Riverine organic matter composition as a function of land use changes, southwest amazon
Download 332.19 Kb. Pdf ko'rish
|
|
S263 Ecological Applications, 14(4) Supplement, 2004, pp. S263–S279 ᭧ 2004 by the Ecological Society of America RIVERINE ORGANIC MATTER COMPOSITION AS A FUNCTION OF LAND USE CHANGES, SOUTHWEST AMAZON M ARCELO
C. B ERNARDES
, 1,4
L UIZ
A. M ARTINELLI , 1
LEX V. K
RUSCHE , 1 J ACK
G UDEMAN
, 2 M ARCELO M OREIRA , 1 R EYNALDO L. V
ICTORIA , 2 J EAN
P. H. B. O METTO
, 1 M ARIA V. R. B
ALLESTER , 1 A NTHONY
K. A UFDENKAMPE , 3
EFFREY E. R
ICHEY , 2 AND J OHN I. H EDGES
2 1
CEP 13400-970, Piracicaba, SP, Brazil 2
3
We investigated the forms and composition of dissolved and particulate or- ganic matter in rivers of the Ji-Parana´ Basin, which is situated at the southern limit of the Amazon lowlands and has experienced extensive deforestation in the last three decades ( ϳ35 000 km 2 ). Our objective was to investigate how extensive land-use changes, from forest to cattle pasture, have affected river biogeochemistry. We measured a series of chemical, biochemical, and isotopic tracers in three size classes of organic matter within five sites along Ji-Parana´ River and eight more sites in six tributaries. The results were compared with C 4 leaf and pasture soils end members in order to test for a pasture-derived signal in the riverine organic matter. The coarse size fraction was least degraded and derived primarily from fresh leaves in lowland forests. The fine fraction was mostly associated with a mineral soil phase, but its ultimate source appeared to be leaves from forests; this fraction was the most enriched in nitrogen. The ultrafiltered dissolved organic matter (UDOM) appeared to have the same source as the coarse fraction, but it was the most extensively degraded of the three fractions. In contrast to Amazon white-water rivers, rivers of the Ji- Parana´ Basin had lower concentrations of suspended solids with a higher carbon and nitrogen content in the three size fractions. However, principal component analyses showed a cor- relation between areas covered with pasture and the ␦ 13 C values of the three size fractions. The highest ␦ 13
Rolim-de-Moura and Jaru´ rivers, which have the highest areas covered with pasture. The lower the order of the streams and the higher the pasture area, the greater is the possibility that the C 4 -derived organic matter signal will be detected first in the faster-cycling fraction (UDOM). The large change in land use in the Ji-Parana´ Basin, replacement of primary forests by C 4 pastures for cattle feeding, that has taken place in the last 30–40 yr, has already changed the characteristics of the composition of the riverine organic matter. Key words: Amazon rivers; black water; deforestation; isotopes; land use; lignin; organic matter; pasture; principal component analyses; white water. I NTRODUCTION Of the states located in the Amazon region, the state of Rondoˆnia in the southern Amazon has experienced the fourth highest rate of deforestation (Instituto Na- cional de Pesquisas Espaciais [INPE] 2001). In this region, large areas of rainforest have been replaced by pasture for cattle in the last 30 years (INPE 2001). Of the total deforested area in the Amazon until 1999, almost 10% (58 000 km 2 ) has been in Rondoˆnia (INPE 2001). The increase in the land use is concentrated along highway BR-364 within the limits of the Ji-Pa- rana´ River basin. As a consequence, in 1986, of all the deforested areas in Rondoˆnia, 59% were located in the Manuscript received 2 November 2001; revised 30 November 2002; accepted 31 December 2002; final version received 29 Jan- uary 2003. Corresponding Editor: J. M. Melack. For reprints of this Special Issue, see footnote 1, p. S1. 4 E-mail: bernardes@geoq.uff.br Ji-Parana´ River basin, with the most intensive land cov- er changes in the central part of this basin (Fig. 1). The conversion of large areas of primary forests into grasslands leads to profound modifications in the struc- ture and functioning of terrestrial ecosystems in the Amazon. Several studies have shown that the intro- duction of C 4 grasses alters carbon and nitrogen stocks and dynamics in soil organic matter (Desjardin et al. 1994, Trumbore et al. 1995, Neill et al. 1997, Camargo et al. 1999). However, these studies do not address whether these changes in organic matter cycling have effects beyond the local site of deforestation. Streams are natural transporters in landscapes, and their flowing waters reflect the biogeochemistry of their watersheds. Thus, higher order streams and rivers are considered to be good integrators of both natural and anthropo- genic processes in their drainage basins, and they have the potential to offer a broad view of the magnitude of biogeochemical changes over a landscape. S264 MARCELO C. BERNARDES ET AL. Ecological Applications Special Issue F IG
Study area and sample locations in Rondoˆnia, Brazil. Few studies have addressed the consequences of land-use changes on aquatic environments in the Am- azon Basin (e.g., Williams and Melack 1997, Neill et al. 2001, Biggs et al. 2002). None of these studies investigated possible changes in the compositional characteristics of organic material of river systems. On the other hand, the forms and composition of different- sized classes of organic matter transported by white- water rivers in largely unaltered basins of the Amazon have been investigated (Hedges et al. 1986, 1994, 2000, Richey et al. 1990, Quay et al. 1992, Devol and Hedges 2001). These studies have analyzed an array of ele- mental (carbon and nitrogen concentrations), biochem- ical (lignin, carbohydrate, and amino acid), stable iso- tope (
␦ 13 C, ␦ 15 N), and radioisotope ( ⌬ 14 C) compositions in riverine organic matter. These measurements were carried out in three size fractions: coarse ( Ͼ63 m) and fine ( Ͻ63 m to 0.1 m) particulate and ultrafil- tered dissolved organic matter (UDOM). The compo- sition and fates of these three fractions throughout the Amazon Basin are consistently distinct, although they share the same ultimate source—the leaves of C 3 forest trees. The coarse fraction is the least degraded and resembles relatively undecomposed tree leaves. The dissolved products of the decomposition of tree leaves percolate through the soil column, where nitrogen-rich August 2004 S265
ORGANIC MATTER OF AMAZON LOWLAND RIVERS F IG . 2. Daily variability of the discharge from 1999 to 2001 in the following rivers: Com- emorac¸a˜o River (COM-2), Pimenta Bueno Riv- er (PB-2), and Ji-Parana´ River (JIP-2 and JIP- 4). Arrows indicate seven times at which water samples were collected. compounds are sorbed and stabilized by soil mineral particles (Aufdenkampe et al. 2001). Consequently, the fine fraction is the richest in nitrogen and enters the river systems primarily via soil erosion. Finally, the dissolved organic fraction is most degraded, being composed of organic, nitrogen-poor substances that are not sorbed onto mineral surfaces. Another important feature is the constancy over time and space of the compositional characteristics of the size classes of riv- erine organic matter. From first-order Andean tributar- ies to the major rivers of the Amazon Basin, no major differences were observed for a given size fraction (Hedges et al. 2000). In the present study, we investigated the forms and composition of riverine organic matter of the Ji-Parana´ Basin. We sampled the Ji-Parana´ River at five different sites along its main stem. We also sampled several tributaries, with drainage basins exhibiting a range of sizes and extents of deforestation. Our main objective was to compare the composition of the organic matter in rivers of the Ji-Parana´ Basin with other rivers of the Amazon Basin. Most previous studies of riverine or- ganic matter composition in the Amazon Basin have focused on turbid white-water systems, which have a considerable portion of their headwaters in the Andes. In contrast, rivers of the Ji-Parana´ Basin have their watersheds in lowland areas, carrying a much lower load of suspended particles than their white-water counterparts (Martinelli et al. 1993). Our second objective was to investigate whether the extensive land-use changes in the Ji-Parana´ Basin over the last 30 yr have altered the organic matter compo- sition of its rivers. In rivers of the Piracicaba Basin, located in the southeast region of Brazil, land-use changes that occurred 70–80 yr ago have left their ‘‘imprint’’ in the dissolved and fine fractions of rivers of that basin (Martinelli et al. 1999). In contrast, most of the changes in land use in the Ji-Parana´ Basin started just 30–40 yr ago. Studies with soils under pastures of different ages have shown that after 3–5 yr of pasture cultivation the signal of C 4 -derived organic matter from pasture grasses is detectable in the soil organic matter (Neill et al. 1997). Therefore, it is possible that the composition of organic matter in the Ji-Parana´ River network might reflect these new organic matter sources. In order to answer these questions, we analyzed a series of chemical, biochemical, and isotopic tracers in the three size classes of organic matter from rivers within the Ji-Parana´ Basin and compared them with values obtained in previous studies in the Amazon Ba- sin. In addition, we compared these results with C 4 plants and pasture soil sources to test for a signal in the riverine organic matter traceable to introduced grass species.
M ETHODS
Sampling sites The Ji-Parana´ Basin, with a drainage area of 75 000 km 2
Amazon Basin (Fig. 1). The headwaters of the Ji-Parana´ River are formed by the confluence of the Comemo- rac¸a˜o and Pimenta Bueno rivers. The Ji-Parana´ River channel has a total length of 972 km and varies in width from 150 to 500 m, whereas the channel widths of the major tributaries range from 30 to 200 m (Table 1). The high and low water periods for these rivers range from December to May, and from June to November, respectively. Water samples were collected seven times between 1999 and 2001 (Fig. 2). Samples were collected at five sites along the main channel of the Ji-Parana´ River, at the mouths of six major tributaries, and at two head- water sites, for a total of 13 sampling stations (Fig. 1). The first four sites were located on the Comemorac¸a˜o (COM-1 and COM-2) and Pimenta Bueno (PB-1 and PB-2) rivers; below their junction there were five sam- pling sites (JIP-1 to JIP-5). Along its course, the Ji- Parana´ receives contributions from four main tributar- ies that were also sampled: the Rolim de Moura (ROL), Urupa´ (URU), Jaru´ (JAR), and Machadinho (MAC) rivers. The drainage area above each sampling site was delineated and individually characterized in terms of total cumulative area, population density, river order, land use, and soil textural characteristics using a digital library built using the Arc/Info geographic information system (ESRI, Redlands, California, USA) (Table 1).
S266 MARCELO C. BERNARDES ET AL. Ecological Applications Special Issue F IG
Compositional averages of the coarse suspended solids fraction of rivers of the Ji-Parana´ Basin. Error bars represent Ϯ1 SD . Site identities: 1, COM1; 2, COM-2; 3, PB1; 4, PB2; 5, JIP1; 6, JIP2; 7, JIP3; 8, JIP4; 9, JIP5; 10, ROL; 11, URU; 12, JAR; 13, MAC. Sample collection and preparation At each site, 50–100 L of water were collected from the river in the middle of the channel at 60% of the total depth using an electric pump. The water sample was sieved ( Ͼ63 m) in field in the order to separate the coarse suspended solid (CSS) fraction, which was immediately preserved with HgCl 2 , to a final concen- tration of 100 M. The fine suspended solid (FSS) fraction ( Ͻ63 m and Ͼ0.1 m) and ultrafiltered dis- solved organic matter (UDOM) fraction ( Ͻ0.1 m and Ͼ1000 daltons) were isolated in the laboratory with a Millipore tangential flow ultrafiltration system (model Pellicon-2; Millipore, Billerica, Massachusetts, USA), using membrane cartridges having a nominal 0.1- m pore size (model Durapore VVPP; Millipore) and a 1000-daltons molecular weight nominal cut off (model PLAC; Millipore), respectively. After filtration, the material was roto-evaporated and then dried to constant mass in an oven, both at 50 ЊC. The average percentage recovery of organic matter in all samples during ultra- filtration was 98 Ϯ 8%, of which an average of 20 Ϯ 6% was recovered as UDOM, from both forested and pasture drainage areas. Lignin oxidation, elemental, and isotopic analysis Lignin analyses were made according to the CuO oxidation procedure of Hedges and Ertel (1982) as modified by Gon˜i and Hedges (1990). Briefly, between 30 to 300 mg of dry sample was oxidized at 155 ЊC for
3 h with CuO under basic (8% NaOH) conditions. The reaction solution was spiked with a nine-compound gas chromatography recovery standard mixture (in pyri-
August 2004 S267
ORGANIC MATTER OF AMAZON LOWLAND RIVERS F IG . 4. Compositional averages of the fine suspended solids fraction of rivers of the Ji-Parana´ Basin. Error bars represent Ϯ1 SD
T ABLE
1. Characteristics of the studied drainage areas of the southwest Amazon Basin. Sites Area (km
2 )
3 /s) Width (m) Population ( ϫ 10 2 ) Density (no. inhabitants/km 2 ) River order
Pasture (%)
Forest (%)
Sand (%) Clay (%) COM-1
COM-2 PB-1
PB-2 JIP-1
JIP-2 JIP-3
132 5894
152 10 118
17 843 32 793
39 461 ···
161 ···
205 ···
641 ···
8 71 16 86 115
239 214
5 269
6 219
680 1516
2509 3.8
4.6 3.9
2.2 37.1
13.5 14.0
3 5 3 6 6 6 6 13 28 8 32 31 39 33 47 65 90 60 62 54 53 44 73 76 66 69 68 67 48 21 19 25 24 24 25 JIP-4 JIP-5 ROL
URU JAR
MAC 60 494
64 294 1349
4209 7275
3250 1332
··· ···
··· ···
··· 259
329 37 95 68 52 2973 3155 148
492 793
182 10.5
9.8 11.0
11.7 10.9
5.6 7 7 5 5 6 5 40 40 66 43 53 22 53 53 26 50 36 68 65 64 64 67 60 41 26 27 27 23 30 51 Notes: ‘‘Area’’ is the cumulative drainage area. Population numbers are from 1996. Q is the annual mean discharge from 1999 to 2001. Pasture (%) and forest (%) are the cumulative percentages of pasture and forest; sand (%) and clay (%), are percentage of sand and clay content in soils (Ballester et al. 2003). Sample site codes are delineated in
S268 MARCELO C. BERNARDES ET AL. Ecological Applications Special Issue F IG
Compositional averages of the ultrafiltered dissolved organic matter (UDOM) fraction of rivers of the Ji-Parana´ Basin. Error bars represent Ϯ1 SD
dine), acidified and extracted with diethyl ether. The lignin extract was diluted in pyridine, mixed with re- gisil plus an absolute recovery standard, and analyzed on a Hewlett Packard 5890 series II gas chromatograph (Agilent Technologies, Palo Alto, California, USA) fit- ted with a DB-1 fused-silica capillary column (J&W Scientific, Folsom, California, USA). Identities of all phenols were confirmed by mass spectrometry of se- lected samples vs. commercial standards. The average analytical precision was Ϯ10% for the reported lignin phenols. Organic carbon and nitrogen concentrations were determined using a Carlo Erba CHN analyzer (Thermoquest, Rodano, Italy). Isotope measurements were performed with a Finnigan Delta-E mass spec- trometer (ThermoFinnigan, Bremen, Germany) fitted with dual inlet and dual collector systems. Results are expressed in ␦ 13
␦ 15 N relative to Pee Dee Bel- emnite (PDB) and atmospheric N 2 standard, respec- tively defined as ␦ 13 C or ␦ 15 N (‰) ϭ ([R sam /
std ] Ϫ 1) ϫ 1000, where R sam
and R std
are the 13 C: 12 C or
15 N/ 14 N of the sample and standard, respectively. Samples were analyzed at least in duplicate with a maximum differ- ence of 0.2‰ between replicates. Statistical analysis Most of our data did not follow a normal distribution. Accordingly we used nonparametric statistical tests. To test for differences among sampling sites we used the Mann-Whitney
(PCA) were performed in order to examine sources of August 2004 S269
ORGANIC MATTER OF AMAZON LOWLAND RIVERS F IG . 6. Plot of organic carbon vs. fine suspended sedi- ments (FSS) for Ji-Parana´ rivers (filled circles), Amazon Riv- er (open circles), white-water tributaries (open diamonds), and black-water tributaries (filled diamonds). Values for the Amazon River and white- and black-water tributaries are from Hedges et al. (1986, 2000). T ABLE 2. Compositional averages of the coarse and fine fractions of organic matter of rivers of the Ji-Parana´ Basin in comparison with the Solimo˜es/Amazon main channel, and their white- and black-water tributaries. Rivers
Size fraction
␦ 13 C O r g a n i c C (mass %) Total N (mass %) C:N Suspended sediments (mg/L) ⌳
organic C) cin:van syr:van (Ad/
Al)v Ji-Parana´ Solimo˜es/Amazon† White-water tributaries† Black-water tributaries† Ji-Parana´ Solimo˜es/Amazon White-water tributaries Black-water tributaries coarse
coarse coarse
coarse fine
fine fine
fine Ϫ29.1
Ϫ28.0 Ϫ28.0
··· Ϫ28.1
Ϫ26.9 Ϫ27.7
Ϫ28.5 8.92
1.08 0.87
1.29 8.04
1.15 1.39
2.32 0.53
0.04 0.04
··· 0.72
0.10 0.16
0.23 18.0
24.5 20.6
··· 11.5
11.1 8.9
10.2 2.3
69.6 36.4
1.5 20.5
262.3 179.1
14.3 7.81
7.37 6.41
7.99 2.48
2.15 1.74
1.46 0.12
0.07 0.06
0.07 0.16
0.10 0.11
0.09 0.94
0.78 0.69
0.85 0.71
0.84 0.86
0.79 0.27
0.24 0.28
0.25 0.72
0.43 0.55
0.54 Notes: Values reported for ␦ 13 C are the deviations (‰) of the 13 C: 12 C of the samples from the same ratio for the PDB standard; cin:van is the ratio of the sum of cinnamyl phenols to vanillyl phenols; syr:van is the ratio of the sum of syringyl phenols to vanillyl phenols; ⌳ is the total yield of lignin-derived phenols normalized to 100 mg of organic carbon in the sample; (Ad/Al)v is the acid-to-aldehyde ratio of vanillyl phenols. White-water tributaries are the rivers Ic¸a, Jurua´, Japura´, Purus, and Madeira. Black-water tributaries are the rivers Jutaı´ and Negro. † Data from J. E. Richey ( unpublished data). variability in the data. Prior to PCA analyses, we unit normalized our data to have an average of zero and a standard deviation of one, since they did not follow a normal distribution. PCA with Varimax rotation was used to investigate the relationships among basin char- acteristics, such as altitude, slope, area covered with pasture and forest, soil texture (percentage of sand, silt, and clay), and compositional characteristics ( ␦ 13 C, ␦ 15 N, percentage of organic carbon, percentage of total nitrogen) of the three size fractions. Only variables with Ͻ20% of missing values were considered in the anal- ysis.
R ESULTS
Spatial variability Most of the sites had low concentrations of bulk coarse suspended solids, generally Ͻ3 mg/L. Samples from site PB-1 were the exception (no. 3, Fig. 3), with a statistically higher average concentration of 8 mg/L. Concentrations of bulk fine suspended solids averaged 15–35 mg/L at most sites. The 5 mg/L average at COM- 1 (no. 1, Fig. 4) was the exception and was statistically lower than the highest concentrations found at ROL (no. 10, Fig. 4). The mass percentages of organic car- bon (OC%) in the coarse fraction varied from 6.3% (JIP-5; no. 9, Fig. 3) to 11.8% (PB-2; no. 4, Fig. 3), but no significant statistical difference among sampling sites was detected, partially because the variability at each sampling site was large (Fig. 3). The same was true for the ultrafiltered-dissolved organic matter frac- tion. There was a variation from 2.7% (URU; no. 11, Fig. 5) up to 14% (JIP-2; no. 6, Fig. 5), but the vari- ability at each sampling site was also large. The variability of the OC% levels in the fine fractions at each sampling site was smaller than in the coarse and ultrafiltered-dissolved fractions with no statistical difference among sampling sites (Fig. 4). An inverse relationship was observed between the OC% in the fine fraction and fine suspended solids concentrations (Fig. 6). The mass percentages of total nitrogen (TN%) in all size fractions did not vary significantly among sam- pling sites, and the variability at each sampling site was smaller in the fine fraction than in the coarse and ultrafiltered-dissolved fractions (Figs. 3–5). The C:N atomic ratio of the coarse fraction varied from 11 (JAR; no. 12, Fig. 3) to 28 (MAC; no. 13, Fig. 3), with most values between 15 and 20 (Fig. 3). In spite of this large variability among sampling sites, no statistical differ- ence was detected among them. The C:N ratios of the fine fraction were less variable (9.5–12.5) among sam-
S270 MARCELO C. BERNARDES ET AL. Ecological Applications Special Issue T ABLE
3. Compositional averages ( Ϯ1 SD
Measure COM-1
COM-2 PB-1
PB-2 JIP-1
JIP-2 Coarse
⌳ (mg/100 g organic C) (Ad/Al)v
4.83 0.42
8.58 0.28
7.46 0.3
8.33 0.25
8.36 Ϯ 4.17
0.26 Ϯ 0.01
8.73 Ϯ 1.66
0.24 Ϯ 0.01
Fine ⌳ (mg/100 g organic C) (Ad/Al)v ND ND 2.3 0.83
ND ND 2.51 0.85 2.51
Ϯ 0.06 0.71
Ϯ 0.12 2.82
Ϯ 0.13 0.63
Ϯ 0.13 UDOM
⌳ (mg/100 g organic C) (Ad/Al)v
ND ND ND ND ND ND ND ND 0.77 2.07 0.83
2.07 Notes: ⌳ is the total yield of lignin-derived phenols normalized to 100 mg of organic carbon in the sample; (Ad/Al)v is the acid-to-aldehyde ratio of vanillyl phenols; ND , not determined. Sample site and size fraction codes are delineated in Methods. Standard deviation was not computed when N Ͻ 3. UDOM is ultrafiltered dissolved organic matter. pling sites with no statistical differences among them (Fig. 4). Finally, the C:N ratios of the ultrafiltered frac- tion had a large variability at each sampling site and among sampling sites, with no statistical differences among them (Fig. 5). The suspended solids concentrations in rivers of the Ji-Parana´ Basin were significantly smaller than the sus- pended solids concentrations found in the Amazon Riv- er and their white-water tributaries, and similar to the concentrations found in black-water tributaries (Negro and Jutaı´) of the Amazon River (Table 2). As a con- sequence, the organic carbon transported in the partic- ulate form was low. Most of the organic carbon (av- erage 72%) was transported in a dissolved form (UDOM): 26% as fine particulate carbon, and only 2% in the coarse fraction. The ␦
C average values of the coarse fraction varied only 2.5‰ between the highest ( Ϫ28.5‰; URU; no. 11, Fig. 3) and the lowest ( Ϫ31.0‰; MAC; no. 13, Fig. 3) values, and no statistical differences were ob- served among sampling sites. The variability at each sampling site was not large, with exception of the MAC (no. 13, Fig. 3). For the fine fraction the same trend was found, the ␦ 13
Ϫ26.7‰ (URU; no. 11, Fig. 4) and Ϫ30.7‰ (MAC; no. 13, Fig. 4). With the exception of two high ␦ 13
UDOM ( Ϫ21.8‰ at ROL; no. 10, Fig. 5; and Ϫ23.4‰ at JAR; no. 12, Fig. 5), the remainder of the values varied from Ϫ26 to Ϫ28‰ (Fig. 5). Due to low suspended-sediment concentrations in some tributaries of the Ji-Parana´ River, the biochemical composition (lignin-derived phenols) could not be de- termined in every sample (Table 3). The carbon nor- malized yields of total lignin-derived phenols ( ⌳, mg/ 100 mg organic C) of the coarse fraction varied be- tween 7 and 9 mg/100 mg organic C, with the exception of COM-1, where the average concentration was ϳ5 mg/100 mg organic C (Table 3). However, this lower value was not statistically different than the others. The average
⌳ values of the fine fraction were smaller than the coarse fraction and varied from 2 to 3 mg/100 mg organic C (Table 3). Finally, the acid:aldehyde ratio of vanillyl phenols, (Ad/Al)v, of the coarse fraction varied from 0.24 to 0.30 (Table 3). The exception was a value of 0.42 found in the COM-1 sampling site. The (Ad/ Al)v of the fine fraction was generally higher than the coarse fraction, varying from 0.63 to 0.83 (Table 3). The biochemical composition ( ⌳ and [Ad/Al]v) of the UDOM fraction was determined only for the sampling sites JIP-1 throughout JIP-4 and only for one sampling period (Table 3). Therefore, it was not possible to test for differences between these sampling sites. Seasonal variability In order to test for seasonal differences we grouped the data in high water period (February, March, and May) and in low water period (June, September, and November) (Fig. 2). In addition, as we did not, in gen- eral, find significant differences among sampling sites, we also grouped samples collected along a river under the same name in order to have enough statistical power to run the statistical test. Therefore, we included in this comparison only rivers with more than one sampling site, as Comemorac¸a˜o, Pimenta Bueno, and Ji-Parana´ (Table 4). Most of the seasonal differences were found in the Ji-Parana´ River. As expected, the bulk coarse average suspended solids concentration was higher in the high water period than in the lower water period. The same was true for the bulk fine average concentration, but in this case the averages were not statistically different (Table 4). The OC% and TN% levels were higher in the low-water period in the three fractions. However, only for the fine fraction was OC% level statistically different, and only for the coarse and fine fractions was the TN% statistically different (Table 4). The average C:N ratios of the three fractions were smaller during the low water, but only in the fine fraction was the difference statistically significant. For each of the three rivers, average ␦ 13 C values of the fine fraction were statistically lower in the low-water period than in the high-water period. Finally, the average ␦ 15 N value of the coarse fraction of the Ji-Parana´ River was statis- tically higher during the low water (Table 4).
August 2004 S271
ORGANIC MATTER OF AMAZON LOWLAND RIVERS T ABLE 3. Extended. JIP-3 JIP-4
JIP-5 ROL
URU JAR
MAC 7.46
Ϯ 0.38 0.24
Ϯ 0.01 6.97
Ϯ 2.35 0.26
Ϯ 0.01 7.87
0.25 8.87
0.23 8.17
0.27 ND ND ND ND 2.76 Ϯ 0.50 0.70
Ϯ 0.16 2.07
Ϯ 0.84 0.78
Ϯ 0.22 2.36
Ϯ 0.22 0.68
Ϯ 0.07 ND ND ND ND ND ND ND ND 0.73 2.72
0.58 2.2
ND ND ND ND ND ND ND ND ND ND Compositional differences among size fractions The bulk fine suspended solids concentrations were statistically higher than the bulk coarse concentrations (Table 5). The average OC% and TN% levels were only statistically higher in the UDOM fraction. The C:N average ratio of the fine fraction was statistically small- er than the ratios of the coarse and UDOM fractions (Table 5). The ␦ 13
decreasing size fraction, with a difference of ϳ2.5‰
between the coarse and UDOM fraction. The average ␦ 15 N value of the UDOM fraction was statistically high- er than the coarse and fine fractions. The carbon nor- malized yields of total lignin-derived phenols ( ⌳, mg/
100 mg organic C) statistically decreased with decreas- ing size fraction and the (Ad/Al)v statistically in- creased with decreasing size fraction (Table 5).
We compared elemental composition (OC% and N%), stable isotopes characteristics ( ␦ 13 C and ␦ 15 N), and biochemical composition ( ⌳ and [Ad/Al]v) of the size fractions of the Ji-Parana´ Basin rivers with poten- tial end members. Based on previous work of Hedges et al. (1986, 2000) and Martinelli et al. (1999), we selected as potential organic-matter sources forest soil organic matter, pasture soil organic matter, tree leaves (C 3
4 type). As there were no significant differences between sampling sites, we grouped Ji-Parana´ sites together and focused on com- positional differences between the three size fractions. The fine and the coarse fractions of the Ji-Parana´ Rivers were characterized by high concentrations of carbon and nitrogen. A plot of these two parameters produced significant correlation coefficients for the particulate organic matter ( r 2 ϭ 0.74 for coarse and r 2 ϭ 0.72 for fine) and no correlation for the ultrafiltered organic matter (Fig. 7, UDOM data not shown). Both lines have a small positive intercept in the OC% axis that is not statistically different from zero, indicating that most of the nitrogen was in an organic form (Hedges et al. 1986). The three fractions had, in general, higher car- bon and nitrogen concentrations than Rondoˆnia soils and smaller concentrations than found in leaves (Fig. 7). The ␦
C and ␦ 15 N values of the fine and coarse fractions were similar to values found in the soil or- ganic matter of Rondoˆnia forest soils. The coarse frac- tion had the most negative ␦ 13
this fraction nearest to tree leaves compositionally (Fig. 8). The UDOM fraction had higher ␦ 13
␦ 15 N values than the particulate fractions, and the ␦ 13 C average val- ue was higher than values found in the soil organic matter (Fig. 8). None of the three fractions had ␦ 13 C values similar to C 4 leaves; consequently, the three fractions plotted distant from the C 4 leaves average in Fig. 8. The same occurred when ␦ 13 C values were plot- ted as a function of N:C ratios (the inverse of the con- ventional C:N is required so that both axes are math- ematically independent; Fig. 9). The fine and UDOM fractions plotted near forest soils, whereas the coarse fraction plotted between tree leaves and forest soils, but both fractions plotted far from soil organic matter of pastures of different ages (Fig. 9). High syringyl: vanillyl and cinnamyl:vanillyl ratios, specific phenols from lignin, were measured in coarse size fractions followed by the fine size fractions. Similar values were found for Amazon rivers and its tributaries (Table 2 and Fig. 10).
The acid to aldehyde ratio of lignin vanillyl phenols, (Ad/Al)v, is a robust indicator of lignin decomposition, with values increasing with progressive degradation by fungi and bacteria (Hedges et al. 1988, Opsahl and Benner 1995). For samples from the Ji-Parana´, the (Ad/ Al)v ratio increases from the larger to the smaller size fraction, corresponding to a decrease of total yields of lignin-derived phenols (Fig. 11A). This trend indicates a progressive loss of lignin during the degradation pro- cess, as was observed for the coarse and fine fractions of the Amazon River (Fig. 11A; Ertel et al. 1986). Fig. 11B shows that as the riverine fractions become more degraded (coarse Ͻ fine Ͻ UDOM) there was a pro- gressive increase in ␦ 13
a standard isotope effect in which carbonaceous resi- dues became progressively 13 C enriched, as less strong- ly bonded 12 C carbons are preferentially lost. S272 MARCELO C. BERNARDES ET AL. Ecological Applications Special Issue T ABLE
4. Compositional averages ( Ϯ1 SD
Bueno, and Ji-Parana´, grouped according the water level. Measure
Comemorac¸a˜o High
Low Pimenta
Bueno High
Coarse N ␦ 13 C ␦ 15 N Organic C (mass %) Total N (mass %) C:N
Suspended sediments (mg/L) 4 Ϫ29.0 Ϯ 0.4 3.8 Ϯ 0.4
9.9 Ϯ 6.9
0.54 Ϯ 0.37
19 Ϯ 2.1
2.88 Ϯ 2.11
6 Ϫ29.1 Ϯ 0.3 4.4 Ϯ 0.2
8.4 Ϯ 2.8
0.43 Ϯ 0.15
19.4 Ϯ 1.2
1.34 Ϯ 0.68
4 Ϫ29.2 Ϯ 0.1 3.5 Ϯ 0.6
12.5 Ϯ 7.5
0.59 Ϯ 0.39
21.0 Ϯ 1.7
7.50 Ϯ 3.60
Fine N ␦ 13 C ␦ 15 N Organic C (mass %) Total N (mass %) C:N
Suspended sediments (mg/L) 3 Ϫ27.7 a Ϯ 0.5
4.4 Ϯ 4.2
8.7 Ϯ 6.4
0.81 Ϯ 0.64
11.3 Ϯ 2.2
27.6 Ϯ 26.9
5 Ϫ28.6
b Ϯ 0.4
3.9 Ϯ 0.8
10.7 Ϯ 2.3
0.80 Ϯ 0.20
13.6 Ϯ 1.3
8.05 Ϯ 8.92
3 Ϫ27.6
a Ϯ 0.4
3.2 Ϯ 1.2
7.4 Ϯ 3.1
0.69 Ϯ 0.29
10.9 Ϯ 1.1
25.4 Ϯ 23.1
UDOM N ␦ 13 C ␦ 15 N Organic C (mass %) Total N (mass %) C:N
1 Ϫ27.6
7.5 14 0.35 40.1 3 Ϫ27.2 Ϯ 0.9 9.7 Ϯ 2.3
3.5 Ϯ 1.7
1.72 Ϯ 1.44
5.6 Ϯ 7.4
1 Ϫ27.2
9.3 10.6
0.42 25.3
Notes: Values reported for ␦ 13 C are the deviations (‰) of the 13 C: 12 C of the samples from the same ratio for the PDB standard; ␦ 15 N are the deviations (‰) of the 15 N: 14 N of the samples from the same ratio for the atmospheric N 2 stable isotope standard. Standard deviation was not computed when N Ͻ 3. ‘‘High’’ denotes the high-water period (February–May) and ‘‘Low’’ denotes the low-water period (June–November). UDOM is ultrafiltered dissolved organic matter. Different letters indicate statistically significant differences between averages. Principal component analysis The communality for each variable, which represents the fraction of each variable that is explained by the retained factor, were typically higher than 65%. The last lines of Tables 6–8 express the percentages of var- iance explained by each factor. For the coarse fraction, PCA explained 80.6% of the variability of the data. The first two factors explained 24.5% and 24.1% of the data variability, respectively. The third and fourth fac- tors explained 17.5% and 14.5%, respectively. The first factor had significant loadings for ␦ 15 N and soil clay content (Table 6). The second factor had high loading for ␦
C and pasture area and soil silt content. The third factor is loaded with OC% and TN% and fourth factor with forest area, altitudes, and slopes. For the fine frac- tion, PCA explained 78.3% of the variability of the data. The first factor alone explains almost 29% of the variability and had significant loadings for ␦ 13
␦ 15 N with pasture area and silt (Table 7). The second factor, which explains 25% of the variability had sig- nificant loadings especially for soil sand and clay con- tents. The third factor explains ϳ15% of the variability and significant loads are OC% and TN%. The fourth factor explains only 10% of the variability and links mainly forest area, and altitudes and slopes. Finally, the UDOM explained 78.8% of the data variability. The first factor explains 23.5% of the variability, and had significant loadings again for ␦ 13
(Table 8). The second factor had high loadings only for soil clay content. The third factor had high loadings for ␦
N, OC%, and TN%. The fourth factor had high loadings for forest area, altitude, and slopes. D ISCUSSION The rivers of the Ji-Parana´ Basin originate mostly in the lowlands of the Rondoˆnia State and drain mainly the Precambrian Brazilian Shield (Fig. 1). Consequent- ly, these rivers have lower sediment concentrations than white-water rivers of the Amazon Basin that have their headwaters in the Andes or in the Andes foothills, and are more similar to other black-water tributaries of the Amazon system such as the Negro and the Jutaı´. This geomorphological context has important impli- cations for the source of organic matter to rivers in the Amazon. For instance, because of the increase in the ␦ 13 C composition of C 3 plants with increasing eleva- tion, the vegetation of the Andes is a source of 13 C- rich particulate matter to the Amazon main channel (Quay et al. 1992). Progressively, this 13 C-rich organic matter is replaced and diluted by 13 C-poor organic mat- ter produced in the Amazon lowlands. Consequently, the
␦ 13 C values of the fine and coarse organic matter August 2004 S273
ORGANIC MATTER OF AMAZON LOWLAND RIVERS T ABLE 4. Extended. Pimenta Bueno
Low Ji-Parana´ High Low
6 Ϫ28.9 Ϯ 0.3 4.2 Ϯ 0.8
8.1 Ϯ 3.8
0.43 Ϯ 0.21
19.0 Ϯ 1.4
4.47 Ϯ 6.21
14 Ϫ28.9 Ϯ 0.4 4.8 a
7.3 Ϯ 4.2
0.40 a Ϯ 0.23 18.3 Ϯ 2.1
3.42 a Ϯ 1.80 19 Ϫ28.9 Ϯ 0.9 6.5 b
9.4 Ϯ 3.5
0.63 b Ϯ 0.25 15.2 Ϯ 2.9
0.85 b Ϯ 0.58 5 Ϫ28.6
b Ϯ 0.5
4.4 Ϯ 0.9
6.5 Ϯ 1.8
0.54 Ϯ 0.17
12.3 Ϯ 1.2
19.7 Ϯ 12.5
14 Ϫ27.5
a Ϯ 0.5
5.2 Ϯ 1.5
6.8 a Ϯ 1.0 0.54 a Ϯ 0.09 12.6 a Ϯ 1.0 25.8 Ϯ 7.1
19 Ϫ28.4
b Ϯ 1.1
5.5 Ϯ 0.8
8.3 b Ϯ 1.1 0.80 b Ϯ 0.21 10.8 b Ϯ 2.4 21.5 Ϯ 10.3
3 Ϫ25.7 Ϯ 3.5 11.2 Ϯ 5.8
5.4 Ϯ 2.4
1.06 Ϯ 0.85
8.7 Ϯ 6.9
10 Ϫ27.1 Ϯ 0.8 7.6 Ϯ 1.0
11.4 Ϯ 5.8
0.58 Ϯ 0.18
21.7 Ϯ 11.8
13 Ϫ26.9 Ϯ 1.7 6.6 Ϯ 2.1
12.6 Ϯ 7.5
1.08 Ϯ 1.12
19.4 Ϯ 10.0
T ABLE
5. Compositional averages ( Ϯ1 SD
Fraction ␦ 13 C (‰) ␦ 15 N (‰) Organic C (mass %) Total N
(mass %) C:N
Suspended sediments (mg/L) ⌳ (mg/100 g organic C) (Ad/Al)v
Coarse Fine
UDOM Ϫ29.1
a Ϯ 1.0
Ϫ28.1 b Ϯ 1.2 Ϫ26.7 c Ϯ 1.8 5.4 a Ϯ 1.7 5.0 a Ϯ 1.5 7.9 b Ϯ 2.6 8.9 a Ϯ 4.4 8.0 a Ϯ 2.3 9.8 b Ϯ 6.7 0.53 a Ϯ 0.27 0.72 a Ϯ 0.25 0.85 b Ϯ 0.88 18 a Ϯ 7 12 b Ϯ 2 18 a Ϯ 11 2.3 a Ϯ 2.9 20.5 b Ϯ 12.5 7.81 a Ϯ 1.71 2.48 b Ϯ 0.72 0.73 c Ϯ 2.27 0.27 a Ϯ 0.05 0.72 b Ϯ 0.14 2.27 c Ϯ 0.31 Notes: Values reported for ␦ 13 C are the deviations (‰) of the 13 C: 12 C of the samples from the same ratio for the PDB standard; ␦ 15 N are the deviations (‰) of the 15 N: 14 N of the samples from the same ratio for the atmospheric N 2 stable isotope standard; ⌳ is the total yield of lignin-derived phenols normalized to 100 mg of organic carbon in the sample; (Ad/Al)v is the acid-to-aldehyde ratio of vanillyl phenols. UDOM is ultrafiltered dissolved organic matter. Different letters indicate statistically significant differences between averages. become more negative downstream (Quay et al. 1992, Hedges et al. 2000). This is one of the major changes in the composition of the particulate organic matter of the entire Amazon Basin, and sharply contrasts with the elemental and biochemical uniformity of the lower main stem which experiences few changes over a river reach of ϳ2000 km (Hedges et al. 1992). The same lack of pronounced changes in the bulk composition of the riverine organic matter was observed within size fractions sampled over a stretch of the Beni River, ex- tending from its headwater near to the city of La Paz (first order streams) to its confluence with the Madre de Dios River in the southeast Amazon region (Hedges et al. 2000). Investigating a 970-km stretch of the Ji- Parana´ River we reached the same conclusion. Differ- ences in the elemental, biochemical, and stable isotope composition were not very different among the main channel or subject to significant downstream changes. On the other hand, we observed compositional differ- ences among the fine, coarse, and dissolved organic fractions from the same water samples. Another im- portant similarity between the Ji-Parana´ River and the major rivers of the Amazon Basin is that both appear to share common organic-matter sources. The three size fractions appear to have a common source: tree leaves and soil organic matter from the tropical rainforest. We reached this conclusion for the Ji-Parana´ Basin based on the fact that the three fractions exhibit compositions near those of the tree leaf and soil organic-matter end members (Figs. 3, 4, and 5). Secondly, the Ji-Parana´ size fractions plotted near the Amazon River particles, suggesting that both systems have similar organic-mat- ter sources. The organic degradation stages of the size fractions in the Ji-Parana´ Basin also appear to be sim- ilar to what has been found in other Amazonian and South American rivers (Martinelli et al. 1999, Hedges et al. 2000, Devol and Hedges 2001, Krusche et al. 2002). The coarse fraction of the Ji-Parana´ River ap- pears to be the least degraded and resembles the re- mains of tree leaves (Fig. 11A). The relatively high syringyl/vanillyl and cinnamyl/vanillyl ratios measured in coarse and fine size fractions indicate that an im- portant amount of lignin in Ji-Parana´ Basin originates from non-woody angiosperm plants (Fig. 10). The fine fraction is more degraded than the coarse fraction and during this process there was a loss of lignin-derived phenols and an increase in the ␦ 13
Although the fine fraction is more degraded, this frac- tion is richer in nitrogen (Figs. 7 and 9). This char- acteristic was also found in other rivers of the Amazon (Hedges et al. 2000) and also in rivers of southeast Brazil (Krusche et al. 2002). This N enrichment is prob- ably due to the fact that nitrogenous organic matter selectively accumulates in the fine fractions with time by preferential sorption on soil minerals before they are eroded into aquatic systems (Hedges et al. 2000, Aufdenkampe et al. 2001). In addition, nitrogen-rich remains of microbial fauna tend to concentrate in fine- grain minerals (Hedges and Oades 1997, Amelung et al. 1999). Finally, the most degraded fraction appears to be the ultrafiltered dissolved organic material (UDOM; Fig. 11).
S274 MARCELO C. BERNARDES ET AL. Ecological Applications Special Issue F IG
(A) Plot of organic carbon vs. total nitrogen for Ji-Parana´ coarse fractions (open squares), fine fractions (open circles), leaf end members (rectangles representing the dis- tribution of C 4 leaves from pastures and C 3 leaves from for- ests), and soils of Rondoˆnia (filled circles). (B) Zoom of region outlined in (A). (C) Plot of organic carbon vs. total nitrogen for the Amazon River coarse (open square) and fine (open circle) fractions. Values are from the following sources: tree leaves, L. A. Martinelli and J. E. Ehleringer (
Manaus and Santare´m soils, E. V. Telles ( unpublished data); Amazon River coarse and fine fractions, Hedges et al. (1986). F IG
Plot of ␦ 15 N vs. ␦ 13 C for the three size fractions (coarse, fine, and ultrafiltered dissolved organic matter [UDOM]; error bars represent Ϯ1 SD ) and the following end members: C 4 leaves from pastures; C 3 leaves from forests, and C 3
distribution of all points). Values are from the following sources: C 4 leaves from pasture were collected in Santare´m (L. A. Martinelli and J. E. Ehleringer, unpublished data); tree leaves were collected in Ji-Parana´, Santare´m, and Manaus (L. A. Martinelli and J. E. Ehleringer,
soil organic matter was collected in forests near Manaus and Santare´m (E. V. Telles,
In general, the compositional differences between the Ji-Parana´ River, the Amazon River, and the Amazon major tributaries were not large enough to indicate dif- ferent organic matter sources among the rivers. The exception is the higher carbon and nitrogen concentra- tions in the coarse and fine fractions of rivers of the Ji-Parana´ Basin in comparison with the Amazon River and its major white-water tributaries (Table 2), includ- ing the Beni Basin (Hedges et al. 2000). Even the heavi- ly sewage-contaminated rivers of the Piracicaba Basin had lower carbon and nitrogen concentration in these fine and coarse fractions (Krusche et al. 2002). We do not have a good explanation for these much higher C and N concentrations in the coarse and fine fractions of Ji-Parana´ rivers. One possible explanation indepen- dent of recent land-cover alterations would be that riv- ers of the Ji-Parana´ Basin drain mainly lowland forests, which constitute a continuous source of organic matter with little diluting mineral matter. The suspended solids concentration is an order of magnitude smaller in the Ji-Parana´ Basin in comparison to the Solimo˜es/Amazon River and its white-water tributaries (Table 2). This explanation is consistent with lower ␦ 13
in the coarse and fine particles of Ji-Parana´ rivers in relation to the white-waters Amazon rivers, which drain high altitude regions of the Andean mountains that are a source of 13 C-enriched organic matter (Quay et al. 1992, Hedges et al. 2000). A similar situation occurs in the Jutaı´ and Negro rivers (black-water tributaries) that drain exclusively Amazonian lowlands and com- bine lower OC% and TN% levels with lower ␦ 13
ues than white-water rivers (Table 2). The fine fractions of the Ji-Parana´ rivers follow the classic inverse rela- tionship between OC% and suspended solids concen- trations (Fig. 6). The majority of the particulate organic matter in the Amazon main stem is associated with mineral grains and is a direct function of the total sur- August 2004 S275
ORGANIC MATTER OF AMAZON LOWLAND RIVERS F IG . 9. Plot of N:C ratio vs. ␦ 13
three size fractions (coarse, fine, and ultrafil- tered dissolved organic matter [UDOM]; error bars represent Ϯ1 SD ) and the following end members: tree leaves from forests; forest soil; soil covered with a pastures of age 3–5 yr, 7– 13 yr, 20 yr, and 80 yr (boxes represent the distribution of all points). Tree leaves were col- lected in Ji-Parana´, Santare´m, and Manaus (L. A. Martinelli and J. E. Ehleringer,
in forests near Manaus and Santare´m (E. V. Telles,
collected at several sites in the Rondoˆnia State (Neill et al. 1997). F IG . 10. Plot of syringyl:vanillyl ratio vs. cinnamyl:van- illyl ratio for the three size fractions (coarse, fine, and ul- trafiltered dissolved organic matter [UDOM]; error bars rep- resent Ϯ1
); coarse and fine fractions of the Amazon River (boxes represent the distribution of all points); and forest tree leaves as end members (boxes represent the distribution of all points). Values for the Amazon coarse and fine fractions and leaf end members are from Hedges et al. (1986). F IG . 11. Plots of acid:aldehyde ratio of vanillyl phenols ((Ad/Al)v) vs. (A) carbon-normalized yields of total lignin- derived phenols ( ⌳) and (B) ␦ 13 C for the three size fractions (coarse, fine, and ultrafiltered dissolved organic matter [UDOM]; error bars represent Ϯ1 SD
fractions of the Amazon River (boxes represent the distri- bution of all points), and forest tree leaves as end members (boxes represent the distribution of all points). Values for the Amazon coarse and fine fractions and leaf end members are from Hedges et al. (1986). face area of suspended sediment particles, i.e., pro- portionately more carbon is attached to smaller parti- cles (Keil et al. 1994). If the same association holds for the Ji-Parana´ Basin, it is likely that, in addition to the lack of dilution of the particulate organic matter by higher suspended particles concentrations, a higher proportion of the particulate organic matter load is transported by smaller particles, with proportionately more surface area than the Amazon main stem and its white-waters tributaries. The land cover of the Rondoˆnia State, and especially of the Ji-Parana´ Basin, has experienced significant changes in the last 30–40 yr. Approximately 35% of the area of the Ji-Parana´ Basin has been altered, the main change being the replacement of the original for- est by pastures, which in 1999 occupied 30% of this basin (Ballester et al. 2003). In some sub-basins, such as the Rolim de Moura, the area covered with pastures reaches almost 70% (Table 1). In the basins of the rivers Urupa´ and Jaru´, approximately half of the forest has been replaced by pasture (Table 1). Less-altered areas are located primarily in the lower portions of the basin, which includes the sub-basins of the rivers Machad- inho, JIP-4, and JIP-5 (Table 1). The fact that essentially all pastures in the Ji-Parana´ Basin are cultivated with C 4 grass species provides an opportunity to test whether large land-use changes have S276 MARCELO C. BERNARDES ET AL. Ecological Applications Special Issue T ABLE
6. Matrix of component loading of principal component analysis for the coarse organic matter fraction. Variable
Factor 1 Factor 2
Factor 3 Factor 4
Communalities (%)†
␦ 13 C ␦ 15 N Organic C (%) Total N (%) Forest (%) Pasture (%) 0.69 0.31
0.79 0.81
0.96 0.87
0.25 0.81
82 77 95 86 87 86 Sand (%) Clay (%)
Silt (%) Altitude
Slope Variance (%) 0.85 0.45
24.5 0.78
24.1 0.30
17.5 0.79
0.92 14.5
86 85 86 86 86 80.6 Note: Only factors larger than 0.2 are shown. † The percentage of variance accounted for by the current number of factors. T ABLE
7. Matrix of component loading of principal component analysis for the fine organic matter fraction. Variable
Factor 1 Factor 2
Factor 3 Factor 4
Communalities (%)†
␦ 13 C ␦ 15 N Organic Total N
Forest Pasture
0.85 0.61
0.85 0.40
0.33 0.97
0.95 0.37
0.79 85 75 94 94 80 84 Sand (%)
Clay (%) Silt (%)
Altitude Slope
Variance (%) 0.83
28.5 0.98
0.97 25.0
14.6 0.85
0.78 10.2
97 96 84 81 75 78.3 Note: Only factors larger than 0.2 are shown. † The percentage of variance accounted for by the current number of factors. already affected the compositional characteristics of the organic-matter size fractions carried into this basin. This is because C 4 grasses have ␦ 13 C values varying between Ϫ11 and Ϫ14‰, while forest C 3 leaves in the tropics have characteristically low ␦ 13 C values, varying from
Ϫ28 to Ϫ34‰ (Farquhar et al. 1989, Martinelli et al. 1998). In ecosystems where C 4 grasses naturally occur, their presence can be detected either in the soil (Cerri and Volkoff 1987) or in river particles (Mariotti et al. 1991, Bird et al. 1994, 1998). In agricultural ecosystems, where an original C 3 forest has been re- placed by some kind of C 4 plants, the C 4 -derived or- ganic matter quickly becomes incorporated into the sur- face soil layers (Cerri et al. 1985, Vitorello et al. 1989, Moraes et al. 1996, Neill et al. 1996, 1997). For in- stance, the ␦ 13
at the Fazenda Nova Vida, 120 km north of the city of Ji-Parana´, changed from
Ϫ28.1‰ to Ϫ24.7‰, Ϫ20.6‰, and Ϫ18.3‰ after 3, 9, and 20 yr of pasture cultivation, respectively (Neill et al. 1996, 1997). New- ly introduced C 4 material appears also to be quickly incorporated in the aquatic systems. A preliminary in- vestigation in this cattle ranch (Fazenda Nova Vida) on the stable isotopic composition of riverine organic matter size fractions of a first order stream, suggests that changes start in small streams. We found a change in the
␦ 13 C of the fine fraction from Ϫ27.2‰ to Ϫ20.5‰ when a small stream (ϳ3 m width) leaves the forest and enters in a pasture in a reach of only 100 m, (L. F. Charbel, unpublished data). The same riverine organic matter dynamic was observed in rivers of the Piracicaba River basin, where the introduction of C 4 agricultural fields occurred 70–80 year ago and an in- crease in the ␦ 13 C values of the size fractions was de- tected (Martinelli et al. 1999). This replacement, from forest to sugar cane or pasture, was mainly associated with faster cycling fraction (UDOM), while the original vegetation remains were mainly associated with slower cycling fractions (CSS and FSS) (Krusche et al. 2002). In the Amazon River and in the Beni River, areas cov- ered mainly by C 4 grasses had the UDOM fraction with the heaviest ␦ 13 C values in relation to the fine and coarse fraction (Hedges et al. 1986, 2000). Conversely, August 2004 S277
ORGANIC MATTER OF AMAZON LOWLAND RIVERS T ABLE 8. Matrix of component loading of principal component analysis for the unfiltered dissolved organic matter (UDOM) fraction. Variable
Factor 1 Factor 2
Factor 3 Factor 4
Communalities (%)†
␦ 13 C ␦ 15 N Organic Total N
Forest Pasture
0.83 0.93
0.27 0.40
0.80 0.81
0.70 0.88
79 75 79 77 83 91 Sand (%) Clay (%)
Silt (%) Altitude
Slope Variance (%) 0.74 23.5
0.99 0.30
20.1 19.8
0.91 0.79
15.4 99 98 78 86 83 78.8 Note: Only factors larger than 0.2 are shown. † The percentage of variance accounted for by the current number of factors. in areas where C 4 grasses are not important, the ␦ 13 C of the UDOM fraction was lightest. Comparisons of ␦ 13
size fractions with those of tree leaves and forest soils (Figs. 8 and 9) did not allow us to detect a major signal of the presence of C 4 -derived organic matter in the main channel of the Ji-Parana´ River. The average defores- tation extent of the all sampled basins is relatively low ( ϳ30%), which may alone account for the low riverine signal. On the other hand, the PCA developed with all data showed correlations between the areas covered with pasture and the ␦ 13 C values of the three size frac- tions (Tables 6–8). In addition, the highest ␦ 13
of our results were observed in the UDOM of the Ro- lim-de-Moura and Jaru´ rivers, which have among the highest areas covered with pasture (Table 1). The lower the order of the streams and the higher the pasture area, greater is the possibility that the C 4 -derived organic matter signal will be detected first in the faster-cycling fraction (UDOM). The fact that we found a strong correlation between pasture area and ␦ 13
suggests that the large-scale deforestation in Rondoˆnia State, started in early 1970, is affecting the origin of the carbon in the Ji-Parana´ River basin. The extent of changes in the carbon isotopic composition of size frac- tions in the Ji-Parana´ Basin should be investigated in more detail. However, the strong association among ␦ 13
that soil texture may have a strong influence on how the C
4 signal moves from the terrestrial to the aquatic system. In the Congo River Basin, Mariotti et al. (1991) noticed that, in areas where there was a predominance of sandy soils, the ␦ 13 C of riverine size fractions was controlled mainly by riparian vegetation. In contrast, in areas of the basin where clay-texture soils predom- inate, the entire vegetation cover of the basin was the most important factor. Riparian vegetation in Rondoˆnia rivers has a predominance of forests with C 3 plants. A preliminary survey of the soil texture of the A horizon in Rondoˆnia State revealed a high percentage of sand in several sub-basins of the Ji-Parana´ Basin (Table 1). Therefore, depending on the soil texture, riparian forest vegetation could be more important than the area of the basin covered with pasture (Mariotti et al. 1991, McClain et al. 1997). C ONCLUSIONS Most of the earlier studies in the Amazon Basin riv- ers focused on rivers that had their headwaters in the Andes or sub-Andean regions (Hedges et al. 1986, 1994, 2000). These rivers are characterized by high suspended solids concentrations (white-water rivers), and their riverine organic matter has two geographi- cally distinct allochthonous sources, the Andean region and the Amazon lowlands. The rivers that we inves- tigated in this study have their drainage basin almost exclusively in the Amazon lowlands, draining mainly the Precambrian shield. These clear to black-water riv- ers have a lower concentration of suspended solids, and lowlands are their only sources of allochthonous or- ganic matter. Therefore, the results of this study extend our knowledge about the composition and dynamics of size fractionated organic matter over South America river types that have not been extensively investigated in the past. The principal differences that were observed for the Ji-Parana´ river system included higher carbon and ni- trogen concentration found in the three size fractions of Ji-Parana´ rivers in comparison with Amazon white- water and black-water tributaries (Table 3). We do not have a definite explanation for these differences. However, some of the characteristics of organic mat- ter from white-water rivers were similar for rivers of the Ji-Parana´ Basin. For instance, we observed com- positional differences between the three size fractions that suggested substantial differences in the stages of degradation between size fractions despite similar or- ganic matter sources. The coarse fraction is the least degraded and its main source appears to be leaves from S278 MARCELO C. BERNARDES ET AL. Ecological Applications Special Issue lowland forests (Devol and Hedges 2001). The fine fraction is mostly associated with a mineral soil phase, but its ultimate source appears also to be leaves from forests. This fraction was also enriched in nitrogen as for other rivers of the Amazon. The ultrafiltered-dis- solved organic fraction appears to have the same source as the coarse fraction, but this is the most degraded fraction. Although the organic matter transported throughout the main channel of the Ji-Parana´ River seems to share the same sources as the Amazon Rivers, principal com- ponent analysis showed high communality for factors explained by the variables ␦ 13 C and pasture area in all three size fractions. The highest ␦ 13
served in the UDOM of the sub-basins with the highest areas covered by pasture. Finally, the large change in land-use in the Ji-Parana´ Basin, replacement of primary forests by C 4 pastures
for cattle feeding, that has taken place in the last 30– 40 yr, has already changed the composition of the riv- erine organic matter size fractions. Any attempt to un- derstand those changes must take into account the fast- er-cycling fraction (UDOM) and the low-order streams. A CKNOWLEDGMENTS We thank John Melack, Trent Biggs, and one anonymous reviewer for providing valuable comments and discussion to the manuscript. This work was supported by FAPESP (Proc. 01/07580-5 and Projeto Tema´tico 99/01159-4) and by NASA– LBA Ecology (CD-06 and ND-09). L ITERATURE C ITED
Amelung, W., R. Bol, and C. Friedrich. 1999. Natural C-13 abundance: a tool to trace the incorporation of dung-derived carbon into soil particle-size fractions. Rapid Communi- cations in Mass Spectrometry Download 332.19 Kb. Do'stlaringiz bilan baham: 
|