|dc.description.abstract||The Miombo woodlands of southern Africa are a globally significant store of carbon (C) and biodiversity. They also provide services for more than 150M people across several of the world’s most economically impoverished countries. The Miombo woodlands are dynamic, with extensive resource loss accompanied by areas of regrowth and increase. Disturbance processes, both from natural processes and widespread anthropogenic activities, are critical in maintaining woody biomass in these ecosystems, although intensity of disturbance varies widely. Increase in woody biomass has been observed in the Miombo, though the drivers of this trend are uncertain and the fundamental constraints on trees and woodlands not well understood. Ultimately, both losses and gains can be difficult to detect and hard to attribute to a particular cause. The aim of this thesis is to use field data and remote sensing to add to understanding of the constraints on tree populations in the Miombo and the impacts of severe environmental disturbances and management interventions on woodland structure.
Tree growth is a crucial demographic rate in African woodlands and plays a key role in shaping woodland structure and C cycling. However, observations of tree growth are relatively lacking in the Miombo and the determinants of tree growth rates are poorly known. In Chapter 2 I use data collected from long-term monitoring of permanent sample plots in Mozambique and Tanzania and linear mixed modelling to estimate tree growth increments and assess the relative importance of different determinants of tree growth. The estimated growth (diameter increment) in these plots was 1.8 ± 0.17 mm/yr. Climate and edaphic factors explained little variation in tree growth. Tree-tree competition was found to be a significant constraint on growth (trees in experiencing competition levels in the top 5% of values grew 1.24 ± 0.08 mm/yr slower on average than those in the bottom 5%) as was stem wounding (wounded trees grew 0.84 ± 0.04 mm/yr slower). Root symbioses (both fungal and bacterial symbionts) which aid in the uptake of nutrients were found to have a strong positive impact on growth, particularly ectomycorrhizal associations which are common to dominant species in the Miombo. The impacts of tree-tree competition and nutrient symbioses are poorly represented in biogeochemical models in these ecosystems but
this analysis suggests they are critical, whilst the subtle impacts of human interaction with trees (through wounding) are also possibly underappreciated.
Tropical Cyclones can have substantial long-term impacts on woodland structure in affected areas and projections indicate that the impacts of Cyclones will increase in southeastern Africa over the coming century. There are few studies which have documented the immediate impacts or long-term responses of woodland ecosystems to this damage. In Chapter 3 I analyse data from a survey of eight permanent sample plots setup explicitly to assess the damage caused by Cyclone Idai to in woodlands in Gorongosa National Park, central Mozambique. It is found that Cyclone Idai caused damage primarily to large trees, thus whilst only 2% of trees were felled these individuals represented 8.5% of overall basal area. The implications of this damage are discussed in context of the constraints on trees in these woodlands, and whilst the damage is severe it is concluded that the outcomes are highly uncertain.
Whilst damage from the cyclone is substantial, detecting change in woodland structure is challenging in these ecosystems. In Chapter 4 I explore the possibility of upscaling field observations of treefall occurrence using data from a small unmanned aerial vehicle (drone) and satellite radar. Drone survey produced comparable estimates of treefall intensity to the PSP observations (in terms of fallen number of stems, fallen basal area and carbon) and allowed survey of 155 ha, capturing widespread damage across the study area. In the study area radar backscattering intensity in C-Band radar reduced in the two years after the cyclone relative to the two years before whilst interferometric coherence increased - both in agreement with radar theory - although backscattering intensity in L-Band radar increased. Whilst significant relationships were identified between change in radar data and the intensity of damage in drone surveys, there appears to be limited ability to map variations in treefall intensity across the wider landscape using this method, or to determine areal impacts on above ground C thereafter. It is concluded that repeat analysis may yield better results however.
In Tanzania, Village Land Forest Reserves (VLFRs, a form of participatory forest management) aim to promote sustainable profit from woodland resources, although the impact of VLFRs on land cover change rates is uncertain. In Chapter 5 I use satellite radar to map deforestation and a degradation across an area of southern
Tanzania from 2010-2018 and statistical matching to compare rates of land cover change within a sample of VLFRs to woodlands under comparable resource pressure outside VLFRs or other protection status. It is found that VLFRs in the majority of cases were very effective in reducing deforestation (with five of seven having rates close to zero) and also reduced degradation rates (though to a lesser degree). Increasing density of woody biomass in forested areas was observed in all VLFRs, but varied widely across the sample (from 0.2 - 1.5 tC ha yr-1) and was in five of seven VLFRs below that observed in woodland areas with no protection status in this region (+0.7 tC ha yr-1). Whilst it appears that VLFR establishment achieved its intended goal of sustainable profit from woodlands resources from 2010-2018, further work is required to understand variation in outcome across the observed sample. This methodology however shows promise in continued assessment of VLFR performance for this purpose.||en