The adsorption and desorption isotherms were fitted using the transformed Freundlich isotherm equation: log Cs = log K_f + 1 /n log Ce, where “Cs” is the concentration of glyphosate adsorbed in the soil (mg kg⁻¹), K_f is the equilibrium constant of Freundlich reflecting the binding affinity of the soil for the herbicide, ‘Ce” is the concentration of glyphosate in the solution (mg L⁻¹) at equilibrium, and 1/n is the degree of linearity of the isotherm. The K (intercept) and 1/n (slope) were calculated using regression analysis in the adsorption and desorption isotherms.

The field dissipation data of glyphosate in soils were fitted into the first-order kinetics model C_t = C₀ e^-kt where ‘C₀” is the initial concentration (µg kg⁻¹ soil) of the herbicide in the soil, “C_t” is the herbicide concentration (µg kg⁻¹ soil) detected in the soil at time t, and k is the first-order rate constant. Computer Assisted Kinetic Evaluation (CAKE) (version 3.3; Tessella, Newton, MA) software was utilized to model the dissipation kinetics and estimate the first-order dissipation parameters.

PROC GLM statement in SAS (version 9.4; SAS Institute, Cary, NC) software was used for the analysis of variance (ANOVA), and mean separation was achieved using Tukey’s HSD test at p ≤ 0.05.

For the range of herbicide concentrations (5–100 mg L⁻¹) and soils, adsorption data from the current experiment were well-fitted into the Freundlich isotherm model (R² > 0.91). Figure 2 demonstrates the adsorption isotherm plots, and Table 2 shows the calculated Freundlich adsorption coefficient (K_ads), slope (1/n_ads), and the corresponding goodness of fit (R²) for glyphosate adsorption in various soil types tested. Among the different soils, the slope (1/n_ads) values ranged from 0.27 to 0.43, and the Freundlich adsorption coefficient (K_ads) values ranged from 14.3 to 30.9. A lower K_ads value indicates a low adsorption affinity of the soil to the herbicide. The soil from Lake Alfred location showed a significantly higher (p ≤ 0.05) adsorption coefficient than the soils from Immokalee and Fort Pierce. The adsorption coefficients for Immokalee and Fort Pierce soil were found to be similar (Table 2).

Desorption isotherms for glyphosate in all the soils also fitted into the Freundlich model (R² > 0.84) (Table 2). Freundlich desorption coefficient (K_des) values of glyphosate were considerably lower in the Immokalee and Fort Pierce soils than the Lake Alfred soil. A lower K_des value indicates lesser retention of herbicide on the soil surface, and consequently, higher release into the soil solution.

The adsorption parameter, K_ads, observed in the current study, agrees with the previous observation from similar soil types, where glyphosate had a K_ads of 34.5 in sandy loam soil (Al-Rajab et al., 2008). The results obtained from the adsorption-desorption study are in the lower range of K_ads and 1/n values obtained for glyphosate from different soil types and crop production systems. There is a large variation in the literature values, with K_ads ranging from 0.6 to 5.0 x 10⁵ and 1/n ranging from 0.26 to 1.26 (Vereecken, 2005). This variability in glyphosate adsorption is attributed to soils’ physical and chemical heterogeneity (Al-Rajab et al., 2008). The lower adsorption values of glyphosate obtained from the present study indicate lower soil adsorption affinity of this herbicide in all three citrus-producing locations. Clay and organic matter content of soils tend to affect glyphosate adsorption positively (Vereecken, 2005). The citrus growing soils in Florida are predominantly sandy (>90% sand) and low in organic matter (Obreza and Collins, 2008) and could be a plausible reason behind relatively lower adsorption and higher desorption of glyphosate in these soil types. Soil organic matter (SOM) seems to play a questionable role in soil adsorption of glyphosate. On the one hand, investigations have shown that soil sorption of glyphosate is positively correlated with SOM content (Piccolo et al., 1996). However, there are reports that it sometimes negatively impacts glyphosate adsorption to soil (Gimsing et al., 2004a).

Also, it has to be noted that adsorption was comparatively higher in soil from Lake Alfred compared to the Immokalee and Fort Pierce soils. Lake Alfred soil has more silt and clay, and phosphorus (P) than the other two soils tested in this study (Table 1). Although most herbicides, due to their non-polar functional groups, are adsorbed in soils primarily by SOM, glyphosate, a small molecule with three polar functional groups (carboxyl, amino and phosphonate groups), is strongly adsorbed by soil minerals (Borgaard and Elberling, 2007). Hence, clay minerals like iron (Fe) and aluminum (Al) oxides in the soil can provide binding or adsorption sites to the glyphosate (Gimsing and Borggaard 2002; Kanissery et al., 2015). Likewise, sand particles coated with Fe and Al sesquioxides (also known as ‘coated sands’) have been found to improve P adsorption and retention in Florida soil (Obreza and Collins, 2008). It is well known that glyphosate and phosphate are structurally analogous and share similar adsorption mechanisms in soil (Gerritse et al., 1996; Gimsing and Borggaard 2001; Wang et al., 2005). The active phosphonate end group in the glyphosate molecule allows it to form inner-sphere complexes on Fe and Al oxide surfaces (Gimsing and Borggaard 2002; Sheals et al., 2002). Hence, significantly higher amounts of phosphorus observed in the Lake Alfred soil (Table 1) could be an indication of better availability of adsorption sites in that soil, owing to the presence of coated sands, and could be a probable reason behind the higher adsorption of glyphosate in this particular soil.

The value of slope (1/n) < 0.5 in the adsorption model (Table 2) indicates a strong non-linear relationship between the amount of herbicide adsorbed onto the soil surface and its concentration in the aqueous solution. The result indicates that less glyphosate will be adsorbed in the soil as the glyphosate concentration in the solution increases, possibly because most of the adsorption sites are bound to glyphosate, and fewer sites are available for adsorption. This observation suggests a high potential for glyphosate leaching when higher concentrations of this herbicide are present in the soil profile.

Soils with a lower adsorption affinity showed greater desorption, as evident by the K_des value (Table 2). Higher desorption or release of glyphosate from Immokalee and Fort Pierce soils suggests that the herbicide’s mobility for leaching losses or its availability for biodegradation and plant uptake in these soils will be higher. The relatively lower desorption in the Lake Alfred soil suggests that glyphosate will be less mobile and more bound to this soil than the other two citrus production soils tested. Generally, glyphosate is non-bioavailable and inactive when bound or adsorbed to soils (Sorensen et al., 2006).

The dissipation parameters of glyphosate obtained from the field study conducted in Immokalee, FL, are presented in Table 3. The dissipation of glyphosate from the surface soil (top 15 cm) over time (Figure 3) adequately fitted into a first-order kinetic model with an R² value of 0.88. The first-order rate constant (k) value was 0.026 days⁻¹. The dissipation half-life (DT₅₀: time for dissipating 50% of initially applied herbicide from the soil) of glyphosate from the surface soil was 25.9 days. The time for dissipating 90% of initial herbicide (DT₉₀) was 86 days. The cumulative amount of precipitation received in the experimental site during the dissipation study duration is shown in the supplemental materials (Supplementary Figure S2).

The half-life of glyphosate applied in the field (∼26 days) was well within the range of half-lives in similar soil types reported previously (Okada et al., 2019); between 9 and 38 days. Glyphosate DT₅₀ can vary from a few days to months and even years, depending upon several factors, including soil type, texture, organic matter content and other soil environmental conditions (Sprankle et al., 1975; Nomura and Hilton 1977; Carlisle and Trevors 1988). The glyphosate dissipation was well-fitted and described using first-order kinetics, showing rapid initial dissipation of glyphosate followed by a decreased glyphosate dissipation rate which are in agreement with previously conducted research (Mamy et al., 2005; Bergström et al., 2011; Ghafoor et al., 2011). Interestingly, the detected concentration of glyphosate peaked at 7 days after application and then began to decline. A similar observation was made by (Salazar López et al., 2016), where glyphosate concentrations reached a maximum in soil 10–18 days after application. Although the bulk of the applied glyphosate quickly dissipated from the soil surface in the current study, traces of herbicide could remain in the top 15 cm layer almost 3 months after the application, as noted from a prolonged DT₉₀. Researchers have reported a DT₉₀ of over 7 months for glyphosate in certain cases, particularly when the soils are rich in silt and clay content (Okada et al., 2019).

Dissipation of herbicides from the soil can occur through microbially or chemically mediated transformations, surface or subsurface runoff, and leaching. Physico-chemical processes like adsorption and leaching in the soil and microbial degradation are considered primary determinant factors for the glyphosate dissipation from the applied area (Laitinen, 2009). The relatively lower adsorption affinity of glyphosate to sandy soils in citrus orchards in Florida observed from the current study could be potentially implied as a reason behind the quick dissipation of glyphosate, owing to its enhanced mobility from the soil surface.

Figure 4 shows the vertical leaching of glyphosate at different depths in the soil columns. Glyphosate was detected up to 40 cm below the surface in the soil profile columns when they were opened and analyzed 40 days after the herbicide application. The primary degradation product of glyphosate, AMPA, was also detected up to the depth of 20–30 cm below the soil surface. About 35% of applied glyphosate was retained in the top 0–10 cm, and 28% of applied herbicide was found accumulated in the 10–20 cm soil layer, and ∼81% of the initially applied glyphosate was detected within a soil depth of 40 cm in the column. Glyphosate was not detected from the samples beyond 40 cm deep from the herbicide-treated columns or any depth samples in untreated control columns. Analysis of leachate for herbicide content was not performed as only an insignificant amount of leachate was collected at the column’s bottom cap. Also, leachate was observed to be turbid and brownish resulting from soil colloids.

Glyphosate has moved vertically up to 40 cm in the soil, as observed from soil columns. A similar trend was observed by Laitinen. (2009), who found glyphosate’s movement up to 35 cm in a soil profile, 38 days after the herbicide application. The loss of mass balance of glyphosate in the soil columns (recovery from various depths accounts for 81% of the initially applied amount) in the current study could be attributed to the herbicide’s microbial transformation. The primary metabolite AMPA was detected at various soil depths in the columns (Figure 4). Most citrus-producing soils in Florida, including the Immokalee soil used in this column study, have low soil organic matter (SOM) (Table 1). As SOM is closely linked to soil microorganisms, the potential for microbial transformation of herbicides is expected to be relatively low in these soils. However, the detection of AMPA in our soil columns, the most common product of glyphosate’s microbial metabolism (Dick and Quinn 1995), indicates the presence of microbial transformation as a contributing mechanism for glyphosate dissipation and loss from the citrus orchards.

Only 35% of the applied glyphosate stayed in the top 10 cm in the column, followed by relatively lower concentrations in the subsequent soil depths. The vertical movement of glyphosate through the soil profile could be associated with its hydrophilicity and high-water solubility. The soil type utilized for this column study (Immokalee soil) is notably very sandy (>90% sand) with low clay content characterized by lower herbicide retention. Therefore, this may have promoted glyphosate leaching from the surface soils, especially when coupled with simulated citrus tree irrigation in the soil columns. On the other hand, the soils from this location have poor drainage properties due to a subsurface organic hardpan (Obreza and Collins 2008) that possibly restricted the downward movement of the herbicide (>40 cm) into deeper soil layers. Comparable reports on relatively low leaching and mobility of glyphosate have been reported from a previous study conducted in sandy soils. Researchers attribute this to the strong adsorption of glyphosate to soils. Furthermore, degradation of glyphosate to its major metabolite, AMPA, was also observed in the studied soil (Zhou et al., 2010). High rates of glyphosate leaching will be expected in tropical sandy soils like citrus production soils in FL due to their low adsorption capacity and relatively high precipitation. However, field investigations and column studies have shown that physical properties, like porosity, compaction etc., greatly determine the mobility of glyphosate in these soils. For instance, Kjær et al., 2005 showed minimal leaching of glyphosate (or AMPA) in sandy soils and is attributed to the absence of macropores in these soils. Sandy soils in the tropics offer a wide range of such properties (Bruand et al., 2005) and must be considered while predicting the leaching and movement of glyphosate in such soils. The fact that the citrus tree root zone in these types of soils generally occurs within 0–45 cm (Leiva et al., 2015) entails further investigations into the retention of this herbicide in the citrus root zone for an extended period, thus enabling the uptake of its residues by roots (Cornish and Burgin, 2005; Tong et al., 2017).