Résumé : In this work, we reconstruct and analyze the micrometeorological dynamics of the transitional forest located south of the Amazon basin. For this, we use time series of micrometeorological variables collected over five years in the transitional forest of Mato Grosso (Brazil). We employ local feature analysis, a recently proposed extension of principal component analysis, to extract the most relevant physical variables from this set. We show in this way that temperature records contain most of the dynamical information in all seasons. Based on this result, the dimensionality of the space spanned by the system's dynamics and the properties of the so defined attractors are obtained. In the dry season, the system presents a robust oscillatory character described by a welldefined limit cycle. In the wet season, the dynamics becomes more irregular but can still be seen as a periodic behavior affected by external noise. These results can help to develop accurate models for the meteorology of the Amazonian transitional forest and can thus lead to a better understanding of this important ecosystem.