Questions: Does the presence of salt marsh vegetation affect the long-term regeneration of the pioneer mangrove species Avicennia germinans in a degraded dwarf forest? Does immobilized coarse woody debris (CWD) affect regeneration similarly? Do larger trees suppress or facilitate intraspecific saplings?. Location: Dwarf mangrove forest in the high intertidal zone on the Bragança peninsula in northern Brazil. Methods: The spatial patterns of A. germinans, the herbaceous halophyte Sesuvium portulacastrum, and CWD were mapped in three sample plots (each 400 m²) during two consecutive vegetation surveys, conducted in 2011 and 2014. Inhomogeneous Poisson and Thomas point process models were used to assess the distribution of A. germinans life-history stages (seedlings, saplings, and adult dwarf trees), conditioned on the presence of S. portulacastrum and CWD. In addition, intraspecific interactions between trees and regeneration were assessed based on crown projection mapping. Bivariate point pattern analyses were used to assess the dependence of advance regeneration on dwarf A. germinans trees and S. portulacastrum. Results: Avicennia germinans saplings and trees were positively associated with S. portulacastrum and CWD, whereas seedlings were located around tree crowns. The density of fruit-bearing trees was positively associated with sapling density, indicating that regeneration relied on locally dispersed propagules. Conclusions: Herbaceous vegetation and CWD have an important ecological function in degraded mangroves by retaining tidally dispersed propagules. Here, we show that herbaceous vegetation does not suppress the growth of seedlings but facilitates mangrove recolonization. Due to limited tidal dispersal, regeneration relies on local propagule supply. In addition to hydrological restoration, the observed vegetation patterns suggest that, in the absence of propagule-retaining vegetation, restoration of high-intertidal mangroves can be facilitated by establishing nuclei of planted trees and installing secured logs.