Performance Monitoring and Sustainable Management of Piped Water Supply Infrastructure in Developing Communities

Water quantity data was collected through the installation of In-situ Rugged Troll pressure transducers on the bottom of water storage tanks within the systems being investigated. These submersible titanium instruments were programmed to measure and record absolute pressure on a 15-min time interval and were calibrated to NIST certification standards at $\pm 0.1\%$ at 9.0 m of water. The continuous monitoring of water levels within system storage tanks was then used to evaluate the availability and reliability of services. Water level data included seven sites in Madagascar and 10 sites in Nicaragua for a period ranging from 400 to 1,373 days [$\mathrm{mean}=735\mathrm{days}$, standard deviation $(\mathrm{SD})=313.8$]. Data was collected from the systems during periodic site visits where the device was removed from the tank, data was extracted from the logger, and the device was reinstalled. Storage tank specifications along with water level data were used to determine the per-capita availability of water over time [$V_{pc}$, $\mathrm{liters}/\mathrm{person}/\mathrm{day}$ ($\mathrm{L}/\mathrm{p}/\mathrm{d}$)], calculated as a function of daily availability of water in storage ($V_T/t_d$), supply flow rate ($Q_s$), and total number of people ($p$) being served

The availability of water in terms of storage volume ($V_s$, ${\mathrm{m}}^3$) was determined by multiplying the average daily water level ($h_{\mathrm{avg}}$, m) by the cross-sectional area of the tank ($A_T$, ${\mathrm{m}}^2$). The supply flow rate was determined by isolating data where the rate of change in water level was greater than zero ($\mathrm{\Delta }h/\mathrm{\Delta }t>0$) and identifying the maximum change in water level ($\mathrm{\Delta }{h}_{\max }$) daily. The daily maximum change in water level is isolated and used to determine the supply flow rate ($Q_s$) into the system, which assumes that there is a 15-min period where outflow ($Q_{\mathrm{out}}$), or the water demand on the system, equals zero. The total water availability is then determined after accounting for and eliminating overflow, using a binary multiplier (1,0)

A probability exceedance function is used to show the percentage of days that each system met various thresholds of per-capita water availability. From this, the system capacity (SC, %) was defined as the average percentage of days (PD) that each system met the 20, 50, and $80\mathrm{L}/\mathrm{p}/\mathrm{d}$ threshold, where $n$ equals the total days in the study

The system reliability ($\eta$) was analyzed daily by determining the percentage of time that the water storage facility was empty. A sensitivity analysis of various tank conditions revealed that water levels of less than 25% full showed the largest variation, which provided better resolution when evaluating system reliability (Ermilio et al. 2014; Hunt 2015). As a result, tank empty conditions were defined as less than 25% full (${\mathrm{PE}}_{25}$). The daily system reliability ($\eta$) was then determined by analyzing the proportion of time that tank conditions were less than the 25% full condition. The total system reliability (${\eta }_{\mathrm{sys}}$) was determined by averaging the daily reliability scores for the entire study period, where $n$ is the total number of days in the study

Water quantity characterization combined the system reliability with per-capita availability to classify the system performance. In these terms, four types of systems were classified: a reliable supply (${\eta }_{\mathrm{sys}}>80\%$) of high-volume water ($V_{pc}>50\mathrm{L}/\mathrm{p}/\mathrm{d}$), a reliable supply (${\eta }_{\mathrm{sys}}>80\%$) of low-volume water ($V_{pc}<50\mathrm{L}/\mathrm{p}/\mathrm{d}$), an unreliable supply (${\eta }_{\mathrm{sys}}<80\%$) of high-volume water ($V_{pc}>50\mathrm{L}/\mathrm{p}/\mathrm{d}$), and an unreliable supply (${\eta }_{\mathrm{sys}}<80\%$) of low-volume water ($V_{pc}<50\mathrm{L}/\mathrm{p}/\mathrm{d}$).

Participatory methods, which included input from local water professionals and partners, were used to investigate SoM characteristics. As a result, different surveys were employed to evaluate SoM characteristics, with input from partners being unique in each country. In Madagascar, program partners used a donor-approved sustainable index tool (Lockwood 2010; Ermilio 2019) that included a series of presence/absence indicators. In Nicaragua, workshops were conducted with partner organizations to develop the survey, and key stakeholder interviews were used to evaluate community resources for managing the system in terms of time and money (Hunt 2015; Ermilio 2019). In both cases, the data were consolidated into five categories of management: human resources (HR), system administration (SA), operation and maintenance (O&M), asset management (AM), and financial management (FM). The analysis of SoM included converting individual categories into a percent score, where $n$ is the total number of indicators within each category

In Madagascar, the sustainable index tool was modified to evaluate management characteristics based on presence/absence indicators. Human resources included indicators for external support, community involvement, and water management personnel. Operation and maintenance included indicators for the frequency of maintenance activities associated with the source intake, storage tank, and treatment systems, as well as capital maintenance, system expansion, and communications with customers. System administration was evaluated based on evidence of reporting and record keeping in terms of customer accounts, complaints, maintenance activities, water consumption, and the presence of a business plan. Asset management was evaluated based on initial investments into the infrastructure, office space, tools, and equipment, as well as watershed management initiatives. Financial management was evaluated based on evidence of monthly income, total income, annual expenses, tax payments, and overall savings. To prevent artificially weighing subindicators, an average of the indicators for each variable was used.

In Nicaragua, the SoM analysis was based on the same five variables; however, the results were monetized with respect to per-capita investments into the water supply infrastructure. Each variable was monetized by calculating time and money spent managing the system and was normalized using the range of results to establish a percent value. Investments in human resources were determined based on the number of committee members, the number of committee meetings annually, the number of assembly meetings, and the election frequency of committee members. Investments in operation and maintenance were monetized using the cleaning frequency of the source intake, storage tank, and treatment system as well as capital investments for large repairs. System administration was monetized using the actual operational expenses associated with managing the system, including payment of administrative staff, payment to local operators and plumbers, as well as repairs and other overhead expenses. Asset management investments were evaluated based on the presence/absence of an office space; storage space for supplies, tools, and equipment; land ownership; and watershed management. These assets were monetized equally across all sites using office space and land valuation. Financial management was evaluated based on records of actual savings with or without interest depending on banking in the area. The total available savings was adjusted for the age of the system and an annual savings was determined. For all the SoM variables in Nicaragua, the total monetized values were adjusted for per-capita values by dividing the investments by the total number of customers being served by the system.

In both cases the variables in Eq. (6) were normalized to a percent score, and then an average of the individual variables was used to determine the percent SoM score. The difference in the averages between the SoM results in Madagascar and Nicaragua suggests that the per-capita monetized SoM test used in Nicaragua was more difficult than the presence/absence SoM test used in Madagascar. Therefore, calibration of the survey was needed prior to conducting further analysis. The calibration process included three steps. A common point of analysis was identified wherein household customer satisfaction (CS) surveys were implemented at selected sites and plots of CS versus SoM were created. Then a probability distribution plot of the two SoM tests was created to confirm normal distribution, and the SoM surveys were transformed using the common point of analysis and the average difference. Refer to the Data Availability Statement for access to additional details about the SoM calibration.

A comparative analysis of water quantity reveals measurable differences in system performance. Probability distribution functions (PDFs) provide detailed performance characteristics and can be used to delineate between systems. On average, the systems in Madagascar provided minimum basic needs of water ($20\mathrm{L}/\mathrm{p}/\mathrm{d}$) during 89.8% of the days investigated with a range from 72.9% to 98.4% ($\mathrm{SD}=10.1\%$), Fig. 2. In Nicaragua, the minimum basic needs threshold ($20\mathrm{L}/\mathrm{p}/\mathrm{d}$) was provided 97.7% of days, with a range from 79.3% to 100% ($\mathrm{SD}=6.0\%$), Fig. 3. Using an upper threshold of $80\mathrm{L}/\mathrm{p}/\mathrm{d}$, the comparison between systems in Madagascar and Nicaragua shows an average of 3.6% ($\mathrm{SD}=7.0\%$) and 67.0% ($\mathrm{SD}=24.7\%$), respectively, which reveals that the systems in Nicaragua are providing a higher volume of water. In fact, on average, the community-managed systems in Nicaragua provided $68.1\mathrm{L}/\mathrm{p}/\mathrm{d}$ more than the privately managed systems in Madagascar ($p=0.0001$).

When the average per-capita availability results are combined with system reliability, further delineation of system performance is accomplished. Using the $50\mathrm{L}/\mathrm{p}/\mathrm{d}$ and 80% reliability thresholds, systems are classified using quadrants, Fig. 4. Category A systems are highly successful in terms of both reliability and availability of water. Category B systems are highly successful in terms of reliability and are providing minimum availability. Category C systems are highly successful in terms of availability but are not reliable, and Category D systems are vulnerable to failure wherein reliability and availability of water services are low. An important identifying characteristic for all the systems investigated is that the average availability of water was above the minimum basic minimum needs ($20\mathrm{L}/\mathrm{p}/\mathrm{d}$), which would have an expected outcome of improved health within a community (WHO 2011; Nicol 2000).

Understanding management characteristics is essential to ensuring the sustainability of water infrastructure. Measurable differences in SoM suggests that strategic initiatives can improve management of water services in developing communities, Fig. 5. A summary of SoM results shows that on average, management in Nicaragua (79.4%, $n=11$, $\mathrm{SD}=0.125$) is stronger than in Madagascar (72.7%, $n=6$, $\mathrm{SD}=0.231$) but that this difference is statistically insignificant ($p=0.438$). The largest difference in management characteristics between countries is within the SA category, where systems in Nicaragua were, on average, 17.6% ($p=0.164$) stronger than the systems in Madagascar. In addition, the AM category for systems in Nicaragua was 13.6% stronger ($p=0.279$) than in Madagascar. In terms of SoM variables, three indicator variables (HR, SA, and O&M) have mean values greater than, and two (AM and FM) have mean values less than, the composite SoM score.

An exploration of these results is used to investigate the influence that management has on water quantity performance characteristics, employing mean and quartile values as well as quadrant classifications as points of comparison (Table 2). The composite mean value for all systems studied (77.1%, $n=17$, $\mathrm{SD}=0.166$) provides a point of comparative analysis when evaluating the impact that above-average (good) management and below-average (poor) management has on overall system performance. The upper SoM quartile value (88.3%) is used to delineate very good management and the lower SoM quartile value (67.6%) is used to delineate very poor management.