Quantifying potable water savings and water quality implications of
decentralised water sourcing options at the SEQ regional scale
Shiroma MaheepalaPrincipal Research Scientist, CSIRO
5 December 2012
Urban Water Security Research Alliance
OUTLINE
• Outline– Background– Aim– Case study– Results– Conclusions
Total water cycle planning and decentralised water sourcing options
• Examining the effectiveness of decentralised water sourcing options is an important aspect of LGA scale TWCM planning
• This is to identify the most sustainable way to achieve– QDC’s 70 kL/year mandatory water savings target– Maximise grid water savings– Minimise environmental impacts
• Decentralised water servicing options– Roof water harvesting– Local stormwater harvesting– Local wastewater recycling
Challenge: decentralised sourcing options• To evaluate the effectiveness of
decentralised sourcing options, it is fundamental to understand – Contributions to grid water savings– Amount of pollutant removal at the
catchment scale• Grid water savings and pollutant
removal rates of decentralised sourcing options can vary spatially depending on: – Storage capacity, inflow and demand
placed on the source• Effectiveness of each scheme should
be evaluated individually, but this is not practical
• Hence, spatial variability is often ignored, but this will introduce errors
Research Questions
• How do we account for the spatial variability of decentralised sourcing options, when quantifying grid water savings and catchment scale pollutant removal?
• How do we assess supply security of the Grid in the presence of decentralised sourcing options?
• Initial focus: roof water harvesting
Methodology
• Stochastic simulation of the storage behaviour of decentralised sourcing options over a 50-year period of climate
• Instead of using average values, use all probable values for storage, catchment inflow, losses and demand
• Use probability distributions derived from observed data to generate probable values
Application of the methodology to roof water harvesting
Observed statistical distributions of tank size, and roof area and losses derived from observed data
Focus areas
• Average annual rainfall: Ipswich 866 mm; Brisbane 1129 mm; Moreton Bay 1313 mm; Gold Coast 1455 mm; Sunshine Coast 1676 mm
Probabilistic representation of water use• Observed data: 2010
Winter SEQ water consumption data from South East Queensland Residential End Use Study, Beal and Stewart (2011), UWSRA Technical Report No. 31
• For Brisbane observed use: 32 to 283 with a mean of 130.4 litres/person/day (excluding 13.3 l/p/d of observed leaks)
• For Brisbane average household occupancy: 2.6 people
Probabilistic representation of water use cont..• For each end use, generated 10,000 probable demand time series:
• Probability for triggering the event derived from the observed diurnal pattern• Probability distributions for volume, frequency of use, flow rate and
duration, derived from the observed data
Brisbane Statistic
Frequency (events per day)
Half flush
Full flush
Tap shower Bath Dishwasher Clothes washer
irrigation
Mean 4.87 4.21 58.70 2.13 0.13 0.55 0.71 0.12
Standard Deviation
3.97 2.68 33.42 1.99 0.28 0.68 0.56 0.19
Skewness 1.67 1.29 1.13 5.11 2.12 1.94 2.93 1.94
Diurnal pattern and end use statistics for Brisbane (data sourced from Beal and Stewart, 2012, UWSRA Technical Report No. 47)
Probabilistic representation of water use: results
Toilet Laundry
Total
Stochastic simulation of storage behaviour: data
• Tank sizes sourced from:• Measured data in SEQ: Biermann et al. (2012):
UWSRA Technical Report No. 66• Home and garden Waterwise Rebate Scheme
(HWRS), provided by Mark Askins of QWC• Connected roof areas sourced from:
• Measured data in SEQ: Biermann et al. (2012): UWSRA Technical Report No. 66
• Tank Losses: literature based values
Stochastic simulation: input probability distributions
0
50
100
150
200
250
300
1 2 3 4 5 6 7 8 9 101112131415161718192021222324252627282930
effective roof area (m^2)
Probability Density Function
Histogram Normal
x26024022020018016014012010080604020
f(x)
0.48
0.44
0.4
0.36
0.32
0.28
0.24
0.2
0.16
0.12
0.08
0.04
0
Fitting effective tank sizes to log-normal distribution
Fitting effective roof areas to normal distribution
Effective tank sizes vary from 2.6 to 30.5 KL
Effective roof areas vary from 25 to 260 square m
Stochastic simulation: time step and number of iterations
Daily simulation overestimates annual yield by 3% and annual overflow by 30% compared to hourly simulation
10,000 iterations is adequate
Tank yield for Moreton Bay: simulation 1962 - 2011
With spatial variability: 43.9 KL/household/year
Without spatial variability: 50.3 KL/household/year (14.6% overestimate)
Tank yield for Brisbane: simulation 1962 - 2011
With spatial variability: 43.4 KL/household/year
Without spatial variability: 50.0 KL/household/year (15% overestimate)
Results for all areas and SEQ
Variable
Cases
Average
Demand
Requested
(kL)
Average
Annual Yield
(kL)
Average
Annual
Overflow (kL)
Average
Annual
Rainfall (mm)
Average
Tank (kL)
Avg. Roof
Area (m2)
Brisbane 62.4 43.4 61.8 1129 4.4 85Moreton Bay 62.5 43.9 67.9 1313 5.5 110Sunshine Coast 64.8 50.3 93.2 1676 5.6 155Ipswich 48.3 34.5 39.2 866 6.7 155Gold Coast 54.7 44.4 77.9 1455 5.6 180SEQ average 58.5 43.3 68.0 1287.8 5.6 137.0
Average
Cases
Average
Demand
Requested
(kL)
Average
Annual Yield
(kL)
Average
Annual
Overflow (kL)
Yield
difference*
Overflow
difference**
Average
Tank (kL)
Avg. Roof
Area (m2)
Brisbane 61.7 50.0 55.1 15% ‐11% 5.6 115.1Moreton Bay 61.7 50.3 61.3 15% ‐10% 5.5 105.1Sunshine Coast 55.6 57.1 86.1 14% ‐8% 5.7 105.7Ipswich 48.2 42.2 31.4 22% ‐20% 7.9 105.1Gold Coast 55.4 49.0 73.1 10% ‐6% 5.7 101.8SEQ average 56.5 49.7 61.4 15% ‐11% 6.1 106.5
*Yield is over-estimated by average case**Overflow is under-estimated by average case
Variable case: stochastic simulation of 10,000 householdsAverage case: 1 household, uses only ‘average’ parameters
Long term expected tank yield vary from 34.5 KL/hh/year in Ipswich to 50.3 KL/hh/year in Sunshine Coast
Long term expected tank yield in Moreton Bay: 43.9 KL/hh/year
Long term expected tank yield in SEQ: 43.3 KL/hh/year
At the SEQ scale, use of average values over-estimated the yield by 15% and under-estimated the overflow by 11%
Comparison of our results with other rainwater tank yield studies in SEQ
• Beal et al. (2012) study based on 2008 water consumption data– 20 kL/hh/y to 95 kL/hh/y with a mean of 50 kL/hh/y
• Chong et al. (2011) study based on 2008 and 2010 consumption data– 25 kL/hh/y to 89 kL/hh/y with a mean of 58 kL/hh/y
• Umapathi et al. (2012) study based detailed monitoring of rainwater use in 20 homes– 40 kL/hh/y
• QWC analysis based on 2011 Brisbane consumption data– 37 kL/hh/y
• Moreton Bay TWCM Plan study, – 57 kL/hh/y
Stochastic simulation of TP, TN and TSS loads: for Brisbane data
TSS TP TNVariable case: overflow load, kg, per house, per year
1.027 0.63 1.101Average case: overflow load, kg, per house, per year
0.856 0.489 0.931Difference, compared to variable case ‐17% ‐22% ‐15%
Assessing regional supply system behaviour in the presence of small-scale sources
• Hypothetical representation of the SEQ supply system to develop the method
• 30 year (1980 – 2010) daily simulation of the regional supply system using eWater CRC’s Source Integrated Modelling System
Impact on the regional supply cont..
• Need to analyse the regional supply system for many different but plausible climate patterns - to account for climate variability and change
• Upscale the time series of supply obtained with stochastic simulation, using kth Nearest Neighbourhood algorithm (eWater CRC)
System Storage behaviour with and without rainwater tanks
without tanks
with tanks
Conclusions• For small-scale sources, storage capacity, inflow and losses
can vary spatially. The demand placed on small-scale sources can also vary spatially. Observed data in SEQ supports this view.
• We examined the effect of not considering the spatial variability for roof water harvesting in SEQ.• Results indicated that the use of average values can over-estimate the
yield by 15%; under-estimate the overflow by 11%; under-estimate TSS, TP and TN loads from the tank by 17%, 22% and 15% respectively.
• Hence, we recommend the use of stochastic simulation to quantify potable water savings and pollutant removal potential of decentralised sources.
Conclusions cont..
• Stochastic simulation applies to other small-scale sources, if there are many small-scale schemes.
• Stochastic simulation/statistical up-scaling/Source IMS has the potential to quantify the grid water supply and catchment pollutant removal potential of small-scale sources. Further work is needed to demonstrate this capability.
• Further work is continuing in Adelaide to examine the optimal mix of water sources for metropolitan Adelaide.• A project funded by the Goyder Research Institute
Urban Water Security Research Alliance
Acknowledgement Co-authors: Esther Coultas and Luis Newmann
Data providers: Cara Beal, Rodney Stewart, Ashok Sharma and Sharon Biermann
Mark Askins, Phillip Chan, Tad Bagdon and Patricia Hurikino of the Queensland Water Commission for providing access to their study, tank data and their valuable advice
www.urbanwateralliance.org.au