Statewide Riverine Flood Vulnerability Assessment

Emmanouil Anagnostou and Xinyi ShenCivil and Environmental Engineering, UCONN

May 11, 2018 CIRCA Forum

TopicsFlood Frequency Analysis (FFA)

Flood Vulnerability Analysis 

Flood Risk of Road Crossings

Regional Flood Frequency (RFFA)

Flood Inundation Early Warning System

CREST‐SVASA new 

numerical modelling framework

Atmospheric Reanalysis

• NLDAS‐2: North American Land Data Assimilation System (1979‐present, 1h/12km) 

Future Weather Projection

• ClimEX project‐150 years reanalysis

0

500

1000

1500

2000

2500

3000

3500

4000

Date

1990

/7/2

1991

/1/1

1991

/7/3

1992

/1/2

1992

/7/3

1993

/1/2

1993

/7/4

1994

/1/3

1994

/7/5

1995

/1/4

1995

/7/6

1996

/1/5

1996

/7/6

1997

/1/5

1997

/7/7

1998

/1/6

1998

/7/8

1999

/1/7

1999

/7/9

2000

/1/8

2000

/7/9

2001

/1/8

2001

/7/10

2002

/1/9

2002

/7/11

2003

/1/10

2003

/7/12

2004

/1/11

2004

/7/12

2005

/1/11

2005

/7/13

2006

/1/12

2006

/7/14

2007

/1/13

2007

/7/15

2008

/1/14

2008

/7/15

2009

/1/14

2009

/7/16

2010

/1/15

2010

/7/17

2011

/1/16

2011

/7/18

2012

/1/17

2012

/7/18

R_Obs RNSCE=0.58(1990‐2012)NSCE=0.72(2002‐2012)

Thompsonville, Connecticut River, MA (9660 mi2)

Stevenson, CT, Housatonic River (1544 mi2)

0

200

400

600

800

1000

1200

1400

1600

Date

3/13/2002

5/24/2002

8/4/2002

10/15/2002

12/26/2002

3/8/2003

5/19/2003

7/30/2003

10/10/2003

12/21/2003

3/2/2004

5/13/2004

7/24/2004

10/4/2004

12/15/2004

2/25/2005

5/8/2005

7/19/2005

9/29/2005

12/10/2005

2/20/2006

5/3/2006

7/14/2006

9/24/2006

12/5/2006

2/15/2007

4/28/2007

7/9/2007

9/19/2007

11/30/2007

2/10/2008

4/22/2008

7/3/2008

9/13/2008

11/24/2008

2/4/2009

4/17/2009

6/28/2009

9/8/2009

11/19/2009

1/30/2010

4/12/2010

6/23/2010

9/3/2010

11/14/2010

1/25/2011

4/7/2011

6/18/2011

8/29/2011

11/9/2011

R R_ObsNSCE=0.62CC=0.81Bias=‐6.5%

Salmon Kill, CT, Housatonic River (29 mi2)

0

5

10

15

20

25

30

35

Date

10/4/1990

2/6/1991

6/11/1991

10/14/1991

2/16/1992

6/20/1992

10/23/1992

2/25/1993

6/30/1993

11/2/1993

3/7/1994

7/10/1994

11/12/1994

3/17/1995

7/20/1995

11/22/1995

3/26/1996

7/29/1996

12/1/1996

4/5/1997

8/8/1997

12/11/1997

4/15/1998

8/18/1998

12/21/1998

4/25/1999

8/28/1999

12/31/1999

5/4/2000

9/6/2000

1/9/2001

5/14/2001

9/16/2001

1/19/2002

5/24/2002

9/26/2002

1/29/2003

6/3/2003

10/6/2003

2/8/2004

6/12/2004

10/15/2004

2/17/2005

6/22/2005

10/25/2005

2/27/2006

7/2/2006

11/4/2006

3/9/2007

R_Obs R

Flood Frequency Maps (2‐200‐yr) of CT   

FFA – error analysis

0119700001203805

0

20

40

60

80

100

120

Basin Area (km2)

122 USGS stream Gauges

Error=|sim‐obs|/obs

Flood Vulnerability Analysis

• Study Area– Naugatuck River – Thomaston Dam in the middle of the River

– Critical Infrastructure at Freight St., Waterbury, CT

Thomaston Dam inflow, Naugatuck River, CT45 events , 9/36‐‐calibration/validation (9 years record)

‐100

0

100

200

300

400

500

Date

2006/01/15:18

2006/05/13:04

2006/06/30:17

2006/07/09:13

2006/07/18:09

2006/07/27:05

2006/08/05:01

2006/11/18:16

2007/04/14:06

2008/02/18:15

2008/03/11:23

2008/12/27:00

2009/06/18:06

2010/01/26:12

2010/12/07:05

2010/12/16:01

2010/12/24:21

2011/01/02:17

2011/03/06:05

2011/04/18:21

2011/08/17:11

2011/08/26:07

2011/09/04:03

2011/09/12:23

2011/09/21:19

2011/09/07:07

2011/09/30:01

2011/10/08:21

2011/10/17:17

2011/10/26:13

2011/11/04:09

2011/10/03:15

2013/01/30:18

2013/05/25:13

2013/06/03:09

2013/06/12:05

2013/06/21:01

2013/06/29:21

2013/06/07:08

2013/06/16:04

2014/04/01:08

R_Obs RNSCE=0.6997CC=0.85Bias=‐6.3%

Synthetic Hydrograph• Timing • A power‐law frequency adjustment

100 101 102 103101

102

103

104

return period (in year)

stre

am fl

ow in

m3 /s

Upstream to Thomaston Dam

Downstream to Thomaston Dam

Hydraulic Simulation by HEC‐RAS• River Profiling using LIDAR derived DEM of 1 m resolution

Inundation ‐ different dam operation scenarios200 yr return flood 500 yr return flood

BuildingTransformer

Road Crossing Flood Risk Assessment ‐ Housatonic River

From Michael Jastremski, HVA

An Example: Salmon Kill watershed

Summary of the Results

56%

315 culverts modeled for Risk of Failure

86 culvertsFail within 25-Year interval

48 culvertsModerate to Severe barriers

(56% of those failing within the 25-Year interval)

From Michael Jastremski, HVA

Summary• A fully‐physically based numerical hydrological framework, CREST‐SVAS, forced with atmospheric reanalysis has been developed to estimate flood frequency in mid‐ and northern‐latitude basins

• CREST‐SVAS has been applied to assess flood‐inundation and flood overtopping risks in conjunction with hydraulic models & LiDAR data

Work in progress• A real‐time early warning system of flood vulnerability based on weather‐forecasting driven CREST‐SVAS, in combination with storm & tide forecasts

• A novel regional flood frequency estimation method combining the strength of CREST‐SVAS numerical simulations and USGS network observations

Und

er develop

ment: Re

al‐

Time Forecast System

Slide 21

SX1 This figure needs to be updated according to the new dataShen, Xinyi, 4/10/2018

RFFA Method According to IUH theory, the runoff of a basin can bedecomposed to that of its sub‐basins in the following form,

（1）

Carrying out Fourier transformation of (1), we have

（2）

Where  stands for the time delay of the ith sub basin to the outlet       and        stands  for the runoff of at the outlet and the ith sub‐basin.

where  is the Fourier transformation of  given by equation (3)

（3）

In a general case illustrated by the right Figure, theequation (4) could be derived from equation (2)

（4）

In order to solve FFA at no data location, the first step is to solve (4)  for  and           .

No Data Location

Gauge Station

To extend the length of observation data, assuming simulationand observation satisfy a power law relationship

（5）

For yellow region, also could use (2) and derive equation (6)

（6）Simultaneous equations (4), (5) and (6) in the low frequencyrange, could solve , and . Finally, RFFA of nodata location could be derived from equation (7).

（7）

01131500(1514)

01138500(2644)

01144500(4092)

01154500(5493)

01170500 (7860)

01172010 (8332)

01184000 (9660) 61.12%

31.88%

66.12%

57.86%

56.89%

57.55%

49.49%

Use daily Obs and Sims data of 7 main station on C.T river

ScenarioⅠ

（5）

（6）

（7）

ScenarioⅡ Eq(4),(6)&(7)

RFFA results

Eq(6)

① Power Law

② Same Frequency

③ IUH Theory

Ⅱ:Ⅰ Ⅱ:Ⅰ:B

Statistical Approaches• Average stand errors of streamstats

• Large error• Short record length• No future projection• Subjective

2 10 25 50 100 500

CT 31.8 32.7 34.4 35.9 37.6 45.0

NJ 49 63.9

NC 25.0 73.3

OK 31 46

Available USGS records in CT (reported in 2004)

A New RRFA method Under Construction

• Combine the strength of simulation and observation

Peak biased

Sparsely available

Record Length Future projection

Simulation √ × ‐1900 (New GLDAS)‐1949 (GLDAS)‐1979 (NLDAS)

2100

Observation × √ ‐1990 ×

01154500+01138500 01170500+01138500

01172010+01138500 01184000+01138500

0.97:1:1.83 2.60:1:1.83

2.93:1:1.83 3.84:1:1.83

Preliminary Results

01154500+01131500 01170500+01131500

01172010+01138500 01184000+0113150001172010+01131500

0.54:1:0.59 1.46:1:0.59

1.64:1:0.59 2.15:1:0.59

Preliminary Results