, migration, types of employment, estimated Gross Domestic Product, expenditure efficiency, and distances to mining cadastre, national protected natural areas, native communities, timber forest concessions, rivers, and roads. The dataset exhibits missing data due to the variability of variable availability over time, with the general intended collection period spanning from 2001 to 2020. Data processing stages performed include the selection of variables of interest, data extraction, data cleaning and processing, and data uploading to the cloud. The study employs a methodological approach that includes understanding the problem, collecting and preparing data, and training and evaluating models. Data were collected from various sources, including indicators of climate, deforestation, human development, and other socioeconomic and environmental factors. The models focused on underlying drivers such as economic, technological, institutional, environmental, and demographic factors, as well as proximate causes like agriculture and infrastructure. A machine learning approach was used to analyze these data, with a particular emphasis on identifying significant patterns and correlations. The developed models revealed that factors such as forest formation, distance to protected natural areas, population, and agricultural infrastructure are key predictors of deforestation. The study found that it is possible to develop effective predictive models for deforestation in the Sierra and Selva regions of Peru, highlighting the importance of underlying drivers in understanding this phenomenon. Additionally, it was demonstrated that models based on underlying drivers can perform comparably to those that include proximate causes. This innovative approach offers a new perspective in the study of deforestation, providing valuable tools for formulating more effective conservation policies and strategies in the Peruvian Amazon region.