Our study site comprised an area impacted by bauxite tailings at the margins of Lake Batata (1°30′S, 56°20′W), which is located in the district of Porto Trombetas, municipality of Oriximiná, state of Pará, in the Brazilian Amazon. This lake lies on the right bank of the Trombetas, a clear water river. Clear-water rivers typically have high transparency, low quantity of organic and inorganic suspended material, a pH from low to neutral, low nutrient availability, and are bordered by igapó vegetation (see Prance, 1979 for terminology regarding Amazonian flooded forests).

As a result from bauxite mining in the nearby terra-firme forest, tailings were continuously discharged into the Lake from 1979 until 1989. When discharge was halted, ca. 600ha of lake (ca. 30% of the area of the lake) and marginal igapó (flood-prone) vegetation were buried by bauxite tailings. The bauxite tailings consist of red, muddy clay composed basically of iron and aluminum oxides and differ from native igapó soil in having a marked compaction, decreased porosity and infiltration capacity and reduced microbial activity (Parrotta and Knowles, 1999). The study site is subjected to eight months of flooding per year. This severe flooding regime represents a major challenge for restoration due to its impact on trees survival and growth.

On the study site, two restoration strategies were adopted. Passive restoration, where plants were allowed to freely colonize the substrate, and active restoration, where saplings of native igapó species were introduced. The areas of passive and active restoration were contiguous and had no differences in substrate, flooding regime and were not subjected to any management except for the sapling introduction in the active restoration area.

To record the natural colonization of the bauxite tailings, we set up a 1.12ha grid, consisting of 28 quadrats, each with an area of 20m×20m. In December 2003, we labeled and identified all plants within the grid, which were then measured for height and basal diameter. Diameter of each individual was measured at the closest height of soil level that did not show any stem deformity (e.g., root above-ground expansions). In the following two years, in December 2004 and 2005, all surviving plants were measured again, and newly germinated plants were labeled, identified, measured and added to the list. Growth was calculated as the absolute increase in basal diameter per year. Since we found out that all newly germinated plants sampled in the second and third years had a basal diameter equal or lower than 0.5cm, we classified all such plants, including those sampled in 2003, as regenerants. Plants with values higher than 0.5cm were classified as recruits.

We calculated the importance value (IV; Müller-Dombois and Ellenberg, 1974) for the recruited plants of each species for all years. This is a measure of species dominance, which is given by the sum of relative frequency, density and basal area. We considered the species that accounted for up to 75% of total IV to be the dominant species and those with IV smaller than 0.1 to be locally rare species.

In December 2003, we established a species introduction experiment in an area adjacent to the plant survey. Saplings were grown in a greenhouse with ad libitum water and nutrients. On the time of introduction in the field, saplings varied from 30 to 50cm depending on the species. Introduced plants were placed in holes of 0.15m of diameter, with 0.2m of depth, opened with 2m tall heavy wooden stakes. These plants received no fertilization or irrigation after planting. During the subsequent two years, saplings were controlled for survival and measured for height and basal diameter as described above.

We used Pearson correlation coefficient to test if species performance (survival and growth) during passive restoration was correlated to the performance during active restoration. Also, using paired t-test, we tested if species differ in survival and growth between passive and active restoration.

The number of species found in this plant community increased from 15 in 2003 to 32 woody species (27 trees, 1 shrub and 4 vines) in 2005 (Table S1). This was mostly due to a very marked increase in the number of regenerants in 2004 in relation to 2003 (5-fold) and in 2005 in relation to 2004 (4-fold). However, the number of recruited species remained rather constant (between 12 and 13). This is apparently more related to slow growth of these new-coming individuals than to mortality rates (only 3.5% for recruits as a whole over the two study years).

These 32 species recorded in the 2005 survey belonged to ten different families, and the Fabaceae was particularly species-rich (7 species, i.e. ca. 21% of the species total) (Table S1). There were no changes in the dominance structure of the community over the years (Table S2), since the dominant species (those that concentrated 75% of the IV) were the same (Buchenavia, Genipa, Dalbergia, and vine sp. 1).

There was an increase (60.6%) in recruit canopy cover between 2003 and 2005 (Table S2), which resulted mainly from a 15.5% rate of seedling recruitment (see Table S1), a very low mortality rate (3.5%), and an overall positive growth rate, particularly for the dominant species (Table 1).

However, only 2.12% of the area was covered by vegetation (canopy cover) after approximately ten years since this substrate consolidated, which is a clear indication that human intervention is needed.

From the species introduced in the experiment, five were also found naturally colonizing our study site (Table 1). There were no correlation on species survival (r=−0.14; p=0.82; df=3) and growth (r=0.04; p=0.94; df=3) between natural colonization and experimental introduction. There were no significant differences on growth (t=−0.36; p=0.74; df=4; paired t-test) and survival (t=1.24; p=0.28; df=4; paired t-test) between active and passive restoration.