Turbidity-suspended particulate matter (SPM) calibration
The turbidity-SPM calibration plots (Fig. 3) demonstrate that the high-resolution turbidity measurements provide a robust proxy for SPM concentrations at sites A (R
2 = 0.764), B (R
2 = 0.801) and E (R
2 = 0.844) over a wide range of SPM concentrations (1–458 mg L−1), turbidity values (1–758 NTU) and flow measurements (0.001–0.489 m3 s−1). The best fit linear regressions at site A and, in particular, site B have a slight offset from the zero intercept, indicating instrumental error in the optical turbidity probe. Despite this, the calibrations for sites A and E remain robust and the R
2 values are consistent with those reported in other studies (Harrington and Harrington 2013; Thompson et al. 2014).
Diel turbidity cycles
The 30-min resolution turbidity time-series for sites A, B and E during 2013 are presented in Fig. 4. At this timescale, the turbidity record is dominated by large-scale fluctuations (>200 NTU) caused by catchment-wide sediment mobilisation during heavy rainfall events, with this variability obscuring any smaller-scale diel turbidity trends. The turbidity record is also impacted by spurious ‘noise’ caused by random isolated turbidity peaks which can result in misinterpretation of the true stream turbidity. An example of this can be seen at site B during June and July 2013 where the mean turbidity is 7 NTU, but a small number of random isolated peaks make the turbidity appear much greater when plotted at this timescale. However, by focusing upon a period of low-flow conditions and smoothing the data with a Savitzky-Golay filter, this noise is removed, and pronounced diel turbidity cycles are revealed (Fig. 5). Shown here for a 20-day period between the 10th and 29th April 2013, sites A and E consistently recorded daily peaks in turbidity values between 21:00 and 04:00, centring around midnight. Turbidity values subsequently declined by ~10 NTU towards the lowest-recorded daytime values which occurred between 10:00 and 14:00, centring on midday.
Considering the time of sunset (20:00) and sunrise (05:45) in mid-April, the timings of these turbidity peaks and troughs are consistent with the hypothesis that nocturnal bioturbation is responsible for generating these cycles. These timings are also consistent with the observations of Harvey et al. (2014) and Rice et al. (2014), who similarly found turbidity peaked at around midnight in the River Windrush, Oxfordshire and River Nene, Northamptonshire, UK, respectively, with both studies linking these cycles to the nocturnal activities of signal crayfish.
Whilst only shown here for a 20-day period in April 2013, these diel turbidity cycles are present under baseflow conditions throughout much of spring and summer at sites A and E. The cycles do, however, weaken during winter when lower-water temperatures and higher-flow conditions have been shown to reduce crayfish activity (Bubb et al. 2002; Johnson et al. 2014). This can be seen in Fig. 6 for a 20-day period in January 2013 where the diel cycles were not evident at sites A or B. Site E retained evidence of diel cycles, but daily oscillations in turbidity (3–4 NTU) were substantially lower than that observed in April (~10 NTU), thus supporting the hypothesis of reduced crayfish activity during the winter. Diel cycles were also masked during precipitation events when larger-scale sediment mobilisation obscures these smaller diel fluctuations, an observation also made by Halliday et al. (2014). This can be seen during the precipitation events on the 12th April (Fig. 5) and 9th January (Fig. 6).
At site B, diel turbidity cycles were noticeably less pronounced during spring than observed at sites A and E (Fig. 5). Considering the close proximity of these three sites (<600 m distance), it is unlikely that wider environmental factors (e.g. temperature/rainfall variations) are responsible for the absence of diel cycles at site B. Similarly, all three sites used the same type of turbidity probe and were serviced regularly, meaning instrument effects are also unlikely to be responsible. A reduced impact from localised bioturbation could, however, explain these differences. Lower water levels (0.03 m mean stage during this period) and denser stream vegetation in mini-catchment B likely make this part of the channel less accessible to larger fish and crayfish, thus reducing the incidences of streambed and channel bank sediment disturbance from nocturnal feeding and burrowing activities. Similarly, a greater incidence of stream channel bank slumping in mini-catchment B likely reduced the availability of suitable burrowing sites compared to mini-catchments A and E where more frequent bank re-profiling and dredging to improve channel drainage results in steeper, cleaner bank faces into which crayfish can burrow. Lastly, the bed substrate in mini-catchment B is largely gravel dominated as opposed to silt dominated in mini-catchments A and E, thus limiting the availability of fine sediment for resuspension.
Impacts on suspended particulate matter (SPM) concentration and load
Based on the smoothed April turbidity data, these diel turbidity cycles translate into a statistically significant increase in median SPM concentration of 9.9 mg L−1 at site A and 9.3 mg L−1 at site E between midday and midnight (Table 1). In other words, median SPM concentrations increased by 55 and 76 % between midday and midnight at sites A and E, respectively, during this spring baseflow period. Converting this into total night-time (18:00–05:30) and daytime (06:00–17:30) SPM loads over the 20-day period reveals 75.4 kg (53.5–106.8 kg based on 95 % confidence intervals of the stage–discharge rating curve and SPM–turbidity calibration) of SPM was exported during the night at site A, with 64.9 kg (46.0–91.9 kg) exported during the day. At site E, 86.7 kg (64.1–117.2 kg at 95 % confidence interval) was exported at night, with 66.5 kg (49.1–90.0 kg) exported during the day. Assuming nocturnal bioturbation is the causal factor of increased SPM concentrations, these night-time SPM loads based on the smoothed turbidity data are ~16 % greater at site A and ~30 % greater at site E compared to what can be considered the baseline, lower bioturbation, daytime loads.
These load estimates are comparable to those obtained by Rice et al. (2014) in the headwaters of the River Nene, UK, who calculated an increase in night-time suspended sediment load of 46.9 % compared to daytime load under baseflow conditions. This reduced to a 20.3 % increase in night-time loads when higher-flow flood days were included due to the reduced importance of bioturbation in initiating sediment mobilisation during heavy precipitation events. In headwater streams exhibiting diel cycles in Iowa, USA, Loperfido et al. (2010) also reported that mean total phosphorus (TP) concentrations were twice as high at night (0.41 mg L−1) than during the day (0.20 mg L−1) which may be linked to elevated night-time SPM concentrations. In our study, we found only limited evidence for this, with midnight median TP concentrations of 0.073 mg L−1 being marginally greater than the 0.065 mg L−1 recorded at midday, although this difference was not significant (Table 1). This can be explained by the relatively weak correlation between TP and SPM concentration during this period (r = 0.34) indicating that the dissolved fraction is more important than particulate material in controlling TP concentration under these low-flow conditions.
With such a strong and regular diel turbidity cycle observable at sites A and E during the spring and summer months, it is important to rule out any potential instrument artefacts being responsible for inducing these trends if there is to be confidence in relating them to bioturbation. Firstly, it is necessary to recognise that turbidity is just a surrogate measure of SPM concentration and other factors, such as phytoplankton blooms, can lead to an increase in turbidity without the requirement for elevated SPM concentrations. However, the SPM-turbidity calibrations were robust over a wide range of values (Fig. 3), so there is confidence in relating turbidity to SPM concentrations.
Temperature is also known to affect the accuracy of turbidity measurements, with lower water temperatures leading to higher turbidity values and vice versa (Loperfido et al. 2010). Thus, an instrumental anomaly in response to diurnal temperature variations could potentially yield the same diel pattern in turbidity as observed at sites A and E. However, as can be seen in Fig. 7, temperature values at site E during April 2013 fell to a minimum just before dawn, several hours after the peak in turbidity around midnight. Additionally, weak correlation (r = 0.11) between temperature and turbidity implies no causal relationship exists between the two parameters. This point is emphasised during the winter when prominent diurnal temperature variations still occur, yet diel turbidity cycles weaken at this time (Fig. 6).
Along with temperature, the influence of probe exposure to sunlight can also be ruled out as a potential driver of these cycles because the turbidity probes used here were located in flow cells inside the bankside-monitoring kiosks and therefore kept in darkness. Additionally, comparison of turbidity values with an additional probe that was located instream at site E revealed that both probes recorded very similar diel cycles in turbidity with respect to both timing and magnitude (data not shown). Correlation of turbidity with other parameters which display diel cycles, including stage (r = 0.02) and dissolved oxygen (r = 0.33) which can both vary diurnally due to changes in temperature, transpiration and photosynthesis, revealed that changes in these parameters were also unlikely to be causal factors as they do not align with the timings of the turbidity peaks and there exists no obvious mechanistic connection (Fig. 7).
However, diel cycles in stream water pH, which peak at around midday and decline to a minimum around midnight, do correspond more strongly with the turbidity trends. This daily pH variability relates to CO2 consumption during photosynthesis by submergent macrophytes and algae during the day decreasing water acidity and CO2 release at night during respiration increasing acidity (Nimick et al. 2005). Whilst correlation between pH and turbidity is relatively strong (r = 0.45), it is difficult to understand why a regular ~0.1 decrease in pH at night would induce a ~10 NTU increase in turbidity. This is because the more acidic conditions at night should favour dissolution, not precipitation, of carbonate material in the stream and would therefore reduce the suspended particulate fraction. Additionally, the pronounced ~0.2 decline in pH on the 26th April in response to a small rainfall event was not matched by a similarly pronounced shift in turbidity values, indicating low turbidity sensitivity to fluctuating pH.
Significance and further research
Whilst the evidence presented here does not conclusively prove that the diel turbidity cycles witnessed in the River Blackwater were driven by bioturbation, the seasonality and timing of elevated turbidity, visual observations of crayfish activity and an absence of robust mechanistic connections with other water quality parameters means this represents a plausible causal mechanism. These findings therefore support a growing body of research highlighting the detrimental impact of non-native signal crayfish on fluvial geochemistry, fluvial geomorphology and native aquatic biota (Crawford et al. 2006; Johnson et al. 2010; Harvey et al. 2011). Their large size (typically 10-15 cm in length), aggressive nature, high population densities (up to 20 per square metre) and ability to rapidly colonise new environments have seen signal crayfish populations spread rapidly across Europe putting an increasing number of freshwater environments at risk (Holdich et al. 2014; Kouba et al. 2014). Further research is required in a wider range of fluvial environments at a variety of spatial and temporal scales to fully understand the importance, extent and underlying processes behind the nocturnal fine sediment mobilisation phenomena reported here.