In a context of global change, soil has been identified as a potential carbon (C) sink, depending on land-use strategies. To detect the trends in carbon stocks after the implementation of new agricultural practices, early indicators, which can highlight changes in short timescales, are required. This study proposes the combined use of stable isotope probing and chemometrics applied to solid-state 13C nuclear magnetic resonance (NMR) spectra to unveil the dynamics of the storage and mineralization of soil carbon (C) pools. We focused on light organic matter fractions isolated by density fractionation of soil water stable aggregates because they respond faster to changes in land use than the total soil organic matter (SOM). Samples were collected from an agricultural field experiment with grassland, continuous maize cropping, and ley grassland under temperate climate conditions. Our results indicated contrasting aggregate dynamics depending on land-use systems. Under our experimental conditions, grassland returns larger amounts of C as belowground inputs than maize cropping, evidencing a different distribution of light C fractions between aggregate classes. Coarse aboveground inputs from maize contributed mostly to larger macroaggregates. Land-use changes with the introduction of ley grassland provoked a decoupling of the storage and/or degradation processes after the grassland phase. The newly derived maize inputs were barely degraded during the first 3 years of maize cropping, whereas grassland-derived material was depleted. As a whole, results suggest large microbial proliferation as shown by 13C NMR under permanent grassland, then reduced within the first years after the land-use conversion, and finally restored. The study highlighted a fractal structure of the soil, determining a scattered spatial distribution of the cycles of storage and degradation of soil organic matter related to detritusphere dynamics. As a consequence, vegetal inputs from a new land use are creating new detritusphere microenvironments that may be disconnected from the dynamics of C cycle of the previous land use. The formation of those different and unconnected microenvironments may explain the observed legacy effect of the previous land use, since each microenvironment type contributes separately to the overall soil C cycle. The effects of the new land use on the soil C cycle are delayed until the different detritusphere microenvironments remain unconnected, and the ones from the previous land use represent the predominant microenvironment type. Increasing knowledge of the soil C dynamics at a fine scale will be helpful in refining the prediction models and land-use policies.