Both water quantity and quality are impacted by climate change. In addition, rapid urbanization has also brought an immeasurable loss of life and property resulting from floods. Hence, there is a need to predict changes in rainfall events to effectively design stormwater infrastructure to protect urban areas from disaster. This study develops a framework for predicting future short duration rainfall intensity and examining the effects of climate change on urban runoff in the Gunja Drainage Basin. Non-stationarities in rainfall records are first analysed using trend analysis to extrapolate future climate change scenarios. The US Environmental Protection Agency Storm Water Management Model (SWMM) was used for single event simulation of runoff quantity from the study area. For the 1-hour and 24-hour durations, statistically significant upward trends were observed. Although the 10-minute duration was only nearly significant at the 90% level, the steepest slope was observed for this short duration. Moreover, it was observed that the simulated peak discharge from SWMM increases as the short duration rainfall intensity increases. The proposed framework is thought to provide a means to review the current design of stormwater infrastructures to determine their capacity, along with consideration of climate change impact.

INTRODUCTION

Since the beginning of industrialization, significant quantities of air pollution materials and greenhouse gases have been emitted into the atmosphere by human activities. Global warming affects the water cycle, causing a significant rise in the temperature and sea level all over the globe (IPCC 2007). The increase in the average temperature and sea level also causes an increase in rainfall intensity and frequency (Houghton et al. 1996). In addition to climate change, rapid urbanization has also brought an immeasurable loss of life and property from floods. In the past, frequent torrential rains and typhoons did not occur in South Korea. However, nowadays, the physical impacts of climate change such as casualties and property damage can be observed. Typhoon Rusa in 2002 caused enormous damage amounting to US$5 billion and the loss of 266 human lives (Lee 2007). These typhoons caused 69% of the total property damage over the Korean peninsula from 1995 to 2004 (Kim & Jain 2011). With these growing concerns, it is important to predict climate changes for the design and construction of drainage infrastructure to effectively protect urban areas from disaster. In order to mitigate the effects of disaster, it is necessary to estimate a proper design rainfall, which considers the increasing trend of rainfall intensity for hydrological planning and design.

Climate change impact studies have been conducted over the years using several methods such as general circulation models (GCMs), sensitivity analyses, and trend analyses. Recently, a number of studies have used GCM for climate change scenarios on rainfall and runoff (Bronstert et al. 2002; Nguyen et al. 2010; Sheshukov et al. 2011). Rainfall scenarios from GCMs depend upon the details of the physical parameterizations in the cloud and surface energy exchange components of the model (Sansom & Renwick 2007). However, climate model simulations using GCMs have some limitations because broad average results from GCMs cannot represent actual climate conditions accurately at the fine scale of urban drainage systems (Willems 2011). Sensitivity analyses have been used to assess the potential impact of changing precipitation intensity on urban drainage infrastructures; however, the future changes are arbitrarily selected (e.g., ±10 percent, ±20 percent, ±30 percent) as had been done by Niemczynowicz (1989).

Trend analysis through linear regression is able to describe realistic non-stationarity of rainfall as it is based on specific local data. Because of its simplicity and ability to detect and quantify non-stationary trends, several studies consider it to be the most suitable method for obtaining future regional scenarios to consider for urban design (Pagliara et al. 1998; Denault et al. 2006). Linear regression has been used by Pagliara et al. (1998) to investigate trends in design storms for data collected from two gauges in Tuscany, Italy. Denault et al. (2006) also studied the potential influence of future increases in runoff on the city's hydraulic structures and ecosystem by performing trend analysis for climate change scenarios using data from four weather stations located in Vancouver, Canada. The KWL (2002) also used trend analysis to investigate the impacts of increased rainfall in the city sewage and drainage systems due to climate change in the region.

Overall, compared with the analysis of changes in rainfall due to climate change, relatively less work has been performed on the analysis of climate change effects on urban stormwater runoff using trend analysis. Moreover, stormwater model calibration was not done in former studies. No previous study has reported on the Gunja Drainage Basin, which may have significant differences in land use, regional climate and trends from those observed in other locations.

The focus of this study is on the estimation of design rainfall intensity to reflect future climate changes. In addition, this study also estimates the effects of the variation of rainfall events on the increase in runoff based on proper calibration of a stormwater model, thus ensuring the validity of the results for future scenarios. The design rainfall intensity and peak discharge from the urban drainage basin are estimated for the target years (2020 and 2050) by using linear regression and implementation of the US Environmental Protection Agency (EPA) Storm Water Management Model (SWMM) version 5.0 (Rossman 2007).

MATERIALS AND METHODS

Study area

The study area is the Gunja Drainage Basin located in Seoul City. It consists of residential and business areas with a total impervious area of approximately 85%. The total area of the drainage basin is 0.964 km2 and the maximum overland flow length is 3,910.6 m with a 1.4% uniform slope. The land use in the watershed consists of impervious areas (residential, commercial, road, school, and public areas) and pervious areas (park and green areas). The classification of the soil according to the Soil Conservation Service (SCS) hydrologic soil groups is mainly type B, which corresponds to shallow loess and sandy loam (Chow et al. 1988). Soil group classification was used for determining the runoff curve number for infiltration.

Climate change scenario

Linear regression analysis for future climate change and rainfall simulation on the study area was performed using the annual maximum series of observation data from the Seoul Weather Station located at the shortest distance to the Gunja Drainage Basin. The station is operated by the Korea Meteorological Administration (KMA 2014). The KMA performs data quality control through a World Meteorological Organization (WMO)-compliant unified data management and control system according to WMO guidance. A considerable length of time series data is usually preferred in order to properly study climate change signals. The authors consider that the rainfall records used in this study are relatively shorter, due to limitation in data availability in the study area. However, these can still be useful in analyzing climate change effects as in previous studies (Denault et al. 2006; Jones et al. 2014; Zhao et al. 2014). Thus, 10-minute and 1-hour rainfall data from 1971 to 2011 obtained from the weather station were used in this study. For the 24-hour duration, data from 1961 to 2011 were used. With respect to this study, linear regression analysis for the climate change scenarios used data from three rainfall durations (10-minute, 1-hour, and 24-hour). According to the data obtained from the WAter Management Information System database (WAMIS 2013), the impervious area ratio in the drainage basin rapidly increased from 43.3 to 75.5% over the period from 1975 to 1985 due to urbanization. Hence, the year 1975 was selected as the base year to represent the pre-developed condition in the study area.

The trend analysis involves two steps. First, the 10-minute, hourly, and daily rainfall duration data gathered from the rain gauge station are analysed using linear regression analysis to extrapolate future climate change scenarios; then, the p-value is used to check the significance of the trend. The main characteristics of rainfall such as intensity, duration, and return period can be represented graphically in the form of Intensity-Duration-Frequency (IDF) curves. An IDF curve is very useful in the design of the various hydraulic structures. It can be constructed through the use of enough (over 30 years) annual maximum data series for rainfall durations ranging from 5 to 1,440 minutes. Limited data were available from the Seoul Weather Station to generate an IDF curve for a 100-year return period. However, in some situations, particularly when only a few years of data are available (less than 20 to 25 years), an alternative method for analysis can be used (Chow et al. 1988). This approach has also been used in previous studies (Wheater & Bell 1983; Denault et al. 2006; Zainudini et al. 2011) where annual maximum values of rainfall or runoff for a given duration were ranked and fitted to a Gumbel distribution, which is known to be the best distribution for such data (Chow et al. 1988).

The Gumbel (or Extreme Value Type I distribution), a probability density function, was used to calculate the frequency factor for determining the design rainfall. Chow (1953) derived the expression for frequency factor (kt) given below 
formula
1
where T is the return period (say 2, 5, 10, 25, 50, 100 years). The frequency factor calculated from Equation (1) is substituted in Equation (2) to calculate the design rainfall, , for a forecast year. 
formula
2
where is the projected rainfall intensity from the regression for a forecast year (mm/hr) and is the standard error of estimate. The is taken from the RMSE (Root Mean Squared Error) given by 
formula
3
The for a forecast year is determined using the following regression equation obtained by fitting historical rainfall data on a log-log scale: 
formula
4
where A and B are constants of the regression equation, and D is the rainfall duration in hours.

For a design rainfall event, fourth quartile 50-percent distribution of the Huff method (Huff 1967) was used for the temporal distribution of the design storm as it was found that the greatest peak flow from runoff modelling occurred for this quartile distribution. Results of the Huff method distribution were used as input data to the runoff model.

Urban runoff model

The SWMM developed by the US EPA is known as one of the best and most well-known urban runoff models used to simulate the hydrological processes in watersheds (Huber & Dickinson 1988; Roesner et al. 1988). Using the results of climate change scenarios, runoff simulation was performed using SWMM for the return period (10-year, 100-year) and rainfall duration (10-minute, 1-hour) of the base year (1975) and forecast years (2020 and 2050). The parameters required in the SWMM model comprised topography, soil, conduit, and climate data. Typically, the following steps are carried out when using SWMM to model stormwater runoff over a study area (Rossman 2007):

  1. Specify a default set of options and object properties to use.

  2. Draw a network representation of the physical components of the study area.

  3. Edit the properties of the objects that make up the system.

  4. Select a set of analysis options.

  5. Run a simulation.

  6. View the results of the simulation.

Flow routing methods included none, steady, kinematic and dynamic wave options. Infiltration losses were estimated using either the Horton, Green-Ampt or Soil Conservation Service Curve Number (SCS-CN) formulae. For this study, the dynamic wave option was used for flow routing. The rainfall data obtained from each climate change scenario were used as input to the model while the rest of the model parameter values used were the same as the values from the previous work for the same drainage basin (Lee et al. 2010) as shown in Table 1. In this previous work, runoff parameters were calibrated using observed data.

Table 1

Calibrated runoff parameters from Lee et al. (2010) 

Parameter Value 
Impervious Manning's, n 0.02 
Pervious Manning's, n 0.04 
Impervious initial loss (mm) 2.54 
Pervious initial loss (mm) 5.08 
Conductivity (mm/hr) 6.25 
Drying time (days) 
Parameter Value 
Impervious Manning's, n 0.02 
Pervious Manning's, n 0.04 
Impervious initial loss (mm) 2.54 
Pervious initial loss (mm) 5.08 
Conductivity (mm/hr) 6.25 
Drying time (days) 

RESULTS AND DISCUSSION

The results of the linear regression analysis were used to estimate the rainfall changes from the annual maximum series of data for the 10-minute and 1-hour durations, after which, runoff simulation using the SWMM model was performed for past and future scenarios. Results from the trend analysis of the annual series of maximum rainfall data are shown in Figure 1. P-value was used to check the validity of the statistical trend. P-value represents the actual level of significance of a trend where a value smaller than 0.10 indicates a level of significance greater than 90%. Statistically significant trends were obtained for the 1-hour and 24-hour durations. The linear regression analysis results from the annual series of maximum rainfall data are shown in Table 2, including the slope, intercept, and p-value for all durations. The values of 10-year and 100-year return period peak discharge (Q10 and Q100, respectively) due to climate change for the base and forecast years are shown in Table 3.

Table 2

Statistics summary for annual maximum time series

Duration Slope (mm/hr/yr) Intercept (mm/hr) p-value (90% significant level) 
10 minutes 0.5932 97.744 0.101 
1 hour 0.3731 46.535 0.076 
24 hours 0.0541 6.423 0.043 
Duration Slope (mm/hr/yr) Intercept (mm/hr) p-value (90% significant level) 
10 minutes 0.5932 97.744 0.101 
1 hour 0.3731 46.535 0.076 
24 hours 0.0541 6.423 0.043 
Table 3

Peak discharge modelling results considering climate change impacts

  Duration (10-min) Duration (1-hour) Duration (24-hour) 
Year Q10 (m3/s) Q100 (m3/s) Q10 (m3/s) Q100 (m3/s) Q10 (m3/s) Q100 (m3/s) 
1975 27.4 41.9 27.1 43.9 7.9 12.1 
2020 43.7 60.8 41.9 57.9 9.6 13.7 
2050 53.3 60.1 48.6 64.6 11.1 15.2 
  Duration (10-min) Duration (1-hour) Duration (24-hour) 
Year Q10 (m3/s) Q100 (m3/s) Q10 (m3/s) Q100 (m3/s) Q10 (m3/s) Q100 (m3/s) 
1975 27.4 41.9 27.1 43.9 7.9 12.1 
2020 43.7 60.8 41.9 57.9 9.6 13.7 
2050 53.3 60.1 48.6 64.6 11.1 15.2 
Figure 1

Linear regression for annual maximum time series.

Figure 1

Linear regression for annual maximum time series.

Although scattered, the quality of data fit to the regression line, in general, was found to be good (p-values <0.1), and the results showed a generally increasing trend during the considered time period, as also evidenced by the positive slope for all durations (Figure 1). These results, which suggest the increasing trend of rainfall intensity for future climate change scenarios as also seen in the developed IDF curves (Figure 2), are compatible with those found in other studies (Denault et al. 2006; Willems 2011). Statistically significant trends are obtained for the 1-hour and 24-hour durations. An increasing trend is also found for the 10-minute duration. However, the p-value for this duration is 0.101, a result that is nearly significant at the 90% level. The statistical summary is also suggestive that the slope increases as the duration becomes shorter. This trend is consistent with the observations of Denault et al. (2006) and Dourte et al. (2013), which imply that the increasing rate of rainfall intensity for the future is higher for shorter rainfall durations. Rainfall characteristics may vary greatly in different regions and can be affected by factors other than climate change. For instance, Berz (1997) anticipates future increases in the frequency and severity of storms for various parts of the world, due to the shifting of climate zones and the increasing intensity of convective processes.

Figure 2

IDF curves for the (a) 10-year and (b) 100-year return period.

Figure 2

IDF curves for the (a) 10-year and (b) 100-year return period.

Based on the results of rainfall analysis, increased intensities of rainfall events are also likely to result in increased runoff volume. Runoff modelling results using SWMM are summarized in Tables 3 and 4, and they show that the modelled peak discharge and total runoff volume for all rainfall durations and both return periods also follow an increasing trend. The modelling results for future scenarios are generally increased when compared to the base year. The greatest increase in peak discharge of approximately 50%, and increase in total runoff volume of approximately 150% was observed for the 10-minute duration with a 10-year return period from the base year 1975 to year 2050. An increasing trend in predicted peak discharge and runoff volume for future scenarios is expected for urban areas with an increasing impervious ratio as had been reported in the literature (Bronstert et al. 2002; Nguyen et al. 2010).

Table 4

Runoff volume modelling results considering climate change impacts

  Duration (10-min) Duration (1-hour) Duration (24-hour) 
Year Q10 (m3Q100 (m3Q10 (m3Q100 (m3Q10 (m3Q100 (m3
1975 10,394 14,352 31,336 50,463 218,623 344,177 
2020 23,183 30,484 58,464 80,301 286,942 410,049 
2050 26,510 33,026 67,433 89,474 331,284 454,829 
  Duration (10-min) Duration (1-hour) Duration (24-hour) 
Year Q10 (m3Q100 (m3Q10 (m3Q100 (m3Q10 (m3Q100 (m3
1975 10,394 14,352 31,336 50,463 218,623 344,177 
2020 23,183 30,484 58,464 80,301 286,942 410,049 
2050 26,510 33,026 67,433 89,474 331,284 454,829 

It should be noted, however, that the trend in Figures 1 and 2 may not be considered a consequence of climate change with absolute certainty. More importantly, it has been acknowledged that the length of record is too short to determine whether this trend will persist into the future or whether it is part of a longer-term cycle. However, despite uncertainty in relation to the effects and magnitude of climate change, it is practical to consider that such trends may well persist into the future since infrastructure is being designed and built today with an intended service life of 50 years or more (Denault et al. 2006). Less water storage capacity in urban basins and more rapid runoff means that higher peak discharge rates and greater total volumes of water discharged during floods are expected.

Development along stream channels and floodplains can also alter the capacity of a channel to convey water. Sediment and debris carried by floodwaters can cause blockage of drainage systems and increase flooding. This can often be observed upstream of culverts, bridges, or other places where debris collects. Small stream channels can be filled with sediment or become clogged with debris, because of undersized culverts, for example. This creates a closed basin with no outlet for runoff (Konrad 2003). Thus, in addition to the effects of climate change on rainfall statistics, these hydraulic factors should also be considered and incorporated in the planning of drainage design.

CONCLUSIONS

In this study, trend analysis was used to predict climate change scenarios. The impact of predicted scenarios on the runoff in the watershed was determined based on the SWMM simulation. The significance of trend analysis was determined using the p-value with a 90% confidence level. Linear regression analysis was found to be significant for the 1-hour and 24-hour duration (p-values less than 0.10), while the 10-minute duration is nearly significant at the 90% level (p-value equal to 0.101). Thus, the predicted scenarios show overall increases in rainfall with climate change, largely brought about by an increase in rainfall intensity. Among all climate change scenarios, the shortest duration (10-minute) scenario appeared to be the most critical situation where the slope of linear regression is steepest and increase in peak discharge and runoff volume is greatest. Increases of high-intensity short-duration rainfall suggest that the design of hydraulic structures should consider design rainfall for shorter durations.

Although some uncertainties are associated with future climate change scenarios, it is prudent to consider the practical implications for infrastructure planning and design, taking into account the possibility of increases in rainfall intensity as indicated by the results of this study. The predicted peak discharge obtained from the proposed framework can provide a means to determine whether the capacity of existing stormwater infrastructure is sufficient for current conditions, and also whether it will be sufficient for future scenarios. Consideration of climate change can therefore lead to reconstruction and/or upgrade of existing infrastructure.

Furthermore, the effects of climate change on urban runoff can be incorporated in the review and update of current standards for stormwater infrastructure design. As a means of remediating the increase in peak discharge, reduction can be achieved by increasing the amount of pervious areas in highly urbanized catchments. This can be done by incorporating Best Management Practice and Low Impact Development into urban planning and design.

As indicated by the results of this study, climate change has an impact on rainfall trends. However, there are other factors that could have had an impact on the recorded rain, which have not been explored in this paper, such as increased urbanization, which may lead to heat islands and result in rainfall changes (Pathirana et al. 2014). This is an area that deserves further attention in future research.

ACKNOWLEDGMENT

This research was supported by a Korea University Grant.

REFERENCES

REFERENCES
Berz
G. A.
1997
Catastrophes and climate change: risks and reactions from the viewpoint of an international re-insurer
.
Eclogae Geol. Helv.
90
(
3
),
375
379
.
Chow
V. T.
1953
Frequency analysis of hydrologic data with special application to rainfall intensities
.
Bulletin Series No. 414
,
Univ. of Illinois Engineering Experiment Station
,
Urbana, IL
.
Chow
V. T.
Maidment
D. R.
Mays
L. W.
1988
Applied Hydrology
.
McGraw-Hill
,
New York, USA
.
Denault
C.
Millar
R. G.
Barbara
J. L.
2006
Assessment of possible impacts of climate change in an urban catchment
.
J. Am. Water Resour. Assoc.
42
(
3
),
685
697
.
Houghton
J. T.
Meira Filho
L. G.
Callander
B. A.
Harris
N.
Kattenberg
A.
Maskell
K.
1996
Climate Change 1995: The Science of Climate Change. Contribution of Working Group 1 to the Second Assessment Report of the Intergovernmental Panel on Climate Change
,
Cambridge University Press
,
Cambridge, UK
.
Huber
W. C.
Dickinson
R. E.
1988
Storm Water Management Model Version 4 User's Manual
,
EPA/600/3–88/001a
,
Environmental Protection Agency
,
Athens, GA, USA
.
Huff
F. A.
1967
Time distribution of rainfall in heavy storms
.
Water Resour. Res.
3
(
4
),
1007
1019
.
IPCC
2007
Issues related to mitigation in the long term context, Climate Change 2007: Mitigation, Contribution of Working Group III to the Fourth Assessment Report of the Inter- Governmental Panel on Climate Change
,
Cambridge University Press
,
Cambridge, UK and New York, NY, USA
.
KMA
2014
Korea Meteorological Administration
: (
accessed 8 July 2014
).
Konrad
C. P.
2003
Effects of Urban Development on Floods
.
USGS Fact Sheet FS-076–03
. (
accessed 6 July 2014
).
KWL
2002
Development of GVRD Precipitation Scenarios
,
Greater Vancouver Regional District
,
Canada
.
Lee
K.
Kim
H.
Pak
G.
Jang
S.
Kim
L.
Yoo
C.
Yun
Z.
Yoon
J.
2010
Cost-effectiveness analysis of stormwater best management practices (BMPs) in urban watersheds
.
Desalin. Water Treat.
19
(
1–3
),
92
96
.
Niemczynowicz
J.
1989
Impact of the greenhouse effect on sewerage systems
.
Hydrol. Sci. J.
34
(
6
),
651
666
.
Pagliara
S.
Viti
C.
Gozzini
B.
Meneguzzo
F.
Crisci
A.
1998
Uncertainties and trends in extreme rainfall series in Tuscany, Italy: effects on urban drainage networks design
.
Water Sci. Technol.
37
(
11
),
195
202
.
Pathirana
A.
Denekew
H. B.
Veerbeek
W.
Zevenbergen
C.
Banda
A. T.
2014
Impact of urban growth-driven landuse change on microclimate and extreme precipitation – A sensitivity study
.
Atmos. Res.
138
,
59
72
.
Roesner
L. A.
Aldrich
J. A.
Dickinson
R. E.
1988
Stormwater Management Model User's Manual, Version 4: Extran Addendum
.
EPA/600/3–88/001b
,
Environmental Protection Agency
,
Athens, GA, USA
.
Rossman
L. A.
2007
Storm Water Management Model User's Manual Version 5.0
,
EPA/600/R-05/040, Revised 2007
RMRL
,
Cincinnati, OH, USA
.
Sansom
J.
Renwick
J. A.
2007
Climate change scenarios for New Zealand rainfall
.
J. Appl. Meteor. Climatol.
46
,
573
590
.
Sheshukov
A. Y.
Siebenmorgen
C. B.
Douglas-Mankin
K. R.
2011
Seasonal and annual impacts of climate change on watershed response using an ensemble of global climate models
.
Trans. ASABE
54
(
6
),
2209
2218
.
WAMIS
2013
WAter Management Information System
: (
accessed 30 October 2013
).
Wheater
H. S.
Bell
N. C.
1983
Northern Oman flood study
.
ICE Proc.
75
(
3
),
453
473
.
Willems
P.
2011
Revision of urban drainage design rules based on extrapolation of design rainfall statistics, Proceedings of the 11th International Conference on Urban Drainage, Porto Alegre, Brazil
.
Zainudini
M. A.
Marriott
M. J.
Mirjat
M. S.
Chandio
A. S.
2011
Establishing intensity duration frequency curves for Sistan and Balochistan provinces of Iran
.
Pak. J. Agri., Agril. Eng., Vet. Sci.
27
(
2
),
115
124
.