Detection of forest resilience to environmental change and quantification of contemporary carbon fluxes over Amazonia using remote sensing

Abstract

Assessment of the interactions between contemporary climate and global vegetation is critical in bioclimatic zones where vegetation is highly vulnerable to environmental change or provide sufficient influence on current climate to be capable of drastically accelerating global warming. This thesis presents an improvement on the quantification of the spatio-temporal variability in the Amazonian carbon [C] cycle, specifically through monitoring changes in gross primary productivity [GPP] and vegetation structure, the latter being a good proxy for above ground biomass [AGB]. This work uses the moderate resolution imaging spectrometer [MODIS] to study the impacts of environmental change on the Amazon forest and to identify factors mediating forest responses during the period 2001-2021. As a result, regions with increased vulnerability to climate change and human induced pressures or that have been rapidly altering Amazonian C fluxes have been identified.

In Chapter 2, the influence of forest fragmentation and temperature change on vegetation structure in Amazonia is investigated by exploiting the anisotropic characteristics of leaves and canopy crowns through the deployment of the anisotropic enhanced vegetation index [AnisoEVI]. A strong relationship is found between AnisoEVI and AGB across the basin, and therefore AnisoEVI can be used as a proxy for long term alterations to vegetation structure and AGB. Firstly, AnisoEVI is shown to significantly decline under intense fragmentation with the largest absolute reduction between interior to dominant forest area density [FAD] classifications. Secondly, this study found a strong relationship between AnisoEVI and temperature change, with widespread losses in vegetation structure between 2003-2018 across the basin. Background climate, soil fertility and fragmentation extent highly mediated the response of forests to temperature change, with western interior forests displaying a greater resilience and southern forests having a greater vulnerability to warming and edge effects. Future forest resilience was studied by applying this relationship using outputs from a climate model ensemble (15 models) of land temperature under 3 shared socioeconomic pathways [SSPs]. By 2100 Amazon forest structure is projected to transition towards a Cerrado-like structure in the worst-case climate scenario. Thus, this work provides unique basin wide estimates of ecological resilience to warming and highlights regions of Amazonian tropical forest that may surpass temperature change thresholds leading to a rapid degradation of vegetation structure.

Chapter 3 comprises the development and evaluation of a novel light use efficiency [] and GPP model utilising the eddy covariance tower and Global Ecosystem Monitoring networks for targeted use within the Amazon. The model, termed the BioChemical Light-Use Efficiency model [BioCLUE], is applied to study the influence of climate and phenology on the seasonal and inter-annual variability in  and GPP. Through the inclusion of the photochemical reflectance index [PRI] within BioCLUE, this new method of estimating  incorporates a direct measure of photo-inhibition rates which implicitly accounts for non-prescribed environmental factors on  such as soil water content [SWC]. Therefore, BioCLUE presents a simple and effective remote sensing-based methodology to monitor the input into the Amazon C cycle by lowering the number of variables required to capture the multitudinous environmental controls upon GPP which can be produced with low latency. Additionally, BioCLUE explicitly prescribes the impact of elevated carbon dioxide [CO2] on , a process currently underrepresented by many remote sensing GPP algorithms. Evaluation and comparison of the model developed here against other GPP products to in-situ estimates demonstrates the proficient performance of BioCLUE. Notably, the performance of BioCLUE is found exceed other products in estimating the observed variability of GPP across the Amazonian hydrological gradient and at sites with greater seasonal variability.

In Chapter 4, the BioCLUE model is applied across lowland Amazon forests to evaluate  and GPP at seasonal, interannual and decadal (2001-2021) timescales. Through the inclusion of PRI in the  parameterisation, BioCLUE is additionally revealed to contain an implicit sensitivity to soil nutrient concentration with photo-inhibition rates reduced in regions with low water stress and high nitrogen concentrations. BioCLUE simulates seasonal variations in  and GPP which are spatially variable in both timing and magnitude with the former primarily governed by vapour pressure deficit [VPD]. However, in regions where moisture availability remains high year-round, the impact of PRI, which is further demonstrated to capture seasonal leaf demographic shifts, becomes an increasingly dominant and significant driver of  seasonality. Analogous to long term trends suggested by dynamic global vegetation models [DGVMs], GPP is estimated in BioCLUE to have significantly increased over the majority of lowland Amazonia during the study period, principally via the stimulation of C assimilation rates from rising atmospheric CO2 concentrations. However, severe droughts associated with El Niño resulted in widespread reductions in both  and GPP with 2015/16 identified here as the most critically impacted period. Furthermore, water table depth [WTD] was found to strongly mediate the forest response to drought with low WTD regions increasingly more resilient to significant rainfall deficits.