GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 9, NO.4, PAGES 503-513, DECEMBER 1995 The magnitude and persistence of soil NO, N20, CH4, and C02 fluxes from burned tropical savanna in Brazil Mark Poth, 1 Iris Cofman Anderson, 2 Heloisa Sinatora Miranda, 3 Antonio Carlos Miranda,3 Philip J. Riggan1 Abstract. Among all global ecosystems, tropical savannas are the most severely and extensively affected by anthropogenic burning. Frequency of fire in cerrado, a type of tropical savanna covering 25% of Brazil, is 2 to 4 years. In 1992 we measured soil fluxes of NO, N20, C~. and C02 from cerrado sites that had been burned within the previous 2 days, 30 days, I year, and from a control site last burned in 1976. NO and N20 fluxes responded dramatically to fire with the highest fluxes observed from newly burned soils after addition of water. Emissions of N-trace gases after burning were of similar magnitude to estimated emissions during combustion. NO fluxes immediately after burning are among the highest observed for any ecosystem studied to date. These rates declined with time after burning and had returned to control levels 1 year after the bum. An assessment of our data suggested that tropical savanna, burned or unburned, is a major source of NO to the troposphere. Cerrado appeared to be a minor source of N20 and a sink for atmospheric CH4. Burning also elevated C02 fluxes, which remained detectably elevated I year later. Introduction more trace gas exchange data and a better understanding of One of the keys to understanding global atmospheric the mechanisms involved in order to select optimal change is to understand the interactions between land strategies for mitigating emissions. management practices and the influence of these actions on Disturbance can greatly alter soil processes and trigger the magnitude and direction of ecosystem trace gas changes in the emission rates of important trace gases. For exchange. Many of the atmospheric trace gases of interest example, grassland fertilizations can stimulate N 20 are the direct or indirect products of biological activity. emissions and depress methane uptake [Mosier et al., Consequently, tropical biomes of high biological activity 1991]. This is the result of the stimulation of soil are important contributors to the global trace gas budget. nitrification and denitrification processes and the inhibiting The emissions data available on what are a wide variety of effects of ammonium on the biological oxidation of tropical ecosystems supports the importance of tropical methane. Flooding caused by new reservoirs has increased ecosystems as sources of NO, N20 [Keller et al., 1983; C02 and CH4 emissions [Rudd et al., 1993] by changing Keller et al., 1986; Seiler and Conrad, 1987; Kaplan et al., oxygen availability to what used to be forest soils. Fire in 1988; Goreau and de Mello, 1988; Johansson and California chaparral dramatically stimulated NO and N 20 Sanhueza, 1988; Luizaoetal., 1989; Garcia-Mendezetal., emissions for over 6 months [Anderson and Poth, 1989] by 1991] and methane [Keller et al., 1986; Hao et al., 1988]. converting nitrogen into more available forms for An understanding of how biosphere/atmosphere trace gas nitrification and temporarily reducing competition for exchanges are modified by various land use practices is nitrogen from plant uptake. Conversely, in burned tall lacking [Keller et al., 1992; Davidson et al., 1993]. grass prairie, little N 20 was released from burned soils Without such data, projections or assessments cannot be because of uptake by the surviving grasses [Groffman et made of the numerous land use options available. We need al., 1993]. Fire is certainly a potent force in t.t~e tropics [Crutzen and Andreae, 1990; Riggan et al., 1993], yet information on the influence of fire on soil emissions of 1U.S. Department of Argiculture, Forest Service Research, Riverside, California. important trace gases is limited [Luizao and Matson, 1989; ZVirginia Institute of Marine Science, College of William and Mary, Hao et al., 1988]. Gloucester point, Virginia. Cerrado, a tropical vegetation type unique to South 3Departmento do Ecol6gia, Universidade de Brasilia, Brasilia, Brasil. America, covers 2 million km2 or about 25% of Brazil. Copyright 1995 by the American Geophysical Union. Fires set by man or by lightning are common in the dry season and have been occurring in these ecosystems for Paper number 95GB02086. thousands of years [Goodland, 1911]. Cerrado is made up 0886-6236/95/95GB-02086$1 0.00 of diverse plant communities covering a continuum from 503 504 POTH ET AL.: SAVANNA SOIL TRACE GAS EXCHANGE IN BRAZIL open tropical grassland to savanna (campo limpio) to Plant Biomass and Fuel Consumption cerradao, a closed canopy semideciduous forest. Five distinct plant communities are recognized [Goodland, 1971; To partially characterize the fire at these sites, we Coutinho, 1982]. The gradient of cerrado vegetation has measured total fuel consumption during the fire by been hypothesized to follow gradients in soil fertility and determining the plant biomass before and after burning. fire use. Fire is very common in grasslands and savannas On the basis of previous prescribed fire experiments, we but is a rare occurrence in the mature closed canopy have defined fuel for the cerrado simply as the living or cerradao forests [Goodland, 1971]. Fuel characteristics and dead parts of the vegetation found from the soil surface to a direct fire emissions rates have been reported for cetrado by height of 2 m. Tree trunks and stems with a diameter Ward et al. [1992]. This paper reports the first measure­ greater than 6 mm are excluded. To estimate fuel, five ments of trace gas fluxes from cerrado soils and the effects transect lines of 15 m were established in each plot. For of burning on changing flux rates, and also assesses the each transect, five subplots (25x25 em) at 3-m intervals importance of fire disturbance of tropical savanna on the were used to sample the fuel. All fuel within the subplot global budget of trace gases. was clipped at ground level, and together with the litter on the soil surface, taken to the laboratory. In the laboratory, Methods fuel was separated into live and dead grasses, leaves, and stems. The fuel was dried at 80°C for 48 hours, allowed to Study Site cool, and weighed. For higher shrubs and trees, the fuel was sampled in a subplot 1-m wide, 5-m long, and 2-m Our study was conducted within the research and ecologi­ high in the middle of each transect. After the fire, a similar cal reserve operated by the Instituto Braseleiro de Geografia transect was used to collect and estimate the amount of e Estatfstica (IBGE), located 20 km south of Brasilia, unburned fuel. District Federal ( 15°55' 58"S,4 7°51 '02"W). The climate is tropical (Koppen's Aw) with mean annual precipitation of Flux Measurements 11 00-1600 mm. This area of the tropics is characterized by distinct wet (October to March) and dry (April to Fluxes of NO, NzO, CH4, and COz were measured September) seasons with 90% of the precipitation falling using a closed box flux technique [Anderson and Levine, in the wet season. The mean relative humidity is 64%, but 1987]. One day prior to making flux measurements at each it can be as low as 18% during the driest periods. The plot, six stainless steel frames, covering an area of 0.581 annual mean maximum temperature is 25°C and the annual m2 were driven 2.5 em into the soil. In order to measure mean minimum temperature 16°C (climate data courtesy of fluxes of either COz or NO, a Teflon-lined, expanded-cell the Reserva Ecol6gica, IBGE, 1980-1989). Soils are polycarbonate box fitted with a fan and with a total volume dystrophic, deep, and well-drained red latisols. of 150.5 L (including the frame volume) was placed over Current research at IBGE has been carried out in three the frame. The base of the flux box was sealed to the experimental sites with different physiognomic forms: frame via a Teflon-covered foam lining. Two openings cerradao, closed canopy semideciduous forest; cerrado, a (0.64 em) in the top of the flux box prevented pressure dense scrub of shrubs and trees (in a strict sense), and changes from developing during sampling. For N zO and campo sujo, a grassland with scattered shrubs [Goodland, CH4 measurements, a flux box with a total volume of 76.9 1971]. Plots are burned every 4 years (in August), or every L was used in order to increase sensitivity. Because we 2 years (in either June, August, or September). Long-term made measurements during the dry season, soils were control plots have not burned since 1976. Plot size ranges relatively dry. After measuring trace gas flux rates of dry from 1 to 5 ha. The overall design includes 37 plots. soils, we added distilled water equivalent to a 1-cm rainfall. The work we report on here was done during late After 30 min, gas flux measurements were repeated. We August of 1992. Burning was carried out by the Reserva sampled a new plot each day. Ecol6gica - IBGE technicians in conjunction with Universidade de Brasilia staff. Fires were usually started at Inhibitors and Soil Amendments 1400 LT. Fires spread quickly and typically were completed within 30 min. To assess the roles of different microbial communities We measured fluxes of NO, NzO, CH4, and COz from in producing the observed fluxes of trace gases, we cerrado sites that had been burned within the previous 2 employed two nitrification inhibitors: acetylene and days. We also made measurements on plots 30 days and 1 allylthiourea (A TU). Acetylene was added to the flux box year after burning and from long-term control plots that to a concentration of 1% and allowed to react with the soil have remained unburned since 1976.