Advantages of Geographically Weighted Regression for Modeling Benthic Substrate in Two Greater Yellowstone Ecosystem Streams

Stream habitat assessments are commonplace in fish management, and often involve nonspatial analysis methods for quantifying or predicting habitat, such as ordinary least squares regression (OLS). Spatial relationships, however, often exist among stream habitat variables. For example, water depth, water velocity, and benthic substrate sizes within streams are often spatially correlated and may exhibit spatial nonstationarity or inconsistency in geographic space. Thus, analysis methods should address spatial relationships within habitat datasets. In this study, OLS and a recently developed method, geographically weighted regression (GWR), were used to model benthic substrate from water depth and water velocity data at two stream sites within the Greater Yellowstone Ecosystem. For data collection, each site was represented by a grid of 0.1 m² cells, where actual values of water depth, water velocity, and benthic substrate class were measured for each cell. Accuracies of regressed substrate class data by OLS and GWR methods were calculated by comparing maps, parameter estimates, and determination coefficient r ². For analysis of data from both sites, Akaike’s Information Criterion corrected for sample size indicated the best approximating model for the data resulted from GWR and not from OLS. Adjusted r ² values also supported GWR as a better approach than OLS for prediction of substrate. This study supports GWR (a spatial analysis approach) over nonspatial OLS methods for prediction of habitat for stream habitat assessments.     

Mussel isotope signature as indicator of nutrient pollution in a freshwater eutrophic lake: species, spatial, and seasonal variability

Stable nitrogen isotope ratios of five mussel species from littoral and pelagic areas were investigated with different trophic states in the eutrophic Lake Taihu, the third largest lake in China. Interpopulation variability for these mussels was relatively small in foot tissues because of the slow turnover time. Seasonal and spatial variations among the δ ¹⁵N values of mussels might be due in part to the natural variation in δ ¹⁵N values of potential food sources and the variation in the amount of human pollutions discharged into various locations of the lake. Although the increase of mussel δ ¹⁵N values was accompanied by the increase of nutrient concentrations in most situations in this study, statistically significant correlations were only 22% of the total correlations in this survey, which might be attributed to the different time-scale variations in nutrient concentrations and isotope signatures and the unknown details of the trophic pathways and metabolism for incorporation of these nutrients.     

Community metabolism in a deep (stratified) tropical reservoir during a period of high water-level fluctuations

As long as lakes and reservoirs are an important component of the global carbon cycle, monitoring of their metabolism is required, especially in the tropics. In particular, the response of deep reservoirs to water-level fluctuations (WLF) is an understudied field. Here, we study community metabolism through oxygen dynamics in a deep monomictic reservoir where high WLF (~10 m) have recently occurred. Simultaneous monitoring of environmental variables and zooplankton dynamics was used to assess the effects of WLF on the metabolism of the eutrophic Valle de Bravo (VB) reservoir, where cyanobacteria blooms are frequent. Mean gross primary production (P _g) was high (2.2 g C m^?2 day^?1), but temporal variation of P _g was low except for a drastic reduction during circulation attributed to zooplankton grazing. The trophogenic layer showed net autotrophy on an annual basis, but turned to net heterotrophy during mixing, and furthermore when the whole water-column oxygen balance was calculated, considering the aphotic respiration (R _aphotic). The high total respiration resulting (3.1 g C m^?2 day^?1) is considered to be partly due to mixing enhanced by WLF. Net ecosystem production was equivalent to a net export of 3.4 mg CO₂?m^?2 day^?1 to the atmosphere. Low water levels are posed to intensify boundary-mixing events driven by the wind during the stratification in VB. Long-term monitoring showed changes in the planktonic community and a strong silicon decrease that matched with low water-level periods. The effects of low water-level on metabolism and planktonic community in VB suggest that water-level manipulation could be a useful management tool to promote phytoplankton groups other than cyanobacteria.     

Temporal changes of soil respiration under different tree species

Soil respiration rates were measured monthly (from April 2007 to March 2008) under four adjacent coniferous plantation sites [Oriental spruce (Picea orientalis L.), Austrian pine (Pinus nigra Arnold), Turkish fir (Abies bornmulleriana L.), and Scots pine (Pinus sylvestris L.)] and adjacent natural Sessile oak forest (Quercus petraea L.) in Belgrad Forest—Istanbul/Turkey. Also, soil moisture, soil temperature, and fine root biomass were determined to identify the underlying environmental variables among sites which are most likely causing differences in soil respiration. Mean annual soil moisture was determined to be between 6.3 % and 8.1 %, and mean annual temperature ranged from 13.0°C to 14.2°C under all species. Mean annual fine root biomass changed between 368.09 g/m² and 883.71 g/m² indicating significant differences among species. Except May 2007, monthly soil respiration rates show significantly difference among species. However, focusing on tree species, differences of mean annual respiration rates did not differ significantly. Mean annual soil respiration ranged from 0.56 to 1.09 g?C/m²/day. The highest rates of soil respiration reached on autumn months and the lowest rates were determined on summer season. Soil temperature, soil moisture, and fine root biomass explain mean annual soil respiration rates at the highest under Austrian pine (R ²?=?0.562) and the lowest (R ²?=?0.223) under Turkish fir.     

Energy-Based Land Use Predictors of Proximal Factors and Benthic Diatom Composition in Florida Freshwater Marshes

The Landscape Development Intensity index (LDI), which is based on non-renewable energy use and integrates diverse land use activities, was compared to other measures of LU (e.g., %agriculture, %urban) to determine its ability for predicting benthic diatom composition in freshwater marshes of peninsular Florida. In this study, 70 small, isolated herbaceous marshes located along a human disturbance gradient (generally agricultural) throughout peninsular Florida were sampled for benthic diatoms and soil and water physical/chemical parameters (i.e., TP, TKN, pH, specific conductance, etc.). Landscape measures of percent agriculture, percent urban, percent natural, and LDI index values were calculated for a 100 m buffer around each site. The strongest relationships using Mantel's r statistic, which ranges from −1 to 1, were found between benthic diatom composition, the combined soil and water variables, and LDI scores (r=0.51, P<0.0001). Although similar, soil and water variables alone (r=0.45, P<0.0001) or with percent agriculture or percent natural were not as strongly correlated (both Mantel's r=0.46, P<0.0001). Little urban land use was found in the areas surrounding the study wetlands. Diatom data were clustered using flexible beta into 2 groups, and stepwise discriminant analysis identified specific conductance, followed by LDI score, soil pH, water total phosphorus, and ammonia, as cluster-separating variables. The LDI explained slightly more of the variation in species composition than either percent agriculture or percent natural, perhaps because the LDI can combine disparate land uses into a single quantitative value. However, the ecological significance of the difference between land use metrics and diatom composition is controvertible, and additional tests including more varied land uses appear warranted.     

典型陆源入海排污对邻近海域富营养化及生物毒性的影响

主要采用等标污染负荷、营养状态指数、生物综合毒性指标和GIS手段,通过对秦皇岛海域3个陆源入海排污口及邻近海域水质因子调查,考察典型陆源入海排污对邻近海域富营养化及生物毒性的影响。结果表明,大蒲河、人造河和洋河入海污染物的等标污染负荷比分别为36.9%、61.1%和2.0%,其中大蒲河和人造河对该海域污染负荷比的贡献达98%,为主要污染源;邻近海域主要超标污染物为PO3-4和DIN,PO3-4浓度与NQI相关分析(r=0.76,P=0.004 6)和GIS空间对比表明,PO3-4可能是造成该海域富营养化的主要因子。NQI和发光细菌发光抑制率的空间分布特征表明,该海域受陆源排污的影响比较明显。     

The influence of suburban land use on habitat and biotic integrity of coastal Rhode Island streams

Watershed land use in suburban areas can affect stream biota through degradation of instream habitat, water quality, and riparian vegetation. By monitoring stream biotic communities in various geographic regions, we can better understand and conserve our watershed ecosystems. The objective of this study was to examine the relationship between watershed land use and the integrity of benthic invertebrate communities in eight streams that were assessed over a 3-year period (2001-2003). Sites were selected from coastal Rhode Island watersheds along a residential land-use gradient (4-59%). Using the rapid bioassessment protocol, we collected biological, physicochemical, habitat, and nutrient data from wadeable stream reaches and compared metrics of structure and integrity. Principal component analyses showed significant negative correlation of indicators for stream physicochemical, habitat, and instream biodiversity with increasing residential land use (RLU) in the watershed. The physicochemical variables that were most responsive to percent RLU were conductivity, instream habitat, nitrate, and dissolved inorganic nitrogen (DIN). The positive correlation of DIN with percent RLU indicated an anthropogenic source of pollution affecting the streams. The biotic composition of the streams shifted from sensitive to insensitive taxa as percent RLU increased; the most responsive biological variables were percent Ephemeroptera, percent Scrapers, percent Insects, and the Hilsenhoff biotic index. These data show the importance of land management and conservation at the watershed scale to sustaining the biotic integrity of coastal stream ecosystems.     

Selected organochlorine pesticides (OCPs) in surface soils from three major states from the northeastern part of India

Eighty-two surface soil samples were collected from forest, grassland, tea estate, wildlife sanctuary, wetland, and roadside areas from the northeastern states of India, viz., Tripura, Manipur, and Assam. Thirteen different organochlorine pesticides (OCPs) were detected from background soils using gas chromatography electron capture detector. Manipur soils were found to be with higher concentration of dichlorodiphenyltrichloroethanes (DDTs), hexachlorocyclohexanes (HCHs), and endosulfan followed by Tripura and Assam. The spearman correlation coefficient shows significant correlation between HCHs, DDTs, and endosulfan isomers (r ²?>?0.5 and p?<?0.05). Additionally, α-HCH, δ-HCH, o,p′-DDE, and endosulfan-sulfate shows good correlation with total organic carbon in soil (r ²?=?0.5, p?=?0.05), indicating that the soil organic matter could enhance adsorption of these compounds, also demonstrating that the present OCPs in the background soil were from similar source. Further principal component analysis evaluates that most of the higher volatile compounds where clustered together in soil. However, after comparing with different states of Indian soil samples, the concentrations of OCPs in the present study areas are much lower and comparable with background soil across the globe.     

Assessment of Spatiotemporal Varying Relationships Between Rainfall,Land Cover and Surface Water Area Using Geographically Weighted Regression

Traditional regression techniques such as ordinary least squares (OLS) are often unable to accurately model spatially varying data and may ignore or hide local variations in model coefficients. A relatively new technique, geographically weighted regression (GWR) has been shown to greatly improve model performance compared to OLS in terms of higher R ² and lower corrected Akaike information criterion (AIC_C). GWR models have the potential to improve reliabilities of the identified relationships by reducing spatial autocorrelations and by accounting for local variations and spatial non-stationarity between dependent and independent variables. In this study, GWR was used to examine the relationship between land cover, rainfall and surface water habitat in 149 sub-catchments in a predominately agricultural region covering 2.6 million ha in southeast Australia. The application of the GWR models revealed that the relationships between land cover, rainfall and surface water habitat display significant spatial non-stationarity. GWR showed improvements over analogous OLS models in terms of higher R ² and lower AIC_C. The increased explanatory power of GWR was confirmed by the results of an approximate likelihood ratio test, which showed statistically significant improvements over analogous OLS models. The models suggest that the amount of surface water area in the landscape is related to anthropogenic drainage practices enhancing runoff to facilitate intensive agriculture and increased plantation forestry. However, with some key variables not present in our analysis, the strength of this relationship could not be qualified. GWR techniques have the potential to serve as a useful tool for environmental research and management across a broad range of scales for the investigation of spatially varying relationships.     

First recorded proliferation of Coolia monotis (Meunier, 1919) in the North Lake of Tunis (Tunisia) correlation with environmental factors

Coolia monotis is a potentially toxic epiphytic dinoflagellate widespread along the Mediterranean coasts, where it is frequently detected year round at low concentrations. However, it only proliferates recurrently in some localities. The North Lake of Tunis is one of the affected areas in the southwestern part of the Mediterranean Sea. This site is one of the most productive aquatic Tunisian areas (Recreational Fisheries and shellfish collecting). In the south part of this area of study, recurrent C. monotis proliferation (5 ×10⁵ cells per liter) took place in late spring and early summer of 2006. During this proliferation, the spatial distribution of C. monotis species, phytoplankton community, and abiotic factors were studied. The composition of the phytoplankton community exhibited a clear dominance of dinoflagellates over other genera. We suggest that proliferation development of C. monotis was linked to climatic conditions, water temperature (r?=?0.24, p?<?0.05) and high concentrations of nitrogenous nutrients, essentially NH₄ ^?+? (r?=?0.18, p?<?0.05) and NO₃ ^??? (r?=?0.21, p?<?0.05).     

Water quality time series for Big Melen stream (Turkey): its decomposition analysis and comparison to upstream

Big Melen stream is one of the major water resources providing 268 km³ year^???1 of drinking and municipal water for Istanbul. Monthly time series data between 1991 and 2004 for 25 chemical, biological, and physical water properties of Big Melen stream were separated into linear trend, seasonality, and error components using additive decomposition models. Water quality index (WQI) derived from 17 water quality variables were used to compare Aksu upstream and Big Melen downstream water quality. Twenty-six additive decomposition models of water quality time series data including WQI had R ² values ranging from 88% for log(water temperature) (P?≤?0.001) to 3% for log(total dissolved solids) (P?≤?0.026). Linear trend models revealed that total hardness, calcium concentration, and log(nitrite concentration) had the highest rate of increase over time. Tukey’s multiple comparison pointed to significant decreases in 17 water quality variables including WQI of Big Melen downstream relative to those of Aksu upstream (P?≤?0.001). Monitoring changes in water quality on the basis of watersheds through WQI and decomposition analysis of time series data paves the way for an adaptive management process of water resources that can be tailored in response to effectiveness and dynamics of management practices.     

Nutrient baselines of Cerrado low-order streams: comparing natural and impacted sites in Central Brazil

The aquatic systems responsible for water supply in the Brazilian Federal District (FD) have been threatened by anthropogenic pressures, especially considering the expressive demographic increase in the region during the last decades. The purposes of this research were: (a) to assess the water quality in streams located in the FD by monitoring physical–chemical variables; (b) to define baselines for these variables among different ecological status categories. The 14 investigated streams were sampled between 2006 and 2009, in the dry (August–September, 2006, 2008, 2009) and rainy (March–April, 2008, 2009) seasons. All sampling sites were classified in four categories (“very impacted”, “impacted”, “in transition” and “natural”) using an adaptation of a rapid habitat assessment protocol. Differences in water quality among sites were generally well predicted in the four ecological status categories defined by the protocol, which showed a gradient in nutrient concentrations from reference sites classified as “natural” (medians: electrical conductivity?=?7.3 μS cm^?1; nitrate?=?0.040 mg L^?1; ammonium?=?0.039 mg L^?1; soluble reactive phosphorus (SRP)?=?<0.001 mg L^?1; total phosphorus (TP)?=?0.006 mg L^?1; ) to those classified as “very impacted” (medians: electrical conductivity?=?87.7 μS cm^?1; nitrate?=?0.247 mg L^?1; ammonium?=?0.219 mg L^?1; SRP?=?0.010 mg L^?1; TP?=?0.035 mg L^?1). Point sources inputs were the main factor for water quality deterioration. The nutrient baselines reported were relatively low when compared to data collected from reference areas in Brazil (e.g., São Paulo State) or temperate regions, especially for TP.     

Macrobenthos response to sewage pollution in a tropical inshore area

Organic sewage pollution is the major stressor that affects benthic communities in the coastal waters. In the present study involving a once-off sampling (July–August 2003) of a sewage treatment plant (STP) outfall and areas 6 km farther into the sea, we tried to estimate the severity of organic pollution on marine macrobenthos over a pollution gradient in the inshore waters (station depths, 5–30 m) off a heavily urbanized tropical city, on the east coast of India. Multivariate ordination analyses revealed two different groups of faunal assemblages. Group I is associated with sites impacted by the sewage outfall and group II with the locations 3–6 km away in the open sea. Polychaetes and amphipods were the predominant fauna with significant taxonomic differences between the assemblages. Despite the homogeneity in sediment texture, the two-fold increase in sediment organic matter near the sewage outfall area supported r-strategists, while group II locations favoured K-strategists. Approximation through benthic opportunistic polychaetes amphipods (BOPA) index and information on the key taxa responsible for the observed assemblage patterns corroborated these findings. Thus, the present findings revealed how organic sewage pollution influences benthic diversity in coastal waters by supporting communities of opportunistic characteristics. We advocate inclusion of community traits and compatible analytical tools (statistical approaches) in studies of similar nature so that the observations could be compared and broad remedial measures could be evolved.     

Use of Maryland Biological Stream Survey Data to Determine Effects of Agricultural Riparian Buffers on Measures of Biological Stream Health

This study was undertaken to determine the importance of riparian buffers to stream ecology in agricultural areas. The original Maryland Biological Stream Survey (MBSS) data set was partitioned to represent agricultural sites in Maryland's Coastal Plain and Piedmont regions. ANOVA, multiple linear regression (MLR), and CART regression tree models were developed using riparian and site catchment landscape characteristics. MBSS data were both stratified by physiographic region and analyzed as a combined data set. All models indicated that land management at the site was not the controlling factor for fish IBIs (FIBI) at that site and, hence, using FIBI to evaluate site-scale factors would not be a prudent procedure. Measures of instream habitat and location in the stream network were the dominant explanatory factors for FIBI models. Both CART and MLR models indicated that forest buffers were influential on benthic IBIs (BIBI). Explanatory variables reflected instream conditions, adjacent landscape influence, and chemistry in the Coastal Plains sites, all of which are relatively site specific. However, for Piedmont sites, hydrologic factors were important, in addition to adjacent landscape influence, and chemistry. Both Coastal Plain and Piedmont CART models identified several hydrologic factors, emphasizing the dominant control of hydrology on the physical habitat index (PHI). Riparian buffers were a secondary influence on PHI in the Coastal Plain, but not in the Piedmont. Between 40% and 70% of the variation in FIBI, BIBI, and PHI was explained by the “easily obtainable” variables available from the MBSS data set. While these are empirical results specific to Maryland, the general findings are of use to other locations where the establishment of forest buffers is considered as an aquatic ecosystem restoration measure.     

Artificial neural network modeling of dissolved oxygen in the Heihe River, Northwestern China

Identification and quantification of dissolved oxygen (DO) profiles of river is one of the primary concerns for water resources managers. In this research, an artificial neural network (ANN) was developed to simulate the DO concentrations in the Heihe River, Northwestern China. A three-layer back-propagation ANN was used with the Bayesian regularization training algorithm. The input variables of the neural network were pH, electrical conductivity, chloride (Cl^?), calcium (Ca²⁺), total alkalinity, total hardness, nitrate nitrogen (NO₃-N), and ammonical nitrogen (NH₄-N). The ANN structure with 14 hidden neurons obtained the best selection. By making comparison between the results of the ANN model and the measured data on the basis of correlation coefficient (r) and root mean square error (RMSE), a good model-fitting DO values indicated the effectiveness of neural network model. It is found that the coefficient of correlation (r) values for the training, validation, and test sets were 0.9654, 0.9841, and 0.9680, respectively, and the respective values of RMSE for the training, validation, and test sets were 0.4272, 0.3667, and 0.4570, respectively. Sensitivity analysis was used to determine the influence of input variables on the dependent variable. The most effective inputs were determined as pH, NO₃-N, NH₄-N, and Ca²⁺. Cl^? was found to be least effective variables on the proposed model. The identified ANN model can be used to simulate the water quality parameters.     

Hydrometeorological variables predict fecal indicator bacteria densities in freshwater: data-driven methods for variable selection

Statistical models of microbial water quality inform risk management for water recreation. Current research focuses on resource-intensive, location-specific data collection and water quality modeling, but this approach may be cost-prohibitive for risk managers responsible for numerous recreation sites. As an alternative, we tested the ability of two data-driven models, tree regression and random forests with conditional inference trees, to select readily available hydrometeorological variables for use in linear mixed effects (LME) models predicting bacterial density. The study included the Chicago Area Waterway System (CAWS) and Lake Michigan beaches and harbors in Chicago, Illinois, at which Escherichia coli and enterococci were measured seasonally in 2007–2009. Tree regression node variables reduced data dimensionality by >50 %. Variable importance ranks from random forests were used in a forward-step selection based on R ² and root mean squared prediction error (RMSPE). We found two to three variables explained bacteria densities well relative to random forests with all variables. LME models with tree- or forest-selected variables performed reasonably well (0.335?<?R ²?<?0.658). LME models for Lake Michigan had good prediction accuracy with respect to the single sample maximum standard (72–77 %), but limited sensitivity (23–62 %). Results suggest that our alternative approach is feasible and performs similarly to more resource-intensive approaches.     

Pilot-study on GIS-based risk modelling of a climate warming induced tertian malaria outbreak in Lower Saxony (Germany)

This project was designed to establish baseline aquatic biological community structure and physical habitat conditions in select wadeable streams within the California Central Valley. A secondary objective was to evaluate possible water quality differences between site types and seasons. Two agricultural and two urban streams were monitored in spring and fall for two consecutive years beginning in the fall of 2002. Bioassessment sampling was conducted according to modified US EPA methods. The study included physical habitat assessment, water and sediment chemical analysis and characterization of the benthic macroinvertebrate community at each site. Water samples were analyzed for selected organophosphate insecticides, pyrethroid insecticides and herbicides, while sediment samples were analyzed for pyrethroids only. All sites had substantial physical habitat and water quality impairments, and the absence of pollution intolerant macroinvertebrates and dominance of pollution tolerant macroinvertebrates were indications of biological impairment. Due to the limited amount of water quality and pesticide data collected, it was not possible to definitively demonstrate any cause and effect relationships between BMI community structure and water quality or pesticide concentrations. Though most physical habitat parameters were similar and EPA physical habitat scores revealed on no significant differences between urban and agricultural sites (P? = ?0.290), a significant difference was seen in substrate embeddedness (P? = ?0.020). Dominant taxon found at all sites were chironomids, amphipods, and oligochaetes. Benthic macroinvertebrate metrics were significantly different between both types of sites (P? = ?0.001) and seasons (P? = ?0.014). Chironomidae taxon and those of the functional feeding group scrapers were greater at urban sites, while those of the functional feeding group filterers were greater at agricultural sites. In addition, the metric groups Chironomidae, filterers, and predators were found in greater numbers in the spring than the fall.     

Heavy metals in the habitat and throughout the food chain of the Neotropical otter, Lontra longicaudis, in protected Mexican wetlands

Top predators like the Neotropical otter, Lontra longicaudis annectens, are usually considered good bioindicators of habitat quality. In this study, we evaluated heavy metal contamination (Hg_tot, Pb, Cd) in the riverine habitat, prey (crustaceans and fish), and otter feces in two Ramsar wetlands with contrasting upstream contamination discharges: Río Blanco and Río Caño Grande in Veracruz, Mexico, during the dry, the wet, and the nortes seasons. Most comparisons revealed no differences between sites while seasonal differences were repeatedly detected for all of the compartments. Higher concentrations of Pb during the dry season and of Cd during the wet season in otter feces mirrored differences detected in the most seasonally consumed prey. Compared with fecal methylmercury values reported for the European otter (0.25–0.75 mg kg^?1) in unprotected areas, the Hg_tot levels that we measured were lower (0.02–0.17 mg kg^?1). However, Pb (117.87 mg kg^?1) and Cd (9.14 mg kg^?1) concentrations were higher (Pb, 38.15 mg kg^?1 and Cd, 4.72 mg kg^?1) in the two Ramsar wetlands. Protected areas may shelter species, but those with water-linked diets may suffer the effect of chemicals used upstream.     

Diurnal variations in bacterial and viral production in Cochin estuary, India

Microbes play a central role in the decomposition and remineralization of organic matter and recycling of nutrients in aquatic environments. In this study, we examined the influence of physical, chemical, and biological parameters on the rate of bacterial production (BP) and viral production (VP) with respect to primary production over a diurnal period in Cochin estuary. Time series measurements were made every 2 h for 12 h (6 a.m.–6 p.m.) during periods of low and high salinities. The light intensity as photosynthetically active radiation, temperature, salinity, nutrients like NO₃–N, SiO₄–Si, and PO₄–P, and chlorophyll a (Chl a) were measured along with BP, VP, and net primary production (NPP). NPP showed a strong positive correlation with light and Chl a (r ²?=?0.56 and 0.47, respectively), while VP showed a strong positive correlation with light, salinity, and Chl a (r ²?=?0.37, 0.58, and 0.37, respectively) and a negative correlation with BP (r ²?=??0.39) at P?≤?0.05. We observed a diurnal pattern in BP but did not have any significant correlation with light. Similar diurnal pattern was seen in VP, the peak of which was in succession with BP, suggesting that virus-mediated lysis plays an important role in loss processes of bacteria in Cochin estuary. The results of our study highlight the light-dependent and physicochemical-dependent diurnal variation in virioplankton production in a tropical estuarine ecosystem.