Representative soil samples were collected at three different locations around each fruit tree (avocado and lemon). A total of 18 sampling sites for soil samples were considered, 10 sites within 200 m and 8 sites within 400 m away from TD25. This was achieved by digging into the ground at depths ranging from 30 to 40 cm and the collected soil put into sampling bags. Avocado root systems may extend from 30 to 60 cm, while lemon fibrous root system is concentrated within 30 to 90 cm [19] [21] . The sampling bags were labelled according to their sampling points for easier identification. Two main sets of samples were collected, those around 200 and 400 m away from TD25. Soil samples were also collected on the suspected point source (TD25) at 5 different points (A, B, C, D and E) to act as the control (Figure 1).

The collected soil samples were sieved through standard soil sieves up to 250 µm in order to obtain a homogenous sample. Two grams of each soil sample was put in beakers into which 3 - 4 drops of hydrochloric acid (HCl), [35% purity supplied by LABCHEM] added in order to break-down the silicate in the soil. For decomposition to take place, 30 ml of nitric acid (HNO₃) [purity 69% - 70% supplied by Himedia] was added to each sample, and then heated on a hot plate for 15 minutes. The solution was then diluted with distilled water up to 100 ml using 100 ml volumetric flasks, shaken thoroughly and filtered using Whatman filter paper (grade 91 qualitative filter paper wet-strengthened, 100 circles, 12.5 cm).

The treated soil samples were analyzed for copper (Cu), cobalt (Co), iron (Fe), manganese (Mn) and lead (Pb) using the PerkinElmer Atomic absorption spectrophotometer (AAS) PinAAcle 500. Triplicate tests were conducted for each soil sample.

Fruit samples were randomly sampled from each of the avocado and lemon trees. Three ripe fruits were plucked from the opposite ends of each fruit tree. The fruit samples were put in polythene bags and labelled. Similarly, 18 sampling stations as outlined in Section 2.1 were considered for avocado and lemon fruits.

In the laboratory, the fruit samples collected, were washed with distilled water to remove any possible contamination. The samples were cut into slices using a sharp clean knife and then oven-dried at 70˚C for 72 hours. Thereafter, the dried samples were weighed and ground into powder using a pestle and mortar [24] . The powdered samples were then stored in new sampling bags in readiness for laboratory analysis.

Wet digestion method was used to analyze the pulverized fruit samples. 2.5 g of the powdered sample was obtained and put into a beaker. Thereafter, 25 ml of nitric acid was added and the mixture was left for 24 hours for digestion to take place. Thereafter, 15 ml of perchloric acid was added to each mixture earlier prepared and heated until white fumes of perchloric acid appeared. 50 ml of distilled water was then added to each beaker, heated for at least 20 minutes and left to cool. The solution was mixed thoroughly and diluted with distilled water up to 100 ml per sample using a 100 ml volumetric flask and then filtered through Whatman filter paper (grade 91 qualitative filter paper wet-strengthened, 100 circles, 12.5 cm).

The analysis of heavy metal content was determined using the PerkinElmer Atomic absorption spectrophotometer (AAS) PinAAcle 500. Triplicate tests were conducted for each sample.

The estimated daily intake of heavy metals depends on the concentration of heavy metals in the fruit (avocado and lemon) being consumed and the average body weight of the population [25] . The average body weight of 60 kg for adults and 15 kg for children are assumed for most populations globally, for dietary exposure purposes. However, the average body weight for adults may differ for other regions significantly from 60 kg. Estimations of food consumption can be determined through individual and household surveys or approximated through statistical national food balance sheets [26] .

The daily intake of heavy metals through the consumption of fruits was calculated using Equation (1) by [25] ;

EDI = C metal × W food B weight (1)

where,

C_metal is the heavy metal concentration in the food sample (mg∙g⁻¹),

W_food is the daily average weight of the consumed food sample (g∙day⁻¹), and

B_weight is the average body weight (kg) with an assumption that the food samples are being consumed by both adults and children.

Multivariate analysis of variance (MANOVA) was used to analyze the data on the heavy metal content (Co, Cu, Fe, Mn, and Pb) amongst multiple groups. This method takes into account the intercorrelations among the dependent variables in the analysis, to provide a better protection against the inflation of the probability of type I error and has a greater sensitivity for detecting differences [27] .

MANOVA tests were then followed by a series of separate ANOVAs, using the Bonferroni-adjusted significance level of α’ = α/g (where g was the number of heavy metals considered in the analysis), to identify whether group differences were present for that variable. Post hoc Tukey tests were conducted as follow-up procedures after significant univariate ANOVA tests with an adjusted significance level of α’ = α/g. This procedure is recommended because it provides a strict control of experiment-wise type I error rate for all possible pairwise comparisons [27] .

When assessing whether heavy metal contents in fruit samples were within the recommended permissible safe limits by the World Health Organization (WHO) and the Food and Agriculture Organization (FAO), Hotelling’s T² test [28] was used. Similarly, the same technique was used for testing the multivariate null hypothesis that the population mean vector of heavy metal content was equal to the mean vector of FAO/WHO permissible limits. Follow-up analyses consisted of constructing 95% simultaneous Bonferroni confidence intervals [28] for all heavy metals.

Preliminary analyses were conducted to assess the validity of the MANOVA and ANOVA assumptions to check for any univariate or multivariate outlier. Univariate outliers were identified as observations with z-scores greater than 3 or less than −3, while multivariate outliers were recognized by their large Mahalanobis values, that is, multivariate outliers’ distances exceeding x² critical value with g (number of dependent variables) degrees of freedom at the 0.001 significance level [27] . In order to check if the use of MANOVA was justified with the data at hand, bivariate correlations were examined and the Bartlett’s test of sphericity was conducted.

For all analyses conducted, neither univariate nor multivariate outliers were detected as all z-scores in absolute values were less than 3, and all Mahalanobis distances were less than the x² critical value with g (number of dependent variables) degrees of freedom at the 0.001 significance level. In addition, the Bartlett’s test of sphericity was significant at the 5% significance level (all p-values < 0.05), suggesting that there were sufficient correlations between dependent variables to justify the use of MANOVA.

All analyses were conducted using the Statistical Package for Social Sciences (SPSS) version 20 with a significance level of α = 0.05.

The analyses in this section were to investigate the heavy metal contents in TD25 and fruit tree soil samples and to check whether the distance from TD25 had any effect on the concentrations of heavy metals in soils. The results of the analysis of the TD25 soil samples at the five sampling points suggest that the concentration levels of heavy metals were significantly different (MANOVA and ANOVA p-values < 0.05; Tukey p-values < 0.01), while the comparison of the heavy metal content for avocado and lemon soil samples at both 200 and 400 m away from TD25 showed that group differences were not statistically significant (all p-values > 0.05).

Concerning the comparison of the soil samples at 200 and 400 m, the results show that the concentration levels of heavy metals in (lemon and avocado) soil samples within 200 and 400 m were not the same (MANOVA p-values = 0.00). A further investigation using the t-tests and the 95% t-confidence intervals (not reported) found that the mean concentrations of Co and Pb in lemon soil samples at 200 m were less than those at 400 m. For avocado soil samples, the mean concentration of Pb at 200 m was lower than that at 400 m, while Mn concentration was higher at 200 m than at 400 m. [24] obtained similar results for soil and mango fruit samples around a mine tailings dam, though the influence of distance on heavy metals from suspected point source was not investigated.

A comparison of heavy metal content in soil samples with that in fruit samples was made to ascertain the level of abstraction of the heavy metals from the soil through the roots of the trees up to the fruits. At both distances of 200 and 400 m away from TD25, significant group differences of the concentration of heavy metals in soil and fruit samples were found (all p-values < 0.05). Following significant MANOVA tests, post hoc analyses, based on the 95% confidence intervals (CI) of the mean differences of heavy metal content in soil and fruit samples were done (see Table 1). In Table 1, the mean differences, and lower and upper bounds of the confidence intervals were all positive indicating that heavy metal contents in (lemon and avocado) soil samples were very high as compared to those in (lemon and avocado) fruits at both locations. [29] investigated Cd, As, Cr and Pb in mangoes and found all the four heavy metals below the acceptable limits. [9] found that levels of concentration for Cd, Ni and Pb in plant samples

were higher than in soil samples. Similarly, [10] found that out of the 7 considered heavy metals in the soil, only Zn was below the WHO permissible limits. The results in the current study could suggest that the levels of abstraction of heavy metals from the soil up to the fruits by avocado and lemon trees were very low (see Table 4 and Table 5).

In this section, heavy metal contents in lemon and avocado fruits were investigated. The MANOVA tests for comparing heavy metals in avocado with those in lemon fruits at both 200 and 400 m were significant (all p-values < 0.05).

Post hoc analyses based on the t-confidence intervals with 95% confidence level (see Table 2) indicated that the concentration levels of Fe, Cu, and Mn in lemon fruits were lower than those in avocado fruits at 200 m away from TD25 as suggested by negative mean differences, lower and upper bounds of the confidence intervals at 200 m. For fruits at 400 m, the mean concentration of Fe in lemon was higher than that in avocado, while the concentration of Cu was lower in lemon as compared with that in avocado. No significant difference was found between the concentration of Mn in lemon and in avocado. Variations in heavy metal ion uptake or translocation from plant root system upwards to the fruits have been reported by [24] (soil and mango); [22] (4 varieties of mangoes); [23] (fruits and vegetables).

Confidence intervals for the mean differences of heavy metal contents for fruits at 200 m and those at 400 m (not reported) suggested that, apart from Fe for both lemon and avocado fruits, the distance from TD25 did not have any effect on the concentration levels for other heavy metals.

The Hotelling’s T² test was significant for all analyses (all p-values = 0.000), indicating that the mean vectors of concentration levels in soil and fruits at both locations were highly statistically different from the WHO/FAO permissible limits. In addition, the 95% Bonferroni simultaneous confidence intervals in Table 3 had lower and upper bounds less than the WHO/FAO limits for all heavy

metals, suggesting that the concentration levels of heavy metals in soil and fruit samples at 200 m (and equally at 400 m, results not shown) were well below the permissible WHO/FAO limits.

The amount of consumer exposure to heavy metals and other health related risks may be expressed in terms of estimated daily intake (EDI). Each heavy metal concentration for both fruits was computed at 200 and 400 m away from TD25 (Table 4 and Table 5) using Equation (1).

The highest concentration of heavy metals in avocado fruits (Table 4 was that for Fe at 200 m (0.53 mg∙kg⁻¹), with an estimated daily intake of 0.22 mg∙kg⁻¹, compared with a standard limit range of 12.50 - 19.60 mg∙kg⁻¹. This result (EDI

of 0.22 mg∙kg⁻¹ for Fe) was well below the allowable limit, an indication that there were no related health risks due to possible heavy metal intake from fruits grown around TD25. A similar study on assessment of EDI for 4 different heavy metals found that Cd and Pb in rice and vegetables, exceeded permissible limits [25] ; while [5] reported health risks among local children due to contamination from Cd, As and Pb. In the current study, all EDI results for avocado fruits grown around 200 and 400 m away from TD25 were below the acceptable limit.

Similarly, the highest concentration of heavy metals in lemon fruits (see Table 5) was for Fe with an estimated daily intake of 0.14 mg∙kg⁻¹, a result far below the permissible limit. The results for the EDI of heavy metals in avocado and lemon fruits grown at 200 and 400 m away from the suspect point source (TD25) were all within the permissible levels according to FAO/WHO guidelines.