Groundwater is often perceived as microbiologically safer than surface water because of natural filtration through soil and geological formations [6]. However, shallow aquifers in rapidly urbanizing areas can be vulnerable to faecal contamination where sanitation infrastructure is inadequate [7][8]. Poorly designed or poorly maintained sanitation systems can provide pathways through which microorganisms reach groundwater [9][10]. In Ghana, groundwater is an important source of domestic water, and contamination of shallow wells can therefore have important public-health implications [11][12].

Because direct detection of every potential pathogen is technically demanding and costly, microbiological water-quality assessment commonly uses indicator organisms [3][13]. Total coliforms are broad indicators of sanitary vulnerability, whereas Escherichia coli is a more specific indicator of recent faecal contamination [4][14]. Indicator organisms do not, however, demonstrate the presence of every bacterial, viral or protozoan pathogen [14][15].

International drinking-water guidelines emphasize the absence of E. coli in water intended for human consumption [4][5]. Membrane filtration is widely used for microbiological water-quality assessment because a measured volume of water can be filtered and organisms subsequently enumerated on selective or differential media [16][17]. In this study, microbiological indicators were used to assess the sanitary quality of selected hand-dug wells.

The study was conducted in Agona Swedru Township, the capital of the Agona West Municipality in Ghana’s Central Region. The Municipality lies approximately between latitudes 5˚30' and 5˚50'N and longitudes 0˚35' and 0˚55'W and covers about 447 km² [18]. The area has a wet semi-equatorial climate with bimodal rainfall [18]. The study focused on selected cemented hand-dug wells used for domestic water supply (see Figure 1).

Figure 1. Geographical map of the study area showing the sampling locations.

A repeated-sampling observational design was used [19]. Ten cemented hand-dug wells were selected purposively and sampled once per week for five consecutive weeks. Thus, the dataset comprised 50 well-week observations. The measured distances between the wells and septic tanks ranged from 3.205 to 32.488 m. The five weekly measurements for each well were averaged to obtain one well-level mean for each physicochemical and microbiological outcome before the correlation and regression analyses (see Table 1). Consequently, the effective sample size for the regression models was 10 wells, not 50 observations. The regression models were treated as exploratory because four predictors were examined with only ten independent wells.

Water samples were collected aseptically in sterile containers, transported under controlled conditions below 4˚C, and analyzed on the day of collection. A total of 50 sampling events were included in the five-week dataset. The study assessed total coliforms and E. coli using culture-based methods.

The total coliforms were cultivated on a medium that has lactose at 35˚C to 37˚C. They were tentatively determined by the production of acid and gasses due to the fermentation of the lactose. The microbiologically procedure involved the following steps [16][17]:

Before the experiment began, the whole bench was sterilized with methylated spirit.100 mL of well water sample was poured into a glass filter holder with filter paper beneath while the valve was closed. The vacuum pump was turned on to filter the water through the membrane filter that retained the bacteria, while this process was ongoing, a harlequin medium in a maccatini bottle was heated and poured into a petri dish.The filter paper was inoculated into the surface of the petri dish containing harlequin agar that is selective for the growth of specific organisms and distinguishes colonies of desired organism. The medium was allowed to solidify and incubated at 37˚C for 24 - 48 hours.After incubation, the colonies grown were counted and identified by the use of illuminated magnifier with a greenish-gold sheen with results were expressed as CFU/100 mL.

Three important controls (blanks, duplicates, and standards/reference controls) were put in place to determine whether the results were reliable and whether contamination occurred during sampling or laboratory analysis. These Quality Control measures strengthened the credibility of the laboratory results.

BLANKS: A sterile sample containing no well water was first used to detect contamination introduced from bottles, reagents, equipment, transport, or laboratory handling.

DUPLICATES: Two analytical portions from the same well sample were analyzed twice to check precision/reproducibility of sampling and laboratory analysis. Sample W1 was analyzed twice and compared the E. coli or total coliform counts.

STANDARD/REFERENCE CONTROL: A sample or organism with a known expected response was used to confirm that the analytical method, media, reagents, and incubation conditions were effective.

2.4.1. Conductivity and Total Dissolved Solids

Multi-parameter electrode was used to determine the conductivity and total concentration of the ions of the cations and the anions present in an aqueous solution. When chloride and sodium mix, cations like calcium and magnesium that gives its water sample its hardness is given away and also the anions like chlorides which combine with calcium/ magnesium or sodium to form different salts like NaCl. It is generally in micro milisiemens per centimetre (µs/cm or ms/cm).

The Total Dissolved solids contain inorganic salt, mainly calcium, magnesium, potassium, sodium, bicarbonate, chlorides, Sulphate and a small part of organic matter that were dissolved in water. TDS test is a qualitative measure of the quantity of dissolved ions and is an indicator test to establish the overall quality of water. The procedure involved [16][17][20]:

1) 50 mL of water was poured into the beaker.

2) The multi-parameter electrode probe was fully immersed in the water.

3) The readings of both conductivity and TDS were taken at the same time.

Blank: Deionized water was used to check for dissolved solids and excessive background conductivity.

Duplicate: The same well-water sample was measured twice for reproducibility. Similar readings were read indicating good precision.

Standard: A certified conductivity standard of known conductivity was used to calibrate or verify the conductivity meter.

2.4.2. Total Suspended Solids (TSS)

The spectrophotometer was used to measure total suspended solids. The spectrophotometer operates on the Beer lambert law: the amount of light absorbed or transmitted by a solution is proportional to the solution’s molar absorptivity and the concentration of the solute. In the process of determining TSS [16][17][20]:

1) Distilled water was first used to zero the spectrophotometer.

2) Sample bottle was cleaned with well water.

3) Well water was poured into the sample bottle and insert into the spectrophotometer.

4) Readings were recorded.

BLANKS: A clean, known-quality water sample is processed through the entire filtration and drying procedure. The blank should have very little or no measurable residue.

DUPLICATE: The same well-water sample was filtered and analyzed twice for analytical precision.

STANDARD/REFERENCE: TSS was determined gravimetrically. Quality control included method blanks, duplicate samples, and verification of the analytical balance using certified calibration weights.

Microbiological data were summarized using well-level means (seeTable 2). Pearson correlation coefficients were used to describe linear associations among total coliforms, E. coli, distance, conductivity, TDS and TSS (see Table 3). Separate multiple linear regression models were fitted for total coliforms and E. coli using distance, conductivity, TDS and TSS as predictors (Table 4). Each model therefore contained four predictors and ten independent well-level observations. The models were considered exploratory because the number of independent wells was small relative to the number of predictors. Statistical significance was assessed at α = 0.05.

The well-level results showed variation in conductivity, TDS, TSS, total coliforms and E. coli concentrations among the sampled wells. These measurements were compared with the applicable drinking-water quality guidance and thresholds used in the study (seeTable 1). The microbiological results were interpreted primarily using the presence or absence and concentration of indicator organisms, with particular attention to E. coli because drinking-water guidelines generally require its absence in a 100 mL sample intended for human consumption [4][5].

Table 1. Laboratory analysis of well water for five consecutive weeks.

*COND—CONDUCTIVITY; TDS—TOTAL DISSOLVED SOLIDS; TSS—TOTAL SUSPENDED SOLIDS; TC—TOTAL COLIFORM; ND—NOT DETECTED. E. coli—Escherichia coli; *A value of 0 indicates no colonies were recorded in the reported observation.

Table 2.Means of all parameters.

Source: Field Work, 2025. *Cond.—conductivity. *ND—NOT DETECTED.

Table 3.Pearson correlation matrix for total coliforms and predictor variables.

Table 4.Multiple-regression coefficients for total coliforms (n = 10 wells).

The full model had R = 0.534 and R² = 0.285, meaning that the four predictors together accounted for 28.5% of the variation in well-level mean total coliform concentrations. The overall model was not statistically significant, F (4,5) = 0.498, p = 0.740. Distance was also not a statistically significant individual predictor (β = 0.232, t = −0.582, p = 0.586). The bivariate Pearson correlation between distance and total coliforms was r = −0.339. These statistics indicate that the small sample did not demonstrate a statistically significant association.

Table 5 shows that conductivity had the strongest positive correlation with E. coli concentration (r = 0.363), indicating that wells with higher conductivity tended to have higher E. coli levels, although the association was weak-to-moderate. Distance, TDS, and TSS showed very weak relationships with E. coli. The predictor variables were also weakly correlated with one another, with the strongest relationship being between distance and TDS (r = 0.297), suggesting little evidence of multicollinearity (see Table 5).

Consistent with this finding, Table 6 shows that conductivity had the largest standardized regression coefficient (β = 0.344), indicating a positive association with E. coli after controlling simultaneously for distance, TDS, and TSS. The similarity between the correlation and regression coefficients is consistent with the relatively low correlations among the predictors. However, the conductivity effect was not statistically significant (p = 0.455) (see Table 6). The overall findings of Table 5 and Table 6 indicate that none of the four predictors, distance, conductivity, TDS, or TSS was a statistically significant independent predictor of E. coli concentration in the 10 wells.

Table 5. Pearson correlation matrix for E. coli and predictor variables.

Table 6. Multiple-regression coefficients for E. coli (n = 10 wells).

The standard error of the estimate was calculated from the reported R² and the variance of the ten well-level mean outcomes, using residual degrees of freedom of 5. The E. coli model produced a negative adjusted R² (−0.498), indicating poor model performance after accounting for the four-predictor model complexity (see Table 7).

Table 7. Model summary for exploratory multiple regression.

Mean total coliform concentrations ranged from 0 to 326.2 CFU/100 mL, while mean E. coli concentrations ranged from 0 to 155.8 CFU/100 mL. W5 had the highest mean concentrations for both indicators (326.2 CFU/100 mL total coliforms and 155.8 CFU/100 mL E. coli). E. coli was detected in six wells, indicating faecal contamination in those wells. The presence of total coliforms in seven wells further indicates microbiological vulnerability of the water sources.

The observed contamination is important because drinking-water guidelines generally require the absence of E. coli in a 100-mL drinking-water sample [4][5]. However, the present study did not directly test for specific pathogenic bacteria, viruses or protozoa. Therefore, the results demonstrate faecal-indicator contamination rather than confirmed occurrence of specific pathogens.

Distances between wells and septic tanks ranged from 3.205 to 32.488 m. The relationship between distance and microbiological contamination was not uniform. W3 and W8, which had distances of 20.051 m and 32.488 m respectively, had no detected total coliforms or E. coli. Conversely, W5 and W9, at 10.164 m and 27.613 m, had relatively high E. coli means. W4, W6, W7 and W10 had shorter distances but comparatively lower E. coli means. This variability is consistent with the non-significant regression results.

Previous studies have reported relationships between sanitation-source proximity and groundwater contamination. Ngasala, Masten and Phanikumar (2019), for example, investigated domestic wells and hydrogeological conditions in peri-urban Dares Salaam and reported widespread E. coli contamination [21]. Such evidence supports the plausibility of sanitation-related groundwater contamination but does not establish that the same mechanism explains the variation observed in Agona Swedru.

1) Only ten wells were investigated, resulting in low statistical power and a high risk of over-fitting when four predictors were included in the regression models.

2) The wells were selected through non-probability purposive sampling, limiting generalizability beyond the sampled wells and study area.

3) Each well was sampled repeatedly over five weeks. The regression analysis used well-level means to avoid treating repeated observations from the same well as independent; consequently, temporal variation was not modelled explicitly.

4) The geographical scope was limited to selected wells in Agona Swedru Township.

5) Specific pathogenic bacteria, viruses and protozoa were not directly tested.

6) Detailed hydrogeological characteristics, groundwater flow and septic-tank leakage were not directly assessed.

7) Detailed analytical quality-control information for the microbiological and physicochemical procedures was not available in the study record used for this revision.

8) The negative adjusted R² for the E. coli model demonstrates poor predictive performance after accounting for model complexity.