Fuel characterization using remote sensing in support of large-scale wildfire management in the Cerrado biome

• Ano de defesa: 2022
• Autor(a) principal: Leite, Rodrigo Vieira
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa; Ciência Florestal
• Palavras-chave em Português: Incêndios florestais - Prevenção e controle; Incêndios florestais - Sensoriamento remoto; Aprendizado do computador; Cerrados; Fotointerpretação Florestal
• Link de acesso: https://locus.ufv.br//handle/123456789/30687; https://doi.org/10.47328/ufvbbt.2022.798

Cerrado is the most biodiverse tropical savanna in the world characterized by a range of vegetation structures from grasslands to forests. However, Cerrado’s stability has been threatened by anthropogenic changes in the fire regime. Managing fuels is one of the most important options we have for reducing wildfire’s negative impacts on society and the environment and amplifying positive ones on fire-dependent flora. Large areas such as Cerrado often require remote sensors to characterize fuels, especially those on-board space-based platforms. Even though spaceborne sensors have been around for over 50 years, new opportunities arise with recently launched sensors with unprecedented characteristics. Yet, Cerrado fuel properties and their potential to be characterized with the new generation of sensors are still little studied. In this thesis proximal and near-surface approaches as a step forward to scale-up Cerrado fuel properties are presented. The thesis is divided into three chapters, where: i) fuel moisture content is predicted using leaf spectroscopy and machine learning algorithms, ii) fuel load is predicted for the whole Cerrado extent using lidar data collected from unoccupied aerial vehicles and the recently launched GEDI spaceborne lidar sensor and ili) a review of the novel spaceborne remote sensors with potential to be used for fuel characterization and the main fuel-related variables for integrated fire management is presented. The results showed the applicability of spectroscopy and lidar data together with machine learning to retrieve important fuel characteristics for fire management. With the increasing availability of data from sensors with new technologies and capabilities, which allow the upscaling of information collected in the field, it is suggested that we are facing a new era for the mapping and monitoring of fuels, which is essential for the development of integrated fire management programs aiming at preserving the Cerrado biome and similar ecosystems worldwide. Keywords: Fire risk, Vegetation fuel, Spectroscopy, Lidar, GEDI, Machine learning, Satellite-borne sensors