Estimating the Ecological Performance of Water and Wastewater Treatment in Africa: A Meta-Analysis

3.2.1 Geographical Location

The mean values of energy use and EIs vary substantially between locations. There is an uneven representation of countries in the available data: South Africa and Egypt have about 84 % of the total estimates. Regarding energy use, GWP, and EP, South Africa occupied 39 %, 45 %, and 45 %, while Egypt had 31 %, 20 %, and 20 % of the estimates, respectively. South Africa had higher values for energy use than Egypt and the overall mean.

South Africa also had higher GWP values than Egypt and the overall mean. However, other countries (excluding Egypt) had the highest value. Regarding EP, Egypt and South Africa had the lowest values compared to the overall mean and other countries. These disparities have been linked to population characteristics, living conditions, economic advancement ⁶, climate change (e.g., ambient temperature), statutory discharge standards, electricity rates, and geomorphological attributes (e.g., elevation, altitude). Including the industrial outlook ²⁴, electricity mix ²⁵, technology and scale, policy and governance issues, and incidents when incentives in the sector are absent or deceitful ²⁶.

South Africa is classified as a chronic water-scarce nation and embraces water reuse to mitigate drought ²⁷. Consequently, in this review, most studies in South Africa utilized different variations and combinations of energy-intensive technologies: ion exchange and softening ²⁸, reverse osmosis, ultrafiltration ²⁸, ²⁹, membrane filtration ³⁰, eutectic freeze crystallization, evaporative crystallization ³¹, and magnesite-lime and ash-CO₂ bubbling ³² for treatment. Another energy-consuming heating method for treatment was studied in Uganda.

According to AQUASTAT ²², South Africa has 923 treatment facilities (the highest number in Africa) and a treatment capacity of 2.414 × 10⁹ m³a⁻¹ (second largest in Africa). The increased development of treatment facilities and awareness of the environmental impacts of different technologies, in general, is demonstrated by the Environmental Performance Index (EPI; specifies performance pointers that show how nations of the world manage environmental problems). South Africa ranks top for wastewater treatment, waste management, and combating climate change in Africa. These also consume energy, nevertheless. Hence, ongoing optimization of energy in WWTPs in South Africa demonstrate that 71 % of the treatment facilities in South Africa can generate power with a possibility of 20–50 % energy savings ²⁷, ³³. In contrast, less energy-consuming technologies like natural and aerated lagoons in Morocco ³⁴, rapid sand filtration ³⁵, and filtration, wetlands ³⁶, and activated sludge ³⁷ in Egypt were also utilized.

Moreover, energy-intensive technologies for water treatment have also been studied worldwide, comparable to the apparently high average in Africa. Again, these variations are attributed to the quality, volume, and legal treatment levels at various locations ³⁸. Studies from China and USA have shown that desalination technologies using thermal and membrane processes are the most energy-intensive, such as reverse osmosis (RO): 2.4–8.5 kWh m⁻³, vapor compression: 8–15.85 kWh m⁻³, multistage flash distillation: 26.42–68.69 kWh m⁻³, multiple-effect distillation: 39.71–105.7 kWh m⁻³, nanofiltration and electrodialysis ³⁹, ⁴⁰. Furthermore, energy use intensity for RO and other energy-intense technologies varied globally, Eritrea 2.33 kWh m⁻³, Kuwait 4.52 kWh m⁻³, Caribbean Island 3.15 kWh m^{−3 41}, South Africa 3.97–4.39 kWh m^{−3 29}, China 6.282 kWh m⁻³, and Saudi Arabia 4.4 kWh m^{−3 42}.

On the other hand, dilapidated and ageing technology and unscientific management of the wastewater industry in China was reported as critical issue for high energy intensity and subsequent emission of pollutants ⁶, which can also apply to facilities in Africa. Likewise, studies have shown that energy use intensity varies with location, maybe due to the prevalence of a particular treatment technology. WWTPs in Canada and France had high energy consumption compared to the USA, Spain, Germany, and Italy ²⁶. Besides, regions with higher water risks, dependence on groundwater or desalination sources, and those that use tertiary treatment for their WWT have relatively higher energy intensities ⁴³.

In the present study, electricity generation and use were reported as the chief contributors to environmental impacts. It can be attributed to the electricity mix dependency on fossil fuels, e.g., coal, oil, and gas. Moreover, high energy use, especially fossil fuels, produces high emissions of hydrocarbons and NO₂ ⁶, ²⁵. Hence, most observations that reported electricity generation and use had electricity mix from coal ²⁸, ⁴⁴ or fossil fuel ²⁹, ³², ⁴⁵. Consequently, the impact category with the strongest influence was GWP for most of these studies. Oil, gas, and coal constitute 39 %, 30 %, and 21 % of energy consumption in Africa, respectively, and 8 %, 40 %, and 28 %, respectively, for electricity production. Subsequently, they contribute 36 %, 22 %, and 34 %, respectively, to CO₂ emissions from fuel sources in Africa.

South Africa and Morocco are vital contributors to CO₂ emissions from coal, while Egypt, Algeria, and Nigeria contribute the most to emissions from oil and gas ⁴⁶. Likewise, regarding total GHG emissions, significant contributors were DRC Congo, South Africa, Nigeria, Egypt, and Algeria ⁴⁷. A similar trend is observed in the present study, as the GWP of facilities from South Africa is higher than those from Egypt. Equally, the indirect GHG emissions from treatment facilities arise mainly from energy consumption during aeration, pumping, wastewater, and sludge transportation, and contribute 14–68 % to the whole carbon footprint ⁴⁸. GWP is linked to N₂O, CH₄, and CO₂ emissions from nitrogen, biochemical oxygen demand (BOD), and chemical oxygen demand (COD) ⁷. GHG emissions for WWTPs also demonstrated inconsistency from USA 0.00–0.56, China 0.13–0.90, and South Africa 0.07–1.22 kg CO₂-eq m^{−3 25}. Several studies have shown a proportional relationship between energy use and GHG emission intensity ²⁶, ⁴³.

EP relates to various nitrogen, carbon, and phosphorus species, such as BOD, total Kjeldahl nitrogen (TKN), NH₄, NO₃, NO₂, , total suspended solids (TSS), and total phosphorus (TP). It is a function of the concentration of pollutants in the effluent, regional environmental considerations ⁴⁹, weather conditions, and seasons ⁵⁰. But most common LCIA methods fail to integrate these local differences, particularly in developing countries ⁴. Furthermore, since eutrophication partly depends on the concentration of nutrients in emissions to water, untreated wastewater likely has a higher EP than treated effluents. Moreso, there is a reported insufficiency in the capacity of sanitation systems and wastewater treatment facilities in Africa. Between 2008–2019, the total volume of municipal WW generated (× 10⁹ m³) was 77.3, 32.2, 7.6, 5.6, 3.3, 3.1, and 1.3 in Egypt, South Africa, Morocco, Libya, Tunisia, Ghana, and Senegal, respectively ²².

Fig. 5 displays the proportions of treated, untreated, and direct use of untreated municipal WW for irrigation purposes in Africa compared to other countries. The ratio of untreated WW discharged into the environment is higher in most African countries, hence a higher chance of eutrophication. From treatment facilities in Africa, the discharge standard is less stringent, and the level of compliance is low compared to developing countries, which is indicated by the EP values. Likewise, variations exist in the effluent discharge standards of developing countries. Nigeria has a BOD (mg O₂ L⁻¹) limit of 30–50, Tanzania 30, and 50 for Ghana, Uganda, and Malaysia, respectively. For DS (mg L⁻¹), Thailand and Tanzania had as high as 3000 and low as 200 in Nigeria. Other parameters, such as COD and soluble solid (SS), showed wide discrepancies ⁵¹. These inconsistencies in parameters reflect in the discharges and, consequently, EP values.

3.2.2 Water Source

The mean values of energy use and EIs in the treatment of raw water and wastewater are displayed in Fig. 6. As mentioned earlier, only mean values are commented on. The treatment of raw water consumed more energy than wastewater. When disintegrated, industrial consumes more energy than municipal WW treatment. Like the variations observed in various locations, energy consumption in the water sector varies due to groundwater characteristics, climate, seasonal temperature, rainfall, water requirement, volume of water, and treatment technologies ⁴⁰. Furthermore, the discharge standard and treatment scale are responsible for a significant disparity in energy use outcomes and EI intensities ⁶.

The flow rate, the equivalent population, the dilution factor, and the plant layout influence the intensity of energy use ²⁶. Meanwhile, most of the observations under raw water and industrial WW were the treatment of water with high total dissolved salts, such as industrial WW from mine drains ²⁹, ³¹, ³², seawater ²⁹, and saline water ²⁸. Raw water treatment is energy-intensive, yet desalination systems consume more. However, the energy use values in this review fall within the range of different desalination technologies listed earlier, up to 105.7 kWh m^{−3 40}.

In general, the critical source of energy consumption in water or wastewater treatment is the nature of the pollutant to be removed. Before supply, potable water is purified to strict physiochemical standards devoid of pathogens, which might not necessarily apply to effluent discharge ⁴². Although raw water might not contain as many pollutants as wastewater, the degree of purification needed to treat raw water to potable standards is higher. Subsequently, raw water of saline origin consumes more energy and resources. Similarly, industrial WW often contains recalcitrant pollutants such as heavy metals and phenolic compounds, which often require advanced technologies for their removal ⁵².

Raw water treatment, depending on the source, consumes more energy than wastewater. On average, municipal water treatment spends 0.2–8.5 kWh m⁻³ in Australia and 0.07–5.47 kWh m⁻³ in California and for recycled water treatment 2.8–3.8 kWh m⁻³ in Australia, 0.33–3.1 kWh m⁻³ in California, while wastewater consumes 0.44–1.1 kWh m⁻³ in Australia and 0.38–1.22 kWh m⁻³ in California ⁴². Similarly, industrial WW consumes more energy than municipal WW treatment ²⁵. Thus, as observed in the present study, potable water production consumes more energy than municipal and industrial WW treatment, respectively.

Generally, in centralized WWT systems, with certain exceptions, the energy use intensity is inversely proportional to the increase in the capacity ⁶, ²⁶. This is attributed to more stable and automated operational conditions, the use of efficient equipment, and a more experienced workforce at larger facilities ²⁶. On the other hand, systems using cycle activated sludge systems (CASS) processes consume more energy than anaerobic/anoxic/oxic (AAO) and anoxic/oxic (AO) processes, respectively. But there is less energy use during the construction and demolition stages of AO and CASS systems. However, CASS has complex operational procedures, such as aeration with high energy consumption ⁶.

Similarly, a membrane reactor (MBR) has the highest energy use compared to biological nutrient removal (BNR) and conventional activated sludge (CAS) systems due to the energy requirement of aeration units and losses due to fouling and clogging ²⁶. Regarding sludge management, aerobic stabilization requires more energy than anaerobic digestion but depends on plant size and pollutant removal efficiency. Dewatering and mechanical centrifugation also take a high chunk of the energy demand of treatment facilities ²⁶.

Meanwhile, the treatment technologies, sludge handling, and disposal methods influence GWP and EP values ⁵, ⁴⁸. Water and sludge treatment processes are accountable for the direct GHG emissions from treatment facilities. They contribute 23–83 % to the overall carbon footprint while 1–13 % come from offsite sludge disposal ⁴⁸. Likewise, the direct emission profile of treatment systems depends on influent characteristics, dissolved oxygen, and water temperature ⁶. This indicates why WWT can have higher GWP than raw water because wastewater has a high load of biomass (BOD, COD), thus, higher direct emissions of CH₄ and N₂O species. Likewise, the superior energy use intensity of raw water amounts to greater indirect emissions from energy consumption. Also, the GWP like energy use intensity decreases with expanding the scale of treatment ²⁵. The GWP of AAO systems is higher than CASS and AO because of their higher energy use intensity ⁶.

Furthermore, though criticized for high CH₄ emissions, anaerobic technologies have lower GHG emissions than other technologies. Upflow anaerobic sludge blanket (USAB) configurations, when compared to modified Ludzak-Ettinger (MLE) and Bardenpho, showed higher direct but lower overall GHG emissions. This is mainly because anaerobic technologies are more energy-efficient and allow for energy recovery. The energy recovery offsets the total emissions ⁴⁸. Similarly, sludge drying contributes 22–59 % to total GHG emissions, while anaerobic digestion of sludge reduces it by about 12–38 %. However, landfill sludge disposal has higher GHG emissions than incineration, composting, and agricultural use ⁴⁸.

EP for sanitation systems were at the peak for SBR than biofilters, soil infiltration, and dry toilet systems. This was associated with nutrient concentration and discharge pattern ⁵⁰. CASS and AAO processes are supposed to be more efficient than AO, which reflects that AO has the highest EP intensity. With increasing capacity, the nature and concentration of pollutants lessen. Thus, EP can increase with sudden expansion due to the diminishing efficiency of treatment processes ⁶. However, decentralized (source-separation) systems had higher EP than centralized ⁵. Recycling sludge as a phosphorus product has a higher EP than digested sludge. EP can be reduced using decentralized recovery systems, optimization of chemical use in sludge management, accounting for ammonia emissions and avoided fertilizers ⁵.

3.2.3 Meta Regression Results

To further elucidate the homogeneity in the pooled variables, meta-regression is used to identify the influential factors. It also identifies the moderating effect of these factors with the corresponding magnitude and direction. The outcome of the meta regression analysis is presented in Tab. 1, together with the results obtained for energy use, GWP, and EP. All regression results are presented in the reduced form. Under the energy (same for GWP, EP) column, the estimate and standard error results from the ordinary least square (OLS) HCCM procedure are shown. Only coefficient estimates significant at p-value ≤ 0.1 have been included in the reduced form. This explains the empty cells in Tab. 1.

Concerning the model information, N represents the number of observations. The R-squared indicates the variation percentage defined by the model. The adjusted R-squared statistic (Adj. R-squared) is like the R-squared, but the former is insensitive to the number of variables contained in a model. Also reported are the logarithm likelihood (Log-likelihood), Fisher test statistic (F-stat.), Akaike's Information Criterion (AIC), and Bayesian Information Criterion (BIC) for the model.

In this section, comments on the results are only centered on the signs and significance level of parameter estimates since the absolute magnitude is trivial ¹³. The consequences of factors influencing the estimates of energy use and EIs are subsequently argued by pertinent literature comparison. Again, the purpose of the meta-regression is to test if the variables affect the values of energy use and EIs from primary LCA studies.

In the regression models, energy use has an adj. R-squared of 63 % and the F-stat p-value significant at 1 %, indicating a significant joint impact on energy use by the variables considered. Likewise, a significant combined effect on EP was shown at an F-stat p-value of 1 % with adj. R-squared 68 %. However, the regression model for GWP is insignificant, but some variables separately influence GWP, as indicated in Tab. 1.

For the sources of water, the estimates indicate a statistically lower energy use for municipal WW treatment than for industrial WW (p-value < 0.01), while raw water treatment expends more than industrial and municipal WW (p-value < 0.001). Compared to municipal WW treatment, the energy use is statistically lower for municipal by 60 kWh m⁻³ and higher for raw water treatment by 1.5 kWh m⁻³. At the same time, GWP and EP are higher (p-value ≤ 0.01) for municipal WW treatment. GWP and EP are statistically higher for municipal by 1.42 kg CO₂-eq m⁻³ and 0.001 kg -eq m⁻³, respectively. Hence, these suggest that the source of water treated has an influence on energy use and EI results. This aligns with the visuals in and the arguments for Fig. 6. The treatment technology, influent characteristics, and effluent requirements were critical contributors to this as explained earlier.

Similarly, South Africa has a negative influence on energy use for the geographical locations, significant at 1 % level. Thus, energy use is significantly lower for South African studies than in Egypt. The estimates indicate that treatment processes in South Africa consume about 75.77 kWh m⁻³ less than in Egypt. But the visual from descriptive statistics in Fig. 4 opposes this trend. It is unexpected because both the arithmetic and pooled means showed that energy use in South Africa is higher than in Egypt. Nevertheless, GWP values from Egypt are lower than those from South Africa and other countries, while EP values are lower in South Africa (p-value < 0.1) compared to Egypt and other countries (p < 0.01), respectively. Compared to Egypt, GWP values are higher in South Africa and other countries by 3 and 9 kg CO₂-eq m⁻³, respectively, while EP values are lower in South Africa by 0.006 kg -eq m⁻³ and higher by 0.2 kg -eq m⁻³ in other countries than Egypt. Hence, the geographical location influences energy use and EI results. Most of these discrepancies were attributed to energy/electricity mix, electricity rates, demography, economic and industry outlook, and the geomorphology of countries.

For stages considered, the studies reporting only the complete life cycle (C/O/D) of the treatment facilities had significantly lower (p-value < 0.001) energy use values compared to those considering construction and operation phases only. However, emission estimates were lower in studies reporting O and C/O/D than C/O stage by 5 and 2 kg CO₂-eq m⁻³ for GWP and by 0.006 and 0.002 kg -eq m⁻³ for EP. Moreover, a meta-analysis study ⁶ ignored the effects of these boundaries in the analysis. Meanwhile, the construction phase contributes more than 5 % ⁷ and up to 50 % ⁵³ to the overall environmental impact of non-intensive technologies and large plants, respectively. Thus, the present study has gone further to elaborate the variations due to the boundaries considered.

In contrast, the LCIA method showed no significant impact on energy use outcomes and EIs in the primary studies considered. The comparison of LCIA methods in assessing WWTPs exhibited no significant variation in GHG emissions, EP, and resources ⁵⁴. Likewise, a study on virgin and recycled plastic found consistent results for GWP, acidification and eutrophication for five LCIA methods (Eco–indicator 95&99, CML, EPS, and EDIP) ⁵⁵. However, there was inconsistency in the comparison between CML and e-Balance for assessing WWTPs ⁵⁶.

Expectedly, the choice of modeling software and LCIA method had no significant influence on energy use. This could be predicted since the energy use values are debatably not primary outcomes of LCA studies. Although, it could be argued that software also contains inbuilt databases like the Ecoinvent, where data on energy use for a unit process can be obtained for LCA studies. Nevertheless, the influence seems insignificant. Regardless, the GWP values from other software (excluding SimaPro) are significantly higher (p-value < 0.01) than from GaBi, whereas EP values from SimaPro are substantially higher (p-value < 0.01) than from GaBi. Hence, the choice of software influences the GWP and EP outcomes.

Lastly, the publication year has a significantly negative (p-value < 0.01) impact on GWP and EP values, respectively. The estimates imply a decrease in GWP by 0.2 kg CO₂-eq m⁻³ and EP by 0.0002 kg -eq m⁻³ per annum. Hence, the publication year affects outcomes of GWP and EP. The reason for this might not be apparent but is not unconnected to the rising awareness of sustainability and various steps being put in place over the years to reduce the environmental impacts of the water sector. However, several studies, as seen in Sect. 3.2.1 and Sect. 3.2.2, showed that energy use is directly proportional to GWP but does not apply in this instance. This could be due to the lack of energy and resource recovery scenarios in the observations in the present study. Only two studies reported energy recovery via biogas ³⁷, ⁴⁵. However, biosolids and organic manure were also recovered ³⁵, ³⁶, ⁴⁵, ⁵⁷, ⁵⁸.