Reconstrução de séries temporais de NDVI para uma área de caatinga.

Abstract

Orbital Remote Sensing (SRO) enables the monitoring of land cover from vegetation indexes. gap-filling analysis in time series of these indices is generally done with high temporal resolution and low spatial resolution sensors. However, in semi-arid regions, the local and daily scale comprehension of the phenomena that ocurrs in the land cover, such as Caatinga vegetation, is very important due to its heterogeneity and multiple human actions. Therefore, the objective of this study was to reconstruct NDVI series (Normalized Difference Vegetation Index) for an area of Caatinga vegetation obtained by the TM, ETM + and OLI sensors of the Landsat 5, 7 and 8 satellites, respectively, by testing the application of curves based on mathematical functions to fill in gaps, defining among them the one that presents the best fit for the study area, which is located between the cities of Sumé and Monteiro, in the Cariri region of Paraíba and is represented in this work by 19 points, being 4 of Caatinga preserved and 15 that have some antropic alteration. In order to choose the best adjustment equation, the LAB Fit software was used, taking into account the reduced chi-square (χ²) and determination coefficient (R²) parameters. For this application, 4 points of the study area were used, for the period from 1994 to 2017. The three-parameter Cauchy function was chosen for NDVI adjustment and was tested using two criteria: status of land cover change (altered or not by anthropic action and annual rainfall). The function fit well even after the coverage change, for most of the analyzed points (85%) and performed well for the rainy years, that is, those that presented annual precipitation above the average (514 mm) for the analyzed period adjusting well in 83% of these years, obtaining an average R² of 0.82. The adjusted function can be used in areas that are similar in relation to the climate and vegetation present in the points observed in this work. This approach can be used as a tool to produce estimates of biomass at local and regional scale, in addition to enabling more sensitive analyzes of vegetation behavior in the Brazilian semi-arid, such as its phenological cycle, its relation with the carbon sequestration of the atmosphere and the effect of climate change on the Caatinga.