Site description The study was conducted in the Katka-Kochikhali area within the Sundarbans mangrove forests of Bangladesh. The zone (Das and Siddiqi, 1985), because of the proximity of the Bay of Bengal, this area also experiences high tidal actions and high salinity (Rashid, 2007). Erosion and accretion combined with tidal surges lead to large scale variation in habitats within the mangrove forest (Rashid, 2007). Our study focused on three habitat types: (i) swamp forests, which receives regular tidal flooding twice a day, (ii) grasslands, i.e., raised land occupied by grasses;this area usually does not receive regular tidal flooding but it receives periodic flooding during high tides and (iii) sand dunes,raised land facing seawards. General vegetation and environ-mental parameters of the Sundarbans mangrove forests are described in Biswas et al. (2007), Iftekhar and Saenger (2008), Karim (1988) and Rashid (2007). Sampling protocol We identified all three habitats of the Katka-Kochikhali delta on maps and consulted with the local forest department for validation. Then, within each habitat type, we placed our quadrants randomly using a random number generator. We followed the species areacurve to determine quadrat size and used 10 m10 m quadrants for swamp forests and 5 m5 m quadrates for grasslands and sanddunes. Following this protocol, we sampled 40 quadrates in the swamp forest, 40 quadrates in the grasslands and 20 quadrates in the sand dunes. The unbalanced sampling design was due to logistical constraints. For a balanced design, we needed to sample grasslands and sand dunes either in another forest delta or in some habitats affected by human interference. However, by choosing sites with in one delta, we controlled the landscape level variation and also the potential variation due to anthropogenic activities. Within each quadrat we recorded all the plant species encountered and visually estimated their percentage cover. Our first author (SHR) identified all the plant species and collected sample specimens that were further verified and preserved at the Bangladesh National Herbarium (BNH), Dhaka. We followed the protocols of Tomlinson (1986) and Jayatissa et al. (2002) for nomenclature.Secondly, for the soil seed bank study, we selected four random quadrats from the swamp forest (out of 40), four quadrats from the grasslands (out of 40) and two quadrats (out of 20) from the sand dunes. Within each of the selected quadrats we collected two soil cores for detailed seed bank analysis. Therefore, our total soil seed bank sample of soil cores was 40 (i.e., 16 from the swamp forest, 16from grassland and eight from the sand dune). In the same quadrat,but at a different location within the quadrat, we sampled soil cores twice: in 2004 (January–February) and in 2005 (March–April). We used a PVC soil corer (diameter 8.7 cm) for collection of soil samples at a depth of 10 cm from the soil surface. This resulted in a total soil volume of 297 cm3per soil core, i.e., per sample. Seed bank analysis: germination chamber and green house experiment After the collection of soil cores from the forest, all soil samples were stored in the laboratory, with temperatures maintained above the freezing point. Shortly after, they were processed and analyzed at the Hohenheim University, Germany.We divided each soil sample into two halves. The first half was placed in a germination chamber. We sieved the soil samples using 0.5–2 mm sieves and separated the coarse seeds. The separated seeds were then placed inside a double layer of filter paper and sorted within a plastic box in the germination chamber at temperature of 170C to 230C, with aeration. We watered them as required. Many species possess tiny seeds and it was likely that they may have passed through the sieves. Therefore, we also placed the sieved soils in the germination chamber and watered them.These sets were treated at different temperature regimes rangingfrom80C to 350C, because, in a preliminary assessment of seeds mixed with soil, some seed showed germination success in a thermal gradient of 80C to 350C. The second half of the soil sample was directly placed in thegreenhousewithoutanysieving.Wetransferredthesoilsampleintosmallpotsandwateredthem.Allsmallpotswereplacedwithinabigtray. The greenhouse temperature was maintained between 203.350C during the night and 302.150C in the daytime. After a week, we observed germination both in the separated seeds and in the soil mixtures. We counted the number of germinated seedlings and identified them at 2-day intervals. This phase of the experiment continued for 6 months. While calculating species richness and density in a soil sample, we pooled germination results from the germination chamber, i.e., seeds that were directly placed and in soil mixtures and germination results from the greenhouse. For correct identification of the seed bank species, all the germinated seedlings were grown until they flowered. Thus, we transferred all the germinated seedlings from the germination chamber to the greenhouse. The full experiment in the greenhouse continued for 18 months. Statistical analyses In order to test whether species richness of the soil seed bank and aboveground vegetation differed significantly between three habitats, we conducted two separate non-parametric Kruskal–Wallis tests because of unequal replicates and violation of normality assumption, using habitat types as the grouping variable and species richness as the test variable. Quadrat sizes were unequal (i.e. 25 m2and 100 m2).