The Caatinga is the largest nucleus of Seasonally Dry Tropical Forests (SDTF) in the Neotropics. The leafing patterns of SDTF vegetation are adapted to the current environmental and climate variability, but the impacts of climate change tend to alter plants’ phenology. Thus, it is necessary to characterise phenological parameters and evaluate the relationship between vegetation and environmental drivers. From this information, it is possible to identify the dominant forces in the environment that trigger the phenological dynamics of the Caatinga. In this way, remote sensing represents an essential tool to investigate the phenology of vegetation, particularly as it has a long series of vegetation monitoring and allows relationships with different environmental drivers. This study has two objectives: (i) estimate phenological parameters using an Enhanced Vegetation Index (EVI) time-series over 20 years, and (ii) characterise the relationship between phenologic dynamics and environmental drivers. TIMESAT software was used to determine four phenological parameters: Start Of Season (SOS), End Of Season (EOS), Length Of Season (LOS), and Amplitude (AMPL). Boxplots, Pearson’s, and partial correlation coefficients defined relationships between phenologic dynamics and environmental drivers. The non-parametric test of Fligner–Killeen was used to test the interannual variability in SOS and EOS. Our results show that the seasonality of vegetation growth in the Caatinga was different in the three experimental sites. The SOS was the parameter that presented the greatest variability in the days of the year (DOY), reaching a variation of 117 days. The sites with the highest SOS variability are the same ones that showed the lowest EOS variation. In addition, the values of LOS and AMPL are directly linked to the annual distribution of rainfall, and the longer the rainy season, the greater their values are. The variability of the natural cycles of the environmental drivers that regulate the ecosystem’s phenology and the influence on the Caatinga’s natural dynamics indicated a greater sensitivity of the phenologic dynamics to water availability, with precipitation being the limiting factor of the phenologic dynamics. Highlights: The EVI time series was efficient in estimating phenological parameters. The high variability of the start of season (SOS) occurred in sites with low variability of end of the season (EOS) and vice versa. The precipitation and water deficit presented a higher correlation coefficient with phenological dynamics. Length of Season (LOS) and amplitude (AMPL) are directly linked to the annual distribution of rainfall.

RESUMO De modo geral, a cobertura vegetal do semiárido paraibano se encontra bastante antropizada; exemplo disto é a bacia hidrográfica do Rio do Peixe, totalmente inserida na região semiárida do Nordeste brasileiro a qual tem sofrido, ao longo dos anos, as consequências da ocupação humana. Neste contexto, o presente trabalho tem como objetivo realizar uma análise multitemporal de variáveis biofísicas derivadas de imagens do satélite Landsat 5 e técnicas de processamento digital de imagens e geoprocessamento, para o período de 23 anos. Os resultados apontam para uma recuperação da cobertura vegetal, mais evidente nos trechos mais declivosos da bacia hidrográfica, fato relacionado ao êxodo da população rural ocorrido no período de estudo. Ficou patente, também, que o albedo da superfície foi o parâmetro que melhor se correlacionou com as variações da biomassa vegetal. Contudo, as variações da cobertura da vegetação não foram suficientes para, isoladamente, explicar as variações na temperatura da superfície e o saldo de radiação, ao longo dos anos analisados.

The study of energy, water, and carbon exchanges between ecosystems and the atmosphere is important in understanding the role of vegetation in regional microclimates. However, they are still relatively scarce when it comes to Caatinga vegetation. This study aims to identify differences in the dynamics of critical environmental variables such as net radiation (Rn), evapotranspiration (ET), and carbon fluxes (gross primary production, GPP) in contrasting recovered Caatinga (dense Caatinga, DC) and degraded Caatinga (sparse Caatinga, SC) in the state of Paraíba, northeastern Brazil. Estimates were performed using the Surface Energy Balance Algorithm for Land (SEBAL), and comparisons between estimated and measured data were conducted based on the coefficient of determination (R²). The fluxes were measured using the Eddy Covariance (EC) method for comparison with the same variables derived from Moderate Resolution Imaging Spectroradiometer (MODIS) data aboard the Terra satellite. The estimates showed higher Rn values for the DC, indicating that this area should have greater energy availability for physical, biological, and chemical processes. The R² between daily Rn estimates and observations was 0.93. The ET estimated using the SEBAL showed higher differences in relation to the observed values; however, it presented better spatial discrimination of the surface features. The MOD16A2 algorithm, however, presented ET values closer to the observed data and agreed with the seasonality of the Enhanced Vegetation Index (EVI). The DC generally showed higher ET values than the SC, while the MODIS data (GPP MOD17A2H) presented a temporal behavior closer to the observations. The difference between the two areas was more evident in the rainy season. The R² values between GPP and GPP MOD17A2H were 0.76 and 0.65 for DC and SC, respectively. In addition, the R² values for GPP Observed and GPP modeled were lower, i.e., 0.28 and 0.12 for the DC and SC, respectively. The capture of CO₂ is more evident for the DC considering the whole year, with the SC showing a notable increase in CO₂ absorption only in the rainy season. The GPP estimated from the MOD17A2H showed a predominant underestimation but evidenced the effects of land use and land cover changes over the two areas for all seasons.