australia; impact of rainwater tank and on-site detention options on stormwater management in the...
TRANSCRIPT
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
1/64
Impact of Rainwater Tank and On-siteDetention Options on Stormwater
Management in the Upper ParramattaRiver Catchment
Peter Coombes, Andrew Frost and George Kuczera
Department of Civil, Surveying and Environmental Engineering
University of Newcastle
on behalf of
TUNRA
University of Newcastle, Callaghan, NSW, 2308, Australia
For
Upper Parramatta River Catchment Trust
August 2001
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
2/64
SUMMARY
The principal objective of this study is to determine by how much do rainwater tanks reducethe amount of on-site stormwater detention (OSD) storage required to satisfy Upper
Parramatta River Catchment Trusts (UPRCT) policy. In pursuit of this objective four taskswere performed:
Calibrate the DRIP point rainfall model to a pluviograph record at Parramatta.
Using DRIP generate a synthetic 1000-year pluviograph record representative of theUpper Parramatta River (UPR) catchment.
Modify an existing allotment water balance model to include OSD storage.
Evaluate using continuous simulation the performance of a range of rainwater tank and
on-site detention options for several allotment scenarios over a 1000-year syntheticpluviograph record.
Drip rainfall model calibration and validation
The DRIP point rainfall model underpins this study because the rainfall regime is the primaryfactor controlling OSD outcomes. Hence its calibration and validation are key tasks. Severalkey findings arose from the calibration and validation of DRIP:
1. It was originally envisaged that DRIP would be directly calibrated to a long pluviographat Parramatta. However, owing to the very short pluviograph record at Parramatta this
approach was deemed infeasible and an alternative approach was developed.
2. DRIP was directly calibrated to the 53-year Ryde Pumping Station pluviograph record.The Ryde record was the longest available record for gauges located in and near the UPRcatchment. The calibrated model was used to simulate statistics not used in thecalibration. Such statistics ranged from annual rainfall distributions to IFD curves. For allthe validation statistics considered, DRIP simulations were found to be statisticallyconsistent with the observed statistics. This result engendered confidence in DRIPsability to simulate the entire rainfall regime from very short to annual timescales.
3. It was found at Ryde that both the DRIP and observed IFD curves produced short
duration 100-year intensities about 25% greater than those predicted by AR&R.Examination of observed IFD curves for Guildford showed similar underestimation byAR&R, whereas for Bankstown AR&R IFD curves unsatisfactorily reproduced theoverall shape of the observed IFD curves but did manage to reproduce the right tailsbetter than at Ryde or Guildford. It is difficult to avoid the conclusion that the AR&RIFD curves are in significant error and need revision.
4. The Ryde pluviograph can be used as a master site to transfer DRIP to other shorterpluviograph records within the UPR catchment. The justification for use of the Ryderecord as the master site in future work is that it has a similar annual rainfall distributionas Parramatta, is similarly distant from the coast, and has AR&R IFD statistics that only
marginally differ across a range of sites in the UPR catchment.
ii
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
3/64
OSD and rainwater tank performance
Using an allotment water balance model purposely modified for this study, 1000 years of
continuous simulation using the synthetic Ryde record was undertaken for four case studies atthe allotment scale: Single dwelling; duplex; townhouse and apartment developments.
For each case study the performance of the UPRCT OSD policy along with various rainwatertank scenarios with and without detention storage was considered.
Table A summarizes the performance of UPRCTs OSD policy for different allotmentscenarios.
Table A. Performance of UPRCT OSD policy.
Allotmentscenario
Imperviousfraction (%)
Totalarea(m2)
UPRCTOSD
storage(m3)
PSDL/s
ARI atwhichPSD is
exceeded
OSD storage forPSD to be exceeded
at 100 yr ARI(m3)
Single dwelling 58 600 28.2 4.8 63 55
Duplex 83 600 28.2 4.8 12 67
Townhouse 75 1858 87.3 14.9 22 165
Apartment 67 1200 56.3 9.6 15 119
Three important findings are noted:
1. The PSD is exceeded for ARIs well below 100 years. An exceedance was defined as an
overflow event in which the volume of stormwater was greater than 2 mm times theallotment area. A corollary of this is that the OSD storage requirement to achievecompliance with the 100 year PSD is almost double that of the UPRCTs current OSDrequirement.
It is noted that these results are highly sensitive to the choice of time of concentration. Inthis study a time of concentration of 2 minutes was adopted to be consistent with theexperimental observations of Stephens and Kuczera (1999). If the widely used time ofconcentration of 5 minutes were adopted the complying OSD volume reduces from 55 m3to 29 m3 for a single dwelling allotment. However, the authors cannot find experimentalevidence in support of the 5 minute value and therefore did not consider it.
2. The PSD depends on the allotment type and impervious fraction. It is suggested the OSDpolicy discriminate according to allotment type and its impervious fraction.
The effectiveness of rainwater tanks as a stormwater management measure was found toincrease with housing density. As the proportion of the allotment area (roofs) contributing tothe rainwater tank increased the peak allotment discharge for a given ARI decreased furtherbelow the peak discharge for an allotment with no OSD or rainwater tanks.
Rainwater tanks used to supply in-house uses were found to have storage volumes availablefor stormwater retention at the beginning of over 90% of annual maximum storm events. In
combination with a policy to manage or limit directly connected impervious areas rainwater
iii
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
4/64
tanks could produce similar stormwater management performance as the current UPRCTOSD policy.
The average percentage of rainwater tank volume that can be counted as OSD site storage ispresented in Table B for each allotment scenario.
Table B. Average percentage of rainwater tank volume that can be counted as OSD sitestorage
Volume of rainwater tank counting as OSD storage (%)
Scenario No airspace in tank 50% airspace in tank
Allotment 42 65
Duplex 50 72
Townhouses 40 53
Walk up apartments 32 51
The rainwater tank scenarios in which no air space was provided for stormwater detention
demonstrated a reduction in required OSD storage volume equivalent to about 41% of therainwater tank capacity. In contrast, the rainwater tank scenarios with air space for stormwaterdetention demonstrated a reduction in required OSD storage volume equivalent to 60% of therainwater tank capacity.
The averages presented in Table B are indicative of the rainwater tank contribution to OSDstorage. The actual OSD contribution of rainwater tanks depends on the tank volume, theinclusion of airspace and the allotment type. It is recommended that the site-specific valuespresented in Tables 12, 17, 22 and 27 be used.
The true benefits of rainwater tanks for stormwater management may be obscured by
focussing on peak discharges at the allotment scale. Rainwater tanks reduce volumes ofstormwater discharged into the larger catchment, whereas OSD tanks merely detain thestormwater. It is recommended that the UPRCT undertake a study to analyse the stormwaterperformance of catchments in which OSD and rainwater tanks are distributed according totheir actual location within the catchment.
iv
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
5/64
TABLE OF CONTENTS
1. Introduction ........................................................................................................................ 1
2. Generation of a Synthetic Pluviograph Record..................................................................2
2.1 Introduction ................................................................................................................2
2.2 The DRIP Model ........................................................................................................ 2
2.3 DRIP Calibration To Parramatta: Regionalisation Approach .................................... 3
2.3.1 DRIP data requirements ..................................................................................... 3
2.3.2 Choice of sites .................................................................................................... 4
2.3.3 Sampling variability of validation statistics ....................................................... 4
2.3.4 Master site Observatory Hill Sydney .............................................................. 4
2.4.5 Target Site - Parramatta/Parramatta North......................................................... 6
2.5 DRIP Calibration to Ryde Pumping Station Pluviograph ..........................................9
2.5.1 Calibration and Validation ................................................................................. 9
2.5.2 Comparison with AR&R IFD .......................................................................... 13
2.5.3 Comparison of DRIP Storm Statistics.............................................................. 14
2.6 Discussion ................................................................................................................ 16
2.6.1 Adequacy of AR&R IFD curves ...................................................................... 16
2.6.2 Application of DRIP to other sites within the UPC catchment........................ 18
2.7 Conclusion................................................................................................................ 19
3. Design of Rainwater Tanks .............................................................................................. 21
3.1 Australian Standards ................................................................................................ 21
3.2 NSW Department of Health ..................................................................................... 21
3.3 Water Authorities ..................................................................................................... 21
3.4 Local Councils.......................................................................................................... 22
3.5 Design Details .......................................................................................................... 22
4. Cost Models...................................................................................................................... 24
v
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
6/64
4.1 Installation of a Rainwater Tank .............................................................................. 24
4.2 Installation of an OSD Tank..................................................................................... 24
5. The Allotment Water Balance Model .............................................................................. 26
5.2 Indoor Water Use ..................................................................................................... 28
5.3 Pluviograph Rainfall Generation.............................................................................. 29
5.4 First Flush Separation............................................................................................... 30
5.5 The Rainwater Tank ................................................................................................. 31
5.6 The OSD Tank ......................................................................................................... 31
5.7 The Infiltration Trench ............................................................................................. 32
5.8 Pervious Area ........................................................................................................... 32
5.9 Impervious Area....................................................................................................... 33
6. Allotment Case Studies .................................................................................................... 35
6.1 Single Dwelling Study ............................................................................................. 36
6.1.1 Stormwater impacts.......................................................................................... 36
6.1.2 Water supply impacts ....................................................................................... 40
6.1.3 Costs ................................................................................................................. 41
6.1.4 Summary .......................................................................................................... 41
6.2 Duplex Study............................................................................................................ 42
6.2.1 Stormwater impact ........................................................................................... 42
6.2.2 Water supply impacts ....................................................................................... 45
6.2.3 Costs ................................................................................................................. 45
6.2.4 Summary .......................................................................................................... 45
6.3 Townhouse Development Study .............................................................................. 46
6.3.1 Stormwater impact ........................................................................................... 46
6.3.2 Water supply impacts ....................................................................................... 49
6.3.3 Costs ................................................................................................................. 49
6.3.4 Summary .......................................................................................................... 49
vi
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
7/64
6.4 Three-Storey Walk-Up Apartments ......................................................................... 50
6.4.1 Stormwater impact ........................................................................................... 50
6.4.2 Water supply impacts ....................................................................................... 53
6.4.3 Costs ................................................................................................................. 53
6.4.4 Summary .......................................................................................................... 54
6.5 Case Study Conclusions ........................................................................................... 54
7. References ........................................................................................................................ 56
vii
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
8/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
1. INTRODUCTION
The Upper Parramatta River Catchment Trust (UPRCT) commissioned the authors to
determine by how much do rainwater tanks reduce the amount of on-site stormwater detention(OSD) storage required to satisfy UPRCTs policy. UPRCT required that continuoussimulation be used to evaluate the performance of the rainwater tanks. The rationale for usingcontinuous simulation arises from the intrinsic limitation of the design storm approach tospecify initial conditions. It is now accepted that a design storm typically represents a burst ofextreme rainfall embedded in a longer storm event. The pre-burst rainfall may significantlyaffect the performance of the rainwater tank during the design burst. Only continuoussimulation can rationally simulate these complex dependencies.
In pursuit of this objective UPRCT requested four tasks be performed with the aid of modelscurrently under development in the hydrology research program at the Department of Civil,
Surveying and Environmental Engineering at the University of Newcastle:
Calibrate the DRIP point rainfall model to a pluviograph record at Parramatta.
Using DRIP generate a synthetic 1000-year pluviograph record representative of theUpper Parramatta River (UPR) catchment.
Modify the allotment water balance model to include OSD storage.
Evaluate the performance of a range of rainwater tank and on-site detention options forseveral allotment scenarios using the 1000-year synthetic pluviograph record.
This report is organised as follows:
Section 2 describes the application of the DRIP point rainfall model. Owing to aninadequate pluviograph record at Parramatta, a revised methodology is implemented.Validation using statistics not used in the calibration is used to check the performance ofthe calibrated DRIP model. DRIP intensity-frequency-duration (IFD) curves are comparedagainst Australian Rainfall and Runoff (AR&R) and observed IFD curves. A discussion ofthe results is presented followed by recommendations.
Section 3 overviews design requirements for rainwater tanks emphasising the three zones
for peaks mains demand reduction, rainwater storage and stormwater detention. Section 4describes the cost models used to evaluate the economic performance of the variousoptions.
Section 5 provides an overview of the allotment water balance model.
Section 6 presents the results of 1000 years of continuous simulation using the allotmentwater balance model for four case studies at the allotment scale: Single dwelling; duplex;townhouse and apartment developments. For each case study the performance of theUPRCT OSD policy is documented. In addition various rainwater tank scenarios with andwithout detention storage are considered. Combinations of OSD and rainwater tanks
which ensure UPRCTs PSD is only exceeded with an ARI of 100 years are presentedalong with an economic analysis.
Final 26/06/2003 7:58 AM 1
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
9/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
2. GENERATION OF A SYNTHETIC PLUVIOGRAPH RECORD
2.1 Introduction
A 1000-year synthetic pluviograph series was generated for two raingauge locations in theUPR catchment using the event-based rainfall model DRIP. UPRCT nominated theParramatta gauge as representative of the study region. However, on account, of majorlimitations in the pluviograph record at Parramatta, the calibration approach was revised. Twoindependent approaches were implemented:
1. DRIP was calibrated using pluviograph data from Observatory Hill Sydney and acomposite daily record based on the Parramatta and Parramatta North gauges.
2. DRIP was calibrated using pluviograph data from the Ryde Pumping Station gauge.
In both cases, the synthetic series were validated using a variety of rainfall statistics and werecompared against Australian Rainfall and Runoff (AR&R) IFD curves. Following acomparison of the results a recommendation is made with regard to choice of calibratedmodel and extension of DRIP to other parts of the UPR catchment.
2.2 The DRIP Model
DRIP (Disaggregated Rectangular Intensity Pulse) is a stochastic rainfall simulation packagecurrently under development at the University of Newcastle and the University of Adelaide.The DRIP model is event-based and is capable of representing the inter-event time, storm
duration, average event intensity and the within-storm temporal characteristics of pointrainfall. It can be used to simulate long sequences of rainfall events at time-scales down toless than 6 minutes. DRIP is able to satisfactorily reproduce rainfall statistics important inurban design given a long length pluviograph. A full description of DRIP can be found inHeneker et al. [2001].
The current version of DRIP incorporates a hidden state Markov model to simulate theoccurrence of dry and wet climate states. Frost et al. [2000] show that storm characteristicsare different between the dry and wet climate states. They demonstrate that inclusion of ahidden state Markov model is necessary to be able to reproduce annual rainfall statistics.
The preferred method for calibrating DRIP is to use a long-term pluviograph record at the siteof interest. Unfortunately long-term pluviograph records often are not available.
In response to this problem, research currently underway at the Universities of Newcastle andAdelaide is investigating ways of regionalising DRIP. The most promising approach involvestransfer of information from a long-term master pluviograph site to the target site. Twotechniques have been developed and are currently being assessed for different Australianclimate zones:
1. Transfer from the master site to a short (10 to 20 yrs) pluviograph record at the target site.
2. Transfer from the master site to a medium length (approx 50yrs) daily rainfall record atthe target site.
Final 26/06/2003 7:58 AM 2
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
10/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
2.3 DRIP Calibration To Parramatta: Regionalisation Approach
Owing to the short pluviograph record at Parramatta, direct calibration of DRIP was notpossible. The only possible calibration strategy involves transfer from a master pluviographsite.
2.3.1 DRIP data requirements
DRIP calibration using the regionalisation approach involves two steps:
1. A long master pluviograph record at a site near the desired target site is used to calibrateDRIP.
2. A target site rainfall gauge is then used to calibrate scaling factors that scale the stormcharacteristics at the master site to reproduce those at the target site.
A schematic of this calibration process is shown below in Figure 1.
1. Calibrate DRIP Parameters
2. Calibrate scaling parameters
Master Site
Long PluvioTarget Site
Short Pluvioor
Long Daily
Figure 1. Drip calibration schematic
The scaling is defined as follows: Let xM and xT be random variables corresponding to thesame rainfall characteristic at the master and target sites respectively. The rainfallcharacteristic may be storm duration, dry spell or average intensity conditioned on duration.
The scaling factor k
for a month k is defined as
xT = kxM (1)
The probability density function of xT is given by
=
k
TM
k
TT
xp
1)x(p (2)
where pM( ) is the density of the master random variable.
Final 26/06/2003 7:58 AM 3
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
11/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
2.3.2 Choice of sites
The master site chosen for this study was Observatory Hill Sydney (066062). This site waschosen due to the pluviographs long length and its close proximity to the target Parramattasite. Indeed Observatory Hill Sydney has the longest pluviograph record in the Sydney region.
This is important as there have been distinct dry and wet climate epochs. A short record mayonly sample climate variability from one climate state and hence bias the long-term extremerainfall distributions.
At the target site, it would be ideal if there existed a pluviograph of sufficient length to beable to calibrate the scaling factors. However, the longest Bureau of Meteorology pluviographwithin the Parramatta area (Parramatta North) contained less than three years of data. Thiswas considered insufficient to calibrate the DRIP scaling parameters. Therefore daily rainfallrecords were used to calibrate the scaling parameters.
The use of daily records at the target site only allows calibration of the intensity and dry spell
scaling factors. At Observatory Hill Sydney the average storm duration is of the order of 4hours. As a result it is expected that the daily record contains virtually no information aboutstorm durations. At this stage in the development of the DRIP regionalisation it is necessaryto assume that the storm duration probability distribution is the same at the master and target
sites; that is, the scaling factor for storm duration k=1 for all months.
Parramatta and Parramatta North daily rainfall records were the two obvious choices for thetarget site data. The Parramatta (066046) and the Parramatta North (066124) sites havecontinuous daily rainfall records spanning from 1909-1960 and 1965-1998 respectively. Dueto the proximity of the sites, it was assumed that the daily distribution of rainfall was the sameat both sites. Hence, the two records were added to one another to produce a single augmented
Parramatta daily rainfall record to be used as the target site rainfall.
2.3.3 Sampling variability of validation statistics
Repeated simulation using the calibrated parameters was undertaken to quantify samplingvariability for validation statistics. This is essential if validation statistics are to bemeaningfully compared against DRIP statistics. DRIP simulated a record with the samelength as the observed record 1000 times. For each replicate statistics of interest wereextracted. The 1000 statistics were then ranked to obtain the median and 90% confidencelimits. If the calibrated DRIP model is the correct model of the rainfall process, then there is a90% chance that the observed statistic will fall within the simulated 90% confidence limits. In
fact the 90% confidence limits should be wider than reported because uncertainty in the DRIPcalibrate parameters is not accounted for in the current version.
In this report the DRIP model is judged not to be inconsistent with the observed data if theobserved statistics lie within the 90% confidence limits. Note that because the DRIP modelhas not been calibrated to the validation statistics, median values do not necessarily follow theobserved data.
2.3.4 Master site Observatory Hill Sydney
The DRIP model was calibrated to the 87 years of pluviograph data available at Observatory
Hill. Several validation plots for DRIP simulation at the master site are shown below inFigures 2,3 and 4.
Final 26/06/2003 7:58 AM 4
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
12/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
Figure 2 shows that observed annual rainfall is reproduced well by the DRIP simulation.More importantly for this study, Figure 3 shows that short timescale aggregation statisticssuch as daily and hourly means and standard deviation are also reproduced by DRIPsimulation quite well. Figure 4 compares observed and DRIP-simulated IFD curves. For theshorter durations a close match is found. However, for the 72-hour duration, the DRIP
simulation produces a downward bias. This was considered unimportant for the current studybecause 72-hour durations are much greater than catchment response times - significantflooding is triggered by storm events lasting from around 15 minutes to 6 hours. Overall theDRIP simulation was considered to accurately reproduce the rainfall statistics apparent at themaster site.
0
500
1000
1500
2000
2500
3000
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.9
Observed
Simulated Median
Sim 90% Confidence Limit
AnnualRainfallTotal(mm)
Percent
Figure 2. Sydney observed annual rainfall versus DRIP simulation
0
0.1
0.2
0.3
0.4
0.5
0
0.5
1
1.5
J F M A M J J A S O N D
Sydney Hourly A ggregated Statistics
Mean Observed
90% Confidence Limit
Mean Simulated
Std Dev Observed
90% Confidence Limit
Std Dev Simulated
MeanRainfall(mm)
Standard
D
eviationRainfall(mm)
Month
0
4
8
12
16
J F M A M J J A S O N D
Sy dne y Da i ly Aggr e ga te d S ta t is t ics
Mean Observed
Mean Simulated
Std Dev Observed
Std Dev Sim ulated
Sim 90% C onf idence Limit
Rainfall(m
m)
Month
Figure 3. Sydney observed versus simulated (a) hourly and (b) daily mean and standard
deviation of rainfall.
Final 26/06/2003 7:58 AM 5
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
13/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
0.1
1
10
100
1000
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.99
1.00hr Obs
12.00hr Obs
72.00hr Obs
Median Simulated
Sim 90% Co nfidence Limit
Intensity(mm/hr)
Percent
Figure 4. Sydney observed versus simulated intensity-frequency-duration curves
2.4.5 Target Site - Parramatta/Parramatta North
DRIP scaling parameters for dry spell and conditional average intensity were calibrated using
the augmented Parramatta record. The scaling parameters produced were then used in a
repeated DRIP simulation. Figure 5 shows a comparison between the simulated and observed
annual rainfall totals.
Although the observed annual rainfall values do not lie on the simulated median curve, the
majority of the observed values are within the 90% confidence limits for the model
simulation. It is, therefore, concluded that DRIP adequately simulates the annual rainfall
distribution at Parramatta..
Figures 2 and 5 highlight differences between the Observatory Hill Sydney and Parramatta
locations. The Parramatta gauge has a median annual rainfall of the order of 950 mm whereas
Observatory Hill Sydney has an annual median of about 1200 mm. This difference,
attributable to coastal effects, forces one to question the assumption that the storm duration
probability distributions are identical at both sites.
As there is no pluviograph of sufficient length in the Parramatta area, comparison of DRIP
simulated statistics with observed variables with a timescale less than 24 hours was
impossible. However, observed daily statistics could be calculated using the daily rainfall
records.
Figure 6 compares the simulated daily means and standard deviations against those calculated
from the augmented Parramatta daily record. The DRIP simulated median matches the
observed daily mean for the majority of months. From September onwards DRIP slightly
overestimates the daily mean, which can explain why the simulated annual rainfall is slightly
Final 26/06/2003 7:58 AM 6
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
14/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
overestimated in Figure 5. The simulated standard deviation of daily rainfall is not as accurateas the daily mean, with some observed values lying outside the confidence limits. Figure 7shows a close match between the observed and simulated probabilities of a dry day.
0
500
1000
1500
2000
2500
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.99
Observed Parra/Parra Nth
Median Simulated
Sim 90% Confidence Limits
AnnualRainfallTotal(mm)
Percent
Figure 5. Parramatta observed versus simulated annual rainfall.
0
2
4
6
8
10
12
14
16Parramatta Daily Aggregated Statistics
Observed Target Daily MeanMedian Mean Daily RainObserved Target Daily StddevMedian Daily Std Dev90% Confidence Limit
J F M A M J J A S O N D
Rainfall(mm)
Month
Figure 6. Parramatta observed vs simulated daily rainfall mean and standard deviation.
As mentioned previously, due to the lack of pluviograph data in the Parramatta area, observedstatistics based on timescales less than 24 hours cannot be calculated. This presents a problem
in validating the DRIP results on the timescales that are most important to this study, around15 minutes to 6 hours. However, design IFD curves calculated using the methods described in
Final 26/06/2003 7:58 AM 7
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
15/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
Australian Rainfall and Runoff (AR&R) can be used to provide a check against simulatedvalues. Figure 8 shows a comparison between the simulated and AR&R IFD curves for arange of durations. The observed 24 hour IFD curve was also calculated using the daily data.
50
55
60
65
70
75
80
J F M A M J J A S O N D
Parramatta Dry Probability
Observed Target Daily Dry prob90% Confidence Limit
Median Daily Dry Prob
DryProbability(%)
Month
Figure 7. Parramatta observed vs simulated probability of a dry day.
1
10
100
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.99
30min AR&R1hrs AR&R3hrs AR&R12hrs AR&R
24hrs AR&R90% Confidence LimitsMedian SimulatedObserved Corrected Daily
In
tensity(mm/hr)
Percent
ARI = 1yr
10 yrs
20 yrs
50 yrs
100 yrs
200 yrs
500 yrs
5 yrs
2 yrs
Figure 8. Parramatta observed vs simulated vs AR&R IFD curves.
Final 26/06/2003 7:58 AM 8
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
16/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
The AR&R and observed 24-hour IFD curves show good agreement. It is noted that theobserved IFD produces higher extreme intensities than does the AR&R curve. However,these differences are judged not be significant. Good agreement between the observed and theupper half of the DRIP 24-hour IFD curve is noted. DRIP overestimates the 24-hourintensities for more frequent events with an Average Recurrence Interval (ARI) less than 2
years with the observed curve lying just below the lower 90% confidence limit.
For durations less than 24 hours there are no adequate pluviograph data to check the DRIPIFD curves. As the ARI increases beyond 2 years the DRIP IFD curves consistently producehigher intensities than do the AR&R curves. For example the 100-year 30-minute intensitypredicted by AR&R is 100mm/hr whereas DRIP predicts 150mm/hr. It is difficult to ascertainwhich is closer to the truth. However, three points deserve to be made:
1. For durations less than 12 hours, the DRIP IFD curves closely matched the data fromObservatory Hill Sydney. It is stressed that DRIP is not calibrated to IFD data but ratherindividual storm event data. This engenders confidence in DRIPs ability to simulate
extreme rainfall sequences.
2. The AR&R maps for 2 and 50-year short duration intensities are based on limitedpluviograph data and reflect the exercise of judgement by experiencedhydrometeorolgists. Bias in such circumstances is inevitable.
3. The DRIP IFD curves show weak positive skew for Parramatta, whereas the AR&Rcurves assume zero skew. At high ARIs such differences in skew exacerbate differences.Interestingly the DRIP 24-hour IFD shows virtually zero skew which is consistent withthe 24-hour data. However, at smaller timescales, DRIP simulates skewed IFD curves.The AR&R procedure for deriving IFD curves assumes that one skewness applies to all
durations. This assumption may have been necessary at the time of publishing theAR&R maps because of limited pluviograph data. However, such expediency does notassure the assumption is correct.
2.5 DRIP Calibration to Ryde Pumping Station Pluviograph
In view of doubt about the representativeness of the master site at Observatory Hill Sydney, itwas decided to directly calibrate DRIP to a medium length pluviograph record considered tobe more representative of the Parramatta site. In the search for a suitable site UPRCTprovided the list of pluviograph sites, presented in Table 1, located within 10 km ofToongabbie, the centre point of the UPR catchment. From this list Ryde Pumping Station
gauge was chosen for two reasons:
1. It had the longest record, namely 53 years.
2. Despite the fact that it is located outside the UPR catchment in a region with higherrainfall intensities, it is located sufficiently far from the coast to have annual rainfallstatistics similar to those at Parramatta.
2.5.1 Calibration and Validation
The DRIP model was calibrated to the 53 years of pluviograph data available at Ryde
pumping station. This gauge is operated by Sydney Water. The data was provided by
Final 26/06/2003 7:58 AM 9
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
17/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
Australian Water Technologies. Several validation plots for DRIP simulation at the mastersite are shown below in Figures 9 to 11.
Table 1. List of pluviograph located within 10 km of Toongabbie.
Station Pluviograph Name Years of record as at Oct 20'00
566037 Ryde WPS 52.0567079 Guildford (Pipehead) 29.3
566036 Potts Hill Reservoir 18.7
567092 South Prospect 17.7
567076 Castle Hill Stp 17.0
567104 Northmead Bowling Club 10.5
567111 Westmead Hospital 10.2
566081 Carlingford Bowling Club 10.0
567112 North Parramatta (Burnside Homes) 10.0
567147 Baulkham Hills Swimming Pool 10.0
567148 Kings Langley (Nsw Soccer Federation) 10.0567151 Toongabbie Bowling Club 10.0
566086 Parramatta (Masonic Club) 9.9
567106 Rouse Hill (Api Country Club Kellyville) 9.9
566084 Quakers Hill Stp 9.9
567084 Quakers Hill Stp 9.8
567113 Blacktown (Ashlar Golf Club) 9.8
567146 Greystanes (Cumberland Golf Club) 9.7
567149 Cumberland State Forest (Ibm) 9.7
567150 Blacktown (Dog Pound) 9.7
567152 Merrylands West (Canal Road) 9.4567153 Minchinbury (Pinegrove Memorial Park ) 9.1
567162 Lalor Park (Vardys Rd) 7.8
567083 Prospect Reservoir 7.7
567167 Schofields (Fyfe Rd) 7.2
567075 Blacktown Survey Depot 3.6
567097 Kellyville Stp 2.2
Figure 9 shows that observed annual rainfall is reproduced well by DRIP simulation. It isnoted that the annual distributions for Ryde and Parramatta are similar.
More importantly for this study, Figure 10 shows that short timescale aggregation statisticssuch as daily and hourly means and standard deviation are satisfactorily reproduced by theDRIP simulation. Figure 11 shows that the probability of observing a dry hour or day ismatched closely by simulated values.
Although the simulated aggregation statistics match the observed values well, extremerainfalls on timescales from around 15 minutes to 6 hours are of most importance to thisstudy. Therefore, the simulated IFD curves for a range of durations were compared to thoseobserved in Figure 12.
For the durations shown, simulated IFD curves from 30 min to 3 hours match the observed
values well. For durations greater than 12 hours, the simulated median values underestimatethose observed for long recurrence interval periods. Nonetheless the observed IFDs lie within
Final 26/06/2003 7:58 AM 10
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
18/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
the simulation 90% confidence limits suggesting that sampling variability can account for theobserved discrepancies. It is stressed that DRIP is not calibrated to IFD data but ratherindividual storm event data. This engenders confidence in DRIPs ability to simulate extremerainfall sequences.
0
500
1000
1500
2000
2500
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.99
Ryde Pumping Station Annual Rainfall
Observed
90% Confidence Limits
Median Simulated
90% Confidence Limits
AnnualRainfall(mm)
Percent
Figure 9. Ryde observed annual rainfall versus DRIP Simulation
0
0.2
0.4
0.6
0.8
1
1.2
1.4Ryde Hourly Aggregated Statistics
Obs Hourly Mean
Obs Hourly Stddev
Median Simulated
90% Confidence Limit
1 2 3 4 5 6 7 8 9 10 11 12
Rainfall(mm)
Month
0
2
4
6
8
10
12
14
16Ryde Daily Aggregated Statistics
Obs Daily Mean
Obs Daily Stddev
Sim Median
90% Conf Lim
1 2 3 4 5 6 7 8 9 10 1 1 12
Rainfall(mm)
Month
Figure 10. Ryde observed versus simulated (a) hourly and (b) daily mean and standard
deviation of rainfall.
Final 26/06/2003 7:58 AM 11
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
19/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
50
60
70
80
90
100
2 4 6 8 10 12
RydeWPS Dry Probability
Dry Prob (%)Daily Dry Prob (%)Median Simulated
90% Confidence Limit
DryProbability(%
)
Month
Figure 11. Ryde observed versus simulated dry probability curves
1
10
100
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.9
0.50hrs Obs1.00hrs Obs
3.00hrs Obs12.00hrs Obs24.00hrs ObsMedian Simulated90% Confidence Limit
Intensity(mm/hr)
Percent
Figure 12. Ryde observed versus simulated IFD curves
Final 26/06/2003 7:58 AM 12
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
20/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
2.5.2 Comparison with AR&R IFD
Design IFD curves calculated using the methods described in AR&R can be used to provide acheck against simulated values. Figure 13 compares the observed, median simulated andAR&R IFD curves for a range of timescales.
1
10
100
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.99
0.50hrs Obs1.00hrs Obs
3.00hrs Obs12.00hrs Obs24.00hrs Obs
Median DRIP SimulationAR&R Estimate
Intensity(mm/hr)
Percent
200yrs100yrs
50yrs
20yrs
10yrs
5yrs
500yrs
2yrs
1yr = ARI
Figure 13. Ryde observed vs simulated vs AR&R IFD curves.
Figure 13 shows that the AR&R IFD curves underestimate the observed IFD statistics at Rydein the right tail. This appears to be due to consistent underestimation of the log-standarddeviation (which is the slope of the IFD curve on log-normal probability paper) rather thandue to shifts in log skewness. The observed IFD curves show only weak evidence of non-zerolog skew, which is consistent with the data at Observatory Hill Sydney.
The consistent underestimation by AR&R is due to two factors:
1. The AR&R IFD curves are derived from a regional procedure that spatially interpolatesbetween gauge locations with good record lengths. Regionalisation, by its nature,smoothes spatial variability and hence can introduce systematic error.
2. The AR&R curves were derived from a database ending in the late 1970s and early1980s. This database had a good coverage of 24-hour bulk gauges but few long-term
Final 26/06/2003 7:58 AM 13
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
21/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
pluviograph records in the Sydney area. With the availability of up to 20 years more datait is probable, indeed expected, that differences will arise particularly in the right tail ofthe IFD curve.
2.5.3 Comparison of DRIP Storm Statistics
Given the proximity of the Ryde and Parramatta gauges and the almost identical annualrainfall distributions one would expect similar statistics for storm characteristics such as dryspell, storm duration and storm depth. Table 2 does not bear out this expectation.
The reason for this unexpected outcome is related to the apparent unsuitability of theObservatory Hill Sydney gauge as a master pluviograph for the UPR catchment. In calibratingDRIP to the daily rainfall record at Parramatta it was necessary to assume the storm durationdistributions at Observatory Hill Sydney and Parramatta were identical. This is why in Table2 the mean and standard deviation for storm durations are identical for these sites. In contrastRyde has a mean storm duration which is 55% that of Observatory Hill Sydney.
If Parramatta has a storm duration probability distribution similar to that of Ryde (areasonable assumption) then the imposition of Observatory Hill storm duration characteristicswill distort the Parramatta dry spell and storm depth statistics in order that the daily rainfallcharacteristics at Parramatta be preserved. Hence Parramatta has substantially longer dry spelland greater storm depths than Ryde.
Given the evidence presented it is concluded that:
1. Observatory Hill Sydney is not a suitable master pluviograph site for the UPR catchment;and
2. Ryde is a more suitable master pluviograph site.
Table 2. DRIP simulated storm statistics
Stormcharacteristic
Statistic ObservatoryHill Sydney
Parramatta Ryde
Dry spell, hr Mean 44.8 56.0 40.5
Standard deviation 74.0 94.0 85.5
Storm duration, hr Mean 4.07 4.07 2.25
Standard deviation 5.87 5.87 4.08
Depth, mm Mean 6.88 7.16 4.55
Standard deviation 17.0 16.7 13.7
Despite the conclusion that the DRIP calibration to Parramatta using Observatory Hill Sydneyas the master site is not reliable, it is worth comparing the DRIP IFD curves for Parramattaand Ryde shown in Figure 14. These plots differ from those presented in Figures 4, 8, 12 and13. In the earlier figures the IFD plots were based on replicated samples with record lengthsequal to the historic record length. The objective was to quantify the sampling uncertainty inorder to make a judgement about how well the observed data matched the DRIP model. Incontrast the IFD curves in Figure 14 are based on a 1000-year simulation and thus accuratelydefine the IFD curves.
Final 26/06/2003 7:58 AM 14
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
22/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
1
10
100
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.99
0.5hrs Ryde
1hrs Ryde3hrs Ryde
9hrs Ryde
24hrs Ryde
0.5hrs Parra
1hrs Parra3hrs Parra
9hrs Parra
24hrs Parra
Intensity(mm/hr)
Percent
Figure 14. DRIP IFD curves for Ryde and Parramatta.
What is striking about Figure 14 is the close agreement between the DRIP Parramatta and
Ryde IFD curves. This would suggest that the extreme storm mechanisms at Observatory HillSydney and Ryde are similar.
Despite the similarity of the IFD curves, it would be erroneous to use the Parramatta DRIPmodel to simulate OSD and rainwater tank scenarios. This is because the Parramatta DRIPstorms are longer on average than the Ryde DRIP storms. Noting that IFD statistics aretypically obtained from bursts within storms the Parramatta DRIP storms would on averagehave higher antecedent rainfalls prior to the critical burst than the Ryde DRIP storms. Thisconclusion was borne by the finding that OSD requirements were bigger for Parramatta DRIPsimulation than for Ryde DRIP simulation.
The curves in Figure 14 show that as duration decreases log skewness weakly increases. Forthe 24-hour duration the IFD curve is virtually a straight line implying the log skew is zero, asis the case for Observatory Hill. With decreasing duration the upward curvature becomesmore pronounced implying log skew is becoming more positive, albeit marginally so. Thiscomplicates extrapolation of IFD curves to high ARIs for short durations because there arevirtually no pluviograph sites with records sufficiently long to meaningfully identify the righttail of the IFD curve see Table 1.
Final 26/06/2003 7:58 AM 15
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
23/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
2.6 Discussion
2.6.1 Adequacy of AR&R IFD curves
The credibility of the OSD simulations in this study depend critically on the credibility of
DRIP to simulate extreme storms. Figures 12 and 13 show that at Ryde the DRIP IFD curvesare consistent with observed IFD curves but overestimate the AR&R IFD curves. It is,therefore, important to establish whether the IFD underestimation by AR&R at Ryde is anisolated occurrence or a reflection of a deeper problem. To this end a comparison of AR&RIFD curves with observed IFD curves was undertaken at three other gauges: Guildford,Bankstown and Liverpool.
Figures 15 to 17 shows the IFD curves for Guilford, Bankstown and Liverpool which hadrecords lengths of 29, 23 and 9 years respectively. For Guildford AR&R substantiallyunderestimates the extreme intensities for all durations except 3 hours and poorly reproducesthe overall shape of the observed distributions. Indeed there is a distinct upward curvature in
the observed IFD curves which contradicts the zero log skew assumption in AR&R. Thecomparison for Bankstown is more favourable with AR&R tending to fit the upper tails of theIFD curves reasonably well. However, the AR&R reproduction of the overall shape of thedistribution was unsatisfactory for durations greater than or equal to 3 hours. The recordlength for Liverpool is too short to make a judgement about the adequacy of fit.
1
10
100
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.99
0.50hrs Obs1.00hrs Obs3.00hrs Obs
12.00hrs Obs
24.00hrs Obs
30min AR&R1hrs AR&R3hrs AR&R
12hrs AR&R24hrs AR&R
Intensity(mm/hr)
Percent
200yrs100yrs50yrs
20yrs
10yrs5yrs
500yrs
2yrs
1yr = ARI
Figure 15. Comparison of AR&R and observed IFD curve for Guildford gauge.
Final 26/06/2003 7:58 AM 16
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
24/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
1
10
100
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.99
0.50hrs Obs1.00hrs Obs
3.00hrs Obs
12.00hrs Obs24.00hrs Obs30min AR&R
1hrs AR&R3hrs AR&R
12hrs AR&R24hrs AR&R
Intensity(mm/hr)
Percent
200yrs100yrs
50yrs
20yrs
10yrs5yrs
500yrs
2yrs
1yr = ARI
Figure 16. Comparison of AR&R and observed IFD curve for Bankstown gauge.
1
10
100
.01 .1 1 5 10 20 30 50 70 80 90 95 99 99.9 99.99
0.50hrs Obs1.00hrs Obs3.00hrs Obs12.00hrs Obs24.00hrs Obs30min AR&R1hrs AR&R3hrs AR&R12hrs AR&R24hrs AR&R
In
tensity(mm/hr)
Percent
200yrs100yrs
50yrs
20yrs
10yrs5yrs
500yrs
2yrs
1yr = ARI
Figure 17. Comparison of AR&R and observed IFD curve for Liverpool gauge.
Final 26/06/2003 7:58 AM 17
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
25/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
The data for Guildford and Bankstown tend to confirm the Ryde results which suggest thatthe AR&R IFD curves may be systematically in error. As previously discussed the AR&RIFD method is a regional method based on extensive daily rainfall data and limitedpluviograph data ending in the late 1970s. It is not altogether surprising that with theavailability of more data discrepancies become apparent.
Of significance is the positive log skew, particularly at shorter durations, which tends toamplify differences in the right tail of the distributions. Figure 18 presents a log-Pearson IIIplot to the 1-hour Guildford annual maximum intensity data. The log skew was 0.85 with astandard deviation of 0.46! Note the large uncertainty in the right tail. For the 100-yearintensity the expected intensity is 93 mm/hr with 90% confidence limits of 60 to 176 mm/hr.If a log-normal distribution were fitted to the same data the 100-year intensity the expectedintensity is 75 mm/hr with 90% confidence limits of 60 to 100 mm/h, a result more consistentwith the AR&R values reported in Table 3. This underscores the inherent difficulty in fittingto 3-parameter distributions to short data.
It needs to be emphasised that DRIP does not fit distributions to annual maximum rainfalls. Itsimulates and disaggregates individual storm events. Given that DRIP fits probabilitydistributions to about 190 events per year the storm probability distributions are well-definedwhen compared to the log Pearson III in Figure 18. Within DRIP a 100-year annual maximumburst is likely to be the result of the joint occurrence of several random variables whichthemselves are not extremes. This suggests that DRIP does not extrapolate observed data as inFigure 18 but uses a well-identified probability structure to simulate extremes.
ARI (yrs)
1.000
1.300
1.600
1.900
2.200
2.500
Log10Flow
1.5 2 5 10 20 50 100 200
Gauged flow
Exp parameter quantile
Expected prob quantile
90% limits
Figure 18. logPearson III fit to 1 hour annual maximum intensity at Guildford.
2.6.2 Application of DRIP to other sites within the UPC catchment
It was concluded in Section 2.5.3 that the DRIP calibration at Ryde provides a more
representative master site for the UPR catchment than does Observatory Hill Sydney. That
said, it is noted that the Ryde calibration represents the rainfall regime at Ryde and not overthe whole UPR catchment. Table 3 compares AR&R intensities for several locations within
Final 26/06/2003 7:58 AM 18
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
26/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
UPR catchment. It can be seen that there is little variation in intensity over the UPRcatchment. The Ryde intensities are similar to those at Parramatta and marginally lower thanat Cumberland State Forest.
This suggests that the Ryde gauge could be confidently used as the master gauge to calibrate
DRIP to other pluviograph records in the UPR catchment such as those listed in Table 1. It isimportant to note that the proposed calibration involves a transfer from the Ryde masterpluviograph record to a shorter target pluviograph record. Such a calibration enables scalingfactors to be estimated for all three DRIP storm characteristics: dry spell storm duration andconditional intensity. This removes the restriction on storm duration scaling factors thataffected the calibration of DRIP to the Parramatta daily record. This provides a rationalframework for applying DRIP to locations within the UPR catchment at which there are nolong pluviograph records.
Table 3. Selected AR&R intensities in UPR catchment
Duration
hrs
ARI
yrs
Ryde Blacktown
(south westUPR
catchment)
Toongabbie
(middle UPRcatchment)
Parramatta Cumberland SF
(north east UPRcatchment)
1 100 77 67 70 72 76
1 50 69 60 63 65 68
3 100 41 35 37 39 45
3 50 36 32 34 35 37
12 100 18.1 15.4 16.7 17.4 18.9
12 50 16.1 13.8 14.9 15.5 16.8
The evidence based on the observed IFD curves for Ryde and Guildford indicates that theAR&R IFD curves underestimate extreme intensities. This has implications for UPRCTsOSD policy because an increase in extreme intensities is likely to result in increased sitestorage requirements. Table 4 provides an indication of the change in extreme intensitiesarising from the use of DRIP in conjunction with pluviograph records longer than thoseavailable to AR&R. The table shows that the AR&R intensities for 1 and 3 hour durations atthe 100 year ARI are about 20% less than what DRIP predicts for Ryde. Because Table 3demonstrates that Ryde AR&R intensities are broadly representative of the UPR catchmentone can expect that use of DRIP will increase extreme intensities at the shorter durations by
about 25%.
Table 4. Selected Comparison of DRIP and AR&R intensities
Durationhrs
ARIyrs
DRIP intensity atRyde (from Fig 14)
AR&R intensity as % of Ryde DRIP intensity
mm/hr Ryde Parramatta Toongabbie
1 100 92 84 78 76
3 100 47 87 83 79
2.7 Conclusion
Several important conclusions result from the calibration and validation of DRIP:
Final 26/06/2003 7:58 AM 19
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
27/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
1. It was originally envisaged that DRIP would be directly calibrated to a long pluviographat Parramatta. However, owing to the very short pluviograph record at Parramatta thisapproach was deemed infeasible and an alternative approach was developed.
2. DRIP was directly calibrated to the 53-year Ryde Pumping Station pluviograph record.
The Ryde record was the longest available record for gauges located in and near the UPRcatchment. The calibrated model was used to simulate statistics not used in thecalibration. Such statistics ranged from annual rainfall distributions to IFD curves. For allthe validation statistics considered, DRIP simulations were found to be statisticallyconsistent with the observed statistics. This result engendered confidence in DRIPsability to simulate the entire rainfall regime from very short to annual timescales.
3. It was found at Ryde that the DRIP and observed IFD curves produced short duration100-year intensities about 25% greater than those predicted by AR&R. Examination ofobserved IFD curves for Guildford showed a similar underestimation by AR&R, whereasfor Bankstown AR&R IFD curves unsatisfactorily reproduced the overall shape of the
observed IFD curves but did manage to reproduce the right tails better than at Ryde orGuildford. It is difficult to escape the conclusion that the AR&R IFD curves are in error,possibly of the order of 25% for 100-year storms with durations less than 3 hours.
4. The Ryde pluviograph can be used as a master site to transfer DRIP to other shorterpluviograph records within the UPR catchment. The justification for use of the Ryderecord as the master site in future work is that it has a similar annual rainfall distributionas Parramatta, is similarly distant from the coast, and has AR&R IFD statistics that onlymarginally differ across a range of sites in the UPR catchment.
On the CD supplied with this report, a data file is provided with a 1000-year simulated
pluviograph record based on the calibration to the Ryde Pumping Station record. In additionFORTRAN90 source code is provided to enable UPRCT to generate its own synthetic rainfallrecords of any desired length. Prior to simulation the code must read a DRIP parameter file,which contains the results of the DRIP calibration. To this end DRIP parameter files forObservatory Hill Sydney, Ryde Pumping Station and Parramatta have been included on theCD.
Final 26/06/2003 7:58 AM 20
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
28/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
3. DESIGN OF RAINWATER TANKS
A dual water supply solution uses rainwater from tanks to supplement the mains water supply
for toilet flushing, laundry, hot water and outdoor uses. The use of a dual water supply systemenables the designer to maximise the water supply and stormwater management benefits ofthe rainwater tank.
The design requirements imposed by Australian standards, the NSW Department of Health,water authorities and local government for dual water supply using rainwater tanks and mainswater are reviewed. This is followed by presentation of a design method.
3.1 Australian Standards
The Standard AS/NZS 3500.1.2: Water Supply - Acceptable Solutions provides guidance for
the design of rainwater tanks with dual water supply (rainwater and mains water). Itcategorises cross connections between mains water supply and premises with a rainwater tankto be low hazard, thereby requiring a non-testable backflow prevention device. Rainwatertanks with dual water supply must maintain an air gap, and be designed and connected inaccordance with Figure 19.
3.2 NSW Department of Health
The NSW Department of Health does not prohibit the use of rainwater for drinking or otherpurposes. The Department recommends proper use and maintenance of rainwater tanks andprovides a monograph Guidance on the use of rainwater tanks [Cunliffe, 1998] to assist with
this task. The focus of NSW Department of Health guidelines is drinking water quality.
Overflow
Height of water above
invert of overflow(AS 3500.1.2
and AS 2845.2)
Water supplyto house
Pump &
pressure
vessel
Rainwater Tank
Figure 19. Design details to prevent backflow for a rainwater tank with mains water top up.
3.3 Water Authorities
Water authorities cannot prohibit the use of rainwater from tanks on private land. Theirprimary concern is to maintain the quality of mains water. Accordingly, water authorities may
Final 26/06/2003 7:58 AM 21
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
29/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
require the installation of an adequate backflow prevention device to prevent contamination ofmains water by tank water if the existing water meter does not already have a backflowprevention device.
3.4 Local Councils
Local councils have varying policies on the installation of rainwater tanks. Tanks arestructures, and their erection may require development consent. However, many councils havedeclared rainwater tanks to be exempt development (which does not require consent)provided that certain requirements relating to tank size, height and siting are satisfied. Alltanks should be installed in accordance with The New South Wales Code of Practice:Plumbing and Drainage [Committee on Uniformity of Plumbing and Drainage Regulations inNSW, 1999].
If a development application is required to install a rainwater tank, details should be providedas to the:
Location of the tank and relationship to nearby buildings;
Configuration of inlet/outlet pipe and the overflow pipe;
Tank capacity, dimensions, structural details and proposed materials; and
Purposes for which the tank is intended to be used.
Local councils cannot prohibit the use of water derived from a rainwater tank. However,where a council is a water supply authority it can require the installation of an adequate
backflow prevention device.
3.5 Design Details
In order to maximise water savings and stormwater management benefits, tank capacityshould be between 5 kL and 20 kL for each residential dwelling. The required capacity willdepend on number of persons in the household, water use, rainfall and roof area. Design ofthe rainwater tank (Figure 20) should make provision for:
A minimum storage volume (to ensure that water supply is always available);
A rainwater storage volume; and
An air space for additional stormwater management.
The minimum storage volume is the maximum daily water use that is expected from the tank(about 250-750 litres). If the volume of stored water falls below the minimum storage volume,the shortfall can be overcome by topping up the tank with mains water to the required level. Asimple float valve system can be installed to do this automatically.
The rainwater storage volume is the total volume available in the tank to store rainwaterbelow the overflow pipe. The air space between the overflow pipe and the top of the tank can
be used to provide stormwater detention, thereby delaying the delivery of excess roof waterto the drainage system. The rainwater storage volume and the overlying air space both
Final 26/06/2003 7:58 AM 22
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
30/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
provide stormwater management benefits providing both retention and detention. Therequired volume for the air space will vary according to the selected average recurrenceinterval (ARI) design storm.
The configuration of plumbing required for rainwater tanks is shown in Figure 20. Water
supply from the rainwater tank (such as for outdoor, toilet, laundry or hot water uses) isdirected to the household via a small pump and pressure vessel. When tank water levels arelow, such as during hot, dry periods, the tank is topped up with mains water via a tricklesystem. The trickle top up system will reduce the daily peak demand on the mains waterdistribution network. In the event of pump or power failure the rainwater tank can bebypassed.
Mains top up volume
Rainwater space
Air space used for
detention
Overflow
Trickle top up from
mains supply
Pump &
pressure
vessel
Water supply to
house
Float
Pump & pressure
vessel
Water supply for
irrigation, toilet,
laundry and hot
water uses
Mains water su l for other inhouse uses drinkin
Rainwater
tank
Float
Trickle top up
with mains
water as
required
Bypass in event
of ower failure
Stop valve
Figure 20. Design details for a dual water supply system using rainwater and mains water.
Final 26/06/2003 7:58 AM 23
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
31/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
4. COST MODELS
Costs and benefits have been derived to allow comparative analysis of costs and present
values of OSD and rainwater tank solutions. The costs and benefits reported below are used ina present equivalence analysis [Coombes et al., 2000e and 2001] with a 6% interest rate tocalculate the present value of the OSD and rainwater tank scenarios.
4.1 Installation of a Rainwater Tank
The cost to install a dual water supply system including a 10 kL rainwater tank and a Davypump at the Maryville demonstration site has been reported in Coombes et al. (2000). Thecosts for different rainwater tank sizes are shown in Table 5.
Table 5. Cost to install a rainwater tank system
Cost to install each tank size ($)Item5 kL 10 kL 15 kL 20 kL
Tank 470 670 840 1090
Pump 270 270 270 270
Plumber and fittings 500 500 500 500
Float system 200 200 200 200
Concrete base 200 200 200 200
GST 160 180 200 230
Total 1800 2020 2210 2490
The lifecycle costs of the rainwater tank solutions are:
The pump costs $0.002 per day to operate and has a 10 year life; and
The rainwater tank has a 50 year life.
The operating and maintenance costs for the pump have been assumed to be $0.10 per kL ofrainwater consumed.
The price of water is assumed to be $0.92 per kL and the sewage charge is assumed to be$0.2025 per kL of mains water consumed.
4.2 Installation of an OSD Tank
The costs to install an OSD tank has been derived from Figure H1 in the On Site DetentionHandbook [Upper Parramatta River Catchment Trust, 1999] and are shown in the followingTable 6.
Table 6. Cost to install an OSD tank
Cost ($) per m3 of OSD tank
Volume (m3) 10 20 30 40 50 100
Cost ($) 812 551 406 377 352 279
Final 26/06/2003 7:58 AM 24
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
32/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
Tschantz (2000) reported failure rates of over 90% for OSD systems due to partially orcompletely blocked outlet controls resulting from a lack of routine maintenance. Amaintenance cost of $75 per year has been assigned to the OSD scenarios.
Final 26/06/2003 7:58 AM 25
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
33/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
5. THE ALLOTMENT WATER BALANCE MODEL
Optimum use of rainwater, stormwater and wastewater at the urban allotment are the urban
water supply, sewage and stormwater management strategies of the future. A reliablehousehold water balance model is required to allow accurate assessment of demand reduction,infrastructure provision and water reuse scenarios.
The daily water balance model AquaCycle developed by Mitchell et al. (1997) makes asignificant contribution to the understanding of the urban water balance. However,assessments of the impact of rainwater reuse on stormwater and water supply infrastructureprovision requires analysis at much shorter time steps. For example; a stormwater catchmentresponse time is likely to be between 15 minutes and 6 hours.
A water balance model has been developed to simulate the performance of source control
measures (rainwater tanks, landscaping, gravel trenches and on site detention) in allotmentsand clusters. The model works at 2-minute intervals for rainfall and water consumption andone-second intervals for all stormwater discharges. A schematic of the model is shown inFigure 21.
RainfallPluvio data, DRIP or
rainfall infill generator
Roof
Impervious
area
Pervious
area
Mains waterFirst flush
device
Rainwater
tank
Outdoor
uses
Indoor usestoilet, hot water
Other Indoor uses
Infiltration
trench
On site
detention
Street drainage system
Spill
ET
Infiltration
to Soil
Trickle
top up
Figure 21. Allotment water balance model
A new probabilistic behavioural household water use model [Coombes et al., 2000a] has been
established to account for water use in the model and a pluviograph rainfall infill generator
has been created to allow continuous simulation of the household water balance at small timesteps.
Final 26/06/2003 7:58 AM 26
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
34/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
The household water balance model has also been combined with the DRIP point rainfallmodel to develop design curves for the performance of rainwater tanks and on site detentiontanks in the Upper Parramatta River Catchment.
This section provides a brief description of the components of the water balance model.
5.1 Outdoor Water Use
Outdoor water demand is a large and highly variable portion of total household demandranging from 8% to 53% of total water use in the Lower Hunter region (Tables 7 and 8). TheHunter Water Corporation (HWC) [Berghout, personal communication, 1999] has monitoredindoor and outdoor water use in over 130 houses located in the 9 water supply zones listed inTables 7 and 8 during the period 1986 to 1998. This data has provided the foundation for thedevelopment of the household water use simulation models.
Table 7. Water demand zones and related statistics used for the Lower Hunter region
M in M a x A v e M in M a x A v e M in M a x A v e
In n e r S E N e w c a s t le 2 .2 1 8 .9 4 2 2 1 1 2 5 1 1 4 4 1 6 1 1 0 5
H a m ilto n M a y f ie ld 2 .2 1 8 .9 4 2 2 1 1 2 5 1 1 2 8 1 2 6 7 5
L a m b to n J e s m o n d 2 .2 1 8 .9 4 2 2 1 1 2 5 1 1 1 9 1 2 6 6 3
N W W a lls e n d 2 .2 1 8 .9 4 2 2 1 1 2 5 1 1 5 6 2 7 1 1 5 0
L a k e M a c q u a r ie E a s t 2 .7 7 8 .9 4 2 2 1 1 2 7 1 3 3 2 2 4 5 1 2 4
L a k e M a c q u a r ie W e s t 3 .2 4 8 .9 4 2 2 1 1 2 7 1 2 5 7 2 3 6 1 5 0
M a it la n d 2 .4 7 8 .8 4 7 .8 2 4 1 2 5 1 1 1 0 7 3 2 9 2 1 3
C e s s n o c k 2 .0 7 8 .8 4 7 .8 2 4 1 1 8 6 5 6 2 1 3 1 3 3
P o r t S te p h e n s 3 .4 4 9 .2 4 5 2 3 1 2 6 1 2 5 4 3 6 8 2 1 8
M o n t h ly A v e .
R a i n D a y s
Z o n e
A v e . m o n t h ly
e x h o u s e
d e m a n d ( L /d a y )A v e . ra i n
( m m /d a y)
D a i ly m a x .
t e m p . (0C )
Using the HWC outdoor consumption data Coombes et al. (2000a) developed a new outdoordemand model that yielded satisfactory performance with R2 varying from 0.51 to 0.68 for thezones in the Lower Hunter region. The model differs from the current genre of outdoordemand models in that daily outdoor use is modelled probabilistically using behavioural rules.These rules describe the probability of outdoor water use on a given day with the probabilityaffected by the meteorological variables, daily rainfall depth, days since the last rainfall anddaily maximum temperature, and the long-term average variables, daily average rainfall andmonthly average daily demand.
Table 8. Water demand zones and related statistics used for the Lower Hunter region
Zone
n ouse
(L/d)
x ouse
(L/d)
ve. an
(mm/d)
ncome
($/C)
an ays
per month
rowt
(%/A)
ve. ay
Temp Soil Type
Inner Newcastle 337 105 2.21 429 11 1.03 21 sandy loam
Hamilton Mayfield 454 75 2.21 330 11 0.67 21 sandy loam
Lambton Jesmond 290 63 2.21 323 11 2.09 21 sandy clay
NW Wallsend 498 150 2.21 289 11 3.26 21 silty clay loam
Lake Mac. East 553 124 2.77 286 13 1.24 21 sandy clay loam
Lake Mac. West 517 150 3.24 286 12 2.42 21 silty clay loam
Maitland 351 213 2.47 284 11 1.16 24 sandy clay loam
Cessnock 426 133 2.07 270 6 0.38 24 sandy clay loam
Port Stephens 441 218 3.44 264 12 2.42 23 Sand
Final 26/06/2003 7:58 AM 27
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
35/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
The model was calibrated to HWC outdoor data for each of the nine local water supply zonesshown in Tables 7 and 8 using the SCE-UA global optimisation method [Duan et al., 1994].An example of calibration to outdoor use data for the Mayfield area of the Newcastle zone isshown in Figure 22. The outdoor demand model was able to reliably simulate the strongseasonal trends experienced in the Mayfield area - such seasonality in outdoor demand is
typical for all the zones in the Lower Hunter Region.
Mayfield outdoor water use
0
1000
2000
3000
4000
5000
6000
7000
0 12 24 36 48 60 72 84 96 108 120 132
Months since December 1986
D
emand(L/month)
Predicted
Observed
Observed versus predicted outdoor water use at
Mayfield
0
1000
2000
3000
4000
5000
6000
7000
0 1000 2000 3000 4000 5000 6000 7000
Predicted demand (L/month)
Observ
eddemand(L/month)
R2= 0.67
Figure 22. Observed and predicted monthly outdoor water use at Mayfield.
5.2 Indoor Water Use
The water balance model uses indoor water use data from the HWC [Berghout, personal
communication, 1999], the Figtree Place development [Coombes et al., 2000] and the
Stringybark Grove development [Cox et al., 1998] to establish water use categories and
patterns (Figures 23 and 24).
Hot water
39%
Toilet
25%
Laundry
23%
Other
13%
Figure 23. Household water use categories
The water balance model uses the household water use categories and proportions shown in
Figure 23 to allocate indoor water use from the rainwater tank.
Final 26/06/2003 7:58 AM 28
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
36/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
0
50
100
150
200
250
300
350
400
450
500
0 2 4 6 8 10 12 14 16 18 20 22 24
Hours since midnight
Wateruse(lit
res/hour)
Figure 24. Water use patterns from the Stringybark Grove development
The water use patterns from Stringybark Grove, shown in Figure 24, have been transformedinto a normalised water use versus normalised time relationship, shown in Figure 25, toenable the water balance model to simulate diurnal indoor and outdoor water use patterns.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalised time (time/day)
Normalised
wateruse
(consumption/dailyconsumption)
Figure 25. Normalised diurnal water use pattern.
5.3 Pluviograph Rainfall Generation
The water balance model can use rainfall from pluvio or tipping bucket rain gauges orgenerate synthetic pluviograph rainfall from daily rainfall data. An outline of the pluviorainfall generator is shown in Figure 26.
In addition the water balance model has been linked with the DRIP point rainfall model. Thisentails reading a DRIP parameter file, which contains parameters calibrated for the site understudy, and then generating synthetic storms for the required length, In this study, 1000 yearsof synthetic storm data was generated.
Final 26/06/2003 7:58 AM 29
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
37/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
Pluvio or tipping bucket
rain gauge
Daily rainfall
Infill missing
rainfall data
Make rainfall
intensity andtemporal pattern
tables
Month?
Missing
pluvio
data?
Random selection of
temporal pattern and
intensity
yes
Intensities Temporal
patterns
Determine storm
duration
Pluvio rainfall file
No
Figure 26. Pluvio rainfall generator
The pluviograph rainfall generator requires a daily rainfall file with no missing data. This fileis used as a reference file to find and infill missing storm events based on temporal patternsand rainfall intensities derived from pluviograph or tipping bucket rain gauge records. The
rainfall intensities and temporal patterns are randomly selected from files sorted by month,which is used as a surrogate for season.
Rainfall intensity and temporal patterns can be sourced from Australian Rainfall and Runoff[Institution of Engineers, Australia, 1987] if pluviograph or tipping bucket rainfall data isunavailable.
5.4 First Flush Separation
Many authors [including Jenkins and Pearson, 1978, Mitchell et al. 1997 and Cunliffe, 1998]describe the first flush as a fixed amount of roof runoff requiring separation. Design and
modelling of the performance of first flush devices has been dominated by this belief.However the number of dry days preceding a rainfall event, rainfall intensity and rainfalldepth is an indicator of roofwater quality [Yaziz et al., 1989]. Design of first flush separationdevices and simulation of their performance will need to reflect the dynamic nature ofroofwater quality and quantity.
The first flush pits at Figtree Place were designed to separate the first 2 mm of roofwater frominflow to the rainwater tanks. The design of these first flush pits was based on rainfallvolumes from design storms described in Australian Rainfall and Runoff [The Institution ofEngineers, Australia, 1987]. The first flush pits proved to be so efficient that no inflow to therainwater tanks resulted. Design of first flush separation devices needs to maximise
conservation of roof water and minimise contaminant transport to the rainwater tank whilstaccounting for variation of roof water quality and quantity from all storms.
Final 26/06/2003 7:58 AM 30
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
38/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
The design of the first flush device (Figure 27) in the water balance model includes an inletfrom the roof, a chamber to capture the first flush of rainwater allowing it to leak through asmall hole in the base of the chamber, a mesh screen to separate debris and an overflow to therainwater tank.
Rainwater
tank
Meshscreen Rainwater
from roof
rainwater leaks
through smallhole
Overflow to
tank
Figure 27. Diagram of a first flush separation device
5.5 The Rainwater Tank
The design of a rainwater tank in the water balance model is schematised in Figure 28.Rainwater overflows from the first flush device into the rainwater tank, water is drawn fromthe tank for household use and the tank is topped up with mains water if the water level is lessthan the minimum water level. Overflow from the rainwater tank is routed via a pipe system
to an OSD tank, infiltration trench or the street drainage system. When the rainwater tank isfull spills are directed to impervious areas.
Mains water to maintain Overflow from first flush pit
minimum water level
Supply for irrigation,
hot water and toilet flushing
minimum water level
Minimum storage volume:
one day's water supply
Overflow to OSD,
infiltration trench or
street drainage
Spill to
impervious
area
Figure 28. Schematic of a rainwater tank
5.6 The OSD Tank
The design of the OSD tank used in the water balance model is schematised in Figure 29.Stormwater from roofs or rainwater tanks or infiltration trenches, impervious and pervious
Final 26/06/2003 7:58 AM 31
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
39/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
areas is directed to the discharge control pit in the OSD tank. The OSD tank has twochambers: a discharge control pit and site storage chamber.
Stormwater from roofs or infiltration
trench or rainwater tank, impervious
and pervious areas
Outlet to street
drainage
Discharge
control pitSite storage
chamber
Overflow
Infiltration to soil via
weep holes
Stormwater
transfer between
chambers
Spill to street
drainage system
Figure 29. Schematic elevation view of OSD tank used in the water balance model
Stormwater stored in the discharge pit is discharged via an orifice to the street drainagesystem. If the pit is full stormwater overflows into the site storage chamber. The site storagechamber discharges to the discharge control pit when the discharge control pit empties.Stormwater infiltrates via weep holes to the soil below the discharge control pit and spills tothe street drainage system when the chamber and the pit are full.
5.7 The Infiltration Trench
A schematic of the infiltration trench used in the water balance model is shown in Figure 30.Stormwater from roofs or rainwater tanks is directed to the infiltration trench where it isstored in the gravel void spaces for infiltration to the surrounding soil. When the infiltrationtrench is full it overflows to the street drainage system or to an OSD tank.
Inflow from roofs or
rainwater tanks
Overflow to street
drainage system
Infiltration to soil
Gravel fill with a
given void space
Figure 30. Schematic of the infiltration trench used in the water balance model
5.8 Pervious Area
A schematic of the pervious area water balance is shown in Figure 31. A soil type can bechosen from 4 categories: 1 sand, 2 silty sand, 3 sandy clay and 4 clay. Each soil type has a
given initial and saturated infiltration rate, and porosity.
Final 26/06/2003 7:58 AM 32
-
8/3/2019 Australia; Impact of Rainwater Tank and On-Site Detention Options on Stormwater Management in the Upper Parra
40/64
Impact of Rainwater Tank and On-site Detention Options on Stormwater Management
The model has two storages: surface or depression storage and soil moisture storage in theeffective root zone of chosen vegetation. Rainfall falls on the soil surface, is stored in thedepression storage, infiltrates to the soil moisture store and is transferred to the atmospherevia evapotranspiration. When the depression storage is full it overflows to the OSD tank orthe street drainage system.
Evapotranspiration
Depression storage
Soil moisture store in the root