2.1 Sampling area
A total of 10 sections of the basin were chosen as sampling areas based on the main sources of contamination by domestic, industrial, and agricultural residual discharges, such as the Melchor Ocampo sugarcane mill, the agave crop and greenhouse area of the municipality of San Gabriel, Jalisco, and the domestic residual discharges of Autlán de Navarro and El Grullo cities (Figure 1). Each section had an extension of 5 km along with the streamflow, and the separation between each of the sections was at least 5 km; we considered that each sampling section had an independent subpopulation, based on the home range of the otters (Gallo-Reynoso 1989).
The sampling sections present different land uses associated with different productive activities, such as irrigation and temporary agriculture, urban settlements, and canyon area. The sampling sections (S) are found within different geoforms, sections where the slope is steep (canyons) and others where it is not (valleys), which is an important aspect that can determine the productive activities surrounding each study section (Table 1). The sampling sections S2, S3, S6, S9, and S10 are located in areas with irrigated or seasonal agriculture and urban settlements. The other sections (S1, S4, S5, S7, and S8) are located mainly in canyon areas with vegetation predominantly of deciduous tropical dry forest. The Trigomil dam is located approximately 11 km north of sampling section 1. This dam is the largest in the Ayuquila-Armería basin, with a maximum volume of 250 Mm3, and fragments the longitudinal continuity of the river (Meza-Rodríguez 2006).
Table 1. Morphology physical and land use of the surface water sampling sections in the Ayuquila-Armería river basin.
Sampling sections
|
Sub-basin
|
Location in basin
|
Geoform type
|
Stream order (Strahler)
|
Land use
|
Corcovado (S1)
|
Ayuquila
|
Middle
|
Canyon
|
7
|
DTDF
|
El Chacalito (S2)
|
Ayuquila
|
Middle
|
Valley
|
7
|
Sugarcane
|
El Aguacate (S3)
|
Ayuquila
|
Middle
|
Valley
|
7
|
Sugarcane
|
Ventanas (S4)
|
Ayuquila
|
Middle
|
Canyon
|
7
|
DTDF
|
Zenzontla (S5)
|
Ayuquila
|
Middle
|
Canyon
|
7
|
DTDF
|
San Miguel (S6)
|
Tuxcacuesco
|
Middle
|
Valley
|
8
|
Sugarcane
|
La Croix (S7)
|
Tuxcacuesco
|
Middle
|
Canyon
|
6
|
Agave and greenhouses
|
Zacualpan (S8)
|
Armería
|
Lower
|
Canyon
|
7
|
DTDF
|
El Chical (S9)
|
Armería
|
Lower
|
Valley
|
7
|
Sugarcane
|
Armería (S10)
|
Armería
|
Lower
|
Valley
|
8
|
Sugarcane
|
DTDF = Deciduous tropical dry forest.
The pesticides under study were selected because they are commonly used in agricultural activities and marketed within the Ayuquila-Armería basin (Rodríguez-Aguilar et al. 2019). The selected pesticides were 2,4-D, acetochlor, ametrine, atrazine, carbendazim, carbofuran, diazinon, dimethoate, emamectin, glyphosate, imazalil, λ-cyhalothrin, malathion, methomyl, metoxuron, molinate, parathion, picloram, pyraclostrobin, and thiabendazole. These pesticides are used to control different types of organisms and they are representative of different chemical classes such as fungicides, insecticides, and herbicides. Due to the different productive activities and crop heterogeneity in the basin, the evaluation of the temporal and spatial distribution of pesticides should be accomplished considering pesticides with different physicochemical characteristics.
2.2 Sampling collection
In each study section, four samplings over two years were carried out. Two samplings during the dry season (February 2018 and 2019) and two during the wet season (November 2018 and 2019). The neotropical otter’s feces were identified based on a methodology published by Aranda-Sánchez (2012). The sampling sections were covered entirely on foot, looking for feces in the rocks and logs inside the streamflow, as well as in areas surrounding it (10 m on each side of the river). The samples were identified and taken following the procedure described by Elliot et al. (2008), collecting them manually with latex gloves. To avoid feces samples contamination during the fieldwork, immediately after collection, they were placed in polyethylene bags with hermetic closures.
During transport and fieldwork, the samples were placed in an icebox, keeping them at 4 ºC to increase preservation time and decrease microbial activity. At the laboratory, the samples were dried in an oven at 40 ºC for 5 h and subsequently screened in sieves cascaded to a 2 mm opening, separating the bone material, fish scales, and exoskeletons, to obtain the finest material and facilitate the pretreatment process and extraction of pesticides from samples. Finally, the sieved material was stored in dark at -20 ºC until treatment.
Composite samples of each sampling section were prepared by combining collected feces on the same date and sampling section. Thus, four composite samples from each sampling section were obtained. So, the sampling sections were treated as the primary sampling units. The volume of the composite samples varied according to the number of feces collected in each sampling section.
2.3 Sample pretreatment, solid-phase extraction, and pesticides determination.
Pretreatment of feces of neotropical otter samples consisted of leaching of pesticides from samples into an aqueous phase. To leach the analytes in an aqueous phase, 1 g of each composite sample was placed in a 15 ml centrifuge tube. Subsequently, a mixture of acetonitrile and ultrapure water (1:9, vol: vol) was added to the centrifuge tube and placed in an ultrasound bath for 15 min, all solvents were of analytical grade and obtained from Sigma-Aldrich®. After that time, the tube was immediately centrifuged for 15 min at 25 ° C and 3500 rpm, in a LaboGene® model 1580R centrifuge. Finally, 5 mL of aqueous phase supernatant was diluted to 50 mL with ultrapure water for SPE extraction.
Sample clean-up and analytes concentration/extraction were carried out using Supel-select HLB 500 mg/12 mL solid-phase extraction (SPE) cartridges obtained from Supelco. Before pesticide extraction, the cartridges were conditioned passing 5 mL of MeOH and 10 mL of ultrapure water through them. The 50 mL leachate of the sample was passed through the cartridge at a 2 mL min-1 flow rate. After that, the cartridge was washed with 10 mL of ultrapure water and vacuum drying for 5 minutes. The elution of the analytes was carried out using 5 mL of MeOH. Finally, the sample was evaporated to 1 mL at 30 ºC under a gentle N2 flow. The final extract was placed in a 2 mL vial for pesticide determination using liquid chromatography equipment coupled to a tandem mass spectrometry detector.
Quantification of the 20 pesticides in feces of neotropical otter samples was carried out utilizing a column Zorbax Eclipse XDB-C18 Rapid Resolution 2.1 mm internal diameter x 50 mm long, 3.5 mm particle size connected to a liquid chromatograph model 1200 coupled to tandem mass spectrometer detector model 6430B, all from Agilent Technologies. The software used for data acquisition was MassHunter Workstation Acquisition and MassHunter Workstation Quantitative Analysis. For this purpose, two analytical methods were used, one for 2,4-D and glyphosate, and the other for acetochlor, ametrine, atrazine, carbendazim, carbofuran, diazinon, dimethoate, emamectin, imazalil, λ-cyhalothrin, malathion, methomyl, metoxuron, molinate, parathion, picloram, pyraclostrobin, and thiabendazole. The separation and detection conditions of the two analytical methods used are described in Sierra-Díaz et al. (2019).
2.4 Statistical analysis
Statistical analysis was performed using a Kruskal-Wallis test to assess whether the number of pesticides detected per sample differs by season. Also, an ANOVA test to define whether the number of pesticides detected per sample differs by sampling was addressed. The assumption for normality was determined using the Shapiro–Wilk test.
Statistical analysis was performed to estimate significant differences in the concentrations of each pesticide by season and sampling. For pesticides with detection frequency between 20% and 99% in the total samples, a statistical analysis through left-censored data with non-parametric Kaplan-Meier test and parametric with Maximum likelihood estimation was carried out, using the software R project version 3.6.2, with survival version 3.1-8 and NADA version 1.6-1 package (Therneau 2019). This statistical analysis allows working with data that are below the quantification limit, giving greater robustness to the results obtained; instead of truncating or replacing the results below of the quantification limit, either by zero or a division of the limit, causing biased results and raising the probability of making errors at the conclusions and decision-making. For pesticides with detection frequencies less than 20%, according to the theory of censored data analysis, their analysis with statistical tests is not possible, and only the range, mean, and quantification frequency were reported (Hewett and Ganser 2007; Helsel 2012; Fox 2015).
For the pesticides with a detection frequency of 100% a parametric statistical analysis ANOVA, and non-parametric Kruskal-Walis tests were performed to evaluate significant differences in the concentration of the pesticides by season and sampling. The significantly non-normal data (p <0.05) were normalized with the lognormal distribution. The pesticide concentrations that did not present a normal distribution were transformed with a natural logarithm or analyzed through a non-parametric test.
Statistical analysis as ANOVA, Kruskal-Wallis, and left-censored data with the non-parametric Kaplan-Meier (K-M) test and parametric with Maximum Likelihood Estimation (CENMLE; function “cenmle” from “NADA” package) was carried out, to determine if the spatial variables (location in the basin, geoform type, and stream order) referring to the main sources of contamination and the physical conditions of the sampling sections influence the pesticides concentrations. In analyzes censored data, the percentage of detection of pesticide concentrations determines which type of statistical test presents the best fit for data analysis. In this sense, it is recommended that pesticides with detection percentages <50% are analyzed through the non-parametric K-M test, while pesticides with a percentage >50% are analyzed with the parametric MLE test (Hewett and Ganser 2007; Helsel 2012).