2. Study Area
The Barataria Basin is a 6600 km2
interdistributary basin located between the natural levees of the Mississippi River and Bayou Lafourche, an abandoned distributary of the Mississippi River that was cut off from the river in 1903 to prevent flooding [7
] (Figure 1
). The aquatic portion of the basin is dominated by forested wetlands in the upper-basin, emergent wetlands in the mid-basin, and open water in the lower-basin adjacent to the Gulf of Mexico. Hydrologically, we divided the area into three sub-basins. The upper-basin is fresh and dominated by cypress swamps and, to a lesser extent by fresh marshes, and is separated from the middle-basin by the embankment of US Highway 90 (US HWY 90). The upper-basin receives the majority of agricultural and urban stormwater runoff entering the basin. Stormwater runoff from uplands surrounding the upper-basin drain into Lac des Allemands via Bayou Chevreuil, which drains the most northwestern portion of the basin, and from Bayou Boeuf, which drains a sub-basin to the south. Water then flows from Lac des Allemands down basin through Bayou des Allemands, which is the only outlet to the mid-basin. The mid-basin is confined between Highway 90 and the Gulf Intracoastal Waterway (GIWW). Water flows from Bayou des Allemands to Lake Salvador and then through two outlets to the lower-basin. Lake Salvador also receives flow from the Lake Cataouatche subbasin, which receives stormwater runoff from the west bank of the New Orleans metropolitan area, and since 2002, Mississippi River water from the Davis Pond river diversion. Wetlands of the mid-basin are mostly fresh to low salinity emergent marshes and small areas of forested wetlands. Water leaving the mid-basin flows through a series of shallow water bodies, including Bayou Perot, Bayou Rigolettes, and Little Lake before entering Barataria Bay. Wetlands of the lower-basin are dominated by brackish and saline marshes, and water is exchanged with the Gulf of Mexico through four deep tidal passes.
The hydrology of the basin is characterized by a low astronomical tide, well-mixed and shallow water depths with long residence times, and a high degree of modification by canals and impact by agricultural runoff. The astronomical tide range is about 30 cm at the coast but essentially zero in the upper-basin; however, there is longer term water level variability due to factors such as heavy rainfall runoff, frontal passages and seasonal water level changes in the Gulf of Mexico. Water bodies in the basin are shallow ranging from 1 to 3 m and generally well-mixed and unstratified [25
]. Water turnover is one to four times per year in the upper-basin and a just few days in the lower-basin near the tidal inlets [19
]. Mississippi River water enters the lower-basin from the Gulf of Mexico during high discharge [23
]. Wetland hydrology of the basin has been extensively modified due to canals and water control structures. Additionally, most agricultural runoff is channelized directly into water bodies rather than flowing overland through wetlands, as it did when the system was natural. All of these factors impact the water quality of the basin. Land loss has been low in the upper-basin but is very high in the lower-basin [5
Under natural conditions, the basin received regular inputs of river water from crevasses, minor distributaries, and overbank flooding from both the Mississippi River and Bayou Lafourche [2
]. These inputs ceased when Bayou Lafourche was cut off from the Mississippi River in 1903, and with construction of continuous levees along the River after the large 1927 flood, which now prevent annual floodwaters from entering the basin [30
]. Minor amounts of river water enter the basin via the GIWW, and since 2002 there is episodic input of river water via the Davis Pond Diversion, which has the capacity to discharge up to about 300 m3
of river water. The natural levees along the perimeter of Barataria Basin were forested prior to European colonization but have almost all been cleared and converted to agricultural lands, as well as urban and industrial development. There are about 33,850 ha of farm fields above US HWY 90 that drain into the basin, almost all devoted to sugar cane. Agricultural runoff is the major cause of water quality impairment in the upper-basin [12
]. Under natural conditions, most runoff from uplands bordering the basin flowed through wetlands as sheet flow or via shallow meandering bayous. The long retention time and interface with wetlands lowered nutrient concentrations of overlying water. Since European colonization, there has been pervasive alteration of wetlands hydrology due to the dredging of canals for stormwater drainage, resulting in runoff with high nutrient concentrations bypassing wetlands and flowing directly into waterbodies [16
The two-way ANOVA model with interaction was fit, and the interaction between station and time period was highly significant for every response (p
<< 0.001 for CHL, SD, TP, TIN and TON, and p
< 0.01 for Nitrate). This indicates that the main effect for time period varied over the station location, as illustrated in Figure 2
, a feature that we will further discuss below, along with the summary in Table S2 (see supplementary online material)
. With such strong interaction effects, we refrain from over-interpreting the main effects, as masking or a more complex effect structure exists. However, it should be noted that in all cases the main effect for location was highly significant (p
< 0.0001), indicating strong spatial variation in the responses across stations. All of the pairwise comparison statements below have a Bonferroni family-wise protection at 0.05 for each response and are reflective of differences found within Figure 2
. Since we imposed the logarithmic transformation for the ANOVA on the water quality responses, Figure 2
displays the geometric means. As geometric means have unitless standard errors, the associated standard error bars provided in Figure 2
are those of the arithmetic means.
Mean chlorophyll a concentrations generally ranged from >20 to ~70 µg/L in the upper-basin, which is strongly influenced by upland runoff, and decreased to <20 µg/L in the mid- and lower-basin. Chlorophyll a levels were significantly higher in the upper-basin, with Lake des Allemands (LA), Bayou des Allemands north (BAN) and south (BAS) significantly higher than the other stations down basin and Bayou Chevreuil (BC) up basin (p < 0.05, protected). The decrease at BC was most likely due to light limitation due to high turbidity and shading from trees. Chlorophyll a was significantly higher during the Pre- and Post-transects compared to Seaton for stations BAN and Lake Salvador (LS), as well for all of the stations in the lower-basin. Although chlorophyll a concentrations were higher in Lake Cataouatche (LC) during the Seaton transects compared to Post-diversion, the difference was not significant.
Turbidity was significantly higher in the upper-basin, as reflected by low Secchi depth values, and water clarity increased down basin. Although the main effect trend of clarity was highly significant and apparently linear along the distance gradient, most of the mean Secchi depth differences that were found across the time periods were in the upper-basin. Specifically, for Bayou Chevreuil (BC), a significant ranking for Secchi depth was found (in decreasing order): Post-diversion, Pre-diversion, and finally Seaton. Pre- and Post-diversion Secchi depths were significantly greater than Seaton in the upper-basin, and Post-diversion Secchi depths were greater than Pre-diversion at LA, Little Lake (LL), and Barataria Basin north (BBN). Post-diversion mean Secchi depth was significantly greater than that found by Seaton in LC (p < 0.05, protected). All of these trends suggest increased water clarity over time.
Nitrate concentrations were generally less than 0.3 mg/L and, with the exception of station Bayou Chevreuil (BC), there were no significant differences between stations or transects. There was a spatial trend of higher concentrations at BC compared to Lac des Allemands (LA), and then increasing concentrations peaking in the mid-basin, and then decreasing to the Gulf of Mexico (GOM). Total inorganic nitrogen had a similar spatial pattern as nitrate.
Opposite to what was observed for nitrate and inorganic nitrogen, total organic nitrogen was lower in BC compared to LA. After LA, there was a significant decreasing trend for stations approaching the GOM (p < 0.05, protected). Overall, total organic nitrogen was significantly higher during the Seaton transects compared to the Pre- and Post-diversion transects at most stations, including Lake Cataouatche (LC; p < 0.05, protected).
Total phosphorus concentrations ranged from ~0.2 to 0.3 mg/L at BC and generally decreased down basin. Total phosphorus was significantly higher for the Seaton transect in the lower-basin, with significantly higher mean TP concentrations compared to both Pre- and Post-diversion transects, especially at Barataria Bay south (BBS) and GOM (p < 0.05, protected). There were significantly higher mean phosphorus values during the Pre-diversion transects relative to Post-diversion, and no differences to compared to Seaton. There were highly significant differences for total phosphorus in the upper-basin (BC, LA, and BAN; p < 0.05, protected).
We aimed to understand how the station locations are hydrologically connected, within each time period. To achieve this, objectively and visually, we first clustered stations that were similar to each other and plotted these clusters onto the primary principal components of the water quality variables. Figure 3
(left panels) shows the k-means clustering results by projecting the stations onto the first two principal components (PCs). Each point in the plots represents a station. Different colors and symbols are used to separate the samples from different clusters. The number of clusters for each plot was chosen to maximize the average silhouette width. The percentages within the parenthesis are the proportion of total variance explained by the corresponding PC. The first two PCs explain over 80% of the total variance in all three cases. In the Seaton data, the optimal number of clusters is three, which is less than the Pre- and Post-diversion cases. This is probably due to the small sample size for the Seaton data as compared to those for Pre- and Post-diversion. Thus, these biplots give a good impression of the actual clustering. Note that the optimal numbers of clusters for Pre-diversion and Post-diversion data are nine and seven, respectively. In order to have a reasonable and fair comparison, we choose eight clusters for both of these cases. The average silhouette width for eight clusters is very close to its maximum value in both cases. The Pre-diversion, Post-diversion, and Seaton clustering show goodness of fit. The ratio between between-cluster variation and total variation, which is similar to the R-square in regression, is 93.5%, 92.1%, and 68.9%, respectively. Values closer to 100% yield better separation.
The means of the station groupings yield a number of interesting spatial and temporal patterns among Seaton, Pre-diversion, and Post-diversion transects. All three time periods show an enriched upper-basin with turbid waters, high nutrients and chlorophyll a levels. In contrast, lower-basin stations generally have clearer water, lower nutrients, and lower chlorophyll a. In the bayou stations draining into Lac des Allemands, chlorophyll a is lower and nutrients are higher and water is somewhat more turbid. However, it is likely that light limitation, due both to high turbidity levels and shading by swamp forests lining the narrow bayous are the main factors leading to lower phytoplankton biomass. Stations downstream of Lake Salvador generally have low chlorophyll a and nutrient levels and greater water clarity although there seems to be minor enrichment after the opening of the Davis Pond diversion but this was not significant.
When clusters were displayed on maps of the basin with sampling stations included, clustering shows strong spatial patterns (Figure 3
right). The arrows in Figure 3
(right panels) show how clusters are connected hydrologically and reflect water flow in the basin. The clusters separated clearly from upper- to lower-basin. This finding reflects water flowing down basin from highly enriched water of the upper-basin to cleaner water of the lower-basin. When these spatial patterns are combined with water flow patterns, it is clear that the clusters reflect both the hydrologic connectivity of the basin from fresh to saline as well and biogeochemical processing of nutrients as water flows down basin.
Lake Cataouatche is the area that has changed the most with the opening of the Davis Pond river diversion. When the Seaton transects occurred, Lake Cataouatche received considerable agricultural and urban runoff. The lake was turbid with relatively high nutrient and chlorophyll a
levels similar to the upper-basin. With the opening of the Davis Pond diversion, the Mississippi River became the main source of freshwater to the lake. Diverted river water first flows through a large wetland receiving basin where most river sediments settle so the water flowing into the lake is relatively clear. In addition, dense submerged aquatic vegetation beds promote sediment settling and inhibit resuspension. The mean Post-diversion Secchi disk depth was >100 cm, the clearest water in the basin. Nitrate was the main inorganic nitrogen form with a mean concentration <0.2 mg/L compared to 1–2 mg/L in the river. Nitrate is rapidly reduced by denitrification and plant uptake in the Davis Pond wetlands and shallow submerged sediments of the lake [33
]. The turnover time of the lake prior to the Davis Pond diversion was 1–2 months, but when the diversion is running, it is capable of replacing the entire volume of Lake Cataouatche in three days [35
]. Thus, phytoplankton growth is likely limited both by low inorganic N and rapid flushing time.
CHL, TON, and TP are positively correlated. SD is negatively correlated with the other four variables. TIN is weakly correlated with CHL, TON, and TP. See Table 1
The first principal component (PC1) in all three cases mainly reflects the difference between SD and the sum of CHL, TON, and TP. See Table 2
TIN is the most influential factor in second principal component (PC2) for all three cases. See Table 2
For the Seaton case, cluster 1 and cluster 2 samples have positive PC1 scores (implying high SD and low CHL, TON, and TP values), while cluster 3 samples have negative PC1 scores (implying low SD and high CHL, TON, and TP values). See Figure 3
Cluster 2 samples have positive PC2 scores (implying high TIN), while cluster 1 samples have negative PC2 scores (implies low TIN). See Figure 3
For pre- and post-diversion cases, cluster 4 and 6 in pre-diversion correspond to cluster 4 and 5 in post-diversion. Both clusters have the largest PC1 scores, which implies high SD and low CHL, TON, and TP values.
The PC2 loadings for TIN in pre- and post-diversions have opposite signs. The top cluster in pre-diversion implies low values in TIN, while the top ones in post-diversion implies high TIN values.
The TSI scores for the Seaton, Pre-diversion, and Post-diversion transects yielded very similar spatial patterns, with eutrophic conditions (positive) in the northern basin and mesotrophic (negative) in the lower basin (Figure 4
; Table 3
). Scores for stations upstream of Lake Salvador (LS) were generally greater than one. The highest scores were for Bayou des Allemands north (BAN), Lac des Allemands (LA), and bayous receiving agricultural runoff (LA, BAN, and LC). These stations had high nutrients, chlorophyll a
and turbidity with low Secchi disk depths (see Figure 2
). Downstream of Lake Salvador, TSI scores for the three periods were generally less than −1, indicating more mesotrophic waters with greater clarity and lower chlorophyll a
and nutrient levels (BP, LL, BBN, BBS and GOM). Post-diversion scores in this lower region were somewhat elevated compared to Pre-diversion scores, suggesting a slight tendency towards more enrichment. Lake Cataouatche (LC) scored greater than zero for the Seaton transects while the Post-diversion score was about −1. As noted above, the shift was likely due to relatively clearer river water (after sediments had dropped out) entering the lake, the sediments having been retained in the Davis Pond wetlands as well as rapid reduction of NO3
due to high denitrification rates.
The TSI results for the Seaton re-analysis were very similar to the original analysis. The analysis for the Pre- and Post-diversion data sets yielded results that were very similar to the Seaton re-analysis. This indicates that the trophic status of the basin waters has remained relatively unchanged over the period of the two studies. Upper-basin stations are in the eutrophic to hyper-eutrophic range while Lake Salvador and lower-basin stations are mesotrophic. It is interesting to note that Lake Cataouatche was eutrophic for the Seaton study, but was mesotrophic for the Post-diversion transects. This reflects clear water due to river sediments being deposited in the Davis Pond outfall area, the rapid reduction in NO3 most likely due to denitrification and plant uptake, and the rapid flushing of the lake.