JEQ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 25 January 2007
Published in J Environ Qual 36:416-425 (2007)
DOI: 10.2134/jeq2006.0185
© 2007 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Qian, Y.
Right arrow Articles by Chin, D.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Qian, Y.
Right arrow Articles by Chin, D.
Agricola
Right arrow Articles by Qian, Y.
Right arrow Articles by Chin, D.
Related Collections
Right arrow Water Quality
Right arrow Surface Water Quality
Right arrow Water Conservation

TECHNICAL REPORTS

Surface Water Quality

Seasonality of Selected Surface Water Constituents in the Indian River Lagoon, Florida

Y. Qiana, K. W. Migliaccioa, Y. Wanb, Y. C. Lia,* and D. Chinc

a Univ. of Florida, Tropical Research & Education Center, Soil and Water Science Dep. and Biological and Agricultural Engineering Dep., IFAS, Univ. of Florida, 18905 SW 280th St., Homestead, FL 33031
b South Florida Water Management District, 3301 Gun Club Rd., West Palm Beach, FL 33406
c Univ. of Miami, Dep. of Civil, Architectural, and Environmental Engineering, Coral Gables, FL 33124

* Corresponding author (yunli{at}mail.ifas.ufl.edu)

Received for publication May 8, 2006.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Seasonality is often the major exogenous effect that must be compensated for or removed to discern trends in water quality. Our objective was to provide a methodological example of trend analysis using water quality data with seasonality. Selected water quality constituents from 1979 to 2004 at three monitoring stations in southern Florida were evaluated for seasonality. The seasonal patterns of flow-weighted and log-transformed concentrations were identified by applying side-by-side boxplots and the Wilcoxon signed-rank test (p < 0.05). Seasonal and annual trends were determined by trend analysis (Seasonal Kendall or Tobit procedure) using the U.S. Geological Survey (USGS) Estimate TREND (ESTREND) program. Major water quality indicators (specific conductivity, turbidity, color, and chloride), except for turbidity at Station C24S49, exhibited significant seasonal patterns. Almost all nutrient species (NO2–N, NH4–N, total Kjeldahl N, PO4–P, and total P) had an identical seasonal pattern of concentrations significantly greater in the wet than in the dry season. Some water quality constituents were observed to exhibit significant annual or seasonal trends. In some cases, the overall annual trend was insignificant while opposing trends were present in different seasons. By evaluating seasonal trends separately from all data, constituents can be assessed providing a more accurate interpretation of water quality trends.

Abbreviations: APHA, American Public Health Association • EPA, U.S. Environmental Protection Agency • ESTREND, Estimate TREND • FAC, flow-adjusted concentration • IRL, Indian River Lagoon • LOESS, locally weighted scatter plot smoother • MLE, maximum likelihood estimation • SFWMD, South Florida Water Management District • SK, Seasonal Kendall • SLE, St. Lucie Estuary • USGS, U.S. Geological Survey


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
HISTORICALLY, trend analysis was conducted to evaluate change in water quality over time using various parametric or nonparametric techniques (Lettenmaier et al., 1991; Lietz, 2000; Donohue et al., 2001; Richards and Baker, 2002). Seasonality is often the major exogenous effect that must be compensated for or removed to discern trends in water quality since seasonal variation is observed for most water quality constituents in surface waters (Hirsch et al., 1982; Hirsch et al., 1991; Graham and Li, 1994; Champely and Doledec, 1997; McCartney et al., 2003). Previous studies generally dealt with seasonality by using nonparametric techniques (e.g., Seasonal Kendall test) or other methods that incorporated seasonal terms during trend analysis of water quality constituents (Hirsch et al., 1991; Helsel and Hirsch, 1992). The Seasonal Kendall (SK) test nullifies the seasonality by conducting the Mann-Kendall test on each season separately and combining the results (Cavanaugh and Mitsch, 1989; Yu and Zou, 1993; Lietz, 2000). Richards and Baker (2002) evaluated trends in water quality using analysis of covariance in which seasonality was represented as a pair of sine and cosine functions. Seasonal fluctuations can also be minimized by applying deseasonalized data (e.g., subtracting seasonal medians from all data within the season) into appropriate models (e.g., generalized least squares regression, multiple regression, or Sen's t test) during trend analysis (Helsel and Hirsch, 1992; Doering, 1996; Hess et al., 2001; Kahya and Kalayci, 2004). Others explored seasonal variation of water quality by evaluating the changes in the direction of median values of water quality constituents (Ravichandran, 2003).

While most of the historical studies focused on evaluating overall trends in water quality using all data, this could lead to loss of some information with respect to water quality trends and interpretations. This is because trends using all data do not necessarily coincide with trends observed when using subsets of the data representing specific seasons (Helsel and Hirsch, 1992; Kahya and Kalayci, 2004). First, concentrations of some water quality constituents in different seasons may vary by orders of magnitude, thus preventing trend detection. Second, trends in different seasons may oppose each other and consequently result in an insignificant annual trend (Helsel and Hirsch, 1992). By performing a trend analysis using separated seasonal subsets instead of all data, more detailed and useful information can be obtained. An example of dividing data into groups based on some methodology (e.g., precipitation, flow-regime, etc.) for analyses is the study conducted by White et al. (2004). Their results indicated that evaluation of flow-regime based datasets separate from all data provided more information on the transport of constituents and their respective sources (nonpoint or point).

This study was designed to provide a consistent framework for systematically evaluating seasonality using statistical techniques. Our goal was to provide a methodological example for dealing with seasonality during trend analysis and information on how seasonal factors influence water quality in the study region using a two season subset: wet season and dry season. The objectives were to: (i) analyze selected water quality for seasonal patterns of the flow-weighted concentrations of constituents representing different seasons, and (ii) investigate the seasonal trend for the flow-weighted concentrations of constituents, considering all data and seasons separately. Seasonal pattern refers to the different concentration levels in different seasons, while seasonal trend refers to the magnitude of changes between seasons.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Site Description
The study area is in the Southern Indian River Lagoon (IRL) located in the southeast coast of Florida. Three water quality monitoring stations are selected, namely C24S49 (27°15'42.57'' N, 80°21'32.63'' W), C25S50 (27°28'10.19'' N, 80°20'16.76'' W), and C44S80 (27°06'37.44'' N, 80°17'07.09'' W) (Fig. 1). This area was selected because the IRL estuary system is considered one of the most biologically diverse ecosystems in North America, with approximately 2200 identified species (Gilmore, 1977; Swain et al., 1995; Chamberlain and Hayward, 1996; Graves et al., 2004; SFWMD, 2004). The ecological balance of this unique estuary system is threatened by stormwater runoff, navigation, loss of essential marshland, and agricultural and urban development (Walker, 1991; Chamberlain and Hayward, 1996; Doering, 1996).


Figure 1
View larger version (35K):
[in this window]
[in a new window]

  		Fig. 1. Location of the South Indian River Lagoon/St. Lucie Estuary (IRL/SLE) watershed and primary canals, the South Florida Water Management District (SFWMD) structures in this area.

 
The Southern IRL watershed contains all of Martin County and St. Lucie County, and a small portion of Okeechobee County in Florida, occupying about 2200 km2. Major land use types include agriculture (46%), urban and transportation (16%), wetlands (13%), and upland forest (14%) (SFWMD, 1995). The dominant crop in the watershed is citrus, though other crops (e.g., vegetables, nursery, and sod) are also present. Most soils in this watershed are Spondol (mainly Electra, Immokalee, and Myakka) characterized by nearly level, poorly to somewhat poorly drained sandy soils with dark, loamy subsoil layers. Hence, nutrient leaching may have a considerable influence on movement of pollutants to adjacent drainage canals (Graves et al., 2004).

The watershed is drained into the estuarine system through major canals including C-25, C-24, C-23, and C-44 (Fig. 1). Canals C-24, C-23, and C-44 are connected to the St. Lucie Estuary (SLE), which is the largest tributary to southern IRL. The C-44 canal also conveys regulatory releases from Lake Okeechobee to the South Fork of the St. Lucie River. During the dry season, agricultural irrigation partly relies on direct withdrawal from these canals, which significantly reduces flows to the IRL/SLE watershed. During the wet season, increased inflows to the estuary produce rapid fluctuations of salinity, nutrients, and suspended solids (Chamberlain and Hayward, 1996; Graves et al., 2004).

Data Source
The South Florida Water Management District (SFWMD) (http://www.sfwmd.gov/) has been, on a monthly basis, monitoring the water quality of flows entering the Southern IRL since 1979. Water quality data from three monitoring stations, i.e., C44S80, C24S49, and C25S50 (Fig. 1) were analyzed in our study. Seasonal pattern exploration and seasonal trend analysis were performed on major water quality indicators (specific conductivity, turbidity, color, and chloride concentration), and nutrient concentration of nitrite-N (NO2–N), ammonia-N (NH4–N), total Kjeldahl nitrogen (TKN), orthophosphate (PO4–P), and total phosphorus (TP) between 1979 and 2004 to evaluate seasonal variations in water quality in the IRL area from 1979 to 2004. Surface water quality was monitored using grab samples. Samples were collected and analyzed using the methods summarized in Table 1. Replicate samples were collected on a quarterly basis.


View this table:
[in this window]
[in a new window]

  		Table 1. Selected water quality constituents and corresponding analytical methods.

 
Definition of Seasons
Rainfall is the primary climate factor that defines seasons in south Florida. In the present study, seasons were defined based on long-term rainfall record in the Southern IRL watershed (from 1965 to 2000, collected and maintained by SFWMD). Differentiation between different seasons was evaluated using plots of short-term (5-d) mean precipitation variation and selecting the ‘break points’ between seasons (Hiroyuki and Tomomi, 2004). The break points were identified as the section of the time series that exhibited the steepest slope when plotting precipitation versus time.

Seasonal Analysis
Seasonal analysis including evaluation of seasonal patterns and seasonal trends was conducted for two seasons (dry and wet). A simple flow chart for evaluating data sets that are described with a seasonal component is shown in Fig. 2.


Figure 2
View larger version (21K):
[in this window]
[in a new window]

  		Fig. 2. Flow chart for evaluating seasonal influences in time series water quality data (LOESS refers to locally weighted scatter plot smoother, and ESTREND to Estimation TREND).

 
Analysis of Seasonal Pattern
Because seasonal water quality variation is partly due to seasonal differences in flow, water quality data were divided into seasonal groups and flow-adjusted to account for seasonal differences in flow (Harned et al., 1981). The seasonal patterns were evaluated using boxplots and the Wilcoxon signed-rank test. Boxplots provide a visual interpretation of the seasonal data and the Wilcoxon signed-rank test determines if significant differences are present between seasons (Helsel and Hirsch, 1992). Specifically, the Wilcoxon test determines whether the median difference between paired observations equals zero, with no assumptions regarding the probability distribution of the data. Seasonal patterns of water quality constituents were explored using the following five steps: (i) Censored data were replaced with one half of the detection limit, similar to previous studies (Cavanaugh and Mitsch, 1989). Censored data refer to values recorded as below limit of detection (She, 1997). Due to the development of laboratory techniques, long-term water quality data often show multiple censoring levels. Multiple nondetects were set to one half of the highest detection level for that constituent. (ii) Water quality data were flow-adjusted to determine the flow adjusted concentration (FAC) as follows (Harned et al., 1981; Bouraoui et al., 1999):

Formula 1[1]
where c(t) is the constituent concentration (mg L–1, specific conductivity in dS m–1, turbidity in NTU, and color in PCU), t is time (day), q(t) is the flow rate (m3 s–1), and T (day) is the duration of the season. (iii) Mean FAC was calculated for each defined season for each year over the period of 1979 to 2004. (iv) The mean FAC concentrations were log-transformed to account for the log normal distribution of water quality data and to minimize the effect of outliers within the data (White et al., 2004). (v) The log-transformed FACs were illustrated in side-by-side boxplots. Wilcoxon signed-rank tests were applied and seasonal significant differences were evaluated (significance level p < 0.05).

Analysis of Seasonal Trend-ESTREND Seasonal Kendall Procedure
For detecting trends in each season (seasonal trends) as a separate dataset, data were split into two subsets for each water quality constituent, namely dry and wet season subsets, based on the definition of seasons. In addition, annual trend analyses were performed using all data (data sets combining the dry and wet season subsets). All trend analyses were conducted at a significance level of p < 0.1 using the statistical procedures provided by the program Estimate TREND (ESTREND) (Schertz et al., 1991) library.

Because censored data may influence results of trend analysis, especially for nutrient species (Haan, 2002; White et al., 2004), appropriate procedures were selected based on censored data percentage and censoring level(s) (single or multiple) of each data set for trend testing. That is, for water quality data with less than 5% censored data, the uncensored SK procedure was adopted; while for a data set with a high percentage of censored data and/or multiple detection limits, Tobit procedure was used, similar to Lietz (2000).

ESTREND uncensored SK procedure is based on the distribution-free, nonparametric SK test (Hirsch et al., 1982; Helsel and Hirsch, 1992), which compares relative ranks of the data from the same season based on the null hypothesis (H0) that a test statistic of zero indicates no trend. In this procedure, all censored values are substituted by one half of the reporting limit (Schertz et al., 1991). Few censored values (<5%) are not likely to bias the results of the trend testing significantly even if multiple detection limits are present (Schertz et al., 1991). For seasonal trend testing, ESTREND uncensored SK procedure treated each seasonal subset as one season (without seasonal variability within subset) by specifying one season in the procedure (Schertz et al., 1991).

The effect of flow was minimized in the uncensored SK procedure by using locally weighted scatter plot smoother (LOESS) (Cleveland, 1979) with log-transformed flow versus log-transformed constituent concentrations. Data were log-transformed to account for the non-normal distribution and the effects of outliers (White et al., 2004). After flow adjustment, the residuals were then tested for trend. An example for log-log LOESS flow adjustment is illustrated in Fig. 3Go to 5. Figure 3 shows the raw data points and the LOESS curve, which is a smoothed trend line derived from the data using the LOESS procedure. The difference between the LOESS curve (i.e., trend line) and the raw data (Fig. 4) indicates a symmetrical variation around zero. The normality of the LOESS residuals (Fig. 5) indicates an approximately linear relation between LOESS residuals and the quantiles of the standard normal deviate.


Figure 3
View larger version (15K):
[in this window]
[in a new window]

  		Fig. 3. Total phosphorus concentration representing wet season from 1979 through 2004 at Station C25S50 and flow (in log scale) plot with the LOESS (locally weighted scatter plot smoother) smooth line, x and y axis are in log scales.

 

Figure 4
View larger version (18K):
[in this window]
[in a new window]

  		Fig. 4. Locally weighted scatter plot smoother (LOESS) residuals for total phosphorus and flow (in log scale) plot, including a dotted line at the zero value of the residuals and a LOESS smooth line through the data.

 

Figure 5
View larger version (13K):
[in this window]
[in a new window]

  		Fig. 5. Locally weighted scatter plot smoother (LOESS) residuals for total phosphorus against the quantiles of the standard normal distribution. Residuals should follow a straight line if they were normally distributed.

 
The trend is further defined by the rate of change over time, which is referred to as the trend slope (Sen, 1968; Schertz et al., 1991). The trend slope for the uncensored SK procedure is calculated using the median slope of all pairwise comparisons. Thus, the trend slope provides a numerical value representing the change in constituent concentrations over a specified period.

Analysis of Seasonal Trend-ESTREND Tobit Procedure
The ESTREND Tobit procedure is based on a maximum likelihood estimation (MLE) method, which was developed by Cohn (1988) to account for the bias due to multiple censoring levels of data. In this procedure, trend is evaluated by estimating the parameters of a linear regression model using MLE method based on the null hypothesis (H0) that the trend slope of the test is zero (Schertz et al., 1991). Data was log-transformed to account for the log-normal distribution and minimize the effect of outliers before fitted to the Tobit model.

For detecting annual trend (using all data), the regression model can be expressed as (Helsel and Hirsch, 1992):

Formula 2[2]
For determining seasonal trend (using seasonal subset), the regression model is (Helsel and Hirsch, 1992; Schertz et al., 1991):

Formula 3[3]
where C is the concentration of water quality constituents (mg L–1); T is decimal time (year); {alpha}1 is the slope; {alpha}0, {alpha}2, {alpha}3, and {alpha}4 are the regression parameters; the pair of sine and cosine functions represents the seasonal term; and log(flow) represents the flow term (flow rate in m3 s–1).

The trend slope calculated for the Tobit test is similar to that previously described for the SK test. Hence, the results of the trend slope provide a numerical value related to temporal change of a water quality constituent.


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Definition of Seasons
In south Florida, there are two dominant seasons: wet summer and dry winter. From 1965 to 2000, the mean annual rainfall was 1306, 1312, and 1356 mm in the C-24, C-25, and C-44 basins, respectively. The greatest monthly rainfall occurred in June and September averaging 192 and 183 mm, while the least rainfall occurred in December with a mean of 49 mm. The mean monthly precipitation increased rapidly in late May and decreased rapidly beginning in November (Fig. 6).


Figure 6
View larger version (16K):
[in this window]
[in a new window]

  		Fig. 6. Time series of mean monthly precipitation at Canals C-24, C-25, and C-44.

 
In this study, 5-d mean precipitation time series plots were adopted to describe the seasonal variation (Fig. 7). Figure 7 shows that the slopes from 20–25 May, 25–31 May, 20–25 October, and 5–10 November, were steeper than others. After comparison, the period from 27 May to 7 November (165 d) was defined as the wet season with over 66% of annual precipitation, while 8 November to 26 May (200 d) was defined as the dry season. Although other factors such as wind, evaporation, flow, and agricultural activities also influence the seasonal variation of constituents in water quality, rainfall was considered to be the dominant factor affecting the seasonality of water quality in this watershed.


Figure 7
View larger version (18K):
[in this window]
[in a new window]

  		Fig. 7. Time series of mean 5-d precipitation averaging on three stations. Rainy season was defined based on this figure from 27 May to 7 November.

 
Seasonal Patterns
The descriptive statistics for selected constituents in water quality samples collected at Stations C24S49, C25S50, and C44S80 from 1979 through 2004 are given in Tables 2, 3, and 4.


View this table:
[in this window]
[in a new window]

  		Table 2. Descriptive statistics of constituents in water quality samples collected in the dry and wet seasons at Station C24S49, 1979 through 2004.

 

View this table:
[in this window]
[in a new window]

  		Table 3. Descriptive statistics of constituents in water quality samples collected in the dry and wet seasons at Station C25S50, 1979 through 2004.

 

View this table:
[in this window]
[in a new window]

  		Table 4. Descriptive statistics of constituents in water quality samples collected in the dry and wet seasons at Station C44S80, 1979 through 2004.

 
The IRL water quality data are presented using boxplots in Fig. 8 and 9 (p-values derived from the Wilcoxon signed-rank test, significance level p < 0.05). Results of the Wilcoxon signed-rank test (Table 5, significance level p < 0.05) suggest that almost all the selected constituents exhibit significant seasonal patterns over the period 1979 through 2004, except turbidity and NO2–N at Station C24S49.


Figure 8
View larger version (25K):
[in this window]
[in a new window]

  		Fig. 8. Seasonal patterns of major water quality indicators at Stations (a) C24S49, (b) C25S50, and (c) C44S80. The y axis represents log-transformed flow-adjusted concentration (FAC) of constituents (specific conductivity in dS m–1, turbidity in NTU, color in PCU, and chloride in mg L–1).

 

Figure 9
View larger version (25K):
[in this window]
[in a new window]

  		Fig. 9. Seasonal patterns of nutrient species at Stations (a) C24S49, (b) C25S50, and (c) C44S80. The y axis represents log-transformed flow-adjusted concentration (FAC) of constituent (mg L–1).

 

View this table:
[in this window]
[in a new window]

  		Table 5. Seasonal patterns for water quality constituents from 1979 through 2004 (significance level p < 0.05) using the Wilcoxon signed-rank test.

 
Specific Conductivity, Turbidity, Color, and Chloride
At all three stations, specific conductivity and chloride had a significantly greater concentration in the dry season compared with the wet season (Fig. 8). For example, the median of specific conductivity in the dry season was 1.316 dS m–1 while the wet season had a median of 0.976 dS m–1 at Station C24S49. Similarly, the median concentration of chloride was 263.0 mg L–1 in the dry season and 191.6 mg L–1 in the wet season at C24S49 station. This seasonal pattern is likely due to the need for irrigation of citrus land in the dry season when flow is low and kept in the canal for frequent irrigation by farmers.

In contrast, color in the dry season was much lower than that in the wet season at all three monitoring stations (Fig. 8). High color in the wet season is likely attributable to frequent rainfall events that may speed up the transport of dissolved organic materials (humic and tannic acids, and peat from vegetative decay) into surface water (Graves et al., 2004). While in the dry season, the subsurface flow from infiltration is slower due to the less rainfall.

Among the three monitoring stations, no seasonal pattern was observed for turbidity at C24S49 while significantly greater concentrations in the dry and wet season were observed at C44S80 and C25S50, respectively (Fig. 8). Additional information is needed to interpret the site-specific differences of seasonal patterns for water turbidity, which has historically been linked to the presence of suspended matter such as clay, silt, or plankton (Graves et al., 2004).

Nutrient Species
For almost all the nitrogen and phosphorus species, nutrient concentrations in the wet season were significantly greater than in the dry season as indicated by the median concentration listed in the descriptive statistics tables (Tables 2, 3 and 4) as well as the percentiles in the boxplots (Fig. 9). The Wilcoxon test also clearly identified seasonal differences in the IRL data sets. Nutrient species were almost always significantly greater in the wet season (93%). For example, in the dry season the median concentration for PO4–P and TP at Station C24S49 were 0.105 and 0.154 mg L–1, respectively. In the wet season, the median concentration of these two forms of phosphorus increased to 0.234 and 0.306 mg L–1, approximately doubling the concentration levels of the dry season. Similarly, other nutrient species such as NO2–N, NH4–N, and TKN also had much greater median concentration in the wet season compared to the dry season at the monitoring stations. The greater nutrient concentrations during the wet seasons were also observed by Chamberlain and Hayward (1996) for the estuary.

Similar seasonal patterns were also found by others and could be attributable to the fluctuation of water flow (Donohue et al., 2001; Petersen et al., 2001). For color and most nutrient species, water discharge has been observed to account for the bulk of seasonal variability (Doering, 1996). Greater nutrient concentrations during the wet season are probably indicative of nonpoint sources of nutrients. For example, Plotnikoff and Michaud (1991) indicated that the seasonal patterns of total inorganic nitrogen were probably influenced by agricultural fertilization during the wet season in Portage Creek in northwest Washington. In our study area, agricultural fertilization, particularly for citrus, occurs mostly before the wet summer. This may also contribute to the higher nutrient concentrations in the wet season. Other factors may also have influenced nutrient results; especially, color may alter results when colormetric laboratory techniques are used.

Seasonal Trends
Although most of the water quality variables showed seasonal patterns (Fig. 8 and 9), a few exhibited significant positive or negative trend over the selected time period (1979–2004) (Table 6). At Station C24S49, annual negative trends were observed for NO2–N, NH4–N, TKN, PO4–P, TP, and chloride. At Station C25S50, specific conductivity and chloride showed annual negative trends, while PO4–P showed an annual positive trend. At Station C44S80, annual positive trends for NH4–N, PO4–P, and TP were observed whereas an annual negative trend was observed for TKN.


View this table:
[in this window]
[in a new window]

  		Table 6. Trend of water quality variables at Stations C24S49, C25S50, and C44S80 from 1979 through 2004.

 
For some water quality variables and sites, significant trends were detected in one season, but not both seasons. This seasonal difference in the identification of a significant trend has implications for annual trend evaluations. Significant annual trends were present for NH4–N and chloride at C24S49 and specific conductivity and chloride at C25S50, all of which exhibit negative trends in the dry season and no trend in the wet season, leading to overall negative annual trends. However, while color at C24S49 and turbidity at C25S50 both showed positive trends in the dry season and no trend in the wet season, no annual positive trend was identified. Hence, the trend identified when using all data may mask seasonal trends.

Another occurrence in trend analyses was the presence of opposite trends for different seasons. For example, TP at C44S80 showed a positive trend in the dry season and a negative trend in the wet season. This result could be attributed to changes in pollutant loads such that point source inputs (which dominate the dry season flow or when base flow dominates) have increased over time while nonpoint source inputs (associated with the wet season or when leaching dominate) have reduced over time. The implications of not conducting seasonal analysis for this constituent would be the loss of information on nutrient trends, and their potential sources (i.e., point and nonpoint).

Water quality trends were sometimes observed only for annual data with no seasonal trends being significant. This was true for TKN at C24S49, for which a significant annual negative trend was observed despite insignificant negative slopes both in the dry and wet season. Similarly, PO4–P at C25S50 and NH4–N and PO4–P at C44S80 exhibited significant annual positive trends accumulated from insignificant positive slopes in the dry and wet seasons. This implies that while seasonal analysis is important, annual or use-of-all-data trend analysis must also be completed to fully interpret water quality data.

A third type of result occurs when the dry and wet seasons show strong seasonal trends in identical or even opposite directions, and the overall annual trend varies (either positive, negative, or no trend). For example, no significant annual trend was observed for TP at C25S50 though strong positive trends were observed both for the dry and wet season. For TP at C44S80, however, an overall annual positive trend existed despite the fact that the dry season had a positive trend and the wet season had a negative trend. These results imply that it is useful to perform seasonal trend analysis to characterize the individual seasonal processes, which may not be revealed by using annual data trend analysis.

Analysis of seasonality could provide useful information with respect to the identification of potential nutrient sources. Choi and Koo (1993) indicated that the contribution of nonpoint sources to the external input of TP in the wet season is much greater than in the dry season for a study conducted in Hoedong Lake, Korea. Others have shown that changes in agricultural management were primary causes of observed nutrient trends (Richards and Baker, 2002). Mitchell et al. (2001) linked a positive trend of nitrate concentration observed in surface water to the increase in agricultural activity in surrounding environment. Greater nutrient concentrations in the wet season, obtained by seasonal pattern analysis, might suggest that nonpoint sources contribute a greater concentration of nutrients during this season. The relative importance of point and nonpoint sources could likely be measured by comparing the direction of annual trends with that of the seasonal trends. For example, Station C24S49 TP trend analysis indicated a negative trend for all data, while seasonal results were insignificant. This implies that neither point nor nonpoint sources have increased significantly over the study period. Alternatively, dry and wet seasonal trends for C25S50 TP were both significantly positive, while all data (annual) was not significant. This result suggests that nonpoint (wet season dominant) and point (dry season dominant) sources might both have increased over the evaluated time period. For Station C44S80 TP, significant positive trends were identified for the dry season and all data (annual), while significant negative trend was identified for the wet season. This might imply that point sources have increased while nonpoint sources have decreased during the selected period. To verify these results, further investigation into the nutrient sources and their contributions must be completed. However, it should be noted that decreasing or increasing detection limits in time series data might result in false trends. The Tobit test is an alternative that could be used to account for the bias introduced by multiple censoring levels.


   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
A distinct wet season (27 May to 7 November with over 66% of annual precipitation) and dry season (8 November to 26 May) characterizes the climate in south Florida. Seasonality is evident for the major water quality indicators (specific conductivity, turbidity, color, and chloride) and nutrient species (NO2–N, NH4–N, TKN, PO4–P, and TP) at three SFWMD monitoring stations in the southern IRL watershed from 1979 through 2004. All the selected constituents exhibited significant seasonal patterns except for turbidity and NO2–N at Station C24S49. Almost all nutrient species exhibited an identical seasonal pattern of significantly greater concentration in the wet season than in the dry season. Seasonal differences were observed in trends, which suggested that the overall annual trend could be misleading and season analysis is necessary. Since trends in water quality variables over time may exist for seasons (separated from all data), it is useful to perform and present the trends for all data and each season separately. Results from trend analyses provide water managers with more information for guiding watershed strategies to meet water quality goals. Analysis of time series water quality data that includes this type of seasonal subset and annual trend analyses provides more information than a simple all-data analysis.


   ACKNOWLEDGMENTS
 
This research was supported by the South Florida Water Management District and the Florida Agricultural Experiment Station. The authors wish to thank Barbara Herrmann Welch and Meifang Zhou for their contribution to this study.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Qian, Y.
Right arrow Articles by Chin, D.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Qian, Y.
Right arrow Articles by Chin, D.
Agricola
Right arrow Articles by Qian, Y.
Right arrow Articles by Chin, D.
Related Collections
Right arrow Water Quality
Right arrow Surface Water Quality
Right arrow Water Conservation

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Vadose Zone Journal
Soil Science Society of America Journal Journal of Plant Registrations The Plant Genome