Windthrows change forest structure and species composition in central Amazon
forests. However, the effects of widespread tree mortality associated with
wind disturbances on soil properties have not yet been described in this
vast region. We investigated short-term effects (7 years after
disturbance) of widespread tree mortality caused by a squall line event from
mid-January of 2005 on soil carbon stocks and concentrations in a central
Amazon terra firme forest. The soil carbon stock (averaged over a 0–30 cm
depth profile) in disturbed plots (61.4
Tropical forests contain about 44 % (383 Pg C) of the approximately 860 PgC stored in forests worldwide, with soils accounting for 32 % of the total carbon stocks (Queré et al., 2009; Lal, 2004). Global emissions due to changes in land use and soil cultivation are estimated to be 136 PgC since the industrial revolution (Lal, 2004; Houghton, 1999). However, there are few estimates of emissions by the decomposition and mineralization of organic carbon in soils following natural disturbances (Lal, 2004), presumably because we assume there is a balance between rapid losses that follow disturbance and recovery between disturbances at the larger spatial scales.
Study area (white inset) on the left side of the Rio Cuieiras,
Amazonas, Brazil
The effects of large-scale natural disturbances (i.e., wind disturbances) on carbon stocks and cycling due to the increase of litter inputs promoted by widespread tree mortality, the fraction of this carbon that persists in soil organic matter, and how long it is stabilized are poorly known in both in tropical and temperate forests (Foster et al., 1998; Turner et al., 1998). In temperate forests, newly exposed soil due to wind disturbance can cover from ca. 10 % (Peterson et al., 1990) up to 60 % of the surface (Beatty, 1980; Putz, 1983). In a three-species temperate forest in Slovakia, no organic carbon was lost at two windthrow sites within 3.5 years after disturbance, but shifts occurred within organic layers and mineral soil toward decomposed organic matter (Don et al., 2012). In Amazonian forests, where windthrows are a major natural disturbance (Nelson et al., 1994; Chambers et al., 2013), such effects have not yet been investigated.
Wind disturbances are frequent in the western and central Amazon, (Nelson et
al., 1994; Espírito Santo et al., 2010; Negrón-Juárez et al.,
2010). In this large region, windthrows are associated with torrential rains
and very strong winds (16 m s
In the tropics, winds break and uproot trees causing strong soil disturbances (e.g., increasing leaves and wood debris and changing morphology and nutrient availability; Schaetzl et al., 1989; Lugo, 2008). Treefall gaps can also change microclimate conditions such as light intensity and create a variety of microsites, which can be separated into canopy, trunk, and root/uprooted sites (Putz, 1983). These microsites have important features that drive soil and vegetation recovery after disturbance (Putz, 1983; Schaetzl et al., 1989; Vitousek and Denslow, 1986). They can differ in microbial activity (Batjes, 1996) and enhance the colonization of fast-growing species that help in the assimilation of nutrients and soil carbon, which in turn can contribute to quickly restore the forest canopy through succession (Putz, 1983). This rapid recycling of nutrients potentially enhances the resilience of tropical forests to natural disturbances (Schaetzl et al., 1989; Ostertag et al., 2003; Lugo, 2008). However, how complex and hyperdiverse tropical forests such as the Amazon will respond in a scenario of higher frequency of extreme weather events (Coumou and Rahmstorf, 2012; Cai et al., 2014) is still not clear.
We assessed the effects of wind disturbances on soils of a large terra
firme forest in central Amazon. We hypothesized that windthrows forming large
canopy gaps ( Are there differences in soil carbon stocks between disturbed
and undisturbed areas, and how do possible variations compare to other
tropical and temperate forests worldwide? What is the importance of soil
texture (clay content) on soil organic carbon content in wind disturbed
areas? Does tree mortality intensity influence soil carbon stocks?
Average concentrations of soil organic carbon content (SOC), soil carbon stocks (SCSs), bulk density (BD), and clay, silt, and sand average concentrations in transect 1 (E1), transect 2 (E2), and transect 3 (E3). Values in brackets represent the standard error of the mean.
This study was conducted in a large terra firme forest, ca. 100 km
from Manaus, Amazonas, Brazil (Fig. 1). We sampled soils from the
Estação Experimental de Silvicultura Tropical (EEST) of the
Instituto Nacional de Pesquisas da Amazônia (INPA) and from a contiguous
forest, adjacent to the Ramal-ZF2 road. The forest adjacent to the Ramal-ZF2
road is owned and administered by the Superintendência da Zona Franca de
Manaus (SUFRAMA). Mean annual temperature in this region was 26.7
The soils of the Amazon region are old and complex, with type and texture influenced by local topographical variations. At the studied region, the relief is undulating with altitude ranging from 40 to 180 m a.s.l. Soils on upland plateaus and the upper portions of slopes have high clay content (Oxisols), while soils on slope bottoms and valleys have high sand content (Spodosols; Telles et al., 2003) and are subject to sporadic inundations (Junk et al., 2011). The yellow Oxisols are found primarily on plateaus and slopes. In general, the soils are well drained and have low fertility, low pH, low cation exchange capacity, high aluminum concentration, and low organic carbon (Ferraz et al., 1998; Telles et al., 2003).
Comparison of the entire 0–30 cm depth profile for
Soil carbon stock (SCS) as a linear function of
In January of 2005, a single squall line event propagating across the Amazon caused widespread tree mortality over large areas (Negrón-Juárez et al., 2010), including ca. 250 ha of terra firme forest in the study area (Fig. 1). Tree mortality directly caused by this event was quantified at landscape level through the correlation of plot-based measurements and changes on the fractions of green vegetation (GV) and non-photosynthetic vegetation (NPV) calculated from Landsat images – see Negrón-Juárez et al. (2010) for a detailed method description. This metric, validated by Negrón-Juárez et al. (2011), allowed us to sample soils across an extent tree mortality gradient 0–70 %, including from small- to large-sized gaps and patches of old-growth forest not affected by the 2005 windthrows (Marra et al., 2014).
We sampled soils during the dry season (July–September) of 2012 (7 years
after disturbance) according to the degree of disturbance intensity measured
as tree mortality (%). In total, 16 plots with dimensions of 25 m
In each of our 16 selected plots, we sampled six soil profiles 5 m from each other. We took samples from three depths (0–10, 10–20, and 20–30 cm) using an auger. For soil bulk density, samples were also
collected in the three depths in one or two profiles per plot using 5 cm tall cylinders with a volume of 98 cm
Before performing soil analyses, we removed leaves, twigs, and roots from our
samples. Samples were then sieved, dried, and homogenized by grinding
(< 2 mm). The soil carbon content was determined in a combustion analyzer at
the Centro de Energia Nuclear na Agricultura (CENA-USP), Piracicaba, Brazil.
Bulk density samples were dried at 105
Before performing statistical tests, we tested our data set for normality
and homoscedasticity. To address our first question we use factorial ANOVA
and compared undisturbed/low-disturbance plots (tree mortality < 5 %,
hereafter referred as undisturbed forest) with those that experienced higher
disturbance intensities (tree mortality
Estimates of soil carbon stock (SCS) from this and other studies conducted in different tropical, subtropical, and temperate forests.
Soils from the disturbed forest had higher mean values of SCS and SOC than
those from the undisturbed forest. This was true for all three depths we
sampled (Table 1). SCS values averaged over 0–30 cm were 61.4
The soil clay content in the entire study area ranged from 2.0 to 71.5 %
averaged over 0–30 cm depth. This large variation in soil texture led to a
large variation in the concentration of soil organic carbon (SOC) and soil
carbon stocks (SCSs). The SOC in the upper samples (0–10 cm) had values
ranging from 0.29 to 6.62 % and mean of 2.57
Along the entire sampled area (disturbed and undisturbed forest), the SCS
was positively correlated with soil clay content (Fig. 3a) and with tree
mortality intensity (Fig. 3b). When constraining the tree mortality gradient
into three disturbance categories defined as tree mortality intensity
(%), we found no differences in SCS (
As expected, our results were between those values found in the two soils
types (Oxisols and Spodosols) evaluated in a previous study also conducted
at the EEST (Telles et al., 2003), in which SCS values for 0–10 cm were
reported as 14.9
Soil carbon stock (SCS) at sites with different soil clay content and tree mortality intensity.
The soils from our study area also had different SCS values from those
reported for other regions of the Brazilian Amazon (i.e., same/similar soil
types; Table 2). For the 0–10 cm profile, when comparing to old-growth
forests in the Pará state, the mean SCSs of our undisturbed and disturbed
forests were lower and similar, respectively (Trumbore et al., 1995; Camargo
et al., 1999). In the 0–30 cm depth profile, our undisturbed forest had
similar SCS to that reported for other regions. When including other soil
types, our disturbed forest had SCS values (61.4 Mg ha
When comparing to forests worldwide (i.e., different soil types), both our undisturbed and disturbed forest had lower SCS values (Table 2). We only found higher SCS values than that reported for the 0–30 cm depth profile from an equatorial forest in Senegal, Africa (Batjes, 2001). For the 0–10 cm depth profile, our disturbed forest had SCS higher than that reported for an old-growth coastal hill dipterocarp forest in Singapore (Ngo et al., 2013) and a 68-year-old secondary coastal temperate rain forest in southeast Alaska (Kramer et al., 2004), both in different soil types. In contrast, our disturbed forest had lower SCSs than those reported for other temperate forests in Europe (Don et al., 2012) and North America (Huntington and Ryan 1990; Kramer et al., 2004). This was true for both non-harvested and harvested forests, in which nutrient exportation via logging has an opposite effect than that of wind disturbances (nutrient inputs).
Soil clay content was positively correlated with the SOC (Pearson's
Due to the proximity of our plots, we assume climatic and geological aspects
to be constant. Thus, the importance of soil texture on carbon stocks in our
study site reflects a local pattern. Here we focused on assessing the
effects of the existing Amazon tree mortality gradient (Espírito Santo
et al., 2010; Chambers et al., 2013) on SOC and SCS, which is why we
excluded valleys and selected plots along transects crossing forest patches
with different disturbance intensity. Nonetheless, apart from indicating
significant increase of SCS due to inputs of organic matter from tree
mortality, our data show that clay-richer soils originally had higher SCS
(0–30 cm depth profile) compared to soils with lower clay content (Fig. 5).
Soils from areas where tree mortality was <10 % and clay content
This comparison confirms that the widespread tree mortality caused by the 2005 windthrows increased the SCS in our study area. A higher frequency and intensity of wind disturbances in plateau areas also suggests that the higher SCS in these portions of the relief, apart from those related to abiotic factors (e.g., soil texture, topography and erosion), might also reflect differences of vegetation dynamics. Although the soil clay content is an important aspect and greater inputs of carbon can be expected in more clayey sites, significant inputs can also occur in more sandy sites, for instance, when strong wind gusts reach lower parts of slopes and valleys.
Although we observed an increase of SCS in areas affected by the storm, it is notable that the fresh necromass produced by widespread tree mortality events is not fully incorporated into the soil. Under this assumption, the fast decomposition of carbon stored in roots and other woody material probably contributes most to the observed increases in SCS. Carbon inputs from belowground material, which is already incorporated to the soil, might be specially related to the increase of SCS in the 10–20 and 20–30 cm depth profiles.
Seven years after the windthrow event, the SCS at 30 cm depth was
approximately 13.7 Mg ha
Amazon soils typically have a great variation in texture and nutrient availability related to physical and chemical properties (Quesada et al., 2010, 2011), which can influence basin-wide variations in forest structure and function (Quesada et al., 2012). Our results indicate that in central Amazon terra firme forests, vegetation dynamics can also influence soil attributes at the landscape level. In this region, the observed organic carbon enrichment derived from widespread tree mortality might also be related to the fast establishment and growth of pioneer species in heavily disturbed areas (Chambers et al., 2009; Marra et al., 2014).
In contrast, according to Lin et al. (2003), the Fushan Experimental Forest, which has experienced frequent windstorms, did not regain any nutrients following disturbance. This, in turn, has limited local tree growth (shown as lower canopy height) and, consequently, decreased carbon input into the soil. Thus, more intense mortality regime can also be expected to change forest dynamics, and eventually decrease SCS and nutrient cycling. The effects might depend on forest stature, successional stage (i.e., floristic composition and forest structure attributes such as tree density, basal area, and biomass), and tree mortality intensity, often controlled by the speed and duration of wind gusts (Lugo et al., 1983; Garstang et al., 1998). In our study area, fast vegetation regeneration could even reduce short-term losses of carbon associated with the 2005 windthrows, which had an estimated emission (assuming the carbon from all felled trees emitted to the atmosphere at once) of ca. 0.076 PgC, equivalent to 50 % of the deforestation during that same year (Higuchi et al., 2011; Negrón-Juárez et al., 2010).
The size of gaps in which we observed significant increase on soil carbon content (gaps from 0.1 up to 17 ha) indicates that windthrows – apart from influencing tree species composition, forest structure, and forest dynamics (Chambers et al., 2013; Marra et al., 2014) – also change soil attributes. The nutrients released in this process might have an important feedback on vegetation resilience and recovery following disturbance. To determine how much of the added soil carbon is stabilized in a long term, future studies should assess soil carbon stocks and soil organic carbon along a chronosequence including wind-disturbed terra firme forests with different time since disturbance. Since wind is a major disturbance agent in western and central Amazon, more precise estimates of soil carbon stocks need to consider and reflect differences in tree mortality regimes at the landscape level.
We gratefully acknowledge the workers from the EEST/INPA for giving support with the fieldwork, and the lab team of the CENA-USP and the Laboratório Temático de Solos e Plantas (LTSP/INPA) for giving support with the soil analyses. We also acknowledge the SUFRAMA for allowing us to access part of the study area. At last, we acknowledge Edzo Velkamp, an anonymous referee, and Hermann F. Jungkunst for providing valuable comments during the revision of this article. This study was financed by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) within the project SAWI (Chamada Universal MCTI/N14/2012, Proc. 473357/2012-7) and the INCT – Madeiras da Amazônia. It has also been supported by the Tree Assimilation and Carbon Allocation Physiology Experiment (TACAPE), a joint project between the Biogeochemistry Processes Department of the Max Planck Institute for Biogeochemistry and the Laboratório de Manejo Florestal (LMF/INPA). Robinson I. Negrón-Juárez was supported by the Office of Science, Office of Biological and Environmental Research of the US Department of Energy under contract no. DE-AC02-05CH11231 as part of Next-Generation Ecosystems Experiments (NGEE Tropics) and the Regional and Global Climate Modeling (RGCM) Program. The article processing charges for this open-access publication were covered by the Max Planck Society. Edited by: E. Veldkamp