Black carbon (BC) inventories for cities are scarce, especially in developing countries, despite their importance to tackle climate change and local air pollution. Here, we draw on results from a case study in a Brazilian city to discuss the challenges of compiling a BC inventory for different activity sectors. We included traditionally inventoried sectors, such as industries and on-road transportation, other less reported sectors (food establishments and aviation), and open burning of household solid waste (HSW), typically found in developing countries. We present a machine-learning technique (Random Forest) as a novel approach to obtain HSW burning activity using a set of spatial predictors. The BC inventory was based on PM2.5 emissions weighted by the fraction of PM2.5 emitted as BC and developed for the year 2018. We also reported the disaggregated spatial PM2.5 emissions for the same combustion sources, and documented the databases used for activity data and emission factors (EF). The total estimated BC and PM2.5 emissions amounted to 57.88 and 234.75 tons, respectively, with on-road vehicle exhaust emissions and industrial combustion as the main BC sources (63 and 22%, respectively). For PM2.5 emissions, on-road transportation (exhaust and non-exhaust) contributed 48%, followed by industrial combustion (21%) and food establishments (20%). Population density, number of vacant lots, and property tax values were identified as the most important features to predict the HSW fire activity. A comparison with other inventories revealed that the BC emission profile of Londrina is similar to the profile reported for Greater Mexico City, another Latin American city. Thus, the methodology used in this study could be extended to other cities with similar local BC sources. Finally, we highlight that the lack of local activity data, representative EF, and even methodology may undermine the development of reliable BC inventories, and intensive research should be conducted to characterize the emission sources.