Tropical savannas of Africa and Latin America are dominated by very acid and infertile soils, which limit the productivity of most crops and forages. Liming and fertilizers could overcome these constraints but frequently are too expensive and have limited adoption. In the Amazon rainforest, under similar soil conditions, indigenous knowledge developed millennia ago lead to the creation of the so called Amazonian Dark Earths: highly productive and sustainable soils built through mixing infertile native soils with charcoal (biochar), fish bones, nut shells etc. There is a growing interest to replicate these Dark Earths in other tropical environments where low fertility soils are predominant. Here we present some results from a long-term field experiment established in 2002 on an acid soil savanna in the Eastern Colombian Plains. The experiment assesses the effects of additions of different doses of biochar to soils on plant yields and on net emission of greenhouse gases to the atmosphere. Annual rainfall is 2200mm and average annual temperature is 26°C. Plots were established by burning native vegetation. Lime was applied (2000 and 500 kg ha-1 for the crop and pasture plots respectively). One month later, three levels of biochar were applied to the plots: Zero, (control), 8 ton biochar ha-1 and 20 ton biochar ha-1. The biochar was produced locally from wood of mango trees using traditional methods. The biochar was ground to <2mm, broadcasted on the soil surface and then incorporated by disking to 5 cm depth. Four months later, at the beginning of the rainy season, maize (Cultivar H-108) was sown as well as Brachiaria dictyoneura (var Llanero). Native savanna was allowed to re-grow on some of the plots. Experimental plots are 20m2 each and each treatment has 3 replications within a randomized complete block design. Periodically, the grass and the native vegetation (mostly native grasses) were cut to a height of 10 cm simulating grazing and the biomass produced in each interval was registered. Maize was harvested at full grain maturity. Gas exchange between the soil and the atmosphere was monitored monthly over a three year period using the closed chamber method. Gas samples were analyzed for CO2, CH4 and N2O using a gas chromatograph with ECD and FID detectors. Results: Additions of even low doses of charcoal (Biochar) to soils results in a net cumulative increase in total biomass of maize, imprioved pasture and native savanna vegetation. Yields of maize were similar in all treatments during the first year but significantly increased by biochar in the two subsequent years. In the third year, yields increased from 5.7 ton/ha (control) to 6.6 and 7.3 ton/ha for the low and high dose of biochar. Forage production from B. dictyoneura increased by 26% and 55% in the second year relative to the control in the low and high biochar plots respectively. Total biomass production on the native vegetation trials was slightly increased from 5.78 ton/ha (control) to 6.14 ton/ha in the high biochar dose, but was similar to the control at the low biochar dose. The reason for the increase in plant productivity can be attributed to increases in soil pH, CEC, K availability, and possibly higher water retention in the soil. In addition to increased productivity, the use of biochar resulted in a net reduction of net annual emissions of methane and nitrous oxide from soils as well as net increases in soil carbon. During the initial year, annual methane sinks by soils were increased on average 200 mgCH4.m-2 in all high biochar plots relative to the controls, while N2O emissions were reduced on average 15 mg N2O.m-2. Most of the applied C in the biochar has long residence times in the soil and consequently constitutes a feasible option to store large quantities of C in the soils on the long term. Overall, the use of biochar results in a net decrease in the integrated Global Warming Potential from the studied soils. The rediscovered use of biochar increases crop and plant yield on very unfertile soils and constitutes a new tool to mitigate climate change.