Wildfires are a natural phenomenon that strongly impacts the environment. Many terrestrial ecosystems depend on fire to maintain their ecological equilibrium and biodiversity, but new destructive fire patterns, often associated with land management practices and rapid climate change, have been degrading soil and water resources, increasing erosion by wind, precipitation and floods, decreasing biodiversity and contributing to desertification. Furthermore, pyrogenic emissions from biomass burning are an important source of atmospheric pollution and they impact the radiative balance of the troposphere, strongly contributing to the greenhouse effect. The objective of this research was to investigate the impact of climate variability on geographic, ecological, seasonal and inter-annual distributions of fires and correspondent pyrogenic emissions, across a variety of ecosystems. With this purpose, 10 years of world, monthly, 1°x1° gridded data, from the Global Fire Emissions Database, were compared with land-cover data, from the Goddard Institute of Space Studies, and with weather data, from the European Centre for Medium Range Weather Forecasting, the Global Precipitation Climatology Centre and the Global Hydrology Resource Centre. Overall, the climate parameters significantly correlated with carbon emissions were air and soil temperature, air and soil humidity, rainfall, wind speed and lightning density during the fire season, and also precipitation and snow cover up to 6 months before the fire season. Good statistical quantitative models of carbon emissions (correlations above 70%, and up to 95%, between estimated and predicted values, with residuals normally distributed) using humidity, temperature or lagged rainfall as predictors, were found almost exclusively in tropical grasslands, shrublands and woodlands, especially in Africa, where fire behaviour was more regular. In boreal and temperate forests and woodlands, where fire patterns were irregular and fire returning periods were larger, there were not enough fires, in 10 years of data, to obtain useful predictive statistical models. The fire models presented here, together with the quantitative statistical relationships found between climate and fire patterns, in different land ecosystems, are apt to be used in predictive climate models, land management, fire risk assessment and mitigation of climate change.