Use of absorption optical indices to assess seasonal variability of dissolved organic matter in amazon floodplain lakes

. Given the importance of dissolved organic matter (DOM) in the carbon cycling of aquatic ecosystems, information 10 on its seasonal variability is crucial. In this study we assess the use of available absorption optical indices based on in situ data to both characterize the seasonal variability of DOM in a highly complex environment and for application in large-scale studies using remote sensing data. The study area comprises four lakes located at the Mamirauá Sustainable Development Reserve (MSDR). Samples for the determination of coloured dissolved organic matter (CDOM) and remote sensing reflectance (Rrs) were acquired in situ. The Rrs was used to simulate the response of the visible bands of the Multi-Spectral Instrument 15 (MSI)/Sentinel and used in the proposed models. Differences between lakes were tested regarding CDOM indices. The results highlight the role of the flood pulse in DOM dynamic in the flood plain lakes. The validation results showed that the proposed model using a CDOM as proxy of S 275-295 during rising water is worthwhile, demonstrating its potential application to Sentinel/MSI imagery data for studying DOM dynamics on large scale studies.


Introduction
Floodplain is a type of wetland characterized by a mosaic of landscapes which oscillates periodically between aquatic and terrestrial systems.This oscillation represents a key aspect in the biogeochemistry, ecology and hydrology of floodplain lakes (Junk et al., 1989;Moreira-Turq et al., 2004).Among other effects, the flood pulse (sensu Junk et al., 1989) affects the proportion of autochthonous and allochthonous sources contributing to the dissolved organic matter (DOM) pool in floodplain lakes throughout the year (Melo et al., 2019).DOM represents the largest pool of organic carbon in the aquatic environment and it has an important role in the ecosystem carbon budgets (Seekell et al., 2018;Tranvik et al., 2009;Richey et al., 2002).Besides that, DOM also controls light availability in the water column, playing a vital role in primary productivity of aquatic ecosystems and consequently fisheries and other food webs (Hastie et al., 2019;Maia and Volpato, 2013;Volpato et al., 2004).DOM concentration in the environment is usually determined by the concentration of dissolved organic carbon (DOC) (Coble, 2007).However, simple measurement of DOC concentration can limit the study of the seasonal variation in the DOM quality (e.g.composition) and origin since it is related only to the bulk of DOM (Jaffé et al., 2008).Qualitative parameters are needed to better understand DOM dynamics such as ultraviolet (UV) and visible absorption measurements and fluorescence, which are an alternative for high costly laboratory analysis (Li and Hur, 2017).Helms et al., (2008) have shown that the spectral slope calculated in the range of 275 and 295 nm (S275-295) is an indicator of DOM molecular weight and a tracer of degradation processes.The absorption coefficient of coloured dissolved organic matter (CDOM) at 350 nm (aCDOM (350)), aCDOM spectral slope (S275-295) and DOC concentration have been used to study the rates of water exchange between river and floodplain on Mississippi and Atchafalaya River (Spencer et al. 2008;Shen et al., 2012).
In order to study DOM dynamics in wide spatial-temporal scale satellite images have been assessed as a source of optical information about CDOM.Many studies have used Landsat images for investigating aCDOM at the different wavelengths, more commonly at 350 nm, 440 nm and 420 nm (Fichot et al., 2013;Kutser et al., 2005;Zhu et al., 2014;Brezonik et al., 2015).However, only a few studies have looked at the spectral slope of DOM.In a Pan-Arctic study, Fichot et al. (2013) showed that S275-295 can be directly estimated from satellite images using a multi-linear parameterization of MODIS marine reflectance.
However, the reflectance of water in the visible bands may not reflect changes in the spectral slope of CDOM in the UV domain as show by Vantrepotte et al. (2015) by applying Fichot et al. (2013) model in three coastal water regions.Therefore, Vantrepotte et al., (2015) proposed the use of aCDOM as a proxy for S275-295 as it proved to be less affected by water optical quality and atmospheric correction.Nonetheless, both studies (Fichot et al., 2013;Vantrepotte et al., 2015) used MODIS data whose spatial resolution (250-1000 m) restricts the application to inland water studies.In recent years the availability of Multi-Spectral Instrument (MSI) images, on board of the Sentinel-2A (June/2015) and Sentinel-2B (March/2017), has expanded the potential of remote sensing application for DOM monitoring because of its high spatial (10 and 20 m), temporal (5 days) and radiometric (12-bit) resolutions (Toming et al., 2016).
The main objectives of this study are to: i) investigate the variability of aCDOM in floodplain lakes during the receding and rising phases of the Solimões River, ii) examine the potential of S275-295 for distinguishing differences in DOM by comparing it in two hydrograph phases; and iii) propose an algorithm to estimate aCDOM (440) as a proxy for S275-295 using simulated MSI/Sentinel bands to support future application of satellite remote sensing for inland DOM studies.

Study area
The study sites are four lakes located in the floodplain built at the confluence between Solimões and Japurá rivers, near Tefé and inside the Mamirauá Sustainable Development Reserve (MSDR) (Figure 1b), a well-preserved flooded forest under low human pressure (Ayres, 1995;Castello et al., 2009;Mori et al., 2019;Queiroz, 2007).In this area, the seasonal flood is caused by both the rainfalls (in upper Amazon basin and locally-from December to May, with an average of 300 mm/month) and by the annual melt of the Andean cordillera during the austral summer (Junk, 1989).The yearly MSDR flood pulse causes, in average, 12 meters amplitude in the water level between the dry (September to November) and the flood season (May and mid-July) (Queiroz, 2007).The rising of the water level begins in January and goes up to late April while the water receding phase starts in July.During the flood period, which begins in May, the floodplain is totally occupied by water until the beginning of the receding phase (Affonso et al., 2011).The lakes were selected according to criteria defined in Jorge et al. (2017a) to guarantee access to them throughout the hydrological year and sizes compatible to the spatial resolution of the visible bands of the MSI/Sentinel2A (10 m and 20 m) .Additionally, the lakes have intrinsic differences: two of them (Buabuá and Mamirauá) are small perennial lakes surrounded by flood forest, isolated, while the others (Pantaleão and Pirarara) are lakes connected to the Japurá river along the entire hydrological year, with variable size and depth in response to the flood pulse.

Data source
Data were acquired in Buabuá, Mamirauá, Pantaleão and Pirarara lakes by the Instrumentation Laboratory for Aquatic Systems team (LabISA -http://www.dpi.inpe.br/labisa) of the National Institute for Space Research (INPE-Brazil).More details about the fieldwork and measurements are provided in Jorge et al. (2017aJorge et al. ( , 2017b)).
The field campaigns were carried out in March-April and July-August of 2016 which corresponds to the rising and receding water level of Solimões River.Table 1 contains the sampling points and the DOC concentration measured.In total 87 samples were collected among the lakes and seasons.

Remote sensing reflectance
The radiometric measurements to derive remote sensing reflectance (Rrs) were carried out for all sampling points, using three intercalibrated RAMSES-Trios sensors.The sensors measured above water radiance, sky radiance, and water surface irradiance, between 350 and 900 nm.During the measurements, the sensors were positioned with azimuth angles between 90° and 135° in relation to the sun and a Zenith angle of 45° to avoid sun glint effects (Mueller and Fargion, 2002).The measurement framework followed Mobley (1999).All of the measurements were made between 10:00 and 13:00 (local time) and at least 15 samples were obtained for each sample point.The dataset was processed using MSDA_XE (TRIOS, 2018) and Matlab (Mathworks, Natick, MA, USA).The Rrs estimate followed Mobley (1999), with sun glint correction based on each sampling point.The calculated Rrs was used to simulate the reflectance of the MSI bands.For this, MSI Relative Spectral Response (RSR) of the sensor was used (Equation 1): where Rrs_m is the Rrs measured in situ and Rrs (Bi) is the Rrs simulated for the i-th band of Sentinel-2A, in the wavelength range of λm to λn.MSI RSR were taken from the user guide of the sensor (https://earth.esa.int).

CDOM Absorption Coefficient
Water samples were filtered first through Whatman GF/F (0.7 μm) filters (burned at 400 o C) and then through 0.22 μm pore size polycarbonate filter.The filtrated sample was stored in sterilized dark glass bottles and kept refrigerated up to 14 days until analysis.During the analysis, all samples were kept at ambient temperature.CDOM spectral absorbance was measured with a Shimadzu UV-2600 spectrophotometer in the wavelength range between 220 and 800 nm, with increments of 1 nm and converted to aCDOM (λ) according to Equation (2) (Bricaud et al., 1981): where A (λ) is the spectral absorbance of the filtered sample in the specific wavelength λ (nm) and L is the cuvette path length (0.1 m).
The average of aCDOM between 750 and 800 nm was used to correct the residual absorption spectra due to baseline drift, temperature, scattering, and refractive effects (Coble, 2007).

Spectral slope determination
Helms et al., (2008) have shown that the spectral slope calculated in the range of 275 and 295 nm (S275-295) is an indicator of DOM molecular weight and a tracer of degradation processes.In the present study, S275-295 was computed according to the Equation 3 using non-linear fit (Helms et al., 2008;Bricaud et al., 1981).This function describes the aCDOM (λ) behaviour along the electromagnetic spectrum and is expressed as: where S is the spectral slope parameter (nm -1 ) between the wavelength interval of λ -λref and λref is a reference wavelength (nm).
The spectral slope ratio (SR) between the wavelengths intervals of 275-295 nm on 350-400 nm was also computed in the same away describe in Equation 3.

Statistical analyses
The temporal variability of DOM was assessed using aCDOM (440).This wavelength was chosen due to the high CDOM absorption at low wavelengths (Jorge et al., 2017a), being also a region used as reference in remote sensing studies, at least, in the last thirty-six years (Bricaud et al., 1981;Brezonik et al., 2015;Bukata et al.,1995;Werdell et al., 2018).
The coefficient of variance (CV) was also computed for assessing aCDOM (440) variability.Kruskal Wallis test (one-way ANOVA on rank) with a significance level of 95% was applied to test the differences between lakes and hydrograph phases regarding aCDOM (440) values as follows: i) in a first run, the test used the entire data set; ii) in a second run, to test the influence of Buabuá and Mamirauá samples acquired during the rising period, those samples were removed.
The mean S275-295 of the two months representing the same hydrograph phases (e.g.July and August for receding; March and April for rising) was computed for each sampling point to analyse their variability within each lake and phase.All the statistical analysis were performed using the software Matlab (Mathworks, Natick, MA, USA).

Model calibration and validation
The model proposed by Vantrepotte et al. (2015) based on the ratio of aCDOM (412) and parameterized according to three coastal zones was tested to our data set using aCDOM (440), once the band in 412 nm is not present at MSI/Sentinel.A simple power function (Equation 4) was also tested.
where x and y are fitting coefficients of the Equation (4).
Monte Carlo was applied to Equation 4 to find the coefficients and validate the model.Out of 42 S275-295 samples collected during the rising phase, 29 were randomly selected for model calibration.This process was repeated 10 4 times and the Mean Square Error (MSE) and equation coefficients (x and y) were recorded, at each iteration (Figure 2).The chosen model was validated using the 13 remaining samples (not used in the calibration process) and the final accuracy was assessed following the metrics: coefficient of determination (r²), MSE and normalized root mean square error in percentage (%NRMSE).
Once the relationship between S275-29 and aCDOM (440) was modelled, another algorithm was calibrated and validated for estimating aCDOM (440) from simulated MSI Rrs..Based on the recent literature concern on distinguishing CDOM and Non-Algal Particle (NAP) contribution to the Rrs in complex environments (Matsuoka et al., 2009;Matsuoka et al., 2012).This study proposes a new model for estimating aCDOM (440) based on the ratio between near infrared bands for removing NAP contribution from its inorganic fraction.The rational for introducing this ratio is the null signal of CDOM and the dominance of NAP in near infrared range (Kirk, 2011).Previous studies also have shown that the inclusion of bands at wavelengths >600 nm increases the accuracy of the CDOM estimation model (Chen et al., 2017;Zhu et al., 2014).Thus, to determine aCDOM (440), the exponential of the ratio between bands 6 (λcentral wavelength (λcw)=740 nm) and 5 (λcw=705 nm) are subtracted from the exponential of the ratio between bands 2 (λcw=490 nm) and 3 (λcw =560 nm) (Equation 7): where, x, y and z are the coefficients.B2, B3, B5 and B6 are the MSI sensor simulated bands 2, 3, 5 and 6.
Monte Carlo simulation was similarly performed to select the most representative model for estimating aCDOM (440) as a function of Rrs.The validation process also followed the same procedure previously described for the slope.

Seasonal and spatial variability of CDOM
The highest amplitude of aCDOM (440) in the entire data set (e.g.across all sites) occurred in March (1.22 to 5.46 m -1 ) and April (1.60 to 5.97 m -1 ), with averages of 2.56 and 3.01 m -1 , respectively.In July and August, the amplitude was smaller (1.32 to 2.03 m -1 and 1.27 to 2.19 m -1 , respectively) and both averaged below 2 m -1 .No spatial variation was observed in aCDOM (440) within lake.
The water level during the sampling campaign in the rising and receding phase was almost the same (mean=30.04± 1.38 m).
The Kruskal Wallis test using samples from all lakes and dates indicated that there are significant differences (p<0.001) in aCDOM ( 440) between lakes and hydrograph phases.After the removal of Buabuá and Mamirauá samples acquired in March and April (rising), Kruskal Wallis results showed no significant differences in aCDOM ( 440) values (p=0.51).The two runs indicate that DOM at Buabuá and Mamirauá, during the rising phase have a much higher absorption at 440 nm than those of the remaining lakes and months.

CDOM absorption spectra
The entire set of aCDOM spectra (Figure 4) can be divided in two groups.The first group comprises Buabuá and Mamirauá spectra acquired at the rising phase, with aCDOM at 254 nm ranging between 65 and 95 m -1 .The second group is composed by the Pantaleão and Pirarara spectra at rising phase and the samples of all the lakes acquired during the receding phase, with aCDOM at 254 nm almost three times smaller, ranging from 26 to 35 m -1 .It is also noticeable the presence of shoulder between 245 and 290 nm in the absorption spectra (black arrow in Figure 4).Thus, during the rising phase there are differences between the spectra collected in the lakes surrounded by the flooded forest and those near the river.During the receding phase, however, this difference no longer exists, and all spectra have lower aCDOM (254) values.
The analysis of average S275-295 of each hydrograph phase also indicates the existence of differences between both, phases and lakes (Figure 5).The scatter plot displays the presence of two distinct groups: one including Mamirauá and Buabuá samples, and the other, Pantaleão and Pirarara.S275-295 in all samples from Buabuá and Mamirauá are near or under 0.015 nm -1 in rising phase and equal or higher than 0.016 nm -1 in the receding phase.However, S275-295 in all Pantaleão and Pirarara samples are above 0.015 nm -1 in the rising phase and below 0.0155 nm -1 in the receding phase, except for one single sample from Pantaleão.
The significant relationship between S275-295 and SR indicates that these parameters can be tracking similar pools of DOM (Hansen et al., 2016).SR same as S275-295 indicates differences between the lakes surrounded by flooded forest and those near the river (see supplementary material S.1).Since S275-295 also has been applied in remote sensing studies (Fichot et al., 2013;Vantrepotte et al., 2015), the present study has focused in modelling S275-295.

Seasonal relationship between aCDOM and S275-295
The model proposed by Vantrepotte et al. (2015) was tested using the entire data set (coefficients of equation 0.05, 0.10, 3.06 and 0.0), but a power-law function provided a better fit (Figure 6).
The relationship between aCDOM (440) and S275-295 varies between hydrograph phases (Figure 7).As in the receding phase aCDOM values are very similar among lakes and there is no apparent correlation between aCDOM (440) and S275-295, the model was developed only for the rising phase.The selected model to estimate S275-295 from aCDOM (440), developed using Monte Carlo and data from rising phase, shows a satisfactory fit (MSE<0.0001)and is described in Equation 6: Validation results showed a good explanation of the model´s variance (r²=0.8)and predicted values close to observed values (%NRMSE=9.4),indicating the feasibility of estimating S275-295 from aCDOM (440) (Figure 8a).In the rising water, however, predicted S275-295 diverges from the 1:1 line for values above ~0.015nm -1 , indicating that the model is better parameterized for values of S275-295 smaller than 0.015 nm -1 .
Once the relationship between aCDOM (440) and S275-295 was established for rising water, the model for estimating aCDOM (440) based on Rrs was also calibrated for this period (Equation 7).The final model had provided MSE=0.65 m -1 :  (440) = 4.39 + 0.59 -6.67, Model validation (Figure 8b) shows that almost 70% of the estimated values are within the 95% confidence interval and the statistics parameters (r², MSE and %NRMSE) present good accuracy in the estimation of aCDOM ( 440).

Discussion
The variability of aCDOM ( 440 River located in the eastern extreme of the floodplain; therefore they are not affected by the Solimões overland flow in the beginning of the rising phase, receiving a minor input of organic matter as Buabuá and Mamirauá (Table 1).
As the study area consists entirely of a floodplain, that is subject to marked seasonal flooding (about 12 m), during the high water the entire ecosystem is flooded (Affonso et al., 2011).According to Ferreira-Ferreira et al. (2015), the entire area showed in Figure 1 is flooded for periods of up to 295 days in a year depending on the flood peak.In this study, the high-water phase was not sampled considering that previous studies (Affonso et al., 2011) indicated that during the high water all water bodies become interconnected with the main channels and rivers displaying the lowest spatial variability in all limnological variables, including DOC concentration.Actually, DOC coefficient of variation among sampled water bodies dropped from 53.87% in the low water to 20.89% in the high-water of 2009 hydrological year (Affonso et al., 2011).Considering that in the Amazon basin, DOC accounts for 70% of total organic matter and that floodplain areas are relevant sources of DOC to the Solimões/Amazon River (Morreira-Turq et al., 2003), in the present study we assume that it is possible that the differences in CDOM optical properties among Mamirauá/Buabuá and Pirarara/Pantaleão are related to the fact that the flood wave have not reached the eastern margin of the floodplain at the onset of the rising water phase.
In the rising phase, the water coming mainly from the Solimões river undergoes overbank flooding (Figure 1c), overtopping its channel and flowing across the litter through the forest before reaching the lakes (Junk, 1989).The tree-DOM accumulated during the lower water season may be an import source of organic matter to the lakes during this event (Van Stan and Stubbins, 2018).As explained in the previous paragraph, at beginning of the rising phase, the water from Solimões does not reach all the floodplain lakes at the same time.Therefore, in this period, DOM is expected to have significant differences between those lakes surrounded by flooded forests located near the Solimões River and those connected to Japurá River, located in the extreme eastern boundary of the study area.During the rising water phase, the water path to Pirarara and Pantaleão through the flooded forest is small, because they are closely connected to Japurá River.At that time, the Solimões overbank flood of the high-water season, responsible for homogenizing limnologic properties in floodplain lakes have not occurred yet (Abdo and Silva, 2004;Almeida and Melo, 2009;Carvalho et al., 2001;Henderson, 1999;Queiroz, 2007).
Regarding the assessment of S275-295 values in this study (Figure 5), differences were found between both, lakes and hydrograph phases.During the rising phase, Pantaleão and Pirarara have higher S275-295 (>0.015 nm -1 ) than those of Buabuá and Mamirauá (<0.015 nm -1 ), suggesting that DOM at lakes near that river have lower molecular weight (LMW) than those surrounded by forest.Also, high molecular weight (HMW) can be an indicative of allochthonous DOM since it is composed of refractory compounds such as lignin and cellulose.These results agree with previous studies indicating the presence of HMW DOM during rising water (Melo et al., 2019;Shen et al., 2012;Spencer et al., 2008).However, in the present study the authors do not have data to corroborate the optical analyses regarding the origin and molecular weight of DOM.
In our study, no significant correlation between aCDOM (440) and S275-295 was found for the data set including samples acquired in all hydrograph phases.During the receding water phase it is difficult to draw conclusions regarding DOM origin, since the DOM present in the lakes can be old and highly degraded (Wagner et al., 2019).During the rising phase, a significant correlation between aCDOM (440) and S275-295 can be found.This means that high (low) aCDOM (440) values correspond to low (high) S275-295 values, suggesting the presence of HMW (LMW) substances.In this way, it seems that aCDOM (440) and S275-295 are optical absorption indices that can be used to trace different CDOM dynamics between lakes and hydrograph phases in floodplain lakes.Since literature shows that these indices can be estimated via remote sensing data (Brezonik et al., 2015;Vantrepotte et al., 2015), their relationship with remote sensing reflectance (Rrs) was tested.However, because of the differences in CDOM dynamic among hydrography phases, only the relationship between the variables for the data set sampled during rising water could be modelled.Nonetheless, this hydrography phase is a key moment when the floodplain receives large amount of water coming from different Amazon drainage basin habitats which washes the floodplain floor and carries large amount of organic matter accumulated along the hydrological year.
There are several models relating aCDOM (440) and remote sensing data in literature, but being empirical they are environmentally and seasonally dependent (Zhu et al. 2015).Kutser et al., (2016) tried to calibrate a model using data from Estonian lakes, Três Marias Reservoir (Brazil) and a floodplain lake located in Amazon (Curuai Lake).However, they were not able to fit a model describing the entire data set, what indicates that model development depends on DOM quality and degradation dynamics (Hansen et al., 2016).Models available in literature usually use the ratio between green and red bands (Toming et al., 2016;Zhu et al., 2014).In this study, we tested the correlation between aCDOM (440) values and the ratio between the green and red bands, but the results were poor (see supplementary material S.2).Thus, we proposed a new model to estimate aCDOM (440), using additional bands (Equation 7).
Despite the small number of samples, this study shows that it is possible to estimate S275-295 from aCDOM (440) during one crucial hydrography phase (rising phase), notwithstanding their hydrodynamic differences.Both the MSE and %NRMSE (<0.0001 m -1 and 9.40%) computed in this study are in the range of models available in the literature (Vantrepotte et al., 2015;Fichot et al., 2013), showing potential for estimating S275-295 from aCDOM (440).Therefore, a model for aCDOM (440) estimation was also proposed.The aCDOM (440) model also provided MSE and %NRMSE (0.53 m -1 and 15.12%) which is considered as an accurate estimate considering the various uncertainties related to remote sensing methods.Those modelling results, therefore, are encouraging suggesting that MSI images, when available, might be applied for studying DOM properties of the Amazon floodplain lakes during the rising water level.However, the models have limitations, which are: 1) its empirical nature demand calibration for application in other datasets; and 2) the small range of aCDOM sampled (1.2 to 6.0 m -1 ) and S275-295 (0.0142 to 0.0165 nm -1 ), indicating the need of new experiments including a larger number of lakes spread in a wider range of distance from the Solimões bank, a wider span of the rising hydrograph phase and DOM molecular analyses in order to validate the optical indices.

Conclusions
The present study indicates that the use of the optical indices, aCDOM and S275-295 , provided a deeper understanding on the connections between Solimões and Japurá river flood pulse and DOM dynamics in the Amazon floodplain lakes.The results corroborates the findings in the most recent literature and indicates that there is an urgent need of research to explore new types of indices integrating both, optical spectral properties and remote sensing data.
The empirical model relating Rrs and aCDOM (440); aCDOM (440) and S275-295 provided robust statistics indicating the high potential of MSI sensor for estimating S275-295 during the rising water.Even though this study is the first attempt of using simulated MSI data to estimate S275-295 in Amazon floodplain lakes, the results herein discussed seem very promising particularly considering the new generation of satellite-borne sensors with higher temporal resolution and the resources (costs and time) involving DOM analysis in laboratory.
Author contributions.MPdaS, LASdeC, EN and CCFB planned and designed the research.DSFJ and CCFB carried out parts of the field work and conducted a first version of data processing.MPdaS did the statistical analysis and wrote the paper with contributions from all co-authors.
Competing interests.The authors declare that they have no conflict of interest.

Final
model selection (most representative model based on MSE modal value) follows Augusto-Silva et al. (2014) procedure: i) constructing a histogram of MSE; ii) computing of mean and standard deviations of model's coefficients in the most frequent error interval; iii) ranking of coefficients in the range of mean ± standard deviation according to their MSE, iv) selecting the model with the smallest MSE.
), aCDOM spectra and S275-295 along the hydrological year(Figures 1, 3, 4 and 5)  indicated that these parameters are related to the hydrograph phases and lake geographical location in the floodplain.Mamirauá and Buabuá lakes are located in the middle of the floodplain, far from both main rivers, Solimões and Japurá and surrounded by High and Low Várzea Forests(Ferreira-Ferreira et al., 2015 Figure 5).While Pantaleão and Pirarara are lakes located near to Japurá River, subjected to both river inputs and Solimões River flood pulse.The water level in the floodplain is quite similar between the rising and receding seasons, suggesting that the flood pulse is the major factor explaining the variability of those optical variables.The Solimões flood pulse phase is, therefore, the key variable controlling the variability of CDOM index.During the rising water level, the Solimões inflow into the floodplain as overland flow crosses a large area of forest and carries a considerable amount of organic matter in different stages of decomposition into Buabuá and Mamirauá lakes.Pantaleão and Pirarara lakes, however, are far from Solimões, being connected to Japurá

Figure 5 -Figure 7 -
Figure 5 -Dispersion diagram of average S275-295 (nm -1 ) at each hydrograph phase (rising and receding) and in all lakes.The dotted

Table 1 -
Overview of the sampling points.