Soil physical restrictions and hydrology regulate stand age and wood biomass turnover rates of Purus–Madeira interfluvial wetlands in Amazonia

In Amazonia, wetlands constitute about 30 % of its entire basin, of which ancient fluvial terraces located in vast interfluvial regions cover a large portion. Although the increased number of permanent plots in the recent years has contributed to improved understanding of regional variation in forest dynamics across the Amazon Basin, the functioning of large lowland interfluvial wetlands remain poorly understood. Here we present the first field-based estimate for tree ages, wood biomass productivity and biomass turnover rates for eight 1 ha plots in wetland and non-flooded forests distributed along the BR-319 Highway along a distance of about 600 km crossing the Purus–Madeira rivers interfluvial region in central-southwestern Amazon Basin. We estimate stand age, wood biomass productivity and biomass turnover rates combining tree-ring data and an allometric equation based on diameter, tree height and wood density and relate these structural parameters to physical soil and hydrological restrictions. Wood biomass and productivity varied twofold among the plots, with wood biomass stocks ranging between 138–294 Mg ha −1 and productivity varying between 3.4–6.6 Mg ha −1 yr−1. Soil effective depth, topography, structure and mainly soil water saturation significantly affected stand age (64–103 yr) and forest dynamics in terms of annual biomass turnover rates (2.0–3.2 %). On harsher soils characterized by a poor structure, low effective depth and high water saturation, biomass turnover rates were increased and forests stands were younger compared to welldrained sites. We suggest that soil constraints, especially soil water saturation, limit the development of the stand structure, resulting in forests with younger stand ages and higher biomass turnover rates compared to forests growing on welldrained soils. We do not find, however, any relation between physical soil restrictions or hydrology and wood biomass productivity, but there is a trend of increasing wood biomass productivity and phosphorus concentrations at the soil surface. Based on our results we establish hypotheses for different dynamical processes between forests growing on waterlogged and well-drained soils and discuss how these results can be applied in the background of conservation as well as the potential development of forest management plans in this region, which will experience increased deforestation due to the construction of the BR-319 Highway crossing the interfluvial region of the Purus–Madeira rivers.


Introduction
About two million square kilometers of the Amazon Basin falls under wetland classification criteria (Junk et al., 2011).Although the increased number of permanent plots in recent years has contributed to improved understanding of regional variation in forest dynamics across the Amazon Basin (Baker et al., 2004;Malhi et al., 2004;Girardin et al., 2010), most of the studies on forest dynamics focus on non-flooded terra firme forests.Only a few studies focus on forest dynamics of wetland forests such as the seasonally flooded forests along large rivers, mostly in terms of wood productivity (see Schöngart et al., 2010).Other types of forested wetlands, such as large interfluvial wetlands which are found along vast regions in the central and western Amazonia (Sombroek, 2001), are poorly studied and understood.
Little is known about how environmental factors such as soil conditions, climate and hydrology control forest dynamics in interfluvial wetlands.In these ecosystems soil drainage and the depth of the water table vary considerably, creating a patchwork of seasonally flooded or waterlogged and nonflooded forests.This process often leads to the formation of a firm plinthite layer that can change to hardpan if exposed to repeated wetting and drying cycles (Quesada et al., 2011).Therefore, the presence of soil physical constraints, such as the depth to which tree roots penetrate and the presence of restriction layers (e.g., hardpans) may be very important for stand development.In this study we focus on how the heterogeneous and complex hydrology and soil properties influence stand development and forest dynamics.
Forest dynamics expressed in terms of productivity, tree mortality and turnover rates is known to be affected by edaphic and/or climatic disturbances.Soil physical properties are determinant for tree turnover rates (Quesada et al., 2012), while soil fertility, especially phosphorous (P) status, is related with wood productivity (Malhi et al., 2004;Quesada et al., 2012).However, the responses of forest dynamics to climate vary between forest types (Schöngart et al., 2010).In non-flooded terra firme forests diameter growth is strongly related to seasonal precipitation (Brienen and Zuidema, 2005) and severe droughts negatively affect forest dynamics, decreasing wood biomass productivity and increasing tree mortality (Foley et al., 2002;Phillips et al., 2009;Corlett et al., 2011;Lewis et al., 2011).In contrast, the seasonal inundation of floodplain forests of the central Amazonia leads to a reduction of diameter growth (Schöngart et al., 2002), and an enhanced tree growth during the nonflooded period can be attributed to El Niño-induced droughts, since the terrestrial phase is extended during El Niño years compared to other years (Schöngart et al., 2004;Schöngart and Junk, 2007).
In this study we use tree-ring analysis showing how stand development may be affected by varying conditions of hydrology and soil physical constraints along the Purus-Madeira rivers interfluvial region in the central-southwestern Amazonia.In this region smooth topographic variations result in contrasting soil physical and hydrological conditions at local scale.We expect that forests on well-drained soils present a higher wood biomass productivity and wood biomass turnover than sites characterized by waterlogged, anoxic soils.The presented results are the first field-based estimates for tree ages, wood biomass productivity and biomass turnover rates in this vast interfluvial landscape.As predicted by future deforestation scenarios, the region will suffer huge impacts as a consequence of human occupation following the re-paving of the BR-319 Highway planned by the Brazilian Government (Laurance et al., 2001;Soares-Filho et al., 2006;Fearnside et al., 2009;Davidson et al., 2012).We discuss the application of the obtained scientific data on wood productivity, stand age and biomass turnover and their relationship to environmental conditions in the background of the sustainable development of the region to discriminate between areas for strict conservation or management priorities.

Study region
The study was carried out in the interfluvial region of the Purus-Madeira rivers in the Amazonas state, Brazil, that is intersected by the BR-319 Highway connecting the cities of Manaus and Porto Velho, the capitals of the Amazonas and Rondônia states, respectively (Fig. 1).The dominant vegetation type in the northern part of this region is dense lowland rainforests, while in the south a transition towards open lowland rainforests occurs, probably caused by increasing rainfall seasonality (IBGE, 1997).The predominant soil type for the whole interfluvial region is Plinthosol/Gleysol (Martins et al., 2013).On a local scale, frequent variations in the topography of a few meters create temporary pools on the lower and poorly drained areas during the rainy season (Rosetti et al., 2005).The duration and intensity of the rainy season varies strongly from north to south along the interfluvial area, declining from 2800 to 2100 mm annual precipitation with increasing rainfall seasonality (Table 1).

Sampling design
A total of eleven research modules of the PRONEX project "Rapid Assessment for Long Duration Ecological Projects" (RAPELD) was established along the Purus-Madeira interfluvial region as part of the research module network of the Research Program in Biodiversity (PPBio, http://ppbio.inpa.gov.br/sitios/br319).For this study, we selected four out of the eleven modules (Table 1).The average distance between the modules was 140-200 km, spanning an approximately 600 km distance (Fig. 1).In each module we choose two of the ten 1 ha (250 × 40 m) permanent plots previously installed: one at the highest altitude (well-drained) and the other plot at the lowest topography (poorly drained), using SRTM-DEM data (Shuttle Radar Thematic Mapper Digital Elevation Model, NASA Jet Propulsion Laboratory, US Geological Survey) (Miranda, 2005;Farr et al., 2007).Each plot followed an isocline to minimize variation in the topography and soil conditions within a plot (Magnusson et al., 2005).The RAPELD program records all trees with diameter at breast height (DBH) above 30 cm in the installed 1 ha plots.Trees with a DBH of 10-29.9cm are considered on two 10 m-large sections on both sides of the 250 m-long transect in the middle of the plot (0.5 ha) (Magnusson et al., 2005).

Soil data
Soil samples of 30 cm depth were extracted with an auger every 50 m along the central 250 m-long transect in all plots, totaling six samples per plot, and kept in sealed plastic bags for 2-5 days.On arrival at the laboratory, the samples were air dried at ambient temperature and composite samples were prepared, resulting in one sample per plot.Soil texture was then analyzed following standard protocol of total dispersion using sodium pyrophosphate to obtain clay, sand and silt percentages (Donagema et al., 2011).Chemical analysis concentration was also analyzed, following the protocol by Donagema et al. (2011).Soil physical constraints in each plot were scored using the classification index of Quesada et al. (2010) (Table 2), using soil samples of 2 m depth collected by the HIDROVEG project (http://ppbio.inpa.gov.br/sitios/br319).Additionally pits of 2 m depth were dug in modules M01, M05 and M08 for detailed soil descriptions.Soil physical constraints classified by Quesada's index are based on the effective depth of the soil (i.e.maximum depth that roots can attain), topography, soil structure and density as well as anoxic/hydromorphic properties.Anoxic condition (Table 2) is referred to in the present study as water saturation index.This index may be an important edaphic parameter as it appears to be related to vegetation parameters (Martins et al., 2013), and can be easily scored by identification of redox features at different depths.Superficial phosphorus concentration and soil water saturation index scores were available for all 55 plots spread in eleven modules in the whole interfluvial region.All soil samples were collected in cooperation with the HIDROVEG and RAINFOR project and analyzed at the Thematic Soils Laboratory of the National Institute for Amazon Research (INPA).

Topographical and hydrological conditions
The plots were preselected, based on SRTM data to capture a wide range of topographies within and between the selected modules.Afterwards we applied the Height Above the Nearest Drainage model (HAND) based on SRTM data that indicated the vertical distance of the plot in relation to the nearest water-table as an indicator for the hydrological conditions (Rennó et al., 2008;Moulatlet, 2010).We also used the soil water saturation index (Table 2) to indicate the terrain's hydrological condition.The difference between this index and the height above the nearest drainage is that the HAND data describe hydrology based on topography, and therefore will be most effective in terrains with pronounced topographical variations, since the level of the groundwater table may vary with the elevation depending on the soil conditions.The water saturation index, on the other hand, is based on soil features that were developed by long-term underground water fluctuations, such as patches of oxidation/reduction reactions and formation of plinthite hardpans (see Quesada et al., 2010), and is therefore a more reliable indicator for variations in soil water saturation of the study sites.

Field measurements
We considered all trees with DBH > 30 cm in the forest inventory (Schietti, 2013) in order to obtain data on wood densities as well as to estimate ages and diameter increment rates for each tree by tree-ring analysis.Trees with DBH > 30 cm were selected in ascending order of their numbers in the inventory, in order to select trees spread along the plot.Further, we sampled 30 randomly selected trees within the DBH classes 10-29.9cm in each plot.Palm trees (Arecaceae) were not considered in this study, since they do not present tree rings.In the studied plots palm trees comprise 9.8 ± 7.9 % of all individuals and 10.2 ± 8.2 % of the total basal area.
The DBH of all trees was measured by a diameter tape.In case of buttresses, diameter was measured above them to avoid overestimates of basal area and wood biomass.Tree height was estimated using a height measurement device (Blume Leiss BL6) and meter tapes to measure the distance between the tree and the observer.Two wood samples were extracted from the trunk of each tree 10 cm below the DBH (120 cm above the forest floor) to avoid errors for future repeated diameter measurements with the aim to monitor forest dynamics, using an increment borer of 5.15 mm internal diameter.On buttressed trees, coring was done between the buttresses whenever was possible, to avoid sampling biased rings.When it was not possible, the buttressed tree was not sampled.One sample was extracted for wood density determination and was labeled and stored in closed plastic bags to avoid dehydration for three to nine days before analyzing.The second sample was glued on labeled wooden supports for later tree-ring analysis.All wood samples were transported to the Dendroecological Laboratory of the scientific cooperation between INPA and MPIC (Max Planck Institute for Chemistry) in Manaus for further analyses.

Botanical data
All listed species were identified by Priscila Souza and Carolina Levis.After a preliminary identification, the botanical determination was confirmed with the aid of specialists, field guides and by comparing the collected vouchers with specimens at the INPA herbarium (Manaus, Brazil) and virtual herbariums (http://sciweb.nybg.org/science2/vii2.asp).Fertile specimens were deposited at INPA and sterile material will be stored in an adjacent working collection.Plants were identified in accordance with the APG III (Angiosperm Phylogenetic Group III) classifications.For name correction of the taxa the Brazilian Flora Species List was consulted (http://floradobrasil.jbrj.gov.br/2012/).

Biomass and carbon stock estimates
Information on the total basal area and number of individuals per hectare for the diameter classes 10-29.9cm and above 30 cm was obtained from forest inventory data (Schietti, 2013).Since trees in the RAPELD program protocol with DBH of 10-29.9cm are sampled only on 0.5 ha (Magnusson et al., 2005), information on basal area for this diameter class was multiplied by factor two to obtain estimates for the entire hectare.To calculate wood density (ρ) the fresh volume (V fresh ) of each sample was determined by the water displacement method.The object is mounted on a needle and immersed into a recipient filled with water on an analytic balance calibrated to zero.The volume (in mL) of the object is equal to the indicated weight (in g) of the displaced water after submerging the object entirely into the water without touching the side or the bottom of the recipient.After this step of analysis the samples were dried at a temperature of 105 • C to obtain their dry weight (W dry ) (Chave et al., 2005;Schöngart et al., 2005).Wood specific density was than calculated as For each plot we calculated the mean wood density and standard deviation.Since there are no specific allometric models available for the studied forest types we chose an allometric model that uses diameter, tree height and wood density as independent parameters to enhance the quality of the AGWB estimates (Chave et al., 2004(Chave et al., , 2005;;Schöngart et al., 2010Schöngart et al., , 2011)).This way we may reduce any biases caused by possible differences in tree height and wood density among the areas, that can be expected over large distances (Chave et al., 2005;Wittmann et al., 2006;Nogueira et al., 2008a, b;  Feldpausch et al., 2011) due to vegetation type shifts from the north to the south within the interfluvial region.We used the model developed by Feldpausch et al. (2012) (Eq.2), reporting the relative error estimated by the authors for the use of power law regional model for estimating tree height.For the equation described below, the parameters are referred to as aboveground coarse wood biomass (AGWB, in kg), diameter at breast height (DBH, in cm), tree height (H, in m), and wood density (ρ, in g cm −3 ).
As not all trees of the plot were sampled we estimate the AGWB in relation to the basal area of the sampled trees.This was performed separately for the two diameter classes DBH > 30 cm (> 30) and trees with DBH 10-29.9cm (< 30).
For this study we assume that the carbon content (C % ) in the AGWB is about 47 % (Kirby and Potvin, 2007).

Tree-ring analysis and growth modeling
We used tree-ring analysis to obtain reliable estimates for tree ages and mean diameter growth rates which have successfully been applied to estimate wood biomass productivity in different forest types of central Amazonia (Worbes, 1997;Stadtler, 2007;Oliveira, 2010;Schöngart et al., 2010) and also in the Pantanal wetlands (Schöngart et al., 2011).The prepared wood samples for tree-ring analysis were sanded and polished to produce a plain surface that enabled the visualization of the annual rings based on macroscopic wood anatomical analysis.Wood anatomy of tree rings was characterized following Worbes (2002) (Fig. 2).The anatomy of tropical wood may generate uncertainties due to indistinct, false or missing rings that cannot be detected when analyzing 5.15 mm cores.These uncertainties are reduced when considering the average of large number of rings for estimating mean radial increment and productivity.Since we use only the last five rings to estimate tree productivity, samples with indistinct tree rings were not included in the estimate of aboveground wood biomass productivity (AGWBP).Ring widths were measured using a digital measuring device (LINTAB) with 0.01 mm precision attached to a computer with the software Time Series Analysis and Presentation (TSAP-WIN) to determine mean radial increments (Schöngart et al., 2005).On samples which contained the pith, tree age was estimated by direct ring count.For samples with missing pith we estimated tree age by dividing the obtained average diameter increment rates by the measured DBH in the field.The mean tree age per plot was calculated as the average of the ages from all sampled trees with DBH > 30 cm and with DBH of 10-29.9cm, weighted by their relative number of individuals (Ind > 30 and Ind < 30 ) in the plot: where AGE > 30 and AGE < 30 represent the average of the ages calculated per diameter class, Ind > 30 and Ind < 30 represent the number of individuals per diameter class in the plot and Ind total represent the number of all individuals with DBH > 10 cm in the plot.Non-linear regressions were carried out between DBH and tree height using potential equations to produce DBH-height relationships for each plot and module.The relationship between tree age and DBH was fitted to non-linear regression models (Schöngart, 2008): (4)

Estimation of stand wood biomass productivity and biomass turnover rates
To estimate the AGWBP the cumulative diameter growth curve of a tree was combined with the stand-specific DBHheight regression model and the wood density to calculate AGWB by the allometric model of Eq. ( 2) for every year of the entire life span of the tree (Schöngart et al., 2011).We then estimated the current aboveground wood biomass productivity of each tree (AGWBP tree ) calculating the average of the difference between the AGWB for the last five consecutive years (t) (Eq.5): where t is the year corresponding to the last ring formed by the tree.As not all trees of the plot were considered for this analysis we estimated the productivity in relation to the basal area for all sampled trees to indicate the current total productivity of the plot.This was performed separately for the two diameter classes DBH > 30 cm (> 30) and trees with DBH 10-29.9cm (< 30) (Eq.6).
Biomass turnover rates in percentage were calculated by dividing AGWBP by AGWB for each plot: Data on stand age, biomass turnover and wood productivity were related to the scored physical constraints indexes using linear regressions (Quesada et al., 2010), considering that the referred indices are semi-categoric.Data analysis and graphs were performed using R (R-Core Development Team, 2013) software packages stats and graphics.

Forest structure
A total of 554 trees were cored in the eight 1 ha plots, from which 20 individuals were not considered for tree-ring analysis due to very low distinction of the growth rings.The sampling effort of coring per plot was 47 to 87 trees with DBH > 30 cm per hectare, corresponding to 53-96 % of all individuals in this diameter class (Table 3).Tree height was measured for a total of 627 trees, as well as DBH.Tree diameter explained 56-62 % of the variability in tree height (Fig. 3), with exception of module M08 where the multiple R-squared was only 0.31 (p < 0.01) (Fig. 3c).Among the most abundant tree species wood density varied from 0.44 ± 0.11 g cm −3 (Apeiba echinata) to 0.90 ± 0.05 g cm −3 (Licania oblongifolia) (Table 4).Mean wood density of the 20 individuals that were not considered for tree-ring analysis due to very low distinction of the tree rings was 0.49 g cm −3 and 0.46 g cm −3 in plots M01_TN_1500 and M11_TN_1500, respectively, and varied between 0.68 g cm −3 and 0.84 g cm −3 in the other plots.Only one individual was not considered in the plot M11_TN_2500, with a wood density of 0.91 g cm −3 .
The seasonally flooded sites at M01 presented a much lower basal area (16-17 m 2 ha −1 ) compared to the other sites (21-28 m 2 ha −1 ) (Table 3).We also observed lower canopy heights (24-25 m) and lower mean wood densities (0.64-0.65 g cm −3 ) in the seasonally flooded forests of M01 compared to the other sites with canopy heights of 25-28 m and mean wood densities of 0.65-0.72 g cm −3 .

Tree ages, radial increment, forest productivity and biomass turnover
Maximum estimated tree age based on counted tree rings including pith was 214 yr.Another individual presented 239  rings without reaching the pith, resulting in an age estimate of more than 500 yr, extrapolating the mean radial increments in relation to the mean radius of the tree trunk.By this calculation, we estimated that from all the 534 analyzed trees, 33 would be more than 200 yr old.The nonlinear relationship between DBH and tree age from all plots was significant (n = 534, df = 532, F = 108.65,R 2 = 0.29, p < 0.01) (Fig. 4a).However, the strength of correlation was not sufficient to consider the model as a predictor of tree ages in the field only by measuring DBH, independent of the species.Only about 30 % of the variability of tree age among different species and varying growth condition can be explained by DBH, due to differences in growth rates between species and individuals of the same species growing under different environmental conditions.The lowest mean annual radial increments were observed for Pseudolmedia laevis (1.1 ± 0.2 mm yr −1 ) and Eschweilera coriacea (1.1 ± 0.4 mm yr −1 ), while A. echinata presented radial increment rates of 3.4 ± 2.4 mm yr −1 (Table 4).On the stand level mean wood density was negatively related to the mean radial increments (n = 8, df = 7, F = 11.12,R 2 = 0.65, p < 0.05) (Fig. 4b).
Estimates of AGWB resulted in 230 Mg ha −1 with a standard deviation of 65 Mg ha −1 for the whole study region (Table 3).The lower basal area, canopy height and mean wood density in the M01 stands resulted in lower AGWB estimates of 136-138 Mg ha −1 compared to 205-291 Mg ha −1 at the other sites (Table 3).In relation to the basal area of the stand the AGWB stocks varied 7.9-12.3Mg m −2 basal area .The AGWBP differed from 3.4 Mg ha −1 yr −1 to 4.0 Mg ha −1 yr −1 in the seasonally flooded forests of module M01 and was much higher in the other modules ranging between 5.5-6.6 Mg ha −1 yr −1 .The estimated C-sequestration varied between 1.6-1.9Mg C ha −1 yr −1 in the seasonally flooded forests of module M01 and between 2.6-3.1 Mg ha −1 yr −1 in the other plots.Biomass turnover varied between 2.0-3.2 % with higher turnover rates in the seasonally flooded and waterlogged sites (2.4-3.2 %) and lower rates at non-flooded, well-drained sites (2.0-2.2 %) (Table 3).The estimated stand age was lower at seasonally flooded and waterlogged sites (65-96 yr) compared to forests on well-drained sites (94-103 yr).

Soil properties
The soil texture of all plots was mainly composed of silt (52.8 ± 18.8 %), with varying smaller percentages of sand (27.7 ± 15.7 %) and clay (19.4 ± 6.0 %) (Table 5).Available superficial phosphorus concentrations varied between 1.4 and 6.3 mg g −1 soil , with the higher values being found in the flooded forests M01 which might be related to the seasonal inundation by river water.concentration (Fe + ) ranged from 66 to 389 mg g −1 soil .The soil water saturation index varied at the seasonally flooded and waterlogged soils by 3-4 and in the well-drained soils by 1-2.The scored soil physical constraint index varied in well-drained soils between 4 and 8 and reached and index of 8-10 in the seasonally flooded and waterlogged soils (Table 3).

Relationship between forest dynamics and environmental conditions
Forest dynamics, in terms of coarse wood biomass productivity, biomass turnover and stand age, is here hypothesized to vary with soil structure and hydrology.We found no relation between AGWBP and the scored soil physical constraints index (n = 8, df = 6, R 2 = 0.15, F = 1.04, p = 0.34).On the other hand, the soil physical constraints index was positively related to biomass turnover rates (Fig. 5a) (n = 8, df = 6, R 2 = 0.62, F = 9.81, p < 0.05) and negatively related to stand age (Fig. 5b) (n = 8, df = 6, R 2 = 0.64, F = 10.92,p < 0.05).Among all the soil features that compose the soil physical constraints index, the soil water saturation seems to be an important parameter, showing a strong relation with both biomass turnover and stand age.The soil water saturation index was positively related to the biomass turnover rates (Fig. 5c, n = 8, df = 6, R 2 = 0.79, F = 23.31,p < 0.01) and negatively related to stand age (Fig. 5d, n = 8, df = 6, R 2 = 0.70, F = 13.71,p < 0.01).No significant relationships were found between the soil water saturation index and AGWBP (n = 8, df = 6, R 2 = 0.03, F = 0.20, p > 0.05).We find no correlation between AGWBP, biomass turnover rates or stand age and parameters of soil texture and chemistry.However, since flooding at the M01 module is influenced by large river waters, soil chemical properties at M01 do not follow the same patterns as in other modules.Therefore, excluding the M01 plots, we noticed a very strong correlation between AGWBP and available phosphorus in the soil (Fig. 6a) (n =, df = 4, R 2 = 0.77, F = 13.21,p < 0.05).We find no relation between the water saturation index and the phosphorus concentration in our plots, but on a regional scale, there is a positive correlation between the concentration of available phosphorus and the water saturation index (Fig. 6b, n = 42, df = 40, R 2 = 0.16, F = 7.76, p < 0.01).

Discussion
The presented data on tree ages, diameter growth, wood biomass productivity and biomass turnover rates indicate stands of varying forest dynamics in the studied region.The AGWBP of the studied forests varies considerably between 3.4 and 6.6 Mg ha −1 yr −1 (Table 3) and biomass turnover rates differ by more than 50 % between the plots.We find that soil physical constraints considering soil depth, soil structure and soil water saturation significantly affect forest dynamics in the interfluvial landscape.Harsher soil physical conditions characterized by shallower effective depth, higher density and especially more water saturation, resulting in higher by constraint scores, enhance biomass turnover rates and forest dynamics, resulting in younger forest stands stocking on www.biogeosciences.net/10/7759/2013/Biogeosciences, 10, 7759-7774, 2013  seasonally flooded and waterlogged soils.This makes sense when we think that accelerating biomass turnover rates reduces the time for late successional development, resulting in forests with younger stand ages composed by trees species with shorter life spans.So far we do not have data on forest dynamics related to tree turnover considering mortality and recruitment rates, but our results show that the soil constraints, especially soil water saturation properties, accelerate forest dynamics in terms of higher biomass turnover.However, in terms of aboveground wood biomass productivity, we cannot find relationships with physical soil properties or the water saturation index as we expected by our working hypothesis.This is strange at first, since waterlogged sites with very evident hydromorphic soil features (higher water saturation index) indicate alternating waterlogging and drying very common in Plinthsols (Quesada et al., 2010(Quesada et al., , 2011)), which may reduce oxidized iron atoms, making phosphorus available for plants (Chacon et al., 2006) and thus enhancing tree growth.This mechanism may be valid for the study region as we observe a trend between wood biomass productivity and plant-available phosphorus as well as the relation between the regional variation of available phosphorus and the water saturation index (Fig. 6).Therefore, we expected that enhanced hydromorphic features would result in more productive forests, since the water saturation index correlates with stand age and biomass turnover.On the other hand, these stands stocking on poorly developed soils (very hydromorphic, dense and shallow) present higher tree densities with smaller diameters when compared to forests growing on well-drained soils (Schietti, 2013), probably resulting in lower productivity by structural limitation.
Therefore we consider that soil constraints, especially water saturation, limit the structural development of the forest, maintaining a state of accelerated dynamics in terms of biomass turnover rates and reduced stand age.We suggest that harsher soil conditions may act as an environmental filter, creating a trade-off for fast-growing trees with shorter life-cycles, especially in the flooded or waterlogged sites, as in young successional stages of central Amazonian floodplain forests (Wittmann et al., 2006(Wittmann et al., , 2010) ) and the bottomland flooded forests at the terra firme (Ferry et al., 2010).In this case we would expect lower wood densities at waterlogged sites and higher wood densities on well-drained soils characterized by slow-growing tree species (Gourlet-Fleury et al., 2011).Despite the low variation of mean wood density among our plots, we find a significant negative relationship between mean radial increments and wood density between the studied sites supporting this hypothesis.
Our results are very similar to those observed by Ferry et al. (2010) in French Guiana.They observe along a gradient from non-flooded hilltops down to seasonally flooded bottomlands a decrease of basal area, tree height and biomass storages and an increase in tree mortality, but only small differences in wood biomass productivity.The authors argue that bottomland areas, besides being more fertile, present lower biomass due to higher mortality rates associated to a poorly developed stand structure of the waterlogged soils.They discuss that this would favor fast-growing and lightdemanding species in the bottomland areas, compensating the wood biomass productivity.Our study area is very different from the well-known hilltop-bottomland terra firme forests, mainly because the nearest drainage is vertically very close to the surface even at the well-drained areas.But this mechanism might be also valid for the interfluvial landscape.Since increased soil constraints lead to higher biomass turnover rates, however, the poorly developed stand structure cannot compensate for the wood biomass productivity.
We raise the hypothesis that tree species in the studied interfluvial region characterized by complex hydrological conditions present varying climate-growth relationships between waterlogged and well-drained soils.This is a hypothesis which has to be tested in future studies.General global climate models predict a large impact for the Amazon Basin and its carbon stocks and uptakes, mainly due to shifts in the precipitation patterns and hydrological regimes (Dufresne et al., 2002;Betts et al., 2004;Cook et al., 2012).Changes in the Amazon hydrological cycle related to climate are mainly caused by the anomalies of sea surface temperature (SST) in the equatorial Pacific (Sombroek 2001;Foley et al., 2002;Marengo, 2004;Schöngart and Junk, 2007) and the tropical northern Atlantic (Marengo et al., 2008;Tomasella et al., 2011;Yoon and Zeng, 2010).The droughts of 2005 and 2010 were caused by SST anomalies of the tropical northern Atlantic Ocean (Marengo et al., 2008;Lewis et al., 2011) Table 6.Comparison between estimates of AGWB (in Mg ha −1 ) and AGWBP (in Mg ha −1 yr −1 ) of old-growth forests from this study, with estimates from nutrient-rich (várzea) and nutrient-poor (igapó) floodplain forests, and non-flooded sites (paleovárzeas on ancient fluvial terraces and terra firme).The studies of Stadtler (2007), Oliveira (2010) and Schöngart et al. (2010) applied for the estimates of AGWB and AGWBP allometric models considering three independent parameters (diameter, tree height and wood density), while Chambers et al. (2001) and Malhi et al. (2004Malhi et al. ( , 2006) use allometric models with only one (diameter) and two independent variables (diameter, wood density), respectively.SFC (in Mg m −2 BA ) is the structural conversion factor relating AGWB to 1 square meter basal area (BA).Values are presented as means and standard deviation (with exception of those from Quesada et al., 2012) and affected between 2.5 and 3.2 million square kilometers in the Amazon Basin, respectively, including the interfluvial region between the Purus and Madeira rivers (Phillips et al., 2009;Lewis et al., 2011).Applying conventional tree-ring analysis and stable isotopes analysis for characteristic tree species from different ecotypes (evergreen and deciduous species) between well-drained and waterlogged soils in the interfluvial region, we predict that the vegetation period of tree species varies temporarily between different forest types depending on the soil physical constraints and the soil water regimes such as it was observed between floodplain forests and adjacent terra firme forests in central Amazonia with a shift of about 3 months between the growth rhythms of trees between both ecosystems (Schöngart et al., 2004(Schöngart et al., , 2010)).As expected our results indicate a varying wood biomass productivity of 5.6 ± 1.1 Mg ha −1 yr −1 in the interfluvial landscape which seems to be more related to stand structure as to environmental parameters.In Table 6 we compare the results of AGWB and AGWBP in this study with other studies in different regions and forest ecosystems of the Amazon Basin.The AGWB stocks in the studied forests of the interfluvial landscape of the Purus-Madeira region are lower than in terra firme forests, but higher than in the nutrient-poor central Amazonian floodplain forests (igapó).Comparing the wood biomass productivity, our studied forests indicate a similar AGWBP as paleovárzeas (5.4 ± 0.7 Mg ha −1 yr −1 ) growing on ancient fluvial terraces (Irion et al., 2010) and the nutrient-rich várzea floodplain forests (5.7 ± 0.7 Mg ha −1 yr −1 ).In comparison to the seasonally flooded igapó (3.6 ± 1.2 Mg ha −1 yr −1 ) and the central Amazonian terra firme forests (4.1 ± 0.7 Mg ha −1 yr −1 ) our study indicates a higher AGWBP.Only the southwestern Amazonian terra firme forests with 7.0 ± 1.5 Mg ha −1 yr −1 seem to be more productive, however, as Malhi et al. (2004) applied allometric models, which do not consider tree height for the estimates of AGWB and AGWBP, leading to poten-tial biases in the biomass estimates (Schöngart et al., 2010;Feldpausch et al., 2012).Still, compared with recent estimates of the range of wood biomass productivity for Amazonia (2.7-10.3Mg ha −1 yr −1 ) by Quesada et al. (2012), the forests of the interfluvial Purus-Madeira region present an above-average wood biomass productivity.
The obtained results on tree and stand tree ages, mean radial increment rates of dominant tree species and wood biomass productivity, and turnover rates of forests are essential information for the sustainable development of this particular region.Future scenarios predict increased deforestation rates in the next future in this interfluvial region due to the reconstruction and paving of the BR-319 Highway and following settlements and transformation of forests in lands for pasture and agriculture (Laurance et al., 2001;Soares-Filho et al., 2006;Davidson et al., 2012).The State Secretariat for the Environment and Sustainable Development (SDS) of Amazonas state and the Ministry for Environment (MMA) of the federal government implemented various areas of different protection categories within the interfluvial region to control land-use changes and to mitigate the emission of greenhouse gases caused by deforestation (Fearnside et al., 2009) as it is proposed by the United Nations collaborative initiative on reducing emissions from deforestation and forest degradation (REDD).The subsequent REDD+ initiative also considers the carbon stocks of intact forests in the background of conservation and sustainable management.However, the creation of conservation units requires a thorough scientific data base for designing a controlled monitoring of an ecologically and socio-economically sustainable development of this region.In this study we provide reliable estimates of stand ages, forest productivity and wood biomass turnover in relation to environmental factors, which is basic information, and criteria to decide which strategies regarding forest management and protection could be applied in this region considering land-use and climate changes.We suggest that forests on waterlogged sites should be excluded from any management practices since they present a limited structural development due to poorly structured soils.Forest management activities should be restricted to well-drained soils with more developed forests with a similar or even higher AGWBP of 5.6-6.5 Mg ha −1 yr −1 than those forests on waterlogged or seasonally flooded soils.Furthermore, the tree-ring data can subside the initial development of applying site-specific and species-specific minimum logging diameter and cutting cycles for timber species in these environments as suggested by the GOL (growth-oriented logging) concept for the várzea floodplain forests (Schöngart, 2008) also considering regeneration dynamics and population structure.

Fig. 1 .
Fig. 1.Map of the Purus-Madeira interfluvial area that is crossed by the BR-319 Highway.The study sites are indicated by white asteriks.All study sites are sampling modules (M) of the PRONEX-Project "Rapid Assessment for Long Duration Ecological Projects" (RAPELD).

Fig. 2 .
Fig. 2. Examples of distinct tree rings.The classification of wood anatomy of the tree rings is described following Worbes (2002).(a, b) Parenchyma bands limiting ring boundaries and vessel size/distribution variation from earlywood to latewood; (c, d) alternating fiber and parenchyma bands, large in the earlywood and narrow in the latewood; (e, f) marginal parenchyma bands limiting the tree rings.Arrows indicate the boundary of the tree rings.

Fig. 3 .
Fig. 3. Non-linear relationships between tree diameter at breast height and height.M01, M05, M08 and M11 represent each sampling module for which the models were developed.

Fig. 4 .
Fig. 4. (a) Significant non-linear relationship between the number of counted rings and length of the core samples for all analyzed trees with distinct growth rings and (b) significant correlation between estimated mean radial stem growth and mean wood density of the eight 1 ha plots.

Fig. 5 .
Fig. 5. Significant relations between the soil physical constraints index and (a) biomass turnover rates and (b) stand age, as well as the relation between soil water saturation index and (c) biomass turnover rates and (d) stand age.

Fig. 6 .
Fig. 6.(a) Significant relationship between available phosphorus at the soil surface and total aboveground wood productivity (AGWBP) excluding M01 plots, and (b) significant relationship between regionally available phosphorus at the soil surface and soil water saturation index, with triangles representing the eight sampled sites of this study.

Table 1 .
(Rennó et al., 2008)n, hydrology and rainfall patterns of the selected study sites.Elevation above sea level (a.s.l.) is estimated by SRTM data (Shuttle Radar Topographic Mission).Data of the HAND model (Height Above Nearest Drainage) indicate the distance and elevation above the nearest drainage(Rennó et al., 2008); number in parenthesis indicates the flood height of seasonally inundated forests measured in the field.Soil water saturation conditions are described and presented in more detail in Table2.Rainfall data were obtained from the Brazilian Water Agency (Agência Nacional de Águas -ANA).Length of dry season refers to number of months with less than 100 mm of rainfall.

Table 2 .
Quesada et al., 2010)cal soil constraints (withdrawn and adapted fromQuesada et al., 2010).Anoxic conditions were referred to as "water saturation index" in the present study.
Seasonally flooded; soils with high clay content and very low porosity and/or dominated by plinthite Deep saturated zone (maximum height of saturation 50 cm depth); redox features Deep saturated zone (maximum height of saturation > 100 cm depth); deep redox features

Table 3 .
Number of sampled trees from the diameter classes 10.0-29.9cmand> 30 cm in each plot.Stand structure is described by basal area as well as by means and standard deviation of canopy height and wood density.Estimates of AGWB and C-stocks in AGWB and the structural conversion factor (AGWB SCF related to m 2 basal area); σ AGWB indicates the error of the AGWB estimate.Furthermore for each plot, information on aboveground coarse wood productivity (AGWBP) and carbon sequestration per hectare and year (C-sequestration) are indicated, as well as the annual biomass turnover.The soil conditions are indicated by the soil physical constraints and the soil water saturation indices (see Table2for scoring).

Table 4 .
Mean value and standard deviation of wood density (ρ) and mean radial increments (MRI) for characteristic tree species with five or more individuals in the study sites.

Table 5 .
Chemical and physical soil properties obtained from 30 cm depth of composited soil samples for each study site.Numbers in parentheses indicate values obtained for 10 cm soil depth.