The Amazonian savannas are extremely important for the conservation of biodiversity, being composed of vegetation communities of numerous endemic species. However, the Amazonian savannas are poorly studied. Forest areas of the Amazon have been monitored since 1988 when the Amazon Monitoring and Deforestation Project (PRODES) was created to obtain annual gross deforestation rates of the Brazilian Legal Amazon. However, PRODES does not monitor non-forest areas (NF) within the Amazon biome, restricting information about nonforest formations, their environmental diversity and degree of anthropization. Thus, the general objective of this work is to analyze the landscape dynamics in non-forest formation areas in the periods of 2000, 2015 and 2020. This research has as an area of analysis an area of NF (Amazon-Cerrado transition ecotone) located in the municipalities. of Rio Maria, Redemption, Araguaia Forest, Conceição do Araguaia, Santa Maria das Barreiras, Pau D'arco and Santana do Araguaia, southeast region of Pará State, area of recent settlement process. To perform LULC mapping of the land, the Google Earth Engine (GEE) platform was used. It is a catalog of ready-made analytics data with a high performance, intrinsically parallel computing service. When analyzing the results by thematic class, it was observed that the Savannah Park, Agriculture and Others classes presented a higher agreement 90%. The Pasture and Savannah Wooded classes had lower agreement, with 80%. The classes that represented the highest intensity of omission were Wooded Savannah with 10% and Other 7%. Inclusion, had the highest values in the pasture with 13% and Agriculture or Pasture 7%. The overall accuracy of this mapping was 86%. The GEE platform proved to be efficient and agile, which allowed several sorting attempts to be made in the shortest time until the best possible result was achieved with excellent validation results.
|
