The study was carried in the forest concession allocated to Ouesso Forest Industry (IFO) in the Ngombe Forest Management Unit (FMU, Ngombe) in north of the Republic of Congo. The site presents a relatively flat relief with average altitudes between 400 m and 430 m [16] . The FMU Ngombe extends between 0˚27' - 1˚48'N and 15˚20' - 16˚38'E (Figure 1). The site is part of the semi-humid equatorial climate zone. It is characterized by two dry seasons and two rainy seasons per year [14] . Average annual rainfall and average temperatures are 1730 mm and 27˚C respectively.

The FMU Ngombe forests are semi-deciduous and installed on clay soils which develop on the alluvial deposits of the Congolese basin [16] . They are shelter many species typical of old secondary forests such as Entandrophragma cylindricum, Entandrophragma utile, Erythrophleum suaveolens, Triplochiton scleroxylon. A large part of FMU Ngombe is covered with Marantaceae forests [14] . These Marantaceae forests have a undergrowth invaded by giant herbaceous species of the Marantaceae and Zingiberaceae families, which can reach several meters height. The floor shrubby is poorly represented and displays a regeneration deficit [14] [16] .

The soil samples were taken in August 2020 (during the transition between the short dry season and the big rainy season) within two Marantaceae forests plots (plot 1, exploited in 2019, and plot 2, exploited in 2017). In each Marantaceae forests plot (1000 m * 500 m, area of one plot), three non-contiguous plots of 625 m² (25 * 25 m) were installed, at the rate of 6 plots for the two Marantaceae forests plots.

Inside each 625 m² plot, four sub-plots measuring 12.50 m were used for sampling points collection, i.e. a total of 12 collections points per plot (Figure 2). Four layers of soil were sampled per point: 0 - 5 cm, 5 - 10 cm, 10 - 15 cm et 15 - 20 cm. Each sample was composite, i.e. from the mixture of three unit samples taken on the vertices of an equilateral triangle with sides of 1 m [6] [8] [10] . Each sample concerned an area of 100 cm² (10 * 10 cm). The sampled area by Maranaceae forests plots covered accordingly 0.36 m².

The number of viable seeds in the soil samples has been estimated by the germination method [3] [10] [17] . The samples were sieved before the start of the experiment, in order to eliminate plant elements and minerals elements (leaves, smithereens of wood). Germination was realized in a greenhouse at the National School of Agronomy and Forestry (ENSAF) to Brazzaville, in a non-forest area in order to limit the risk of contamination of the samples by the surrounding vegetation.

Relative light intensity was about 30% - 40% of full light. Each soil sample was spread out on a layer of sterilized substrate-sand mixture about 1 cm thick [3] [10] [17] . Watering was daily, and germinations control was bi-weekly. When their stage of development allowed, the seedlings were withdrawned and repotted individually in polyethylene bags to promote their growth and identification.

As all the germinations took place over the course of the first 12 weeks, the experiment was stopped after 16 weeks (i.e. 4 months of experimentation).

Voucher specimens were collected and species were identified by botanists (Gilbert Nsongola, Isaac Dzombo). We followed the taxonomy of the Geneva Herbaria Catalogue (http://www.ville-ge.ch/musinfo/bd/cjb/chg).

The evolution of germination all layers mixed, of the soil seed bank within the two Marantaceae forests plots indicates that in plot 1 (exploited in 2019), the first germination was observed from the ninth week to reach peak at the twelfth and thirteenth week (Figure 2).

Nonetheless, during the last two weeks (twelfth and thirteenth), nearly 29 and 15 new seedlings were recorded respectively. The last germinations occurred from the eighteenth week (Figure 2).

However, in plot 2 (exploited in 2017), the first germinations occurred from the fifth week to reach peak at the twelfth and thirteenth week (Figure 3). We recorded respectively nearly 20 and 30 new seedlings. Germinations regressed globally from the fourteenth up to the sixteenth week with 7 new seedlings (Figure 3).

Nonetheless, the evolution of germinations in terms of soil layers within the two plots demonstrate that in plot 1 (exploited in 2019), the first germination occurred from the eighth week with the 0 - 5 cm layer to reach peak at the twelfth and fifteenth week (Figure 4). This layer is illustrated as the most representative in number of germinated seeds. Moreover, in the 15 - 20 cm layer, 7 germinated seeds were recorded during the eleventh and seventeenth week. The last germinations occurred from the nineteenth week with the two aforesaid layers (Figure 4).

On the other hand, in plot 2 (exploited in 2017), the first germinations occurred from of the fourth week with the 0 - 5 cm and 10 - 15 cm layers (Figure 5), and the germination peak occurred from the thirteenth week with the 0 - 5 cm layer. Also, the first germinations of the 15 - 20 cm layer were illustrated from the tenth week to fade around the seventeenth week with a overall number of 10 germinated seeds (Figure 5).

101 seedlings germinated from soil seed bank of plot 1 and 129 seedlings from plot 2, with respective average densities of 281 and 358 seedling/m². There was no significant difference in the mean number of seedlings between the two plots (Kruskal-Wallis test, p-value = 0.5938). In plot 1, the seedlings belonged to 17 species with 4 species of trees and shrubs (23%), 12 herbaceous species (71%) and one species of liana (6%). The tree and shrub species group only included one commercial species: Nauclea diderrichii (De Wild. & T. Durand) Merr. (Table 1).

On the other hand, in plot 2, the seedlings belonged to 15 species with one species of tree and shrub (7%), 11 herbaceous species (73%) and 3 undetermined species (20%). No commercial tree species was observed. The most abundant trees and shrubs in the soil bank of plot 1 were N. diderrichii (43%) and Harungana madagascariensis Lam. ex Poire. (29%), while plot 2 presented only the species Margaritaria discoidea (Baill.) G. L. Webster (14%).

Nonetheless, the results of the Principal Component Analysis (PCA) display a total variability of the two marantaceae forests plots of 25.56%. Consequently, the two Marantaceae forests plots globally share the same species (Figure 6). This observation to confirm Sorensen’s relatively low similarity value of 0.56% (Table 2).

All soil layers presented a significant number of seeds and taxa (Table 3). Regardless of the plot, the 0 - 5 cm and 5 - 10 cm layers present the highest values of the estimated species richness. Interesting fact, the deepest layer 15 - 20 cm of plot 2 also presented a non-negligible seeds abundance and species richness of 25% exclusive species (Table 3). Species found exclusively in the deepest soil layer were: Margaritaria discoidea (Baill.) G. L. Webster and Sabicea sp.

In the present study, we demonstrated that pioneer species of forest succession such as Margaritaria discoidea, Harungana madagascariensis, Nauclea diderrichii and Trema orientalis were present in the soil layers studied and much more abundant in plot 1 than in plot 2 which only presented the species M. discoidea. This could mean that this last one would be more mature than the first, which could be considered as a plot in process of recolonization.

All soil layers mixed, this study has demonstrated that in plot 1 (exploited in 2019), seeds germinated from the ninth week (i.e. about 2 months after germination of soil samples) to reach the germination peak from the twelfth week (i.e. about 4 months after germination of soil samples). No germination was recorded from the nineteenth week. On the other hand, in plot 2 (exploited in 2017), the seeds started to germinate from the fifth week (i.e. about 1 month after germination of soil samples) to reach the germination peak from the thirteenth week (i.e. about 3 months after germination of soil samples). From the seventeenth week, no germination was recorded. This slight difference germination times could be explained by the fact that in plot 2, exploited well before plot 1, the seeds stored in the soil bank would have lifted their dormancy, thus allowing enough rapid germination compared to plot 1.

The results of the present study differ in the logging company in southeast Cameroon (Pallisco), whose germination peak occurred from the fifth and seventh week. The last germinations occurred during the twelfth week [8] . Remember that this study was limited to the first five layers of soil. Likewise, germination peaks occurred from the third and fourth week to fade from the eighth week [30] .

Regarding germination from the soil layers of the two plots, the 0 - 5 cm and 5 - 10 cm layers were the first to germinate compared to others soil layers. This observation could be explained by the following facts: 1) the studies inherent to the soil seed bank tend to demonstrate that the floristic richness of the soil seed bank decreases rapidly with depth, whatever the biotope considered, and this, up to 50 cm deep [6] [10] [11] ; 2) the large majority of the soil seed bank is concentrated in the top 25 centimeters of soil, with greatest abundance in the top 5 or 10 centimeters [31] . Consequently, the seeds located in the soil layers superficial (0 - 5 cm and 5 - 10 cm) are exposed to light, the latter being a limiting factor for the germination of most species of tropical forests.

In several studies, the spatial heterogeneity of the tropical forest soil seed bank has been emphasized [7] [8] [10] [32] .

Nonetheless, in Cameroon during the study of soil seed bank characteristics at minimum depth (5 cm depth), 88 seedling/m² have been reported [8] . In mature forest and old secondary forest, we report densities ranging from 247 seedling/m² to 330 seedling/m², respectively [10] .

In the present study, the seed densities obtained were higher and ranged from 281 seedling/m² to 358 seedling/m² because soil layers up to 20 cm deep were considered. The estimates reported in the literature can reach much higher values. Seed abundance in tropical soil layers ranges from 25 to 3350 seeds/m² [31] .

This variation could also be due to the diversity of soil seed bank characterization methods, particularly sample size, season of soil sample collect, sample processing and depth of soils sampled [33] . The longevity of seeds in the soil bank could depend of their size and morphology. Long-lived persistent seeds are generally small size with a spherical shape [34] .

In the present study, herbaceous species dominated with varying between percentages 71% and 73%, respectively in plots 1 and 2. Previous studies have also demonstrated relatively similar percentages [8] [35] .

This can be explained by the following mechanisms: 1) herbaceous species produce a large amount of small seeds that naturally escape predation from the seed-feeding animals compared to large seeds; 2) these small seeds progressively accumulate in large quantities in the soil [36] and 3) they can survive for decades in the forester soil [10] . The tree species observed represented 7% and 23%, respectively in plots 2 and 1. In a similar forest in Cameroon, tree species such as: Erythrophleum suaveolens, Musanga cecropioides, Macaranga spp., Margaritaria discoidea, Tetrochidium didymostemon, Zanthoxylum spp. have been reported [8] . In the present study, the tree species found in the soil bank were: Harungana madagascariensis, Margaritaria discoidea, Nauclea diderrichii et Trema orientalis.

Only one tree species Margaritaria discoidea was found in the deepest soil layer of 15 - 20 cm. Margaritaria discoidea is frequent in wooded savannah, dry evergreen forest and dry secondary forest [37] . It is also found in semi-deciduous forest. Several parts of the plant are used in traditional medicine through out the African continent.

In West Africa, we chew filamentous and fibrous bark as purgative and aphrodisiac [38] . In north of Congo, it was reported that this species part of the indicator species of disturbed forests [14] .

Overall, it was observed that the abundance and the floristic diversity seeds were higher in plot 2 (exploited in 2017) than in plot 1 (exploited in 2019). Such a trend has been observed elsewhere [10] [39] , which could translate a fairly rapid recolonization of plot 2 compared to plot 1.

Regarding species richness, the number of observed species (Sobs) found in this study ranged between 15 and 17 species. These values were relatively lower to those observed in other African forest sites (11 to 43 specie) [8] [10] . However, the strong heterogeneity of the soil seed bank requires the use of species richness estimators instead of the number of species observed (Sobs) [40] .

The most surprising discovery in the present study seems to be siginficant decrease of seeds abundance and species richness along soil depth. In the 15 - 20 cm layer, significant amount of viable seeds (about 25%) can still be observed. However, this layer is rarely explored during studies inherent at the soil seed bank. Clearly, deeper soil layers should not be ignored during studies of the soil seed bank in tropical forests.