Abstract

The Brazilian Institute of Geography and Statistics (IBGE) is transitioning its Census of Agriculture operations toward an optimization-driven framework, modeling fieldwork as an Arc Routing Problem (ARP), where the objective is to service a set of edges in a graph that represents the pathway network of an enumeration area. While initial simulations suggested significant efficiency gains, the first field pilot test revealed some discrepancies between the supporting software workflow and human factors associated with the complex reality of rural surveying. This study analyzes the lessons learned from this pilot, focusing on three pillars: geospatial data integrity, software interface usability, and the necessity of operational flexibility.

Data analysis from the pilot identified a coverage gap stemming from the initial pathway network generation. Many enumeration areas contained sub-regions accessible only via "external'' pathways crossing adjacent areas. Not accounting for this fact led to incomplete local networks. Even though these external pathways should not be serviced nor accounted for in the context of the neighboring enumeration areas, they must be considered for traveling purposes. Furthermore, a proximity-based algorithm associated previously known agricultural establishments (data collection units) with the nearest pathway sections, which are not necessarily their actual access points (gates). To address this, the study proposes allowing field enumerators to manually reassign establishments between pathway sections or flag them as located on unmapped sections, leaving topology correction to post-processing using GPS traces.

On the usability front, the pilot exposed limitations in the current mobile application's navigation interface, where enumerators struggled with unclear directional cues. The findings suggest a substantial redesign of the map-centric interface. This new interface should provide a global view of all elements of interest – pathways, agricultural establishment locations, and status indicators – along with a more intuitive visual presentation of and interaction with the planned routes.

Finally, the most significant lesson was the need for flexible routing. The pilot demonstrated that an optimized route must serve as a guide rather than a rigid constraint. The system must allow enumerators to conduct interviews at any establishment, regardless of its pre-assigned route, ensuring that workflow is not hindered by obstructions or technical blocks. To adapt routes to the dynamic nature of fieldwork, we consider introducing a lightweight optimization module directly within the mobile application. This would enable real-time, local route adjustments without requiring constant communication with the central server. Furthermore, the next methodological evolution involves the integration of time-window constraints into the routing engine. This will allow the model to accommodate pre-scheduled interviews, ensuring that optimized paths respect the specific availability of producers. By treating the route as a dynamic and time-sensitive suggestion, IBGE can leverage the efficiency of mathematical optimization while maintaining the human adaptability required for exhaustive territorial coverage. These refinements are essential for the final methodology of the upcoming Brazilian Census of Agriculture.