CUNHA, J. E. B. L.; http://lattes.cnpq.br/7756258383405207; CUNHA, John Elton de Brito Leite.
Resumo:
Low monitoring plus high human and climate pressures make the Caatinga biome one of the most vulnerabte biomes in the world. Time series of remote sensing are vafuable for analyzing LCC in áreas with high seasonalrty, but they require a lot of computationai resources. Earlier studíes mostly use > 30- years time series of vegetatíon indexes at low spatial resolution (1 to 8 km). However, this spatial resolution usually does not allow to identify human actions (impacts) on the environment. Landsat imagery quality (radiometricalfy as weli as geometrically) and availability has improved in recent years and is now ready to support high temporal resolution monitoring and analysis of land surface processes. The objective of this study is to analyze, from sensors of médium spatial resolution, the changes in land cover of anthropic origin in an area of the Caatinga biome. For this purpose, algorithms were used to generate vegetation índices, surface albedo and evapotranspiration from sensor data on the satellites of the Landsat family. To increase the efficiency in generating this information, the algorithms were conducted to operate with low demand for meteorological station data and without human intervention during processing. In addition, a high performance service for orbitai data processing is proposed. The data generated by these algorithms were tested to field observations, demonstrating the possibility of using these algorithms in automatic processes. The techniques of cloud computing and parallelization used in this study were efficient in producing long time series (over 30 years) of these variables in average spatial resolution. The main application developed in this work, used Landsat time series for a period of 31 years at monthly resolution in order to investigate spatial and temporal pattems of hotspots of land cover change in a Caatinga area of the semi-arid region of the Paraiba state, Brazil. A new spectral index - Surface Albedo Index (SAI) - is proposed to improve the
observation of vegetation biophysica!condition and change. SAI, NDV1 and EVI are compared in order to evaluate the suitability of monitoring LCC driven by human actions in contrast to climate induced (drought) alteration. The TSS RESTREND method was successfully applied to Landsat time series for LCC monitoring. It is employed in order to remove the short-term influences of precípitation on land cover physiognomy, thus allowing to assess the ability of the index time series to discriminate LCC in drylands. Google Earth, Rapid Eye images and in situ observations (from October 2017) were used to observe preservation / degradation conditions along the time. Results showthat SAI is able to distinguish between "changed" and "unchanged" land cover with a high accuracy (87%) to detect the year of change. When using the SAI index, TSS RESTREND is suitable to detect LCC in the Caatinga, and its best performance was achieved when the change event occurred in the middle of the time series (1990-2010), with some inaccuracies in dry years. The lower ability of EVI and NDVI in the detection of LCC in the Caatinga biome is explained by their high sensitivity to leaf cover variations (as a result of seasonal or extreme drought conditions). LCC impacts the whole soilplant-atmosphere
system, such as biomass remova!and changes in soil properties as weil as
mícroclimate, due to the direct exposure to radiation, precípitation, and wind. In this regard, SAI is supposed to be more sensitive to man-made alterations of the land surface, due to its ability to capture a higher number of environmental feedbacks.