Woody-biomass is an important indicator for good vegetation cover, and a healthy environment that provides ecosystems services such as supplying, provisioning, regulating, esthetic, and cultural values. Moreover, woody-biomass stores a very good amount of carbon dioxide as about fifty percent of living organisms are built from carbon. The objective of this study is to estimate the above and belowground woody biomass and carbon stock of Kunzila watershed, Northwest Ethiopia as a piece of baseline information so as to gage the changes after the intervention. The watershed covers about 11,200 ha of land with undulating topographic landscapes. More than 230 sample plots of 30 m by 30 m were surveyed and every woody plant (shrub/bush and tree) was measured. Allometric equations were applied to calculate biomass and carbon stock potential at the plot level. The Random forest algorithm was applied to upscale plot values to watershed level. The results indicate that Kunzila watershed offers maximum biomass of 1,756(±249) Mt year−1/watershed and corresponding carbon dioxide equivalent (CO2e) of 3,022(±429) Mt C year−1 /watershed with the current vegetation status; while the average woody biomass and CO2e were estimated about 244(±249) Mt ha−1 year−1 and 425(±429) Mt C ha−1 year−1, respectively. These results indicate that there are high variabilities among different ecosystem types and also the watershed generally offers below its potential. However, there is a high potential to achieve up to triples of biomass and CO2e by implementing recommended intervention plans developed out of the current study. The current approach using machine learning algorithm to upscale from sample plot-level to watershed scale is a very good one and can be applied to other watersheds at scale.