Wetland Location, Climate, and Weed Treatments

The Lochinvah wetland is approximately 30 ha in size, located on the Burdekin floodplain complex, which is an extensive network of permanent and ephemeral palustrine wetlands (Figure 1). Land use across the floodplain is dominated by sugarcane, with some cattle grazing and tree crops (Davis et al., 2014). Rainfall is strongly seasonal with 90% of total annual rainfall occurring during the wet season (November and March). Mean annual rainfall on the floodplain (1887 to 2016) is 1,047 mm (Burdekin Shire Council gauge station, station number 33001). The wet seasons during this study were low compared with long-term data (386 mm, 2014/15; 386 mm; 133 mm 2015/16), within the 5% percentile of historical records for the region. Daily flow recorded at the Northcote gauge station (station number 119101A) on Barratta Creek reveals that this study coincided with an extended period of low rainfall, well below antecedent years prior to this study (Supplementary Material S1).

Four aerial spray applications were completed to target water hyacinth (Eichhornia crassipes): Spray 1—6 December 2013 (30 ha of wetland); Spray 2—30 June 2014 (35 ha); Spray 3—20 January 2015 (30 ha); Spray 3—17 May 2015 (30 ha); and Spray 4—25 September 2015 (30 ha). Each spray was completed in 4 hr using a helicopter (<10 m above the water surface), where glyphosate (and freshwater mixture—10 mg/L ratio) was sprayed as a mist across invasive aquatic plants in the wetland. Sprays 1 and 2 had low aquatic plant coverage (∼5% wetland coverage), restricted to the wetland edges, while Sprays 3 and 4 aquatic plant growth was high (approximately 80% and 60%, respectively).

Water Quality Measurement

A calibrated multiprobe data logger (Hydrolab DS5, OTT Hydromet, CO, USA) was deployed in the near-surface water layer (0.2 m below the surface, attached to a surface float) to monitor diel cycling of temperature, pH, electrical conductivity, and dissolved oxygen (%) measured at 20 min intervals in the middle of the wetland (though only DO and temperature are examined here). The Hydrolab was deployed for approximately 2 to 3 weeks prior to each spray and remained in place until after (several weeks) spraying, or was deployed shortly before (between 1 and 2 weeks) and shortly after (1 to 4 weeks) spraying to record post spray water quality conditions. A Hydrolab was deployed before the spray application in January 2015; however, the logger was damaged by an estuarine crocodile (Crocodylus porosus) with all data (before and after spray) lost.

A second logger configuration (Pendant Temperature, Onset Corporation, Bourne, MA) was deployed to measure water temperature every 20 min with a logger position at a depth of 0.2 m below water surface (surface), and a second hobo logger deployed at 0.1 m above the bottom of the waterhole (bottom). Following Waltham and Sheaves (2017), the surface logger was attached on the underside of a 0.15-m buoy to shield the logger from direct sunlight, while the bottom logger was attached to a concrete block (so the sensor remained at approximately 1.5 m depth during the logging period). This rig was positioned adjacent to the Hydrolab and remained in place between 1 November 2014 and 28 February 2015.

Flow data recorded on the gauge in Barratta creek (recorded as part of the Queensland Department of Natural Resources program), adjacent to the wetland, highlight that the study coincided with a period of low flow (Supplementary Material S1). This low flow period has been shown to reduce the extent and duration of connection across the Burdekin floodplain more broadly (Waltham, 2017). The wetland depth at the logger site was 1.25 m at the beginning of the study, and although depth increased following small rain events, there was a net loss in wetland water level (0.95 m) on completing the study.

Data Analysis

The change in dissolved oxygen saturation in the wetland before and after each spray was examined separately using an independent t test. To accommodate for diel differences in water temperature, the daily average was calculated before and after each spray. These analyses were completed using SPSS (v23.0) using a significant probability of p < .01, and data were checked for normality (qq plots) and homoscedasticity (residual vs. fitted values) and required no transformation (Underwood, 1997).

Thermal exposure risk to aquatic fauna was examined using frequency distribution plots (Wallace, Waltham, Burrows, & McJannet, 2015; Wallace et al., 2017). We used the surface Hydrolab temperature data between 25 November 2013 and 25 January 2014 (<5% coverage), while hobo loggers (surface and bottom data) examined whether the exposure risk changed during a period of high coverage (80% coverage; 25 November 2014 and 25 January 2015). To ensure the two loggers were comparable, a regression was fitted for a separate logging period (8 August 2014 and 3 September 2014, R² = 0.96; p < .01). Based on the regression equation (y = 1.00029x +0.1723; y is the hydrolab logger adjusted value and x is the hobo logger recorded temperature), hydrolab logger was adjusted accordingly, which helped to correct for the slight logger difference. Climate data were sourced from the Scientific Information for Land Owners database ( https://www.longspaddock.qld.gov.au/silo/) where the mean of the daily ambient air temperature for the region was 31.2℃ (±0.3 SE) and the mean humidity was 53% (±6.9 SE) during 22 November 2013 and 25 January 2014, while these were slightly higher between 22 November 2014 and 25 January 2015, that is, 32.2℃ (±0.2 SE) and 54% (±10.2 SE), respectively (Supplementary Material S2).

Dissolved Oxygen

DO% saturation during the four aerial spray periods is presented in Figure 2, with summary statistics in Table 1. Cycling in the first spray (December 2013) ranged between 0% and 160% (Figure 2), with critically low levels occurring in the early morning hours (4 a.m. to 6 a.m.) though (at approximately 10% saturation/hr) and increased during the day reaching maxima between 3 p.m. and 4 p.m. There was no significant difference in DO before and after Spray 1 (independent t test, p = .187, F = 1.793). During the second spray (June 2014), DO reached overnight lows toward 20% and, however, recovered again by mid-afternoon. There was a significant difference in daily average DO before and after Spray 2 (independent t test, p = .01, F = 6.848). Spray 3 (May 2015) occurred during the period when aquatic plant cover was highest, ∼80% surface coverage, DO cycling was less pronounced with daily maxima reaching ∼ 30% (again between 3 p.m. and 4 p.m.) with nighttime minima of 0% (Table 2). There was no significant difference in daily average DO before and after Spray 3 (independent t test, p = .06, F = 3.433). By the final spray, DO was persistently less than 1% saturation and remained so before and after spraying, though there was no significant difference in daily average DO (independent t test, p = .02, F = 8.287).

Figure 2.

Time series plot (20 min logging period) for surface (0.2 m) dissolved oxygen (%) in wetland before and after spray events. Blue vertical lines show spray events, and black bars are daily total rainfall recorded at Burdekin Shire Council station. Photos illustrate each spray application.

Table 1.

Summary Statistics of Dissolved Oxygen Concentrations Recorded During the Four Survey Periods in the Wetland (From Figure 2).

Table 2.

Daytime (6 a.m. to 6 p.m.) Statistics of Dissolved Oxygen Concentrations (DO%) Recorded During the Four Survey Periods in the Wetland.

To provide context field, DO data were compared with Chronic Trigger Value and Acute Trigger Value thresholds (long-term exposure is called chronic and short-term exposure is called acute—both may cause health effects that are immediate or health effects that occur days or years later) determined for barramundi (Lates calcarifer; Butler & Burrows, 2007). In accordance with these thresholds, barramundi would have a lower risk to Acute Trigger Value and Chronic Trigger Value exposure during the first and second sprays, though exposure increased during the third and final spray with effectively 100% of the logger period below these trigger values (Table 1).

Water Temperature

The water temperature logging series (64 days) is presented in Figure 3. There is a difference in the temperature conditions between the first logging period (where plant coverage was low; <5%), compared with the same logging period a year later when aquatic plant coverage across the wetland was high (∼80%). Unfortunately, the first year of logging included surface water temperature (0.3 m below surface), while the second year included both surface (0.3 m) and bottom (0.15 m above wetland sediments) loggers. Notably, there were very little differences between surface and bottom temperature in the second year indicating that the water column was vertically well mixed. The difference in surface temperature between the first year and second year highlights that the thermal risk for fish is in a state of continual fluctuation. For bony bream (Nematolosa erebi, a widely distributed freshwater fish species across the northern Australia, the T_pref threshold, i.e., the preferred temperature point that which beyond fish display acute hypothermic responses for N. erebi is 31℃; Pusey, Kennard, & Arthington, 2004), the surface water temperature exceeded the threshold 16% of the logging period in the first year, increasing in the second year to 33% at the surface and 32% at the bottom. By comparison, the fly-specked hardhead (Craterochephalus stercusmuscarum), another common wetland species in the region (Waltham, 2017), has a critical thermal maximum (T_max is the maximum temperature end point before death) of 33.4℃ (Burrows & Butler, 2007) which was exceeded only in the second logging period at both surface (8%) and bottom (3%) waters.

Figure 3.

The percentage of time water temperature exceeded a given fish temperature threshold in the wetland. The frequency curves were compiled using all 20-min logging data between 22 November and 25 January. Gray line is 2014 (low [∼5%] coverage of invasive aquatic weeds) and black line is 80% weed coverage (black line surface logger, broken black line is bottom logger). The exceedance of thresholds T_pref 31.8℃ and T_max 34.8℃ are shown for bony bream (Nematolosa erebi; black full line surface, broken line bottom logger) and fly-speck hardyhead (Craterocephalus stercusmuscarum; gray full line surface, broken line bottom logger).

Water Quality Considerations in Wetland System Repair

These data outline that spraying using a helicopter to control invasive aquatic plants does not dramatically reduce water quality conditions, here dissolved oxygen, over the following weeks after mitigation. In fact, this treatment is appropriate when cover is minor, and restricted to the wetland margins because it had little consequences on DO availability for fish. This was evident during Sprays 1 and 2, where the diurnal DO amplitude was proximal before and after treatment. In both surveys, the early morning oxygen deficit recovered over the course of the day, meaning that the net amount of aquatic plant photosynthesis compensates the wetland nocturnal respiration deficit. Even during the early morning hours when DO was low, fish were still observed active in the wetland, presumably the ability to deal with the availability of DO is supplemented via surface respiration (Richards, 2011; Yang, Cao, & Fu, 2013), or because of increased ventilation rates that balance oxygen pressure in the arterial blood with that found under normoxic conditions (Heath, 1995).

Problems occurring with spraying aquatic weeds in wetlands are in situations where a higher coverage of aquatic plant material exists (up to 80% recorded at the peak here), in combination with ongoing routine spraying and low rainfall, does the risk of hypoxia become relevant and concerning for coastal wetland function and productivity. These data show under such extreme conditions that overnight DO deficits are not able to recover with daytime photosynthesis. In fact, DO during Sprays 3 and 4 were obstinately below thresholds for fish common in the region (Butler & Burrows 2007). The prolonged anoxic conditions measured during Sprays 3 and 4 are probably resultant of accumulating bioavailable dissolved organic carbon in the wetland, where the rate of oxygen consumption necessary for decomposition exceeds the rate of oxygen diffusion in the water (even including atmospheric diffusion; O'connell, Baldwin, Robertson, & Rees, 2000). Under anoxic conditions, aquatic animals, such as fish, are more susceptible to asphyxiation (Breitburg, 2002), as in the example here for the barramundi (Lates calcarifer), a coastal iconic species (a popular sports fish species) widely distributed across northern Australia (James et al., 2017), and with a diadromous movement ecology where it needs to migrate between fresh and tidal waters to complete lifecycle stages. This fish is affected by poor water quality, particularly dissolved oxygen (Collins, Clark, Rummer, & Carton, 2013; Flint, Crossland, & Pearson, 2015), with fish kills regularly occurring across the floodplain in summer with the first rainfall flow which delivers oxygen deficient “black water” (Butler & Burrows, 2007; Butler, Loong, & Davis, 2009). The challenge of controlling invasive aquatic plants without contributing to effects on DO highlights the enormity of the invasive aquatic plant problem for land managers in the GBR catchments (Arthington et al., 2015; Dubuc,Waltham, Malerba, & Sheaves, 2017), but also apparently elsewhere across the distribution of this invasive aquatic weed (Masifwa, Twongo, & Denny, 2001).

In tropical northern Australia, freshwater fish face aquatic thermal exposure risks and particularly so leading up to and during summer months (Wallace et al., 2015). Temperature data here revealed that wetland fish are exposed to thermal regimes exceeding conditions ideal for optimal growth, even reaching lethal conditions. Similar thermal risks were reported by Wallace et al. (2017) in arid river waters of northern Queensland and also in a coastal freshwater tropical wetland on the Herbert River floodplain (approximately 150-km north of Lochinvah wetland; Waltham, 2017). In our wetland site, the risk was highest at the surface water layer (∼0.2 m) with bottom waters (∼2 m) rarely, if ever, exceeding thresholds. This presents the model that the thermal risk for fish at the surface waters could be overcome by accessing, cooler, deeper wetland waters. However, the thermal refugia afforded in bottom waters would be compromised by the fact that it exposes fish to critically low DO conditions in the bottom waters (Waltham et al., 2013), especially in weed-choked wetlands where the surface vegetation mat (including water hyacinth) averts vertical mixing of oxygenated surface waters with bottom waters (Villamagna & Murphy, 2010). The warmer surface water conditions during the second years’ monitoring contrasts Waltham (2017), where surface water temperature was cooler under excessive aquatic plant coverage. A likely reason here for higher temperature with high plant coverage could be explained by the warmer ambient air temperature experienced in the region (see Supplementary Material S2). In addition to fish, freshwater crustacean species also have thermal exposure thresholds (Stewart, Close, Cook, & Davies, 2013), many of which are an important food source for these type of wetlands.

Controlling Invasive Aquatic Wetland Plants

Attempts to repair and restore wetland ecological function with respect to removal of invasive aquatic vegetation in GBR catchments has concerted on three approaches. The common is aerial spraying (Burrows et al., 2012), while data accompanying this treatment option are not generally available. The most important point for manages administering weed removal based on the data provided here is that when restricted to edges, aerial spraying for invasive plant maintenance presents a low risk of hypoxia exposure to fish. Under a situation where rainfall flow has been low combined with high plant coverage, continuing to implement spraying contributes to water quality implications. Indeed, spraying has the added risk of herbicide and pesticide toxicity to local receiving waters, and also coastal ecosystems some distance away (e.g., coral reef water quality monitoring has detected trace levels of herbicide from land farming; Lewis et al., 2009). Another strategy is mechanical removal where an aquatic harvester progressively removes floating weed species by collecting it using a floating conveyor belt fixed to a barge, with the material collected and dumped in piles on the bank for drying and disposal. While this method has been used in the region, barge operation and maintenance is expensive, in addition to the ongoing need for follow-up spot spraying where the barge is not able to access. The third strategy, which has been recently examined on the Herbert River floodplain (150 km north of Lochinvah wetland), is to remove bund earth walls which reinstates tidal exchange with the upstream freshwater wetland (Waltham, 2017). In that trial, those authors first developed a hydrodynamic model to define the section of the wall to remove that would maximize saltwater ingress to destroy invasive freshwater aquatic plants. After the bund wall was breached, ingress of saltwater quickly killed most of the WONS, which contributed to improvements in DO concentrations (Waltham, 2017). Not only did the saltwater ingress destroy the WONS (in the case of that study: water hyacinth, Eichhornia crasspies, and Hymenachne, Hymenachne amplexicauli), but the reconnection had the dual benefit of permitting estuarine fish species (including barramundi, Lates calcarifer) to again access to the wetland (Waltham, 2017). The cost to remove the bund wall (cost to hire excavator machinery) is a one-time expense, compared with on-going costs in aerial spraying and weed harvesting.