Resumen de Monitoring deforestation and forest degradation linking high-resolution satellite data and field data in the context of un redd+. A case of Tanzania
The main objective of this PhD is to support the development a forest monitoring system in Tanzania so as to report on current and historical emissions which derive from deforestation and forest degradation. The framework of the thesis is specifically focused on the emerging international context of the REDD+ (Reducing Emissions from Deforestation and Degradation) initiative, under which countries may obtain financial grants for demonstrating that they are reducing their carbon emissions from forest lands with respect to their recent historical practice.
The research focused on five focal areas of research: Part (1) reviews the policy background to REDD+. It outlines the rules and choices to be addressed by participatory countries and demonstrates some of the technical problems and options that they can face and adopt in the remote sensing technology.
Part (2) presents the results from the PhD field work in Tanzania. This included the set-up of rapid field data collection and guidelines on protocols to link the field data to the remote sensing data, so as to produce maps of vegetation cover and above ground biomass using very high resolution images. Part (3) demonstrates the improvement to map forests at a fine spatial resolution and with high frequency of acquisitions with the arrival of the new Sentinel-2 satellites. This potential has been tested on an area of dry forest in Central Tanzania.
Part (4) tests a full scale estimate of above ground biomass for the whole of Tanzania, using a combination of remote sensing and field data. The predictive capability was investigated by comparing the results against ground measurements undertaken by the national inventory.
Part (5) investigates the dynamics of deforestation around Dar es Salaam, along with a model to infer future probability of deforestation at the national level. The ability of the model to replicate spatial patterns of deforestation was assessed through ground truthing.
Among the main outcome of this PhD is that estimates of forest change from different sources have wide variance at national level and emissions estimates for the REDD+ process remain unreliable. There are a large number of choices facing a forest monitoring system, in terms of forest definitions and methods, which have an impact on the feasibility of implementation and results. The difficulty of linking remote sensing data to the forest parameter from national surveys has been shown, with recommendations to improve future field data collection. However the synergistic use of remote sensing and field survey data can effectively reduce the costs for mapping and monitoring forest changes and forest degradation. For this, the use of high resolution (5m) satellite image segmentation and texture indices was found to be useful in the production of forest biomass maps. Additionally, the arrival of Sentinel-2 data provides the opportunity to analyse medium high resolution data (<20m) in time series, especially useful for dry areas.
