Uncovering Patterns: Data Mining the Deforestation Frontier in Brazilian Cerrado
Keywords: Data Mining, Deforestation Patterns, LULC Trajectories, MATOPIBA, Deforestation Hotspot
Abstract. The Brazilian Cerrado, recognized as the world's most diverse tropical savanna, has undergone significant land clearance for agriculture, cattle ranching, and charcoal production, leading to considerable loss of its natural vegetation. This deforestation disrupts major aquifers, affecting ecosystems and human populations. The Cerrado Deforestation Monitoring Project (PRODES), active since 2017, systematically maps new deforestation using satellite imagery. This study used GeoDMA, a geospatial data mining tool for analyzing complex spatial databases such as those provided by the PRODES. The study area is the Alto Parnaíba region within the MATOPIBA agricultural frontier, characterized by high rates of crop expansion and burned areas. Using PRODES deforestation data from 2000 to 2023, the study categorizes deforestation patterns into geometric, multidirectional, and diffuse types, associated with large-scale agriculture, smallholder farming, and subsistence agriculture, respectively. The analysis reveals that approximately 50% of land use trajectories are small-scale subsistence agriculture, predominantly in northern Alto Parnaíba. Unchanged large-scale farms are concentrated in the southern part of the study area, linked to annual crops. The transition from small to large-scale farms is significant in the western region. Protected areas, especially strictly protected conservation units, show minimal changes in native vegetation. This study highlights the expansion of large-scale agriculture in MATOPIBA, emphasizing the need for targeted interventions to manage land use changes and prevent further deforestation. Future research should extend this approach to other Cerrado ecoregions to improve the detection and analysis of deforestation processes.