Water Journal : Water Journal April 2011
refereed paper water APRIL 2011 141 water supply options demand of internal use tend towards being empty during dry periods and so provide reduced savings at this time. A second point to note is that the 5-h group has very high water savings during periods where srwc is low, yet these savings diminish rapidly when srwc rises to be at a similar level as the 1-e group when srwc approaches 0.7kL/house/day. The 3-elt group has relatively modest water savings at any time, with this saving reducing to insignificance when srwc approaches 0.57kL/house/day. Tank volume Tank volume is an important factor governing both the duration of the tank's water supply (in the absence of additional rain) and the volume of storage available for additional inflows. Both storage duration and storage availability are dependent on supply to, and demand from, the tank. Although we can expect metered water use in individual properties with rainwater tanks to rise as demand rises in the greater community, large tanks should continue to provide supply longer into dry periods and so continue to provide reductions in metered water use during these periods. The effects of srwc, connection type and tank volume on post-tank √mwu are analysed using a 3-way analysis of variance, the results of which are shown in Table 1. For a variable or interaction to be significant Pr (>F) must be less than 0.05; thus tank volume is an insignificant factor (F=0.0003, Pr(>F) =0.9866) on √mwu when considered in isolation, but is significant in its interactions with other factors. The coefficients column shows the direction and magnitude of each factor/interaction on √mwu post-tank installation. A visual representation of the statistical relationships given in Table 1 are shown in Figure 13a-d. The relationship between tank volume and metered water use shown in Figure 13a is somewhat surprising in that properties with large tanks appear to use more metered water than properties with small tanks during high demand periods. This relationship holds true for 1-e and 3-elt properties; however, it is reversed for 5-h properties, where mwu is less during high demand periods for properties with large tanks (Figure 13d). One possible explanation for these relationships is that high water users choose larger tanks and so empty their tanks quicker than low-demand households with smaller tanks. However, this theory is not supported by a linear regression of tank-volume to pre-tank √mwu which shows no correlation between the variables. A second possible Table 1: F values, probability of insignificance (Pr(>F)) and coefficients for the statistical model (with interaction) of post-tank installation √mwu by connections*srwc*tank volume using Type I (adjusted) sums of squares. Independent variable F value Pr (>F) Coefficient Connections 156.4 <2.2e-16 -0.130 srwc 98.1 <2.2e-16 -0.099 Tank volume 0.0003 0.987 -0.044 Connections: srwc 25.1 5.849e-07 +0.194 Connections: tank volume 46.7 9.617e-12 +0.005 srwc: tank volume 55.6 1.122e-13 +0.094 Connections: srwc: tank volume 6.11 0.0135 -0.013 Figure 12: mwu and 95% confidence bands pre- and post-tank installation relative to the community's single residential water consumption (srwc) for: external-only (1-e); external, laundry, toilet (3-elt); and whole house (5-h) connection types. The water saving lines and confidence bands are the difference between the pre-tank and post-tank lines and confidence bands.
Water Journal March 2011
Water Journal May 2011