Tropical forests are currently experiencing huge pressures from global changes. If tropical deforestation has been rightly the focus of much attention these last decades, forest degradation and its consequences in the tropics has been much less studied and quantified. Yet, forest degradation is pervasive throughout the tropics and leads to a significant loss of ecosystem services. Hence, a major issue is the reversibility of degradation. In many cases, natural successional processes are expected to bring back the system to a state similar to the initial one. However, in some cases, the system may have trespassed a threshold and lost its ability to follow such a classical successional pathway; it may remain in a so-called blocked succession or in an alternative degraded stable state. The Marantaceae forests from Central Africa likely correspond to such stable degraded systems. They cover very large areas in Central Africa and have a very low density of trees with a dense understory composed of giant herbs. The few studies on Marantaceae forests suggest that their large patches (up to 2000 km2) are likely to have originated from old disturbances (>1000 years ago) and have been maintained over long periods through positive feedback mechanisms, e.g. inhibition of tree regeneration by giant herbs. Observations also suggest that current human disturbances, such as logging activities, as well as climate anomalies, contribute to the rapid expansion of these degraded forests, which could have considerable consequences for local populations. In this project, we will study the mechanisms by which Marantaceae forests may originate and be maintained at different spatial and temporal scales. We will combine observations at the local and regional scales (among which historical data) and parsimonious theoretical models to study the long-term dynamics of Marantaceae forests, the conditions under which stability is expected and the mechanisms by which giant herbs monopolize space and may outcompete trees or restrain their development. The project will be structured in four complementary main work packages (WP): WP0 will be dedicated to synthesize previous studies conducted on Marantaceae forests, of which many were reported as grey literature; WP1 will aim at investigating the main ecological mechanisms underlying the dynamics of Marantaceae forests; WP2 will aim at depicting the contemporary (<100 yrs) and long term (> 500 yrs) spatio-temporal dynamics of Marantaceae forests based on remote sensing data and historical ecology approaches; and WP3 will consist in devising mathematical models to understand the conditions of a post-disturbance stability of Marantaceae forests and the relative importance of the drivers of this stability. WP3 will thus bridge the scale gap between the local ecological mechanisms (WP1) and the broad scale spatio-temporal dynamics (WP2) of Marantaceae forests. Overall, our project will generate important knowledge on the dynamics of a system that constitutes an important part of the tropical African wet forests and ofglobal concern regarding biodiversity and carbon sequestration in the second largest tropical forest area in the world. It will also contribute to the increased awareness that the dynamics of ecosystems is not systematically reversible and that external forcings may make them shift to alternative stable states. The ultimate goal of the project is to anticipate the potential consequences of global changes on the dynamics of Marantaceae forests and to provide recommendations for forest conservation and management.