Assessment of rainfall variability, rainwater harvesting potential and storage requirements in Odeda Local Government Area of Ogun State in South-western Nigeria

Abstract

Rainfall variability with periodicity of 5–6 years has been demonstrated for our study area and may be attributed to tropical and extra-tropical factors which operate during different months, seasons and years. Rainfall variability in terms of coefficient of variation ranges from 24 to 39% and 26 to 41% for the seasons and months. The mean increase of 1.63 mm/year and 1.37 mm/year experienced in the dry season months (November–April) and the wet season months (May–October), respectively, is insignificant from a water management perspective. Hoeffding’s D statistics revealed prevalence of non-monotonic trend in all the months and seasons. Recommended minimum and maximum storage capacity requirements for a six-member household to maximize rainwater harvesting are 1 and 6 m³, respectively. The rainwater harvesting potential for the area of study ranges between 18.16 and 27.45 m³ and 15.23 and 30.40 m³ based on the maximum error estimate and coefficient of variation methods. Domestic rainwater harvesting has the potential to meet 27.51–54.91% of non-potable household water demand as well as 78.34–156.38% of household potable water demand for a six-member household. It is highly encouraged as a supplementary water source especially in rural and peri-urban areas to reduce their vulnerability to acute shortage of water infrastructure.

Keywords:

• climograph
• domestic rainwater harvesting
• rainfall variability
• rainwater harvesting potential
• seasonal classification of climate
• standardized precipitation index
• storage requirements
• water demand

Public Interest Statement

This research has revealed the variability in rainfall with periodicity of 5–6 years in Odeda Local Government Area of Ogun State, south-west of Nigeria. It also demonstrated the potential of harvested rainwater in meeting domestic water demands for potable and non-potable applications as well as the corresponding storage water requirements. Primary water treatment is imperative when the harvested rainwater is used for potable purposes. In order to harness this valuable resource, non-weathering roofing materials are recommended for future construction of residential buildings in rural and peri-urban areas with limited water infrastructural facilities and high rainfall.

1. Introduction

Sustainable access to water for potable and non-potable uses continues to pose a huge challenge in developing countries. Sub-Saharan Africa (SSA) alone accounts for 40% of the global population without access to safe drinking water (Sojobi, Owamah, & Dahunsi, 2014). In Africa, it was estimated that 75–250 million people would be exposed to increased water stress by 2020 (Kalungu, Filho, Mbuge, & Cheruiyot, 2014). This worrisome situation is further aggravated by poor water governance, extreme social inequality, population growth and climate change in Africa.

Rainwater harvesting (RWH) has been proposed as one of the options to improve water supply especially in rural and peri-urban areas of low-income countries (Cruddas, Carter, Parker, Rowe, & Webster, 2013; Opare, 2012), areas without reticulated water supply (Ndiritu, Odiyo, Makungo, Ntuli, & Mwaka, 2011), water-scarce, remote and marginalized areas (Nijhof, Jantowski, Meerman, & Schoemaker, 2010), areas where existing water supply is inadequate (Aladenola & Adeboye, 2010), areas with abundant annual rainfall (Ghis & Schondermark, 2013), highly contaminated and saline coastal areas (Samaddar, Murase, & Okada, 2014) as well as arid and semi-arid regions (Abdulla & Al-Shareef, 2009; Branco, Suassuna, & Vainsencher, 2005).

Literature survey revealed several types of RWH which include infield RWH (IRWH), in situ RWH, roof-based RWH (RRWH) and land-based storm water harvesting (Abdulla & Al-Shareef, 2009; Clark et al., 2015; Lebel, Fleskens, Forster, Jackson & Lorenz, 2015; Welderufael, Woyessa, & Edossa, 2011)

Factors militating against the adoption and scaling of domestic rainwater harvesting (DRWH) include use of poor roofing materials and high cost of storage tank (Cruddas et al., 2013; Opare, 2012), huge capital cost of acquisition, installation and maintenance of DRWH systems (Roebuck, Oltean-Dumbrava, & Tait, 2011), limited knowledge of the potentials of RWH (Kohlitz & Smith, 2015), lack of finance, legislation and coordination (Mwenge Kahinda & Taigbenu, 2011), space requirements (Traboulsi & Traboulsi, 2015) and poor quality of DRW (Oke & Oyebola, 2015

In addition, lack of skills (Kalungu et al., 2014), lack of social capital (Esterhuyse, 2012), risk of water-borne diseases (Dobrowksy, Mannel, et al., 2014; Dobrowsky, Van Deventer, et al., 2014; Mwenge Kahinda, Taigbenu, & Boroto, 2007; O’Hogain, McCarton, McIntyre, Pender, & Reid, 2011; Roebuck et al., 2011) and contaminants (Lee, Yang, Han, & Choi, 2010; Mendez et al., 2011; Sturm, Zimmermann, Schütz, Urban, & Hartung, 2009; Zhang et al., 2014) are concerns that need to be addressed to facilitate the uptake of DRWH.

The quantity of rainwater harvested depends on monthly precipitation, roof catchment area and roof run-off coefficient (Woltersdorf, Liehr, & Döll, 2015), while the quality of rainwater harvested depends on roof type, level of atmospheric pollution, geographical location, container size, catchment characteristics, land use practices and local climate.

Several studies have been done on different issues pertaining to rainwater harvesting. For example, with respect to storage, Woltersdorf et al. (2015) recommended tank size of 30 m³ for a roof size of 100 m², while Ndiritu et al. (2011) recommended a storage tank size of 40 m³ for a roof size area range between 75 and 150 m². Imteaz, Adeboye, Rayburg, and Shanableh (2012) recommended a tank size of 7,000 litres to achieve 100% reliability for toilet flushing and laundry.

Likewise, Biswas and Mandal (2014) observed that a 4,000-L concrete tank installed with a roof area of 40 m² was adequate to take care of water demands of four-member household for five-month dry period, while Mwenge Kahinda, Taigbenu, and Boroto (2010) recommended an optimum tank size of 0.5 m³ which achieved water savings of 10–40%.

In order to achieve a good water-saving efficiency and limit financial losses, Roebuck, Oltean-Dumbrava, and Tait (2012) recommended storage tank size limit of 1.2–1.5 m³. Moreover, Imteaz, Ahsan, and Shanableh (2013) recommended design of rainwater tank size to achieve rainwater accumulation potential (RAP) of 0.8–0.9 and as well opined that 100% reliability is unachievable even with 10,000-litre tank with 300 m² roof area (Imteaz, Matos, & Shanableh, 2014).

Besides, Boelee et al. (2013) recommended careful participatory planning, design and management of DRWH storage to minimize associated health risks, while Bocanegra-Martinez et al. (2014) presented multi-objective optimization approach to DRWH. Further, Fernandes, Terencio, and Racheco (2015) proposed a threshold of 0.8 (Annual water demand/Annual harvestable rainwater) to distinguish low-to-high demand RWH applications and recommended low-storage capacity for low-demand applications.

Recent researches have also shown that the quality of DRWH can be improved by point-of-use treatment, integration of water safety plans, quarterly testing and utilization of weather-resistant materials such as ceramic tiles, public education and regular maintenance (Fry, Cowden, Watkins, Clasen, & Mihelcic, 2010; Gwenzi, Dunjana, Pisa, Tauro & Nyamadza, 2015; Kohlitz & Smith, 2015; Kwaadsteniet, Dobrowsky, Deventer, Khan, & Cloete, 2013; Thomas, Kirisits, Lye, & Kinney, 2014; Zhang et al., 2014).

In addition, Efe (2006) suggested primary treatment to take care of pH, TSS, Fe and colour and preference for aluminium roofing sheets compared to other materials such as corrugated, thatch, asbestos and open surface, while Helmreich and Horn (2009) recommended the use of local materials, skills and equipment to reduce cost.

Furthermore, the benefits of domestic RWH have been found to include achievement of 30–87.6% water savings (Amado & Barroso, 2013; Bocanegra-Martinez et al., 2014; Souza & Ghisi, 2012), mitigation of storm run-off and conservation of potable water (Campisano, Gnecco, Modica, & Palla, 2013), sixfold improvement in crop yield when RWH irrigation was combined with fertilizer applications (Biazin, Sterk, Temesgen, Abdulkedir, & Stroosnijder, 2012), financial savings and cost-effective improvement of urban drainage systems (Słyś & Stec, 2014), aquifer recharge (Clark et al., 2015) and reduction of drinking water risks in highly contaminated and saline coastal areas (Samaddar et al., 2014).

Campisano et al. (2013) found that frequent precipitation increases the performance of DRWH and that the water-saving efficiency depends on storage tank size, demand fraction, storage fraction and climate. Also, Chao-Hsien and Yu-Chuan (2014a) observed that DRWH potential depends on climatic, building characteristics, economic and ecological factors and that with respect to climatic factors, quantity of precipitation is the most crucial factor.

In addition, Chao-Hsien and Yu-Chuan (2014b) found that effective roof area and storage capacity for DRWH vary from one climatic region to another and that failure to account for rainfall variability leads to underestimation of storage capacities. Likewise, Nnaji and Mama (2014) observed that RWH potential is a function of rainfall coefficient of variation (COV), level of water consumption and roof area per capita. The authors recommended integration of rainwater systems in bungalow residential buildings in rainforest regions of Nigeria with COV range of 0.85–1.01, where DRWH has the potential to meet 100% domestic water demand.

Also, Bocanegra-Martinez et al. (2014) demonstrated the variability in the amount of harvested rainwater with highest value recorded in September, followed by October and August and recommended year-round storage. Distribution of rainfall on monthly, seasonal and annual scales is important for planning DRWH, agriculture as well as general water applications. Understanding of seasonality pattern of rainfall is very useful for planning DRWH storage. Guhathakurta and Saji (2013) utilized seasonality index in identifying rainfall regimes, while Akinsanola and Ogunjobi (2014) classified annual rainfall based on standardized annual precipitation index (McKee, Doesken, & Kleist, 1993).

Numerous studies have been done on rainwater harvesting, climate variability and rainfall pattern in Nigeria. These research efforts have focused on change detection in rainfall pattern (Abaje, Ndabula, & Garba, 2014; Ogungbenro & Morakinyo, 2014), rainfall seasonality in Niger Delta on monthly and annual scales (Adejuwon, 2012), annual rainfall and temperature variability (Akinsanola & Ogunjobi, 2014), annual and monthly rainfall patterns in Ekiti State (Akinyemi, Ayeni, Faweya, & Ibraheem, 2013), inter- and intra-annual rainfall variability and distribution pattern over North-west Nigeria (Ekpoh & Nsa, 2011).

Furthermore, other research efforts have investigated quality of RWH in Delta State (Efe, 2006), quality of rainwater from different roof materials in Oyo State (Olaoye & Olaniyan, 2012), monthly rainfall trends in Nasarawa State (Ekwe, Joshua, Igwe, & Osinowo, 2014), monthly rainfall distribution in Benin-Owena River Basin (Ikhile & Aifesehi, 2011), socio-demographic aspects of RWH practices in Ibadan (Lade & Oloke, 2015), DRWH practices in Enugu, uses and advantages (Ajayi & Ugwu, 2008), spatiotemporal variation and prediction of monthly rainfall over North-east Nigeria (Bibi, Kaduk, & Balzter, 2014).

In addition, other researches have focused mainly on DRWH potential (Lekwot, Samuel, Ifeanyi, & Olisaemeka, 2012; Nnaji & Mama, 2014), required storage capacity for DRWH (Otti & Ezenwaji, 2013), DRWH technology (Shittu, Okareh, & Coker, 2015), monthly variability in harvestable rainwater and maximum storage requirement (Ubuoh, Ege, Ogbuji, & Onifade, 2012). Oke and Oyebola (2015) advocated mobilization and motivation of house owners.

Our literature survey revealed that factors affecting rainfall variability in South-west Nigeria can be classified as tropical and extra-tropical factors. The tropical factors include inter-tropical discontinuity, tropical easterly jet, sea surface temperature and biogeophysical feedback mechanism, while the extra-tropical factor include El Nino Southern Oscillation (Olaniran, 2015).

Omogbai (2010) demonstrated that sea surface temperature of the tropical Atlantic Ocean and land–sea thermal contrast between sea surface temperature and rainfall stations are responsible for 87% of rainfall variability in South-west Nigeria, while surface location of inter-tropical discontinuity and land surface temperature of rainfall stations are responsible for 7 and 6% of rainfall variability in South-west Nigeria. The author also attributed the sea surface temperature to the combined action of the cold Benguella undercurrent and Ekman transport.

Giannini, Saravanan, and Chang (2003) also reported that land–atmosphere interactions amplify SST-driven signal which is responsible for inter-annual and inter-decadal variability of rainfall, while Nicholson and Grist (2001) reported deep, well-developed equatorial westerlies, Africa easterly jet (AEJ) and Tropical easterly jet (TEJ) to influence rainfall variability in West Africa.

In another study, Akinsanola and Ogunjobi (2014) attributed rainfall variability to local factors such as orography, boundary layer forcing and moisture build up. Study of rainfall variability is very important because it has been found to affect rural water supply and food production in the south-west of Nigeria (Adegoke & Sojobi, 2015; Ganiyu, Akinniran, & Adeyemo, 2013).

Most of the rural and peri-urban areas of Ogun State experience acute water shortage as a result of the poor water supply coverage of Ogun State Water Corporation (OSWC), the agency of government saddled with the responsibility of providing public water supply within the State. This situation is further aggravated by the poor funding of OSWC which limits expansion of water infrastructure and services, regular power outage prevalent within the State as well as the rapid population increase in the rural and peri-urban areas within the State (Ufoegbune, Oyedepo, Awomeso, & Eruola, 2010).

Within the State, <12.5% of the populace have access to weekly regular public water supply (Odjeba, Idowu, Ikenweiwe, Martins, & Sadeeq, 2015), while Agbelemoje and Odubanjo (2001) reported that 3% of residents within the State have access to clean and safe piped water. As a result of this ugly scenario, residents resort to other alternative sources such as private piped borehole, shallow hand-dugwells, rain, rivers/streams and water vendors (Coster & Otufale, 2014; Federal Republic of Nigeria, 2000; Gbadegesin & Olorunfemi, 2007).

Sadly, our literature review revealed that most of these alternative sources are unwholesome for drinking and are contaminated by pathogens which have led to water-borne diseases such as typhoid, cholera, dysentery and hepatitis (Dahunsi, Owamah, Ayandiran, & Oranusi, 2014; Otufale & Coster, 2012). Furthermore, these water sources have been found to be polluted by heavy metals such as uranium, lead (Pb), Nickel (Ni), Chromium (Cr), Cadmium (Cd), Zinc (Zn) and arsenic (Amori, Oduntan, Okeyode, & Ojo, 2013; Dahunsi et al., 2014).

Majority of the residents in Odeda are middle low-income earners who rely on shallow groundwater supply of poor quality (Dahunsi et al., 2014). Consumption of the boreholes and wells in Odeda exposes the residents to chemical toxicity as a result of the contamination of the groundwater by uranium (Amakom & Jibiri, 2010) which is above the safe limit recommended by WHO (2004a).

Uranium chemical toxicity has been known to cause kidney and genetic mutations, developmental malfunctions and cancer in severe cases. Bacteriological assessment of the groundwater also revealed contamination by Coliform and E. coli (Shittu, Akpan, Popola, Oyedepo, & Oluderu, 2010). The drudgery from fetching water has been found to affect the women’s health who spend an average of 1 h daily covering about 1 km to fetch water (Coster & Otufale, 2014; Otufale & Coster, 2012).

Rainwater harvesting has been successfully deployed in the eastern part of Nigeria such as Edo State with appreciable success and is practiced by >80% of the households (Tobin, Ediagbonya, Ehidiamen, & Asogun, 2013), while between 3 and 6.6% in the south-western part of Nigeria (Gbadegesin & Olorunfemi, 2007; Lade & Oloke, 2015). Also, residents in Odeda rely on RWH during the wet season because of the poor quality of the shallow wells attributed to the poor sewage and sewerage and open defecation prevalent in the area (Shittu et al., 2010).

This study is, therefore, embarked upon with a view to encourage the adoption and utilization of rainwater harvesting to reduce the vulnerability of the rural and peri-urban populace to the prevalent poor water supply and also mitigate health risks associated with other water sources. Indeed, rainwater harvesting has been recommended for use to supplement other water sources and as a buffer during emergencies (Aladenola & Adeboye, 2010).

Further, since rainwater harvesting and infrastructure are affected by rainfall variability (Adegoke & Sojobi, 2015; Aladenola & Adeboye, 2010; Food and Agriculture Organization, 2007; Worm & Hattum, 2006), the effects of rainfall variability on the rainwater harvesting potential and appropriate storage requirements were also investigated to address the inadequate water storage and RWH facilities that is rampant in the rural and peri-urban areas (Aper & Agbehi, 2010). Also, trends in the monthly, seasonal and annual rainfall were studied to ascertain if it is increasing or not.

The significance of this research is that it has incorporated rainfall variability in the calculation of rainwater harvesting potential and in the calculation of storage water requirements. In addition, as a supplementary source of water, it demonstrated the percentage of domestic water demands that can be met by rainwater harvesting for potable and non-potable purposes. Furthermore, results from this study will sensitize, encourage and guide engineers/architects in planning for DRWH in the design and construction of residential buildings.

2. Materials and methods

2.1. Study area

Odeda doubles as a town and headquarters of Odeda Local Government Area (LGA) in Ogun State, located in South-western Nigeria as shown in Figure 1. It lies between longitudes 3° 26′ 76″ and 3° 47′ 28″ and latitudes 7° 29′ 88″ and 7° 05° 54″. Being one of the largest LGA in Abeokuta which is the State capital, it has a population of 109, 449 based on the 2006 population census. The town enjoys tropical climate with uni-modal peak rainfall between June and November, average annual and monthly rainfall of 1, 220 mm and 102 mm, respectively, as well as monthly maximum and minimum temperature ranges of 29–36 and 22–35°C, respectively (Kilanko-Oluwasanya, 2009). Southwesterly wind prevails during rainy season, beginning from March to November, while Northwesterly wind dominates during the dry season, beginning from December to March. Geologically, the town is overlaid by crystalline basement which is basically granitic rocks and is being mined commercially for construction purposes.

Figure 1. Map depicting location of Odeda LGA in South-west Nigeria.

Source: Isaac Idowu Balogun.

Public water supply in the town is erratic, highly unreliable and is limited to once per week, while in some areas, such as the GRAs, borehole is not allowed (Kilanko-Oluwasanya, 2009). Owing to the inadequate public water availability, residents rely mainly on self-supply systems such as boreholes and hand-dug wells which are often contaminated (Amori et al., 2013; Kilanko-Oluwasanya, 2009) and limited in depth (Martins, Ajayi, & Idowu, 2000).

2.2. Seasonal classification of climate

Our literature review showed similarity as well as disparity in seasonal climate classification used globally. Four (4) classifications were observed in the literature. In addition, four seasons identified in the literature, namely, were spring (pre-monsoon), summer (monsoon), autumn (post-monsoon) and winter (winter) as shown in Table 1. Classification (1) was adopted for Europe and Asia by Shaw, Beven, Chappell, and Lamb (2010); Hu et al. (2015), Perry (2006) and Vanem and Walker (2013), while classification (2) was adopted for India by Mahajan and Dodamani (2015) and classification (3) was utilized by Sayemuzzaman and Jha (2014) for the USA.

Table 1. Different Global Seasonal classifications of climate from the literature

Class	(1)		(2)		(3)		(4)
Seasons	Months	Seasons	Months	Seasons	Months	Seasons	Months
Spring	Mar, Apr and May	Pre-monsoon	MAM	Spring	Apr, May and Jun	Spring	Feb, Mar and Apr
Summer	Jun, July and Aug	Monsoon	JJAS	Summer	Jul, Aug and Sep	Summer	May, Jun and Jul
Autumn	Sep, Oct and Dec	Post-monsoon	Oct,Nov, Dec	Fall	Oct, Nov and Dec	Autumn	Aug, Sep and Oct
Winter	Dec, Jan and Feb	Winter	Jan, Feb	Winter	Jan, Feb and Mar	Winter	Nov, Dec and Jan

Climate classification (4) was utilized for our study area to depict the seasonal rainfall variability in Nigeria. The four seasons are: spring which coincides with March Equinox comprising February, March and April (FAM), summer which coincides with June Solstice consisting of May, June and July (MJJ), autumn which streamlines with September Equinox [August, September and October (ASO)] and lastly, winter which is described as December Solstice [November, December and January (NDJ)].

2.3. Data collection, methods and analysis

The rainfall and temperature data spanning 18 years from 1995 to 2012 were obtained from Ogun State Water Corporation which has the only weather-monitoring station in the State. Eighteen years were used owing to scarcity of available data which is prevalent in the State and in South-west Nigeria.

Seasonality Index as described by Guhathakurta and Saji (2013) was computed as follows:(1) $$\overline{S}1=\frac{1}{\overline{R}}\sum _{n=1}^{12}\left|X_n-\frac{\overline{R}}{12}\right|$$(1)

where X_n is the mean rainfall of month n; $\overline{R}$ is the mean annual rainfall.

SPI as described by Akinsanola and Ogunjobi (2014) and Adegoke and Sojobi (2015) were computed as follows:(2) $$\text{SPI}=\frac{X-\overline{X}}{\mathit{\sigma }}$$(2)

where X is the rainfall in each particular month, season or year depending on the time scale being used; $\overline{X}$ is the mean rainfall in each particular month, season or year depending on the time scale being used; and σ is the standard deviation of rainfall in each particular month, season or year depending on the time scale being used. The SI classification is shown in Table 2.

Table 2. Seasonality index (SI) (Kanellopoulou, 2002) and annual standardized precipitation index (SPI) (McKee et al., 1993)

Seasonality index		Standardized annual precipitation index
Rainfall regime	SI	Classification	SPI
Very equable	≤ 0.19	Near normal	−0.99 to 0.99
Equable but with a definite wetter season	0.2–0.39	Moderately wet years	1.0 to 1.49
Rather seasonal with a shorter drier season	0.40–0.59	Moderately dry years	−1.0 to −1.49
Seasonal	0.60–0.79	Very wet	1.5–1.99
Markedly seasonal with a long drier season	0.80–0.99	Severely dry years	−1.5 to −1.99
Most rain in three months or less	1.00–1.19	Wet extreme	≥ +2.0
Extreme, almost all rain in one–two months	≥ 1.20	Dry extreme	≤ −2.0

The rainfall data were also subjected to time series analyses on monthly and seasonal scales using statistical tests which include Mann–Kendal test, linear regression, SPI and Hoeffding’s D statistics. Furthermore, the potential of rainwater to meet domestic water demands and storage requirements was also evaluated.

2.4. Trend Analyses of monthly and seasonal rainfall using Mann–Kendall, Linear regression, SPI and Hoeffding’s statistics

Mann–Kendall (M–K) was used to analyse rainfall trend in the study period. For M–K rank statistics, S was computed by replacing the observations x_i’s by their ranks k_i’s such that each term was assigned a number ranging from 1 to n which reflects its magnitude relative to the magnitudes of all the terms. For each element k_i, the number N_i was calculated as the number of k_j terms preceding it, such that k_j > k_i. The parameter t_m, as given by Gebremichael, Ourashi, and Mamo (2014), was calculated as follows:(3) $$t_m=\frac{4\sum _{i=1}^{n-1}{N}_i}{n(n-1)}-1$$(3) (4) $$r_m=\pm r_g\sqrt{\frac{4n+10}{9n(n-1)}}$$(4)

where n is the number of years; r_g is the desired probability point of the normal distribution appropriate to a two-tailed test; and r_m = M–K’s significance test statistics. If t_m lies within the range of ±r_m, then the time series does not contain a significant trend (Kendall, 1975).

Owing to the limitations of Mann–Kendall test in analysing non-monotonic trend, Hoeffding’s D statistic was also used to analyse the rainfall values for residual rainfall. The residual rainfall was obtained by subtracting the predicted rainfall data obtained by from the linear regression equations from the observed values. The probability values for Hoeffding’s D statistic were computed using the equation provided by Blum, Kiefer and Rosenblatt (1961) as follows:(5) $$\frac{(n-1){\mathit{\pi }}^{4}D}{60}+\frac{{\mathit{\pi }}^4}{72}$$(5)

where n is the number of years of data and D is the Hoeffding’s D statistic.

Hoeffding’s D statistic was also used because it is typically used to detect non-linear and non-monotonic associations and has been found to outperform other statistical methods (Fujita et al., 2009).

Hoeffding’s D statistic was obtained utilizing the formula provided by Santos, Takahashi, Naka, and Fujita (2014) as follows:(6) $$D=\frac{(n-2)(n-3)D_1+D_2(n-2)D_3}{n(n-1)(n-2)(n-3)(n-4)}$$(6) (7) $$\text{where}{D}_1={\mathit{\epsilon }}_i\left(Q_i-1\right)\left(Q_i-2\right);D2={\mathit{\epsilon }}_i\left(R_i-1\right)\left(R_i-2\right)\left(S_i-1\right)\left(S_i-2\right);$$(7) (8) $$D3={\mathit{\epsilon }}_i\left(R_i-2\right)\left(S_i-2\right)\left(Q_i-1\right)$$(8)

where R_i is the rank of x_i; $S_i$ is the rank of y_i; and Q_i known as bivariate rank = 1 + number of points with both x and y values < the ith point.

The null hypothesis H₀ of monotonic trend is rejected if P $\{D\}$ > ρ_n where(9) $${\mathit{\rho }}_n=\frac{1}{30}\sqrt{\frac{2\left(n^2+5n-32\right)}{9n\left(n-1\right)\left(n-3\right)\left(n-4\right)\mathit{\alpha }}}$$(9)

where α is the level of significance.

Hoeffding’s measure ρ_n varies from $-1/60\text{to}1/30$ (Santos et al., 2014), which is equivalent to −0.0167–0.033. The acceptable range of α was obtained by inserting the upper and lower limits of ρ_n given above. The acceptable range of α was found to be 1.32–3.87%. The level of significance for our study was 2% which is within the acceptable range.

Furthermore, the annual rainfall was evaluated using Student’s t test. The formula used was given by Bluman (2013) as follows:(10) $$t=\frac{\overline{X}-\mathit{\mu }}{s/\sqrt{n}}$$(10)

where t is the test value, $\overline{X}$ is the mean of observed values, μ is the claimed mean, S is the standard deviation of data and n is the number of years of data. The degrees of freedom (df) of the data = n−1 = 17. The null hypothesis of normal distribution is rejected when t value or p-value of t is > the critical values. The significance level used was 2%.

2.5. Rainwater harvesting potential and storage requirements

Rainwater harvesting potential for our study was calculated using the monthly balance approach. The monthly harvestable rainwater (Q_m) was calculated as a function of the product of mean monthly rainfall $(\overline{R_m})$, roof area (A), percentage of roof area utilized for rainwater harvesting (β) and roof run-off coefficient (C) as given in Equation 5.(11) $$Q_m=\overline{R_m}\times A\times \mathit{\beta }\times C$$(11)

From the literatures, roof area varied from 25 to 200 m² (Biswas & Mandal, 2014; Islam, Islam, & Lacoursiere, 2014; Ndiritu et al., 2011; Otti & Ezenwaji, 2013; Ubuoh et al., 2012; Woltersdorf et al., 2015). Roof size area of 100 m² was adopted for this study as recommended by Woltersdorf et al. (2015) and Sturm et al. (2009) and utilized by Otti and Ezenwaji (2013), while the β value of 0.35 suggested by Shittu et al. (2015) was utilized. This is low because of cost prohibition and poor planning of DRWH systems typical in Nigeria.

Furthermore, roof run-off coefficient (C) varies between 0.75 and 0.95 from the literatures (Fernandes et al., 2015; Roebuck, 2007; Tomaz, 2005; Woltersdorf et al., 2015). C value of 0.8 was adopted for this study as utilized by Otti and Ezenwaji (2013) and Shittu et al. (2015) and accounts for leakage, spillage, infiltration, roof surface wetting and evaporation (Lee et al., 2010) and is within the range of 70–85% of harvestable rainfall suggested by Helmreich and Horn (2009).

Since mean monthly rainfall was utilized, it is imperative to consider the upper and lower confidence limit scenarios beside the mean case scenario owing to rainfall variability and also because mean can hide rainfall variability which occurs in real-life scenarios. Two approaches were utilized in computing the confidence limits, namely: confidence interval about the mean monthly rainfall as well as confidence interval using Coefficient of Variation (COV) of monthly rainfall. For the first approach, the confidence intervals for mean based on maximum error of estimate (MEE) as described by Johnson and Kuby (2012) as well as Bluman (2013) was utilized and was described as:(12) $$\overline{X}+Z(\mathit{\alpha }/2)\left(\frac{\mathit{\sigma }}{\sqrt{n}}\right)=\text{Upper Confidence Limit (UCL)}$$(12) (13) $$\overline{X}-Z(\mathit{\alpha }/2)\left(\frac{\mathit{\sigma }}{\sqrt{n}}\right)=\text{Lower Confidence Limit (LCL)}$$(13)

where $\overline{X}$ = Mean = $\overline{R_m}$; Z (α/2) = Confidence coefficient; $\left(\mathit{\sigma }/\sqrt{n}\right)$ = Standard error of mean and Z (α/2) $\left(\mathit{\sigma }/\sqrt{n}\right)$ = Maximum error of estimate (MEE), σ = Standard deviation of monthly rainfall for each month and n = sample size = 18. The confidence interval adopted in our study was 0.99 which gave a confidence coefficient of 2.58 as shown in Table 8.

Therefore, harvestable rainwater equations for the scenarios of upper confidence limit (UCL) of monthly mean rainfall and lower confidence limit (LCL) of monthly mean rainfall were obtained as:(14) $$Q_{\text{UCL}}=[{\overline{R}}_m+\text{MEE}]\times A\times \mathit{\beta }\times C$$(14) (15) $$Q_{\text{LCL}}=[{\overline{R}}_m-\text{MEE}]\times A\times \mathit{\beta }\times C$$(15)

For the second approach, harvestable rainwater equations for the upper confidence limit (UCL) of monthly mean rainfall and lower confidence limit (LCL) of monthly mean rainfall were obtained as:(16) $$Q_{\text{UCL}}={\overline{R}}_m\times A\times \mathit{\beta }\times C[1+\text{COV}]$$(16) (17) $$Q_{\text{LCL}}={\overline{R}}_m\times A\times \mathit{\beta }\times C[1-\text{COV}]$$(17)

and the results were shown in Table 10.

3. Results and discussion

3.1. Seasonal analyses of rainfall, seasonality index and annual standardized precipitation index

The basic seasonal rainfall and temperature characteristics of the study area had been shown by the climograph displayed in Figure 2 as well as Table 3.

Figure 2. Climograph of seasonal rainfall in Odeda LGA in Ogun State, South-west, Nigeria.

Table 3. Descriptive characteristics of seasonal rainfall in Odeda LGA, Ogun State

Parameters	Spring	Summer	Autumn	Winter
Mean	141.41	214.68	297.44	162.54
Maximum	206.68	353.76	406.86	277.08
Minimum	61.15	71.94	163.85	108.43
SD	45.56	82.67	71.24	48.2
COV (%)	32	39	24	30
Csx	−0.08	−0.19	−0.11	1.17

Based on mean seasonal values, autumn had the highest contribution of rainfall (36.49%) and the least contribution was by spring (17.35%) as shown in Table 3. Contributions by summer and winter were 26.35 and 19.81%, respectively. The maximum seasonal rainfall of 406.86 mm occurred in autumn, while the minimum seasonal rainfall of 61.15 mm occurred in spring. Also, it was also observed that autumn recorded the highest seasonal rainfall throughout the period of study with the exception of year 2000, where summer recorded the highest seasonal rainfall.

The season with the lowest coefficient of variation (COV) of 0.24 was autumn, while summer had the highest COV as shown in Table 3. Based on Hare’s (1983) rainfall variability index (which is COV expressed in percentage terms), rainfall in summer and spring were highly variable with index > 30%, while rainfall in winter and autumn were moderately variable with index between 20 and 30%.

Furthermore, maximum and minimum temperature values were found to be fairly stable across all seasons, with autumn also recording the highest maximum and highest minimum temperature.

The seasonal variation of rainfall for Odeda is described in Figure 3. Autumn recorded the highest seasonal rainfall with the exception of year 2000, in which summer recorded the highest rainfall. Winter recorded the least amount of rainfall for most of the study period. Based on coefficient of slope of linear regression equation of the line graph of seasonal rainfall, summer recorded the highest increasing trend of 6.6965, followed by winter (5.4363), spring (4.5004) and the least by autumn with (1.5942).

Figure 3. Line graph of seasonal variation of rainfall in Odeda LGA, Ogun State, Nigeria.

Seasonality index (SI) also revealed trend in rainfall pattern. Based on SI value of 0.27 obtained for the area for the period of study, the rainfall regime can be described as equable but with a definite wetter season. Trend in rainfall was also analysed using SPI on an annual scale and the result is presented in Figure 4.

Figure 4. SPI for annual rainfall from 1995 to 2012 for the study area.

It was observed that dry years took place between 2001 and 2005, while wet years were experienced between 2006 and 2011 which corroborated results displayed in Figure 3. Similar to what was obtained in Figure 3, severely dry years were experienced in both 2001 and 2005, while extremely wet year was experienced in 2010 based on SPI classification displayed in Table 4.

Table 4. Classification of annual rainfall based on SPI

Classification	Year
Near normal	1995, 1996, 1997, 1998, 1999, 2000, 2002, 2004, 2006, 2007, 2008, 2012
Moderately dry	2003
Moderately wet	2009, 2011
Severely dry	2001, 2005
Extremely wet	2010

Declining trend in rainfall after 2000 was also reported by Wu, Wang, Cai, and Li (2013) and was attributed to low water vapour and higher than normal air temperature (Liu, Luo, Zhang, Wu, & Liu, 2011). The abrupt changes in rainfall were also attributed to changes in regional circulation patterns (Zang & Liu, 2013).

In summary, the SPI graph in Figure 4 indicated an extremely low increasing trend in annual rainfall with a slope of 0.0838 and likewise corroborated the changes observed using SI. Based on SPI classification in Table 2, near normal rainfall took place between 1995 and 2000, moderately wet years were experienced in 2009 and 2011, moderately dry years in 2003, severely dry years in 2001 and 2005 and extremely wet years in 2010.

3.2. Monthly rainfall analyses

Graph of monthly rainfall revealed singular peaks as shown in Figure 5. The existence of a singular monthly rainfall peak contradicts the bimodal monthly peaks reported by Kilanko-Oluwasanya (2009) who reported monthly rainfall peaks in July and August for Abeokuta. Prior to year 2000, September recorded the highest monthly rainfall between 1995 and 1998, while October and August recorded the highest monthly rainfall in 1999 and 2000, respectively. Between 2001 and 2005, October recorded the highest monthly rainfall in 2001, 2003 and 2005, while September and November recorded the highest monthly rainfall in 2002 and 2004.

Figure 5. Graph of monthly rainfall in Odeda LGA, south-west Nigeria.

Between 2006 and 2012, highest monthly rainfall occurred in October in 2006, 2009 and 2010, while August recorded the highest monthly rainfall in 2008 and 2002 and July and September recorded the highest monthly rainfall in 2007 and 2011, respectively. This indicates a progressive shift in maximum rainfall from September in pre-2000 period to October and/or August in post-2000 period.

Table 5 revealed that the months with the highest variability of rainfall were May and June with COV of 41%, marking the beginning of intense rainfall during the rainy season, while the month with the lowest monthly rainfall variability took place in August with COV of 26%, implying that the intense rainfall in August has been reasonably consistent.

Table 5. Descriptive characteristics of monthly rainfall in Odeda, Ogun State, Nigeria.

Parameters	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec
Mean (mm)	45.8	41.3	47.4	52.7	63.4	69.5	81.8	97.9	101.9	97.7	67	48.6
Max (mm)	73.5	65.8	73.7	84.8	107.6	121.7	126.8	140	148.8	150	143.1	76.9
Min (mm)	17	18.2	17	8.2	23.9	23	33	66	56.9	41.4	40.5	31
SD	15.7	12.7	16.8	19.4	26.1	28.8	29.5	25.4	27.7	28.7	26	15.6
COV (%)	34	31	35	37	41	41	36	26	27	29	29	32
Csx	0.38	0.46	−0.15	−0.50	0.12	−0.13	−0.25	−0.26	−0.07	0.04	1.58	0.93

Based on Hare’s rainfall variability index (1993), all the months exhibited high variability with COV (%) > 30% with the exception of August, September, October and November, which exhibited moderate variability between 20 and 30%. This indicates higher increasing rainfall during the dry season months and slightly increasing rainfall during the rainy season months.

Also, yearly variation of monthly rainfall was depicted in Figure 6. The dry season months of November–April (Akinyemi et al., 2013) had an increasing rainfall range of 1.19–2.27 mm/year with a mean of 1.63 mm/year, while the wet season months of May–October witnessed an increasing rainfall range of 0.20–2.28 mm/year with a mean of 1.37 mm/year.

Figure 6. Line graph of monthly rainfall variation for the study period.

This indicated higher increasing rainfall during the dry season compared to the wet season. In addition, comparison of monthly rainfall for the study period as shown in Figure 6 revealed steep decline in monthly rainfall occurred between 2000 and 2005 and then a general trend of increase after 2005. High amount of rainfall which took place in 2000 was also reported by Perry (2006).

M–K test revealed that all the months experienced significant increasing rainfall trend with the exception of August and September as shown in Table 6. For the linear regression equation, positive slope indicates an increasing trend, while negative slope indicates decreasing trend (Tabari, Marofi, Aeini, Talaee, & Mohammadi, 2011). Based on the slope of linear regression as depicted in Table 6, the highest increasing trend of 2.36 mm/year occurred in May, followed by 2.28 mm/year in June and 2.27 mm/year in November.

Table 6. Trend results for Mann–Kendall, linear regression, SPI Tests and Hoeffding’s D statistics for monthly rainfall

	Mann–Kendall				Linear Regression		SPI	Hoeffding’s D Statistics
	N	tm	rm	Trend	Slope	Rcorr	Slope	D	P{D}	ρn
Jan	18	0.18	±0.15	Significant increase	1.55	0.34	0.098	0.1159	4.5517	0.0661
Feb	18	0.27	±0.14	Significant increase	1.19	0.44	0.109	0.0925	3.9058	0.0661
Mar	18	0.31	±0.13	Significant increase	1.57	0.68	0.096	0.0740	3.3952	0.0661
Apr	18	0.31	±0.13	Significant increase	1.74	0.32	0.092	0.0765	3.4642	0.0661
May	18	0.36	±0.12	Significant increase	2.36	0.27	0.093	0.0740	3.3952	0.0661
Jun	18	0.33	±0.13	Significant increase	2.28	0.25	0.085	0.0742	3.4008	0.0661
Jul	18	0.19	±0.15	Significant increase	1.95	024	0.068	0.0878	3.7761	0.0661
Aug	18	0.05	±0.17	Insignificant increase	0.78	0.07	0.032	0.1144	4.5103	0.0661
Sep	18	0.02	±0.17	Insignificant increase	0.20	−0.07	0.001	0.1247	4.7945	0.0661
Oct	18	0.29	±0.13	Significant increase	0.62	−0.09	0.022	0.1258	4.8249	0.0661
Nov	18	0.19	±0.15	Significant increase	2.27	0.15	0.090	0.1001	4.1156	0.0661
Dec	18	0.28	±0.13	Significant increase	1.48	0.35	0.098	0.0885	3.7954	0.0661

Note: R_corr= correlation coefficient.

Based on the M–K test, insignificant increasing trend took place in August and September, although both have slope values of 0.78 and 0.20 mm/year and was also corroborated by their very low SPI values of 0.032 and 0.001, respectively, as displayed in Table 6 and Figure 7. The months with the highest SPI values were February with the highest SPI value of 0.109, followed by January and December (0.98) and March (0.96). In summary, significant increasing trends took place in the dry season months of November, December, January, February and March.

Figure 7. Comparison of monthly SPI from 1995 to 2012.

M–K test revealed that all the seasons exhibited significant increasing rainfall trend with the exception of autumn which exhibited insignificant increasing rainfall trend as shown in Table 7. This was corroborated by the slope of the linear regression which indicated that summer had the highest increasing trend of 6.7 mm/year, followed by winter (5.44 mm/year) and spring (4.50 mm/year), while autumn recorded the least increasing trend of 1.59 mm/year.

Table 7. Trend results for Mann–Kendall, linear regression, SPI Tests and Hoeffding’s D statistics for seasonal rainfall

	Mann–Kendall				Linear regression	SPI	Hoeffding’s D statistics
Seasons	N	tm	rm	Trend	Slope	Slope	D	P{D}	ρn
Spring	18	0.62	± 0.26	Significant increase	4.500	0.102	0.0988	4.0797	0.0661
Summer	18	0.62	± 0.26	Significant increase	6.70	0.083	0.0356	2.3354	0.0661
Autumn	18	0.09	± 0.16	Insignificant increase	1.59	0.023	0.1178	4.604	0.0661
Winter	18	0.35	± 0.13	Significant increase	5.44	0.116	0.0741	3.398	0.0661

Increasing trend in winter rainfall was attributed to North Atlantic Oscillation which causes westerly flows (Perry, 2006). The increasing trend results were also supported by SPI values in Table 7 as well as Figure 8 with autumn recording the lowest SPI value of 0.023 and winter recording the highest SPI value of 0.116 followed by spring with SPI of 0.102.

Figure 8. Comparison of Seasonal SPI from 1995 to 2012.

Figure 9. Seasonal residual rainfall for Odeda LGA from 1995 to 2012.

Further, all the seasons had negative SPI values from 2000 to 2005 as shown in Figure 10, which indicated preponderance of moderately dry seasons but with few episodes of severely dry seasons. SPI was able to detect significant trends in rainfall which corroborated Mahajan and Dodamani (2015) who advocated the use of SPI to detect significant trends in hydrological parameters.

Figure 10. Annual residual rainfall for Odeda LGA from 1995 to 2012

Though M–K test for monthly rainfall showed increase of 0.20–2.36 mm/year and a seasonal increase of 1.59–6.7 mm/season, from water management point of view, the increase is not significant. This was corroborated by Mazvimavi (2010) who reported statistically insignificant seasonal and annual rainfall for Zimbabwe with a COV range of 23–40% similar to COV range of 26–40% and 24–39% obtained in our study for monthly and seasonal rainfall, respectively. The author adduced the general perception of increasing or declining rainfall to the presence of rainfall variability. Likewise, Tapsoba, Haché, Perreault, and Bobée (2004) also reported insignificant changes in rainfall for Togo and Benin located in West Africa.

P {D} values for all the months and seasons were found to be > ρ_n value of 0.0661, therefore we reject the null hypothesis H₀ of monotonic trend and accept the alternative hypothesis of non-monotonic trend. Therefore, it can be inferred that non-monotonic trend was exhibited across all the months and seasons.

Analysis of annual rainfall by the Student’s t test indicated that the t value of 13.05 was > the critical value of 2.718. Therefore, we reject the null hypothesis of normal distribution. Thus, it can be inferred that the annual rainfall trend is non-linear.

3.5. Residual Trend Analysis of seasonal, annual and monthly rainfall

Residual analyses of seasonal rainfall revealed that autumn recorded the least residual rainfall in a dry state between 1995 and 2000, while winter experienced the highest residual seasonal rainfall in a wet state. Likewise, wet states were experienced in summer and spring also between 1995 and 2000 as depicted in Figure 11. A reversal of state was experienced during the period of 2000–2005, where autumn recorded the highest residual rainfall in a wet state, while spring, summer and winter experienced dry states with winter recording the least residual rainfall.

Figure 11. Monthly residual rainfall for Odeda LGA from 1995 to 2012.

A short reversal of state was experienced in 2007, where winter recorded the highest residual rainfall, while a short dry state was experienced between 2008 and 2009. This result depicts an average of five (5)-year periodicity of oscillation between the wet and dry states, indicating non-linearity of the rainfall pattern. This result corroborated the findings of Ibrahim, Balzter, Kaduk, and Tucker (2015) who observed the oscillating pattern of approximately five (5)-year periodicity for rainfall in Sub-Saharan West Africa.

Also, the annual residual rainfall graph displayed also revealed alternate wet and dry states approximately six (6) years which was within the periodicity of 3–7 years reported for Niger Delta, Nigeria, by Ologunorisa and Adejuwon (2003) with significant clyclical pattern as well. The most profound periodicity for the region was five (5) years.

The wet state was predominant between 1995 and 2000, while the dry state was predominant between 2000 and 2006 and there was a reversal to the wet states between 2006 and 2011. This implies a dry state is expected to take place for the next five years, beginning from 2013. This dry state was actually corroborated by World Meteorological Organization (World Meteorological Organization, 2015) which reported a drier-than-normal rainfall in Ogun State in 2013 in line with the periodicity pattern of rainfall observed in our studies.

Residual analysis of monthly rainfall indicated non-linear, non-monotonic trend as well as some periodicity similar to what was obtained in seasonal and annual time scales. They both exhibited alternation between wet and dry states. For most of the months, wet state was observed between 1995 and 2000, dry state between 2000 and 2005 and a reversal wet state between 2005 and 2010. The monthly residual rainfall graph displayed in Figure 11 exhibited similar alternation between the wet and dry states, similar to what was obtained on the seasonal and annual time scales.

Nicholson (2013) identified the factors responsible for inter-annual rainfall variability and were found to include African Easterly Jet (AEJ), Tropical Easterly Jet (TEJ), African Westerly Jet (AWJ) and West African Westerly Jet (WAWJ). According to the author, AEJ is predominant in the months of May–June, before the onset of rain, TEJ was very strong in January–March, AWJ between July and September, while WAWJ was influential in May–September.

Other factors responsible for rainfall variability on monthly, seasonal and annual time scales in the area of study were attributed to non-linear West African Monsoon (WAM) and relief (Barbe, Lebel, & Tapsoba, 2002; Eltahir & Gong, 1996), Indian Ocean SST and anticyclones over NE China (Hastenrath & Wolter, 1992; Quan, Diaz, & Fu, 2003), global inter-hemispheric SST differences (Semazzi, Mehta, & Sud, 1988) and local surface hydrology such as local evapotranspiration which contributes 27% of rainfall in West Africa (Gong & Eltahir, 1996).

Also important are land surface feedback mechanisms such as soil moisture, lowered surface roughness and dust generation (Rowell, Folland, Maskell, & Ward, 1995). From the residual rainfall analyses, it may be implied that different forcing mechanisms operate during different months, seasons and years (Long & Entekhabi, 2000) and may be responsible for the rainfall variability being experienced in the area of study.

3.6. Rainwater harvesting potential and storage requirements

The monthly harvestable rainwater (MHRW) for the three scenarios of UCL, Mean and LCL were displayed in Table 8. For the UCL scenario, maximum MHRW of 3.32 m³ occurred in September, followed by October (3.22 m³) and August (3.17 m³), while the minimum MHRW was in February (1.37 m³). The corresponding values for the mean and LCL scenarios were 2.85, 2.74, 2.74 m³ and 2.38, 2.31 and 2.25 m³, respectively.

Table 8. Monthly harvestable rainwater (MHRW) based on maximum error estimate of $\overline{R_m}$

Limits	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec
QUCL (m^3)	1.55	1.37	1.61	1.81	2.22	2.44	2.79	3.17	3.32	3.22	2.32	1.63
QMEAN (m^3)	1.28	1.16	1.33	1.48	1.78	1.95	2.29	2.74	2.85	2.74	1.88	1.36
QLCL (m^3)	1.01	0.94	1.04	1.14	1.33	1.46	1.78	2.31	2.38	2.25	1.43	1.09

Comparison of the monthly HRW for the different scenarios in Figure 9 revealed that UCL (COV) and UCL (COV) recorded the highest values and lowest values, respectively. The corresponding highest and lowest monthly HRW were 3.62 and 0.80 m³, respectively. Therefore, the recommended maximum storage capacity that should be provided for DRWH is 4 m³, while the minimum storage capacity should be approximately 1 m³ (Table 9).

Table 9. Monthly harvestable rainwater (HRW) based on coefficient of variation limits

Limits	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec
QUCL (m^3)	1.72	1.51	1.79	2.02	2.50	2.74	3.11	3.45	3.62	3.53	2.61	1.80
QLCL (m^3)	0.85	0.80	0.86	0.93	1.05	1.15	1.47	2.03	2.08	1.94	1.14	0.93

In order to estimate monthly water demand per household, there is need to calculate per capita daily water demand. The water demand is separated into two, namely: potable water demand (PWD) and non-potable water demand (NPWD). PWD covers drinking and cooking applications, while NPWD covers bathing, toilet flushing and dishwashing. Sojobi, Dahunsi, and Afolayan (2015) recommended 7.5 lpcd for both drinking and cooking which covered 4.5 lpcd and 3 lpcd recommended by WHO (2004b, 2005) for drinking and cooking, respectively.

A total of 20 lpcd was recommended for non-potable water uses such as bathing, toilet flushing and dishwashing as shown in Table 10 excluding laundry, which is usually done on weekly basis in typical Nigerian settings. Total estimated weekly per capita NPWD was 150 litres. For a 30-day month, the estimated NPWD was 150 × 4 plus additional 40 litres (for two remaining days), which gives 640 litres.

Table 10. Water demand for various applications

Applications	Water demand (lpcd)	Sources
Bathing	6	WHO (2005)
Toilet flushing	10	WHO (2005)
Dishwashing	4	Authors^1
Laundry (weekly)	10	WHO (2005)
Drinking and cooking	7.5	Sojobi et al. (2015)

¹ Gurung and Sharma (2014) recommended a value of 2.5 lpcd which was considered low for a typical Nigerian setting.

Leaving allowance for contingencies of 20%, the estimated per capita monthly NPWD is 768 litres. Allowance for contingencies takes care of unexpected NPWD and PWD from guests and emergencies such as ceremonies. The estimated per capita monthly PWD was estimated to be 270 litres, leaving room for contingencies as well. The estimated weekly per capita water demand of 202.5 lpcd exceeded the weekly minimum water requirements of 140 lpcd recommended by United Nations (UN) based on 20 lpcd for rural communities in developing countries.

Therefore, for a six-member household comprising father, mother and four children, the total estimated monthly NPWD and PWD were 4,608 litres (4.608 m³) and 1620 litres (1.62 m³), respectively. Therefore, total estimated monthly household water demand (HHWD) for a six-member household was 6.228 m³, while total annual HHWD was 74.74 m³.

For the MEE approach, the percentage contributions of total annual water demand that can only be met by DRWH were computed for the three scenarios and shown in Table 11. For total annual NPWD, between 32.84 and 49.64% can be met by DRWH. For total annual PWD, between 93.42 and 141.20% can be met by DRWH.

Table 11. Domestic rainwater harvesting potential (DRHP) based on MEE limits

Scenarios	Total annual HRW (m^3)	% Annual NPWD	% Annual PWD	% Total annual HHWD
UCL	27.45	49.64	141.20	36.73
Mean	22.45	41	115.48	30.04
LCL	18.16	32.84	93.42	24.30

This indicates that for the mean and UCL scenarios, PWD can be sufficiently met with some excess remaining and 93.42% met in the LCL scenario. In addition, this result also revealed that DRWH can only be used to complement the main water supply for the study area when used for non-potable purposes.

For the COV approach, DRWH has the potential to meet 27.51–54.91% of the NPWD, 78.34–156.38% of PWD and between 20.38 and 40.67% of the total annual HHWD as displayed in Table 12.

Table 12. Domestic rainwater harvesting potential (DRHP) based on COV limits

Scenarios	Total annual HRW (m^3)	% Annual NPWD	% Annual PWD	% Total annual HHWD
UCL	30.40	54.91	156.38	40.67
LCL	15.23	27.51	78.34	20.38

4. Conclusions

Rainfall variability has been demonstrated for our study area and it may be attributed to tropical and extra-tropical factors which operate during different months, seasons and years. Rainfall variability in terms of COV ranges from 24 to 39% for the seasons and 26 to 41% for the months. The dry season months (November–April) have been experiencing a mean rainfall increase of 1.63 mm/year with a range of 1.19–2.27 mm/year, while the wet season months of May–October recorded a mean increase of 1.37 mm/year with a range of 0.20–2.28 mm/year. Periodicity of five–six years was observed in the rainfall pattern in our study area which corroborated earlier research findings.

Though the M–K test revealed significant rainfall in most of the months and likewise significant increase in spring and summer, from a water management perspective, the increase was not significant, which corroborates results obtained for some countries with similar rainfall variability. The general perception of increasing or declining rainfall may be attributed to the presence of rainfall variability on monthly and seasonal time scales. Also, Hoeffding’s D statistics revealed prevalence of non-monotonic trend in all the months and seasons.

Taking into account the effects of rainfall variability, the recommended minimum and maximum storage capacity requirements for a six-member household are 1 and 6 m³, respectively.

In addition, based on the maximum error estimate approach, the rainwater harvesting potential for the area of study ranges between 18.16 and 27.45 m³, while based on the coefficient of variation approach, the rainwater harvesting potential ranges between 15.23 and 30.40 m³.

Our results also showed that domestic rainwater harvesting has the potential to meet 27.51–54.91% of non-potable household water demand as well as 78.34–156.38% of household potable water demand for a six-member household.

Domestic rainwater harvesting is highly encouraged as a supplementary water source especially in rural and peri-urban areas to reduce their vulnerability to acute shortage of water infrastructure.

The significance of this research and contribution to the literature is that it has incorporated rainfall variability in the calculation of rainwater harvesting potential and in the calculation of storage water requirements taking into account the effects of rainfall variability which is often neglected in such studies. In addition, as a supplementary source of water, it demonstrated the percentage of domestic water demand that can be met by rainwater harvesting for potable and non-potable purposes. Furthermore, results from this study revealed the periodicity of rainfall pattern which characterizes our study area.

Acknowledgements

Ogun-Osun River Basin Development Authority is appreciated for the supply of the rainfall data used in this study. Likewise, comments from the anonymous reviewers are highly appreciated.

Additional information

Funding

Funding. The authors received no direct funding for this research.

Notes on contributors

Isaac Idowu Balogun

Isaac Idowu Balogun is a senior lecturer in the Geography Department of University of Lagos, Lagos State, Nigeria.

Adebayo Olatunbosun Sojobi

Adebayo Olatunbosun Sojobi is a lecturer in the Department of Civil Engineering of Landmark University, Kwara State, Nigeria.

Bosede Oyegbemijo Oyedepo

Bosede Oyegbemijo Oyedepo was a MSc student in the Department of Geography of University of Lagos.

Our area of research focus revolves around pragmatic solutions to environmental issues such as climate change and their impacts on water resources, environmental and resources management challenges. Rainwater harvesting as supplementary water source is one of the means to reduce the vulnerability of the populace in rural and peri-urban areas to water stress.