Soil respiration and ecosystem carbon flux in semi-arid woodlands: effect of season, precipitation and wildfire

Abstract

Carbon exchange in arid and semi-arid ecosystems of the world has recently been highlighted because the areal extent of these ecosystems significantly influences the global carbon cycle. More information about the drivers of carbon fluxes in these ecosystems is required to advance our understanding of the contribution of dry ecosystems to the global carbon budget. Research on ecosystem carbon flux and soil respiration in semiarid to arid ecosystems is lacking relative to the more commonly researched temperate and well-watered systems. This thesis aims to determine carbon fluxes in example semi-arid woodlands of southern Australia. As is typical of woodlands in rainfall limited environments, the study sites were characterised by vegetation patches interspersed by open areas with little or no vegetation. A primary focus of the research was to characterise the response of soil respiration in the patches to temperature and rainfall with the aim of improving carbon exchange estimates in these ecosystems. The occurrence of a wildfire at one of the sites provided the opportunity to measure carbon flux response to this event which is also a common, but infrequent ecosystem response driver. Two types of woodlands in South Australia under similar climatic condition were investigated: mallee woodland and floodplain woodland. Mallee woodland was dominated by Eucalyptus dumosa, E. incrassata, E. oleosa and E. socialis. Floodplain woodland was dominated by Eucalyptus camaldulensis and E. largiflorens. In January 2014, a wildfire severely damaged the mallee woodland around an eddy covariance flux tower, burning trees and spinifex and consuming aboveground bark and leaf litter. The eddy covariance flux data was used to calculate net ecosystem productivity (NEP), gross primary productivity (GPP) and ecosystem respiration (Reco) in the mallee woodland. Soil respiration rates were measured in-situ each month under trees and in inter-canopy areas in the two woodlands. Before the fire, the mallee woodland was a net carbon sink, sequestering 81 g C m-2 yr-1 in 2013, but became a net carbon source after the fire (- 88 g C m-2 yr-1, data from May 2014 – April 2015). Reco and GPP declined by about 35% and 65%, respectively, after the fire. Net carbon uptake resumed 12 to 15 months after the fire with NEP approaching values similar to those before the fire. Soil respiration was significantly higher under tree canopies (1.5 μmol CO2 m-2 s-1) than in inter-canopy areas (0.9 μmol CO2 m-2 s-1). Soil organic C and microbial biomass C were higher under canopy than inter canopy. Soil respiration (Rsoil, μmol CO2 m-2 s-1) was mainly driven by soil water content (SWC) and thus magnitude and distribution of rainfall, but was not influenced by the wildfire. Temperature (T, °C) influenced Rsoil only when SWC (volumetric ratio) was not limiting in both patches. The relationships established for Rsoil were - inter-canopy: Rsoil = 1.79 – 0.08T – 31.82SWC + 2.38T*SWC, F3, 8 = 8.38, p < 0.01 and under canopy: Rsoil = 3.06 – 0.12 T – 46.96SWC + 3.43T*SWC, F3, 8 = 19.53, p < 0.001. Accounting for differences in area, soil respiration from inter-canopy and under canopy contributed 51 and 32% to total ecosystem respiration with the remaining 17% assigned to autotrophic respiration from aboveground vegetation. The semi-arid climate of the mallee region is characterized by long dry periods that are interrupted by rainfall events differing in magnitude. The net ecosystem carbon exchanges to single large rainfall events were examined using eddy covariance data from the mallee woodland. Three large rainfall events with similar magnitude (35-55 mm), but contrasting previous rainfall were chosen. NEP, Reco and GPP rates 28 days prior to, and 35 days after the three central rainfall events were used. These system scale observations were complemented with 48h soil respiration measures from a water manipulation experiment (30 mm rainfall simulation) at small plot scale in the field following a long dry period (only 4.8 mm rainfall input in 45 days). Changes in ecosystem carbon fluxes responding to the central rainfall event depended on previous rainfall patterns. After 4 weeks with several medium to large rainfall events, the central rainfall event had little effect on NEP, Reco and GPP. In contrast, a large rainfall event following 4 weeks with very little rainfall induced a decrease in NEP and an increase in Reco for about 3 weeks. The strong increase in Reco after the central rainfall event can, at least partly, be explained by an increase (a “pulse”) in soil respiration upon rewetting. GPP rates did not respond to any of three rainfall events. In a semi-arid floodplain woodland, monthly in-situ soil respiration was higher under tree canopies (5.0 μmol CO2 m-2 s-1) than inter-canopy (2.3 μmol CO2 m-2 s-1), and also had higher soil organic C content. Five days of rainfall (total of 62 mm) in summer (air temperature ca. 23C) increased soil respiration rates by 70-85% four days after the last rainfall event when compared to rates before the rainfall. Soil respiration rates remained high over the following two months with no rain, although soil water content in the top soil (0-5 cm) was very low. This suggests that in the months following the rain period, roots and associated microbial activity in the moist subsoil were the major contributors to soil respiration. On the other hand, a single rainfall event of similar magnitude in autumn (air temperature ca. 17C) did not increase soil respiration compared to the previous dry period although top soil water content was three times higher. Therefore, soil respiration in this ecosystem is influenced by a complex interaction between rainfall amount, soil temperature as well as, possibly, distribution of the rainfall. Soil respiration rates in the floodplain were generally higher than those in the Mallee and also those recorded in other floodplain studies with forests, woodlands, grasses or shrubs. While the respiration rates recorded in this study are at the upper end of those reported from other studies. While the floodplain setting in this experiment was in a semi-arid environment, the location with respect to the free water in the adjacent river resulted in a water table within 2 m of the soil surface. The soil profile below 30 cm will generally remain moist from capillary upflow from the water table. In this generally warm environment, carbon turnover will be high, while total carbon in the soil will be low, as measured. Two laboratory incubation experiments were conducted to understand the response of respiration and nutrient availability to drying and rewetting [first experiment], and to labile carbon as it may be released as root exudates [second experiment] in soil from the mallee woodland. Soils (top 0 - 30 cm) were collected from under tree canopies, under shrubs or in inter-canopy areas in unburnt and burnt locations, four months after the wildfire. In the first experiment, the soils were exposed to two water content treatments: constantly moist (CM) and drying and rewetting (DRW). In CM, soils were incubated at 80% of maximum water holding capacity (WHC) for 19 days. In DRW, soils were dried for four days, kept dry for another five days, then rewet to 80% WHC and maintained at this water content until day 19. Soil respiration decreased during drying and was very low in the dry period; rewetting induced a respiration pulse. Compared to soil under shrubs and in inter-canopy areas, cumulative respiration per g soil in CM and DRW was greater under tree canopy, but lower when expressed per g of organic matter content (TOC). TOC, available P, and microbial biomass C (MBC), but not available N were greater under tree canopy than in inter-canopy areas. Wildfire reduced TOC and MBC concentrations only under tree canopies, but had little effect on available N and P concentrations. Wildfire also decreased the pulse of respiration per g TOC in inter-canopy areas and under shrubs. In the second experiment, the soils were incubated at 80% WHC for 24 days without or with addition of 5 g C kg-1 as glucose. TOC, MBC and microbial biomass P (MBP) and available N and P were measured at the start of the experiment; soil respiration was measured continuously. Soil TOC, MBC, N and P availability and cumulative respiration were greater under trees than in inter-canopy areas. Fire decreased TOC and cumulative respiration only under trees and had little effect on available N, MBC and MBP concentrations. The greater increase in cumulative respiration by glucose addition under shrubs and in open areas compared to under trees and, in a given patch, greater in burnt than unburnt soils, indicate lower availability of native organic carbon. In conclusion, this thesis showed that net ecosystem productivity in the mallee woodland was strongly influenced by rainfall and wildfire, with net carbon uptake of the ecosystem beginning approximately one year after the fire. On the other hand, soil respiration in this woodland was not influenced by wildfire and instead was related to soil water content and temperature. Moreover, the previous rainfall pattern determined response of net ecosystem carbon exchange to a single large rainfall event, which should be considered in carbon model development. The observations from the floodplain site indicate that riparian areas adjacent to river systems are highly likely to be have high annual soil respiration rates, especially in warmer environments and therefore have an influence on regional and global carbon cycles that is greater than would be expected based on climate and area alone.