Deciphering earth mound origins in central Brazil
- First Online:
- Cite this article as:
- Silva, L.C.R., Vale, G.D., Haidar, R.F. et al. Plant Soil (2010) 336: 3. doi:10.1007/s11104-010-0329-y
- 163 Views
Mound fields are a common landscape throughout the world and much of the evidence for their origin has been of a circumstantial nature. It has been hypothesized that earth mounds emerge over grasslands by termite activity; alternatively, they might be formed after erosion. We tested whether a mound field in central Brazil was generated by termite activity or erosion. We used soil organic matter isotopic composition, soil chemical, physical and floristic composition to determine the origin of a mound field. If the mounds emerged by termite activity in an established grassland the soil organic matter below the mound should have the isotopic signature of C4 dominated grassland, which contrasts with savanna C3 + C4 signature. Additionally, soil traits should resemble those of the grassland. All markers indicate that the mounds were formed by erosion. The soil isotopic composition, chemical traits and texture below the mound resembled those of the savanna and not those of the grassland. Moreover, most of the species present in the mound were typical of savanna. Concrete evidence is provided that mound fields in the studied area were produced by erosion of a savanna ecosystem and not termite activity. The use of the techniques applied here would improve the assessments of whether analogous landscapes are of a biogenic nature or not.
Several authors have shown that in Brazil these mound fields frequently occur within ecotone zones, towards forest/savanna boundaries (Eiten 1972, 1984; Furley 1986; Araújo Neto et al. 1986; Oliveira-Filho and Furley 1990). This makes the investigation of this ecosystem especially relevant to the understanding of past and present vegetation dynamics such as the riparian forest expansion recently shown in this region (Silva et al. 2008). There are previous descriptions of the vegetation that typically occur in the campos de murundus, where the mounds were found mostly populated by savanna tree species, even though forest species also occur (Eiten 1984; Furley and Ratter 1988; Oliveira-Filho and Furley 1990; Resende et al. 2004). However, the question of which processes created this landscape is not yet settled. The lack of quantitative evidence in support to previous circumstantial explanations for the origins of this ecosystem hinders comparisons both within the cerrado and with analogous vegetation elsewhere. Deciphering earth mounds origins would allow such comparisons, serving the interests of ecologists and soil scientists.
There are two main hypotheses that attempt to explain the genesis of earth mounds in central Brazil. The first proposes that termite nests, often associated with the mounds, were responsible for raising the microtopography above the grassland vegetation (Eiten 1984, 1990; Oliveira-Filho 1992; Ponce and Cunha 1993). These termite nests, being above the flood levels of the surrounding grassland during the wet season, then became populated by flood-intolerant woody vegetation. The second hypothesis proposes that differential erosive processes produced the micro-topography typical of the campos de murundus (Araújo Neto et al. 1986; Furley 1986). Specifically, this hypothesis proposes that the locations where the campos de murundus formed were once typical savannas and that erosion left behind the mounds on which savanna vegetation persisted. The surrounding areas where erosion occurred became increasingly flooded and only sustained flood-tolerant grasses. This last hypothesis does not exclude the possibility that the presence of termite mounds in the slightly sloping savanna created the initial microtopography, which in turn led to differential erosion and mound formation.
The carbon isotopic composition of SOM below the earth mounds, at the level of the surrounding grassland, shows an isotopic signature similar to that of the surrounding grassland.
The soil physical and chemical properties below the mound resemble those in soils of the surrounding grassland.
Mounds that are closer to the neighbouring forest have a greater number of species from the forest compared to those farther from the forest.
Confirmation of the above would support the hypothesis that mounds were formed in grasslands by the build-up of the microtopography via termite activity. On the other hand, rejection of the above hypotheses and the presence of isotopic, chemical and floristic signatures typical of the savanna would support the erosive origin of the mounds.
Materials and methods
Measurements were taken in an undisturbed campos de murundus site on a grassy slope, having lower than 5° inclination and bordering a savanna/gallery forest ecotone, at the University of Brasilia Experimental Station (FAL-UnB), Federal District, Brazil (Fig. 1; 15° 56′ 40.17″ S and 47° 54′ 36.35″ W). This mound field contained mounds averaging 1.1 ± 0.14 m in height above the grassland and an average perimeter of 32 ± 11 m. These mounds often had termite colonies of the species Armitermes cerradoensis (Mathews). The average annual temperature for this region is 22.5°C and average annual rainfall 1,426 mm (1993–2002). Rainfall is seasonally distributed, with more than 80% falling during the months of November to April. Soils in this savanna can be generally characterized as deep, well drained oxisols, with low organic matter and nutrient contents. Soils in the seasonally flooded grassy areas can have higher fertility and organic matter content (Eiten 1990). The area of each mound was estimated by using perimeter measurements, taken by metric tape from the intersection between the bases of the mounds with the grassy area level, considering their shapes as perfect circles. The distance from the forest was also determined for each mound, by measuring the shortest straight-line distance from each of them to the forest border.
During the dry season of 2007 we sampled the soil profile with a soil auger to a depth of 2.5 m or less if the water table was reached or the presence of gravel impeded further sampling. Three soil profiles were acquired in three savanna sites (Fig. 1; S1, 2 and 3), three in the campos de murundus grassland area (Fig. 1; G1, 2 and 3) and five in the mounds (Fig. 1; mounds 1 to 5). At the savanna profile S2 a gravel layer at 90 cm depth impeded the ongoing sampling. The water table was reached at approximately 75 cm depth in the grassland area, preventing the sampling of soil at lower depths. Soil samples were collected every 10 cm of depth and dried at room temperature after which the fine roots were removed by sieving through a 2 mm mesh. The sieved soil was used for analyses of isotopic abundance as well as chemical and physical properties.
Soil stable isotope analysis
Soil chemical and textural characterization
The soil pH was determined in a water suspension (2:1, v.v.). Total nitrogen was determined by the Kjeldahl method (Bremner and Mulvaney 1982). Available phosphorus and potassium concentrations were determined by the method of Mehlich (1953). Exchangeable aluminum, calcium and magnesium were extracted by 1M KCl solution and their concentrations determined by titulation and atomic absorption. Organic carbon content was determined by wet oxidation (Walkley and Black 1934). Soil texture was determined by the international pipette method (Day 1965).
The floristic composition of the nearby savanna and forest areas were gathered from previous studies (Felfili and Silva Júnior 1992, 1993; Felfili et al. 1993; Felfili 1995; Andrade et al. 2002; Silva Júnior 2004; Ribeiro and Tabarelli 2002; Guarino and Walter 2005). A detailed floristic composition inventory was not done for surrounding grassland, as it was mostly dominated by grasses of the genera Axonopus and Andropogon.
Carbon isotope ratios
Soil chemical and physical properties
Fertility and organic matter content tended to be higher on the surface soil, decreasing at lower depths of the soil profile (Fig. 4). However, Ca, Mg and P were the exception. The concentration of these elements either increased at a lower depth or oscillated in concentration throughout the profile.
Soil particle size properties show strong and consistent differences between profiles. Soil texture was sandier on grassland soils compared to mounds and savanna (Fig. 4). At the surface layers (0–10 cm) sand/clay ratios are on average four times higher in grassland soils than in the savanna or mounds.
The floristic composition of the eleven mounds inventoried included 37 families, 64 genera and 83 species, of which 59% were tree species, 24.1% shrubs and 14.5% herbs. The six most frequent families were the Melastomataceae (14 species), Poaceae (six species), Fabaceae (six species), Myrtaceae (five species), Lauraceae and Euphorbiaceae (four species each). Of the species recorded, 61.4% are usually encountered in open savanna vegetation, while 18.1% are characteristic of forests (mostly gallery forests) and 19.3% can be found in both ecosystems. There was no correlation between relative abundance of forest species present in the mound and distance of the mound from the forest (R2 = 0.11, P > 0.05). No trees or shrubs were encountered out of the mounds on the surrounding grassland, where Andropogon sp. and Axonopus sp. (C4 grasses) were dominant.
Carbon isotope ratios of the SOM profile below the mound (∼1 m) averaging −18.5 ± 1.1‰ (± σ) could not be distinguished from those in two of the three savanna cores taken here and were also similar to those of savanna SOM found in a previous study in the same region (Silva et al. 2008). These values are lower than the average δ13C value of SOM from the C4-dominated grassland area surrounding the mound (−15 ± 0.7‰). The savanna isotopic signal below the mound indicates that this site was a savanna and not grassland before mound formation. Therefore, the isotopic data supports the erosion hypothesis. Unfortunately we could not collect samples from the surrounding grassland to depths below 1 m. However, the interpretation of the carbon isotope ratios of SOM deeper in the grassland soil profile may be equivocal since savanna isotopic signatures could have eroded away during the formation of the mound field. The slight increase in carbon isotope ratios with depth at the top 25 cm in all soil profiles is caused by isotope fractionation during diagenetic processes in SOM (Martinelli et al. 1996; Silva et al. 2008). One particular SOM isotope profile in the savanna (S3) was unusual in that it had δ13C values typical of the grassland area and showing a stronger C4 component than the other savanna samples at depths between 0 cm and 150 cm depth. We propose that this profile is showing the process we hypothesize: differential erosion will cause some savanna areas to become grasslands. At depths greater than 150 cm, representing the older vegetation, SOM from this core has the isotopic signature typical of a savanna. After the erosion, however, this site became more populated with C4 grasses as evidenced by the shift in the carbon isotope ratios in SOM at depths shallower than 150 cm. We also note that this savanna soil profile is the closest to the gallery forest (Fig. 1) and probably more susceptible to erosion from runoff. Accordingly, this area may eventually undergo a transition to a campos de murundus.
Soil chemical and physical profiles for the most part matched the savanna profile. Discrepancies between savanna and mound profiles were observed in soil pH, organic matter, total nitrogen, phosphorus and aluminum. The differences, however, did not indicate that mound profiles were more similar to those of grassland than the savanna profile. Rather, they indicated that savanna and grassland were similar and the mound profiles were unusual. To some extent this is expected because, with the exception of phosphorus, nutrients can be highly mobile and their profile may not be conserved during the erosive process that left the mounds behind. Phosphorus, however, can easily bind to organic matter. The near absence of available phosphorus in the mound profile up to a depth of 100 cm might be explained by its binding to organic matter observed at higher concentration in the mounds compared to savanna and grassland (Fig. 4). In addition, fauna activity can affect these nutrient concentrations (Oliveira-Filho 1992). Differential nutrient input through termites, birds and mammals may have increased productivity in the mounds over savanna and grassland, also explaining low P and high N, organic matter and plant density.
In contrast, there was a close match between savanna and mound soil profiles for potassium, calcium and magnesium, with the grassland showing a different profile. The soil texture, which may be a more permanent feature of the soil profile and therefore more faithfully record the mound genesis, showed a close agreement between mound and savanna profile, with the grassland showing a much greater proportion of sand compared to the mound and savanna soil profiles (Fig. 4). The sandiness of the grassland soil compared to those of the mounds is further evidence that erosion processes were important in the formation of the campos de murundus. We propose that erosion carried away the smaller and lighter clay particles leaving the remaining soil, where erosion occurred, rich in the larger and heavier sand particles. This is supported by significant correlations between soil texture and soil moisture found in transects from grasslands (higher relative sand content) to riparian forests (higher relative clay content) at the same site studied here (Munhoz et al. 2008).
The majority of the species found on the mounds are from savanna and only about 18% are of forest origin. No correlations between the presence of forest species on the mounds and their distance relative to the forest were found and the species/area curve within the mounds matches the savanna saturation (Fig. 5). All these observations support the presence of a savanna ecosystem prior to mounds formation, which supports the erosion hypothesis. However, some floristic metrics were quite different compared to the savannas. For example, the basal area in the mounds (14 m2 ha−1) was greater than those of the savanna (7 m2 ha−1), but much lower than in the forest (42 m2 ha−1), and the diversity on the mounds (3.1 H’) was lower than those observed in the gallery forest (3.6 H’) and the neighbouring savanna (3.5 H’). We propose that these contrasting characteristics were developed after the mound formation. In other tropical systems, increased tree density and basal area in small islands may result from habitat fragmentation due to changes in dispersion and reduced seeds predation (Wright and Duber 2001). This along with expected increases in seedling recruitment near conspecific species may in time reduce diversity (Wright and Duber 2001). Likewise, our results suggest that limited mound area, or the fragmentation of the savanna, may lead to higher density and lower diversity of woody plants.
However, the conclusions raised from the comparison of such floristic data have to be taken cautiously. Floristic composition vary greatly within both savannas and forests of cerrado region, thus any extrapolation based on the particular site studied here would be inappropriate. Additionally, because different sampling designs were used on each of these vegetation types (systematic design—continuous strip layout—in the forest; sets of randomized plots in the savanna; and complete census in the mounds), comparing vegetation features may be not as straight forward as one would think. The reason why different designs are used is to produce the most representative data set of each given ecosystem. Considering that each of the above vegetation physiognomies was satisfactorily described (Fig. 5) and that the criteria of inclusion for woody species was the same (>5 cm), here the floristic features of these different ecosystems are comparable (Felfili 1995).
The isotopic and chemical signature of the soil profile as well as its texture and the floristic composition of the mounds all suggest that they were formed by erosion. This site was most likely a savanna site that gradually eroded to the campos de murundus landscape with small tree islands in a matrix of seasonally flooded grassland. Much of the previous evidence regarding the termite origin of campos de murundus was circumstantial and only based on the observation of the presence of termite nests on the mounds. However, it is not known whether the termites became established after mound formation or were the actual cause of the mound formation. One study in particular (Oliveira-Filho 1992) proposes a succession of different termite species in mound genesis, which initiates with Armitermes euamignathus (Silvestri), a species more tolerant of moist conditions than other termites. According to this interpretation, the succession of termite species culminates in Cornitermes snyderi, which is least tolerant of moist conditions. Our isotopic data do not support the above scenario in that the SOM isotopic signature below the mound is characteristic of savanna and not the moist grassland rich in C4 grasses. The evidence for erosive processes in mound genesis has also been circumstantial and based on the observation that campos de murundus are usually found in seasonally flooded gentle slopes. Although our results do not support the termite origin hypothesis, their association with earth mounds may actively alter the physical environment and vegetation dynamics. For instance, termite activity (hollowing out soil, mining clay etc.) may have affected the height or area of the mounds through time, perhaps intensifying differential erosion. Additionally, the predominance of wind-dispersed species on the regional open savannas (Ribeiro and Tabarelli 2002) can be replaced by species that depend on biotic agents for dispersal. The fast response of insects to current changes in climate should also be of relevance towards such dynamics. It has been shown in southern Africa that termitaria density is positively correlated with rainfall (Picker et al. 2007), futhermore, they may reach more than 20,000 years in age, becoming a stable part of the landscape (Midgley et al. 2002) and altering vegetation composition (Rahlao et al. 2008). While a potential association between termitaria and woody species distribution in campos de murundus of central Brazil remains to be investigated, there are similar landscapes in Africa and Australian savannas, as well elsewhere in the Neotropics, which may indeed be caused by termite or other biotic-mediated activity (Cox et al. 1989; Ellery et al. 1998; McCarthy et al. 1998; Brossard et al. 2007). In such places the isotopic profiling of SOM as used here would certainly improve our confidence in the assessments of whether they are truly of a biogenic nature or not.
We thank Doyle McKey (Centre D’Écologie Fonctionnelle Et Évolutive (CEFE)-CNRS), Norton Polo Benito (EMBRAPA-Cenargen), José Carlos Sousa Silva (EMBRAPA-Cerrados) for helpful comments.