With the aim of studying the primary productivity dynamics of subtropical forests with different degrees of intervention or change due to human intervention, we classify the ecosystems of an area in northeastern Argentina, corresponding to a humid subtropical region, according to their temporal variability. A 22-year time series, ranging from 2000 to 2022, of MODIS EVI was assembled into a spatiotemporal cube, and pixels were classified by an archetypal analysis applied to the frequency components of a Fourier power spectrum. Then the most representative pixels of this classification (archetypoids) were selected. A wavelet decomposition was performed on these archetypoids to identify temporal changes in the frequency composition of the time series, and an ARIMA model to identify changes in the noise pattern of the series. Finally, a distributed-lag model with meteorological variables, was applied to these time series to relate the dynamics of the archetypes to the local climate. A stepwise procedure using Gaussian processes for error and autocorrelation and multiple regressions to determine any univariate relationship between meteorological variables and EVI. The meteorological variables considered were temperature, rainfall, and potential evapotranspiration (PET). The procedure started with a null model with white noise, then Gaussian processes were gradually added to model the errors, and then the explanatory variables, which were filtered moving average time series of the meteorological variables. The interaction between the explanatory variables was assumed to be the minimum EVI of the predicted values of each variable according to Lieibig's law of minimum. The procedure was applied as the BIC decreased to find the optimal model. In this study area, three different archetypes were sufficient to describe most of the variability in the time series matrix. Archetype 1 was characterized by woody plantations of exotic species such as pines, yerba mate and tea, archetype 2 by native forests, and archetype 3 represented a mosaic of forests and agriculture/pasture. The analysis of climate and archetypes indicated that archetype 2 had the highest mean EVI values (i.e., primary productivity), followed by archetype 1 and lastly archetype 3. Also, it showed that all three archetypes responded to the same combination of climate variables (temperature and PET), with varying degrees of sensitivity to each variable. Archetype 2 was the least sensitive to changes in these variables, archetype 1 was more sensitive to PET, and archetype 3 was more sensitive to temperature, while also exhibiting the least response time.