Field and laboratory characterization of rotenone attenuation in eight lakes of the Kenai Peninsula, Alaska

Abstract

Rotenone is a pesticide commonly used to eradicate Northern Pike (Esox lucius), an invasive species, in Southcentral Alaska. The present work incorporates a field investigation of rotenone attenuation in eight lakes of the Kenai Peninsula, following a CFT Legumine® treatment in October 2018 and a laboratory simulation to determine persistence under light/dark and sterile/nonsterile conditions representative of Southcentral Alaskan winters. In the field, rotenone degraded within <60 days of application in all lakes, while rotenolone, the primary product of rotenone degradation, persisted for up to <280 days post-treatment at two locations. Prolonged rotenolone attenuation was most likely caused by short days and ice cover between October and April. This hypothesis was supported by a laboratory simulation which revealed photolysis as the dominant process driving the overall degradation of rotenone and that microbial degradation will significantly contribute in the absence of sunlight under simulated “winter” conditions of 4 °C. Degradation model fit comparisons (pseudo-first order, multi-parameter linear, and gamma) indicate the most accurate prediction occurred when modeling all eight lakes grouped together in a single dataset, combined and treated with pseudo-first order model kinetics, based on Akaike information criteria (AIC) scores.

1.0. Introduction

Rotenone is applied as an active ingredient of the commercial formulation CFT Legumine® to eradicate invasive or unwanted fish species. It is extremely toxic to fish with 24 hour lethal concentrations as low as 2 μg/L (Waller et al. 1993) reported in Southcentral Alaska where it is applied to eradicate Northern pike (Esox lucius); an invasive species that poses a serious economic and environmental threat through heavy predation of native salmonid species (Rutz 1996; Rutz 1999; Sepulveda et al. 2013). CFT Legumine® contains an additional five rotenoids, including rotenolone (the primary degradation product) and deguelin (an isomer) which are present at concentrations roughly 30 – 35 % and equivalent to that of rotenone, respectively (Redman, 2020). Due to their recent identification, analytical standards for these rotenoids are either unavailable or cost prohibitive and while CFT Legumine rotenoids will photodegrade in the laboratory their overall environmental fate remains a topic of current research (Redman, 2021). In the laboratory, rotenone will degrade via a combination of photochemical pathways in water (Redman, 2021) and soil (Cavoski, 2007), biological pathways in soils (Cavoski, 2008), as well as base catalyzed hydrolysis (Thomas, 1983; Redman, 2021). Organic carbon normalized sorption coefficients (Koc) for rotenone have been calculated, though not experimentally derived, between 3.6 – 4.0, and while sorption processes were shown to proceed much slower than photochemical processes on soils in the laboratory, previous field studies have observed rotenone sorption to sediments where it may be biologically degraded (Cavoski, 2007; Vazquez, 2012). Prior studies that evaluated the degradation of rotenone reported water temperatures between 8–24 °C (Gilderhus et al. 1986; CDFG (California Department of Fish and Game) 1988a, 1988b, 1989a, 1989b, 1991a, 1991b, 1992a, 1992b, 1992c, 1994a, 1994b, 1995, 1996, 1999a; Dawson et al. 1991; Chadderton et al. 2001; Vasquez et al. 2012; Finlayson et al. 2014), higher than the typical annual average fall and winter temperature of 8.0 and 4.0 °C in Southcentral Alaskan surface waters. Light (Cheng et al. 1972) and temperature (Dawson et al. 1991) have been shown to impact the rate at which rotenone degrades in the environment; the degradation rate of rotenone and its primary degradation product, rotenolone, may be significantly different in Alaska (above 60 °N) compared to treatment sites in more temperate climates due to its colder temperatures, earlier potential onset of ice cover, and lower daily solar irradiation. Prior studies from California and Wisconsin reported half-lives (DT50) ranging from 3.5 – 10.3 d and temperatures as low as 0 °C (Gilderhus et al. 1986, 1988; CDFG 1997; Finlayson et al. 2001). However, in these field treatments, temperatures rose above 8 °C, and the solar irradiation intensity was significantly higher than Alaska in autumn. Previous field experiment locations were as far north as 43° (Finlayson et al. 2014) for sites in the northern hemisphere and as far south as −41° (Chadderton et al. 2001) in the southern hemisphere; these sites reported DT50 ranging from 0.92 – 10.3 d.

Environmental fate processes governing rotenone attenuation are more accurately assessed in the presence of a combined comparative laboratory experiment conducted under controlled conditions. Otherwise, lake environmental factors are simply reported in a correlative context to degradation rate. Degradation of rotenone (and the priority transformation product rotenolone) are normally modeled using a pseudo first order degradation kinetics (Gilderhus et al. 1986, 1988; Dawson et al. 1991; CDFG 1999b; Finlayson et al. 2001, 2014; Boesten et al.). In addition to this approach, Rohan et al. (2015) applied a gamma process model to several previously performed rotenone degradation studies and concluded that the gamma process model is preferential to the pseudo first order kinetics model citing the advantages of a better fit to the data, the ability to incorporate more covariates, and the ability to determine a confidence interval for the DT50 and a time to complete dissipation (Rohan et al. 2015).

The present study provides a combined field and laboratory assessment of rotenone degradation under the high latitude conditions of Subarctic Alaska. The aims of the present work were to 1) assess the attenuation of rotenone and rotenolone during the 2018 treatment of eight lakes on Alaska’s Kenai Peninsula, 2) compare two attenuation models, pseudo first order and gamma process, to describe the rate at which rotenone and rotenolone degrade in the environment, and 3) determine the relative degradation rates of rotenone in lake water under controlled laboratory conditions simulating a factorial of wintertime (4 °C) light/dark and sterile/non-sterile environments. This information is relevant to future predictions of when Alaskan lakes detoxify and is a first step in untangling the complexities of environmental factors that drive rotenone fate in this climate.

2.0. Materials and Methods

2.1. Reagents and Materials

Methanol (MeOH) and LC/MS grade water were obtained from Fisher Scientific (Hampton, NH, USA). Rotenone (98%) standard was purchased from Sigma Aldrich (St. Louis, MO, USA). Rotenolone (12αβ-hydroxyrotenone) was provided gratis by the California Department of Fish and Game (CDFG, Sacramento, California). The rotenone formulation used for laboratory validation was CFT Legumine 5%® w/w and was obtained from Central Life Sciences (Schaumburg, IL, USA). Amber bottles (1 L) were obtained from Fisher Scientific (Hampton, NH, USA). Luer lock syringes (3 mL), PTFE luer lock syringe filters (0.45 µm) and KINMAX® borosilicate glass disposable culture tubes were acquired from VWR (Randor, PA, USA). Amber borosilicate glass autosampler vials (2 mL) and caps were purchased from Fisher Scientific (Hampton, NH, USA).

Fluorescent light fixtures (48 inches, T8, model number 82049–4–6) manufactured by Quorum International (Fortworth, TX, USA) were purchased from the Lighting Gallery (Anchorage, AK, USA). T8 black light blue bulbs (36 W, 4 ft), model number F36T8/BLB were purchased from Alaska Lighting Supply (Anchorage, AK, USA). Sample incubation containers were Wheaton W216905 Clear Glass 8oz straight sided jars, with 70–400 White Polypropylene Poly-Vinyl Lined Screw Cap (Millville, NJ, USA). 2-nitrobenzaldehyde (2-NB) was purchased from VWR (Randor, PA, USA).

2.2. Site Characterization and Sample Collection

Between 8 October and 11 October, 2018, a series of eight interconnected lakes near Tote Road in Soldotna, Alaska were treated with CFT Legumine® (Figure 1). The Alaska Department of Fish and Game (ADFG) applied the CFT Legumine® (50ppb) by spraying the product onto the water surface and using the turbulence from their outboard motor to mix it into the water column. Ruth and Jennifer lakes, a pair of interconnected lakes independent from the treatment lakes, were selected as control lakes for this experiment. The physicochemical characteristics (temperature, pH, dissolved oxygen concentration (DO), specific conductance (SpC), visibility, depth) of the treatment sites were monitored monthly by ADFG using a HYDROLAB® Quanta Multi-Probe Meter for one year prior to the treatment (Table S1).

Figure 1.

The Tote Road northern pike lakes restoration sites in red with unofficial lake names, (ADFG, 2018).

A composite grab sample was collected from three locations at a depth of 0.5 m near the center of each lake prior to rotenone treatment. Groundwater samples were also collected from an external faucet at a residence adjacent to the lake; water was run for five minutes prior to the collection of fresh groundwater.

Following treatment, lake water samples were collected at a depth of 0.5 m from three locations around the center of the lake either as a grab sample or with a Kemmerer sampler. For each lake, a single composite sample was created. A “deep” sample was also collected for all lakes with a depth > 6 m using a Kemmerer sampler bottle. The sampler was lowered to a depth within two meters from the bottom in three locations around the center of the lake. The collected water was then combined into a single 1 L composite sample. Post-treatment groundwater samples were collected identically as with pre-treatment methods. All samples were collected into sterilized 1 L amber glass bottles, immediately chilled on ice, stored out of the light and transported to the Applied Science, Engineering and Technology (ASET) Lab located at the University of Alaska Anchorage (UAA) within 24 hours of collection. Upon arrival to the lab all samples were stored in the dark at 4 °C and analyzed via high performance liquid chromatography tandem mass spectrometry (HPLC-MS/MS, Supporting Information Section S1) within seven days of collection. Water sterilization was achieved with gamma irradiation (25 kGy dosage) at the Oregon State University Radiation Center.

2.3. Microcosm study

To isolate the relative effects of abiotic (sunlight) and biotic (microbial) degradation, a microcosm study was conducted under simulated “winter” conditions. Inside R.W Smith & CO. controlled environmental room, temperature controlled to 4 °C, a single fluorescent light fixture equipped with four blacklight blue bulbs was suspended 29 cm above a shelf. Black plastic sheeting was used to encase the shelf and block out any incident light from reaching the samples. Spectra for the black light bulb was determined using an AvaSpec-2048L-USB2 fiber optic spectrometer (Figure S1).

For all samples, a 1 ppm CFT Legumine® (50 ppb rotenone) solution in lake water (150 mL) from Ruth lake was placed in Wheaton glass jars (8 oz). Light-exposed samples (n=3) were covered with a single sheet of saran wrap, which effectively transmits light above 270nm (Shafaei et al. 2017). Dark samples (n=3) were wrapped tightly with aluminum foil to ensure no light exposure. Non-sterile samples (n=3) consisted of water collected from near the shore of Ruth lake, kept in a refrigerator at 4 °C until use. Sterile samples consisted of the same Ruth lake water, gamma sterilized in one liter amber bottles at Oregon State University and stored at room temperature.

The light chamber was separated into 15 equally spaced positions. Each sample was randomly assigned to a position; no sample was assigned to subplot C2, which was reserved to perform actinometry measurements (see supporting information, Section S2) over the duration of the exposure in order to monitor any longitudinal changes in light flux. An initial flux map of the sample chamber was established by measuring the degradation of 2-nitrobenzaldehyde (2-NB) over the 3 d prior to starting the experiment. At time points 0 and 85 d, a 1 mL aliquot was analyzed for both rotenone and rotenolone in the same manner as the field samples.

2.4. Data Analysis

2.4.1. Statistical Modeling of Rotenone and Rotenolone Degradation

Predictive modeling of rotenone attenuation is important for regulators when compiling proposals for whole-lake treatments. Accurate models will allow regulators to develop guidelines for public accessibility to dosed water bodies, when native fish will be restocked (Gilderhus et al. 1988; Dunker 2009) and when it is safe for normal recreational activities to resume (Dunker 2009; Finlayson et al. 2014). One method for determine the best model is using the Akaike information criteria (AIC):

$$AIC=2K-2\ln \left(L\right)$$ (1)

where K is the number of parameters in the model and L is the maximum likelihood estimation. The Akaike information criteria describes the relative quality of statistical models for a given set of data by acting as an estimator for the out of sample prediction error (Akaike 1974; McElreath 2016; Taddy 2019). It allows for the comparison by assigning each model a score, lower is better, that describes how well a given model predicts the data.

AIC_c is modification of AIC for use with small sample sizes. Burnham and Anderson (2002) suggest using AIC_c when the ratio of number of samples divided by the number of parameters (n/K) is less than 40 (Burnham and Anderson 2002).

$$AI{C}_c=AIC + \frac{2K^2+2K}{n-K-1}$$ (2)

The best model is determined by selecting the model with the lowest AICc

2.4.2. Pseudo first order kinetics

A pseudo first order regression was applied to the linearized percent degradation data, the rate constant (k) is determined from the slope of this regression line. AIC_c was used for comparing different models. In addition to statistical treatment of all sample sites combined, a half-life (DT50, t_1/2) was determined based on statistical treatment of each site individually. These models were evaluated by their coefficients of determination (r²) as not all sites had enough data points to generate an AIC_c score. The pseudo first order DT50 was calculated with the Equation 3:

$$t_{1/2}=\frac{ln2}{k}$$ (3)

2.4.3. Multi Parameter linear model

Covariates of pseudo-first order degradation rate, including: time, water temperature, pH, specific conductance, dissolved oxygen, and secchi depth (visibility) were assessed using multivariate analysis of variance (MANOVA) to determine factor significance at 95% confidence. Stepwise, the covariate with the highest p-value was removed, evaluating the model with MANOVA after each covariant was removed. Once only significant covariates remained, multi parameter linear regression was used to determine the coefficients of the model. An AIC_c score was calculated for this model. Analysis was performed using JMP PRO 13 software.

2.4.4. Gamma model

The gamma process was performed as described by Maheswaran Rohan et al. (2015) to model the degradation of rotenone in the environment. The single effect gamma model is represented by a gamma distribution with a mean (μ). A function of time (days post application), and described by Equations 4&5:

$$\mu =e^{\left({\beta }_0+{\beta }_{1}t\right)}$$ (4)

$$\mu =A{\left(e^{{\beta }_1}\right)}^t \text{where} A=e^{{\beta }_0} \text{is} \text{a} \text{constant}$$ (5)

The DT50 for the single parameter gamma model is represented by Equation 6:

$$t_{1/2}=\frac{\ln (\frac{1}{2})}{{\beta }_1}$$ (6)

Finally, the multi parameter gamma model is defined by Equation 7:

$$\mu =\exp \left({\sum }_0^n{\beta }_n\ast p_n\right)$$ (7)

where ꞵ_n represents the parameter estimate and p_n represents the covariant estimate.

The gamma distribution model allows for the addition of multiple covariates. For this study, the covariance of time, temperature, specific conductance, dissolved oxygen and pH were analyzed using the Wald test. Covariates were determined to be significant if they had a Chi-square value less than 0.05. Initially, a model was created using all covariates and the covariant with the highest Chi-square value was removed. This process was repeated until only significant covariates remained. An AIC_c score was calculated for this model. A multi parameter (time, temperature) gamma model was calculated for each site. Generalized Coefficient of Determination was used for evaluating these models as not all sites contained enough data points to calculate an AIC_c. All calculations for the gamma model were performed using JMP Pro 13 software

2.4.5. Principle component analysis

Multiple environmental factors have previously been postulated in the rotenone and rotenolone degradation process (Thomas 1983; USEPA (U. S. Environmental Protection Agency) 2007). Principal component analysis (PCA) was performed using JMP Pro 13 software (SAS institute, Cary, NC) (1989). The appropriate number of principal components was selected using the Kaiser criterion (Yeomans and Golder 1982).

3.0. Results

3.1. Field study

3.1.1. Physical and Chemical Characteristics of Treatment Sites

Treatment site (Section 2.2, Figure 1) physical chemical characteristics are provided in Table S1. Briefly, lake temperatures were between 5.10–7.65 °C, specific conductance was 0.016 – 0.160 mS/cm at 25 °C, dissolved oxygen concentrations were 4.98 – 9.35 mg/L, pH ranged from 6.56 – 7.38, visibility and depth were 1.0 – 5.6 m and 1 – 10 m, and volumes were 13,568 – 507,824 m³. The mean daily incident shortwave solar energy on the Kenai Peninsula varied greatly throughout the year with the brightest period 4.7 kWh/m² and a mean value (18 October through 21 February) of 1.3 kWh/m² (Thorsen, 2018). The mean pH across all sites was 7.0 and the average zooplankton-B load was 26 ug/L. (Jones et al. 2003). All lakes were monitored through 30 weeks post-treatment.

3.1.2. Assessment of Physicochemical Variation Between Sites

Three principal components (Section 2.4.5) were selected to describe the correlation between the physicochemical variation between sites. The PC1 and PC2 loadings plot revealed a grouping of physical properties (depth, volume, secchi depth, water temperature) along PC1, with chemical properties (DO, pH, SpC) correlated to PC2. This indicates the physical and chemical properties each correlate highly among themselves. The Scores plot (Figure S2) revealed little between-site clustering. This indicates a high between-site variability of physicochemical properties.

3.1.2. Pretreatment and Target Concentrations

Analysis of pretreatment water samples showed all bodies of water to be free of both rotenone and rotenolone. The target treatment concentration of rotenone in all bodies of water was 50 ppb (Section 2.2). The initial concentration in each lake was determined by analyzing the water 1 day post-treatment for both rotenone and rotenolone and then combining these two values to obtain the initial concentration of rotenone. The concentration range of rotenone was 43.9 ± 14.1 ppb in surface samples across all treated lakes (Table S3).

3.1.3. Persistence

Data were expressed as percent remaining, and normalized to 1 d post-treatment concentration values, therefore all expressions of “percent degradation” are relative to abundance at this time point. The 1 d value was selected instead of the calculated initial rotenone concentration due to the potential of incomplete mixing in the first 24 hrs and the logistics of accessibility during sample treatments preventing a 0 d measurement from occurring. Analysis of CFT Legumine® formulation revealed a significant amount (35%) of rotenone was degraded to rotenolone prior to treatment; to correct for the degradation took place during storage rotenone values from one day post treatment were used as the initial values for the degradation study.

Contrary to previous monitoring efforts, approximately 50% of initially applied rotenone was degraded within the first 8 to 11 days in each treatment site (Section 2.2, Figure 1) with the exception of G Lake deep (Figure 2A). The observed increase of the concentration of rotenone in G Lake deep samples was likely the result of incomplete mixing as it was the second largest lake by volume (348335 m³) and depth (9 m) monitored in this study (Table S1). Rotenolone concentrations peaked between 8 – 11 days post-treatment at 6 m depth sample sights and was observed to decrease over the following 8 weeks in all lakes (Figure 2B). It is reasonable that rotenolone accumulation was not observed in all sites as the overall environmental degradation of the rotenoids described by this data includes multiple pathways for rotenone elimination from lake water (eg sorption, photolysis, hydrolysis, biotic degradation, etc.) that may not produce, or may similarly eliminate, rotenolone. Deep samples were only collected from Leisure and Hope Lakes at the week 4 time point as G and Crystal Lakes were unsafe due to thin ice. No rotenone remained in lake water at any treatment site nine weeks post-treatment (60 to 63 d); however, rotenolone continued to persist for up to 150 d in all lakes and remained above the LOQ (1 ppb) 40 weeks post-treatment at 6 m depths in G and Leisure Lakes (~8 and 18% remaining, respectively). This persistence likely resulted from the onset of seasonally low temperatures and ice-coverage typical of Southcentral Alaska winters. At this point ADFG determined that all sites were safe for reintroduction of native fishes and terminated sampling. Over the entire duration of ground water monitoring (149 d) neither rotenone nor rotenolone was detected (< 1 ppb) in well water samples taken from properties adjacent to the treatment sites.

Figure 2.

Percent remaining rotenone (A) and rotenolone (B) vs time. Inserts show the weekly averages for all treatments with error bars representing standard deviation (n = 11). Surface samples represented by solid lines. Deep samples represented by dashed lines.

3.2. Half-lives

3.2.1. Pseudo First Order

A pseudo-first order degradation kinetics was assumed for both rotenone and rotenolone. Aqueous concentrations were natural log-transformed, normalized to percent degradation (from 100% initial) and regressed over time to linearize the degradation curves (Section 2.4.2). This indicated the mean ±SD pseudo-first order DT50, among the individual sites, for rotenone and rotenolone at 14.9±9.2 d and 25.8±11.3 d, with a coefficient of determination (r²) calculated at 0.708±.236 and 0.923±.026, respectively (Table S4).

When applying the transformation and regression to all sites simultaneously, the DT50 for rotenone was 14.3 (d), ranging from 11.0 (d) to 20.4 (d) at the 95% confidence interval. The coefficient of determination (r²) was 0.464. The overall pseudo first order degradation model for the percent degradation of rotenone is described by Equation 8.

$$\ln (\%\hspace{0.17em}degradation\hspace{0.17em}rotenone)=-0.37-0.049t$$ (8)

The DT50 for rotenolone when modeling all treatment sites together was 97.5 (d), ranging from 78.1 – 130.7 (d) at the 95% confidence interval (r² = 0.425). The overall pseudo first order degradation model for the percent degradation of rotenolone is described by Equation 12. Data for equations 8 and 9 is visualized in Figure 3. One lake was excluded from DT50 calculations, G Lake, which developed a 134 % increase in rotenone concentration from the day 1 value. This was presumably caused by a delay in surface water mixing. Analyzing for covariates using MANOVA resulted in time (p-value <0.0001), DO (p-value 0.047) and pH (p-value 0.0149) returning as significant for describing the degradation of rotenone.

Figure 3.

ln(Ct/C0) for Rotenone and rotenolone vs. time (days) with the first order model fitted

$$\ln (\%\hspace{0.17em}degradation\hspace{0.17em}rotenolone)=-0.48-0.007t$$ (9)

Equation 10 describes the multiparameter degradation model for the linearized percent degradation of rotenone (Section 2.4.3) which has an AIC_c score (Section 2.4.1) of 99.

$$\ln (\%\hspace{0.17em}degradation\hspace{0.17em}rotenone)=-4.440-0.05t-0.248\ast D0+0.906\ast pH$$ (10)

Analyzing for covariates using MANOVA resulted in time (p-value <0.0001) and temperature (p-value <0.0001) returning as significant for describing the degradation of rotenolone. Equation 11 describes the multiparameter degradation model for the linearized percent degradation of rotenolone, which returned an AIC_c of 118. Parameter estimates for all equations are provided in the supplemental information (Tables S5-S10).

$$\ln (\%\hspace{0.17em}degradation\hspace{0.17em}rotenolone)=-1.190-0.006t+0.107\ast temperature$$ (11)

3.2.2. Gamma Modeling

Equation 12 describes the single parameter gamma model for rotenone degradation (Section 2.4.4) which has an AIC_c score of 323.

$$\mu =19.55\ast {0.96}^t$$ (12)

With an AICc score of 323, the gamma model had a similar predictive power (AIC_c score of 339) as the model developed by Rohan (2015) from performing a meta-analysis of previous rotenone degradation studies. The predicted DT50 of rotenone is 15.1 d, 10.8 to 24.6 d at 95% confidence interval. Equation 13 describes the single parameter gamma model for rotenolone degradation which has an AIC_c score of 452.

$$\mu =11.92\ast {0.99}^t$$ (13)

The predicted DT50 of rotenolone is 80.8 d, 61.9 to 116.3 d at 95% confidence interval. Figure 4 plots equation 12 and 13 over their data points.

Figure 4.

Rotenone (A) and rotenolone (B) concentration (ppb) vs. time (days) with the gamma model fitted.

Equation 14 describes the multi parameter gamma model (Section 2.4.4) for rotenone degradation which has an AIC_c score of 323.

$$\mu =9.63\ast {0.96}^t\ast {1.07}^{temperature}$$ (14)

The predicted DT50 is 23.0 d, 11.5 to 105.6 d at 95% confidence interval. Equation 15 describes the multi parameter gamma model for rotenolone degradation which has an AIC_c score of 424.

$$\mu =33.38\ast {0.99}^t\ast {1.09}^{temperature}$$ (15)

A multi parameter gamma distribution model for each individual treatment site using time and temperature as covariates was developed. Generalized coefficient of determination (r²) was used to compare these models as not all sites had enough observations to allow for the calculation of AIC_c. Using the generalized coefficient of determination also allows for easy comparison to the linear models for each site, where the coefficient of determination is commonly used to describe the goodness of fit. The mean±SD generalized coefficient of determination was 0.876 ± 0.233 and 0.806 ± 0.102 for rotenone and rotenolone respectively. Parameter estimates for significant covariates are provided in the supplemental information (Tables S9 & S10).

3.3. Microcosm Study

Prior to the microcosm degradation experiment (Section 2.3), the relative light intensity was determined at each illumination position to construct a photon flux map over the sample chamber surface area using a chemical actinometer (2-NB, see supporting information Section S2). The degradation of 2-NB was monitored over three days relative to a reference position C2, the position located in the center of the chamber. The average photon flux across all illumination positions, relative to the reference position (C2), ranged from 75%±6% to 102%±4%. Additional actinometry was performed under sunlight at noon on October 14th, 2019, the relative rate of 2-NB degradation in the chamber to that measured under natural sunlight was 0.0363.

Over the 85 day exposure period, light output at the actinometer reference position increased to 113.47% of the day one values. It was assumed that light output changes over time were consistent across the entire exposure chamber. The relative percent flux calculations were established by normalizing using location C2 as a reference position.

Rotenone concentrations were determined at two time points, 0 and 85 d. The final concentrations were normalized to the initial concentration and corrected for positional flux (Supporting Information Section S2). The four treatment groups, from a factorial design of light or dark exposures in sterile or nonsterile water, were analyzed using the Student’s T-test to comparatively assess means between treatment groups to determine if they were significantly different at 95% confidence. This revealed that differences in all treatment groups were significant (Figure 5).

Figure 5.

Residual rotenone (% of initial application remaining) following an 85-day microcosm incubation. Data are expressed as means ± 95% confidence intervals. Pairwise comparisons (Student’s t-test) were assessed at 95% confidence.

4.0. Discussion

4.1. Rotenone and Rotenolone Persistence

Rotenone degraded to below 50% C0 (1 day post-treatment) at 2 weeks post-treatment for all sites except G Lake Deep, which was most likely due to poor initial mixing during treatment. Between 2 weeks and 4 weeks post-treatment, rotenone levels remained constant before disappearing by 9 weeks post-treatment. The residence time of rotenone, approximately 54 d, is consistent with most previous studies outside of Alaska, which ranged from 33 to 70 d (Vasquez et al. 2012; Finlayson et al. 2014; Rohan et al. 2015) and considerably faster than previous treatment on Stormy Lake in Alaska, (approximately 19 weeks (Massengill et al. 2017)). The present work exhibits the fastest degradation rate observed in Alaska, presumably stemming from unseasonably warm conditions experienced in 2018. Water temperatures were approximately 5.0 °C warmer at the time of treatment than the year prior. The average ambient temperature in Kenai Alaska in October 2018 was 6.6 °C, significantly higher than the historical average of 2.2 °C. An accelerated degradation in warmer temperatures is consistent with the findings in the literature that the rate at which rotenone degrades is correlated with water temperature (Dawson, 1991). The slowing rotenone degradation corresponded with the formation of surface ice, which did not occur until approximately October 23, 2018. The rapid initial decline of rotenone followed by a slower rate of dissipation is consistent with previously observed degradations (Finlayson et al. 2014; Rohan et al. 2015).

Rotenolone was detectable at treatment sites (G Lake deep and Leisure Lake deep) 40 weeks post-treatment and still detectable at all sites except Hope Lake 21.5 weeks post-treatment. This persistence is considerably longer than previous studies, which showed rotenolone fully dissipated between 5.5 weeks (Finlayson et al. 2014) and 10 weeks (Vasquez et al. 2012) post-treatment. This indicated that rotenolone may be more susceptible to persisting over long durations of shortened photo periods and ice cover. Rotenolone shares similar toxicity to mammals as rotenone and has been reported to be approximately an order of magnitude less toxic to aquatic vertabrates; however, no information regarding their relative chronic toxicities is available. Given its greater persistence and a lack of chronic toxicity data, rotenolone may be of ecotoxicological concern and should be carefully monitored post-treatment (Yamamoto, 1970; Finlayson, 2001). Furthermore, it may be very important to monitor hydraulic residence time to disambiguate dilution vs. degradation. All pretreatment ground water samples were below the LOQ (1ppb) for both rotenone and rotenolone at all time points, which indicates there is little potential for groundwater contamination in this area.

4.2. Microcosm Study

The microcosm data (Figure 5) suggest a half-life of 111 d under dark, cold, non-sterile conditions and 482 d for complete detoxification at <1ppb. Sunlight (and it’s inhibition by ice coverage) will presumably have a major impact, as this simulation was conducted with 24 hr daylight. The results of the actinometry based flux mapping showed that use of commercially available fluorescent light fixtures and black light bulbs provides a viable and low-cost method of performing low light intensity degradation experiments useful for representing deep lake or under-ice conditions. Variation across all subplots in the light chamber was limited to ± 25% relative standard deviation (RSD). Using Student’s T-test to compare the relative degradation (Figure 5), each treatment pairing (dark vs. light, sterile vs. non-sterile) revealed a unique residual abundance at the sample collection time. The difference between the light and dark treatments (p<0.0001) indicates that abiotic (photochemical) factors are significant. Additionally, differences between sterile and nonsterile treatment groups (p-value 0.0220 light, 0.0008 dark) suggest that biotic (microbial) degradation may have significance in the degradation of rotenone under “winter” conditions, and demonstrates that rotenone may be highly persistent following seasonal ice-in.

4.3. Principle Component Analysis

The score plot (Figure S2) identified 3 principal components sufficient to capture the variation, accumulating 96.7% (52.9, 29.1, 14.7 for PC1, PC2, PC3, respectively). The PC1 and PC2 loadings revealed a grouping of physical properties (depth, volume, surface area, water temperature) along PC1, with chemical properties (DO, pH, SpC) correlated to PC2. The lack of clustering in the scores plot indicates high inter-site variation of physical and chemical properties among the interconnected Tote Road lakes and that the factors that were analyzed for this study were insufficient to fully explain the variation in degradation rates between the treatment sites. Given the complex inter-dependencies of rotenone DT50 on a wide range of environmental factors, it is not surprising that this study did not identify a set of specific environmental conditions that indicated a strong correlation to the DT50 of rotenone and rotenolone.

4.4. Model Comparisons

The pseudo-first order single and multi-parameter linearized degradation models were found to out-perform single and multi-parameter gamma distribution models for both rotenone and rotenolone, as exhibited by lower AIC_c (109 vs. 99 vs. 323 vs. 323 for rotenone and 152 vs. 118 vs. 452 vs. 424 for rotenolone for models pseudo first order, multi parameter linear and single parameter gamma respectively). Both the pseudo first order model and the single parameter gamma distribution model had similar DT50s 14.3 d and 15.1 d respectively (Table 1). Consistent with the finding of Rohan (2015), the single parameter gamma distribution model predicted a longer DT50 for the degradation of rotenone than the pseudo first order model. Also consistent with previous results, the pseudo first order models predicted half-life fell within the 95% confidence interval of the single parameter gamma model.

Table 1.

Model AICc scores and resulting half-lives (d) for rotenone and rotenolone degradation.

Model	Rotenone		Rotenolone
	DT50 (d)	AICc	DT50 (d)	AICc
Pseudo First Order	14.3	109	97.8	152
Multi Parameter Linearized		99		118
Single Parameter Gamma	15.1	323	80.8	452
Multi Parameter Gamma		323		424

In the multi parameter linearized model for the degradation of rotenone, temperature was not a significant factor (p-value 0.69); however, DO and pH were, with a p-value of 0.0047 and 0.0149, respectively. This may have been caused by the limited time points analyzed before rotenone dissipated. The Focus Work Group on degradation kinetics recommends that no fewer than 6 time points be taken when trying to establish degradation kinetics (Boesten et al.). Due to the faster than anticipated degradation of rotenone, only 4 time points were captured before rotenone dissipated from the system. However, previous laboratory studies have shown DO and pH will significantly impact the rates of rotenone photolysis and hydrolysis, respectively (Finlayson 2001, Redman 2021). Based on these considerations, future work should incorporate additional sampling events and continued DO and pH measurements to potentially improve the capacity to model their effect on rotenone degradation in the environment. No advantage was seen using a multiparameter gamma model for the degradation of rotenone, with no factors other than time significantly explaining the degradation and an identical AIC_c of 323 for both models (Table 2).

When degradation models were established for each individual treatment site, the gamma distribution better explained the variance in the model. The coefficient of determination was 0.876 ± 0.233 and 0.708 ± 0.236 for the multi parameter gamma model and pseudo first order model respectively. The opposite was observed for rotenolone, with coefficients of determination of 0.923 ± 0.026 and 0.806 ± 0.102 for the pseudo first order and multi parameter gamma models respectively.

5.0. Conclusion

The behavior of rotenone and rotenolone in the environment is extremely complex and affected by a variety of factors. The results of our microcosm study show that light exposure (and associated ice coverage) has a major impact on degradation rate, but when temperatures are cold (4 °C) and light is absent, conditions similar to lake water under a layer of ice and snow, rotenone is more persistent and microbial degradation may have a significant contribution to the overall environmental fate. The relative contribution of microbial activity appeared to be lower once samples were exposed to light. Developing a better understanding of how microbial degradation contributes to the overall degradation of rotenone in the absence of light will help to model the way in which rotenone degrades over the long Southcentral Alaska winters.

Based on the AIC_c scores, linear transformation models outperform gamma distribution models when all sites are grouped, for both rotenone and rotenolone. Gamma distribution models still produce similar half-lives, capturing the half-lives determined by the linear transformation models within the 95% confidence interval. When each site is modeled individually, the multi variant (time and temperature) gamma distribution model more accurately describes the degradation of rotenone based upon the coefficient of determination. When modeling rotenolone degradation for each treatment site individually, the pseudo first order model produced a superior result. This result combined with no site clustering found when using PCA reinforces the complexity of modeling the degradation of rotenone and rotenolone.

The results of this study provide a baseline for regulators to develop an accurate, scalable attenuation model for predicting the environmental fate for rotenone and rotenolone in Southcentral Alaska. While alternative models such as hockey-stick, biexponential, first-order double exponential day, and first-order two compartment are frequently used for the risk assessment of pesticides and as input parameters for other fate models (e.g. GLEAMS and USFS), the kinetic fitment of linear transformation and gamma models were appropriate for rotenone and rotenolone degradation in these treatment sites. The gamma distribution model has several advantages over the use of a linearized model, including the ability to work with non-transformed data and easy incorporation of more covariates into the model. In order to improve degradation models, it is recommended that treatment sites be sampled more frequently in order to guarantee the recommended 6 time points are captured before rotenone dissipates from the system. Sediment samples should also be monitored to determine if rotenone and rotenolone are partitioning from the water before degrading. The covariates of water temperature, DO, specific conductance, pH, suspended solids, turbidity, and dissolved organic carbon should be monitored every time a water sample is taken, as these factors may influence the degradation of rotenone and rotenolone.

Highlights.

• Rotenone and rotenolone were monitored in eight lakes of the Kenai Peninsula, AK

• Rotenone degraded within 60 days, rotenolone persisted up to 280 days

• Single- and multi-parameter first order and gamma models were fit to each dataset

• Pseudo-first order model kinetics most accurately describe rotenoid degradation

• Laboratory manipulations under wintertime conditions significantly increased persistence