Detecção automática de desmatamentos no bioma cerrado: desafios para o monitoramento sistemático
Nenhuma Miniatura disponível
Data
2018-03-13
Autores
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Goiás
Resumo
The Cerrado biome has about 30% or 12,070 of the flora species cataloged in Brazil. Accelerated
occupation and loss of biodiversity have made the Cerrado one of the 34 hotspots in the world.
With a high index of endemism, about 30% (645) of the species that make up the list with some
level of threat of extinction in Brazil are located in the Cerrado. The biome still has a remarkable
climatic seasonality, with dry and rainy season. The replacement of Cerrado vegetation for
anthropic uses intensified in the 1970s, basically for agricultural uses. Studies on climate change
point to deforestation as one of the causes of greenhouse gas (GHG) emissions, motivating the
government to create programs to monitor the coverage and use of land and removal of natural
vegetation. Mappings indicate a reduction of vegetation of the Cerrado in 6% in the 11-year
interval (PROBIO in 2002 with 60.5% and Cerrado TerraClass in 2013 with 54.5%). In this way, it
makes necessary the annual monitoring, financing the actions of conservation for the biome. In
order to understand the detection in the Cerrado conversion and to obtain information in a short
time, it was decided to use the Integrated System of Deforestation Alerts (ISDA), since it is the
only annual monitoring of the biome and has a low cost in relation to the others existing. SIAD
was developed in Lapig and has provided conversion rates since 2003. In this way, the objective
was to evaluate the potential and the limitations of detection of deforestation in the Cerrado using
satellite images of moderate resolution and analysis of time series with reference to the SIAD. The
efficiency of the reference map (i.e. PROBIO and TerraClass) was evaluated in relation to the
amount of false deforestation generated. We attempted to understand the variation in the number of
alerts generated as a function of rainfall seasonality from MODIS collections 5 and 6 (product
MOD13Q1) and to analyze the seasonal behavior of deforestation alerts from 2015-2016 using the
BFAST algorithm. Thus, the SIAD generated 101,902 deforestation alerts for 2015-2016 using the
Cerrado TerraClass mapping and MODIS collection 6, of which 3% (3,185) were classified as
deforestation by the visual inspection method. It was verified that the Cerrado TerraClass reference
map generates less false deforestation than the PROBIO, and that, the collection 6 presents greater
sensitivity of change in the vegetation in relation to the collection 5, contributing to the increase of
false deforestation. The seasonality of precipitation is directly related to the number of alerts
generated. However, the BFAST algorithm was effective in filtering the false positives, not
presenting breaks in 16% of the alerts generated for 2015-2016, with 99% agreement with the
visual inspection.
Descrição
Palavras-chave
Citação
FARIA, A. S. Detecção automática de desmatamentos no bioma cerrado: desafios para o monitoramento sistemático. 2018. 95 f. Dissertação (Mestrado em Ciências Ambientais) - Universidade Federal de Goiás, Goiânia, 2018.