Carbon cycling, fire and phenology in a tropical savanna woodland in Nhambita, Mozambique

Abstract

In the savanna woodlands of Southern Africa, locally know as miombo, carbon cycling is poorly quantified and many of the key processes remain obscure. For example, seasonal constraints on productivity and leaf display are not well understood. Also, fire is known to be a key process, with around 50% of the annual global area burned occurring in Africa, but detailed understanding of its ecological effects is lacking. Land use change and woodland degradation are changing the structure and functioning of these tropical woodlands, which cover 2.7 million km2 of Southern Africa and provide ecosystem services which support the livelihoods of over 100 million people. In this thesis I quantify the major carbon stocks of the woodlands in Nhambita Regulado, Gorongosa District in Sofala Province, Mozambique. I also examine processes that affect these stocks, including fire and clearance for agriculture. Furthermore, I quantify the seasonal cycle of leaf display, and its relationship to climate. I conducted a series of experimental burns and found that fire intensity was strongly related to rates of top-kill and root stock mortality. Top-kill rates decreased as tree diameter increased up to 10 cm DBH. After this point increased size did not affect top-kill rates, possibly because of accumulated wounds and rottenness. I then extrapolated these results to long term predictions of tree populations and carbon stocks by modelling the interactions of fire, mortality and tree growth. The model was able to successfully predict woody vegetation structure at two sites with known fire regimes, including a 50-year fire experiment in Marondera, Zimbabwe. The results show that annual fires in miombo suppress all woody vegetation. Low intensity fires every 2.5 years are required to maintain observed stem biomass in Nhambita. High intensity fires lead to high top-kill rates (12%), even among large stems. Manipulating fire intensity rather than frequency seems to be the most practical approach to limiting degradation by fire in these ecosystems. Using a three year time series of hemispherical photographs of the tree canopy, combined with satellite data, I find that tree leaf phenology is not directly related to seasonal rainfall patterns, both in Nhambita and across Southern Africa. Pre-rain green-up is the dominant phenology, from the semi arid savannas of the south of the continent to the wet miombo of the Congo basin. Wet miombo woodlands have longer periods of green-up before rain onset (mean 60 days) compared with dry miombo (37 days). Green up-dates show little interannual variability but large spatial variability. The importance of pre-rain green-up in determining how these ecosystems will respond to changing rainfall patterns is unknown, but is an important area for future study. I quantified carbon stocks in the Nhambita woodlands in the soil (69% of total carbon stocks of 111 tC ha-1), tree stems (19%) and roots (8%) as well as other smaller pools. An allometric relationship between root and stem biomass and stem diameter was developed, and used to evaluate the uncertainties in stem carbon estimation at plot and landscape scale. We find that the uncertainty (95% confidence intervals) at plot scale can be quite large (60% of the mean) but this is reduced to around 25% at landscape scale. Strategies for effective inventories of miombo woodland are presented. Using a chronosequence of abandoned farmland, we estimate that stem biomass recovers from clearance after around 30 years of abandonment. Changes in soil carbon stocks are less well understood and need further work. This thesis concludes by outlining further work needed to model the carbon cycle of these woodlands, as well as discussing the implication of pre-rain green-up for satellite observations of land cover changes and biomass mapping.