Methodology to identify spatial patterns in coffee (Coffea arabica L.) production

Coffee farming, a lifeline for numerous families in the mountainous regions of Latin America, faces challenges due to climate change and production variability, which complicate the use of forecast models at the territorial level. In response to these challenges, territorial inference has gained relevance, especially with the advancement of Geographic Information Systems (GIS), which provide useful tools for territorial analysis. Although spatial models are increasingly applied in GIS, coffee farming, like many agricultural subsectors, is hindered by a lack of information and spatial methodologies. This work proposes a methodology to identify spatial patterns of homogeneous production areas. Data from 140 farms, representing 3,900 members of the coffee grower cooperative of Andes, dispersed over 200,000 ha, were analyzed between 2019 and 2021. The variables used to measure productivity included the number of fruits per tree, the average fruit weight, planting density, and the conversion rate of cherry coffee to dry parchment coffee. A simple linear regression model was employed, and spatial dependency analyses were performed using the global and local Moran’s Index to identify clusters of territorial subdivisions. The data were processed in R language, and the GeoDa™ program was used to obtain the spatial weight matrix. Territorial units with similar characteristics for high quality mountain coffee production were identified through spatial dependency indicators. The methodology can contribute to estimating coffee production in large territories, improving the reliability of information and allowing for more informed decision-making to optimize coffee farming in mountainous areas.