U.S. Geological Survey

Iowa Water Science Center

Iowa City, IA 52240

319-358-3615

daeash@usgs.gov

by David Eash

Long-Term and Recent-Term Trend Analysis Results for Floods, High Flows, and

Low Flows in IowaPresented at the 2016

Iowa Water ConferenceAmes, IA

March 24, 2016

Iowa streamflow statistics study

Compute streamflow statistics for all Iowa

continuous-record streamgages for both the entire

period of record and for the 1984-2013 record

Statistical analyses include annual probabilities for

instantaneous peak flow, high-flow for 1-, 3-, 7-, 15,

and 30-day consecutive periods, low-flows for 1-, 3-,

7-, 14-, 30-, 60-, 90, 120-, and 183-day consecutive

periods, and 23 flow-duration statistics that include 1

– 99% exceedance probabilities

https://pubs.er.usgs.gov/publication/ofr20151214

Study cooperators

Iowa DOT and Iowa Highway Research Board

U.S. Army Corps of Engineers

USGS

Report is an update of two previously published reports that presented statistical summaries of selected Iowa streamflow data through 1988 & 1996

USGS student help with statistical analyses:

Padraic O’Shea Kevin Nguyen

Adrian Simonson Nicholas Montgomery

Stationarity

One of the key assumptions of streamflow probability analyses

Stationarity, is the idea that the flood-, or high-flow-, or low-flow-generating mechanism fluctuates within an unchanging envelope of variability

Stationarity assumes that future streamflow probabilities can be estimated on the basis of past streamflow probabilities

Stationarity Is Dead: Whither Water Management?

http://www.gfdl.noaa.gov/bibliography/related_files/pc

m0801.pdf (Milly and others, 2008)

Trend analyses

Kendall’s tau hypothesis test was used to

analyze for trends for both the entire period of

record (long-term analyses) and for the 1984-

2013 period of record (recent-term analyses) for

184 continuous-record streamgages in Iowa

For this presentation, a subset of these 184

streamgages with complete records for the

period 1964-2013 were selected for evaluation

Trend analysis results are presented for 55 USGS

streamgages in Iowa

Kendall’s Tau trend analyses

Computes a monotonic relation between discharges

and time (for water years or climatic years)

A nonparametric test that indicates the likelihood of

a positive or negative trend with time

Kendall’s tau test compares the rank of each annual

discharge value with the rank of values following it

in the series – if the second value is consistently

greater than the first, the Kentau coefficient is

positive, if the second value is consistently lower,

the Kentau coefficient is negative

Kendall’s Tau trend analyses (cont.)

The Kentau coefficeint value is a measure of the

correlation between the series and time

A Ken p-value threshold of 5% (α=0.05) was used

for the Kendall’s tau test, and Ken p-values less

than or equal to 5% indicate statistically significant

trends (positive or negative)

Results may be sensitive to multiyear sequences of

larger or smaller values on either end of the record

Kendall’s tau trend tests were performed using

USGS SWSTAT and PeakFQ programs

Location of 55 streamgages

Drainage areas ranged from 25.3 to 85,600 mi2

Mean drainage area size = 3,053 mi2 and median = 846 mi2

Complete streamflow records from 1964 to 2013

Record lengths of 55 streamgages

8 14 17 9 51 1

50-59 60-69 70-79 80-89 90-99 100-109

110-119

Nu

mb

er o

f st

ream

gag

es

Years of record for trend analyses

Mean = 74 years

Median = 73 years

Magnitude of streamflow statistics for

Cedar River at Cedar Rapids 05464500High flows

Discharge

(ft3/s) Low flows

Discharge

(ft3/s)

Peak flow (WIE, RI=50) 81,700 50% exceeds 2,270

High 1-day (LP3, RI=50) 80,400 NWS 75% exceeds 1,150

High 3-day “ 74,100 action 90% exceeds 694

High 7-day “ 57,600 95% exceeds 544

High 15-day “ 42,200 for 99% exceeds 345

High 30-day “ 30,700 Low 90-day (LP3, RI=50) 331

1% exceeds 26,200 Low 60-day “ 299

2% exceeds 20,400 Low 30-day “ 269

5% exceeds 13,100 Low 14-day “ 250

10% exceeds 8,960 Low 7-day “ 239

25% exceeds 4,550 Low 3-day “ 229

Low 1-day “ 205

stage

flooding=

24,600 ft3/s

Trend analysis results for high flows

for period of record for 55 streamgages

Streamflow

No. of significant

trends (p<0.05)

Percent with

significant trends

Number with

positive trends

Number with

negative trends

Peak flow 10 18 8 2

High 1-day 9 16 9 0

High 3-day 9 16 9 0

High 7-day 12 22 12 0

High 15-day 20 36 20 0

High 30-day 25 45 25 0

1% exceeds 10 18 9 1

2% exceeds 16 29 16 0

5% exceeds 30 55 30 0

10% exceeds 44 80 44 0

25% exceeds 46 84 46 0

Trend analysis results for low flows for

period of record for 55 streamgagesStreamflow

No. of significant

trends (p<0.05)

Percent with

significant trends

Number with

positive trends

Number with

negative trends

50% exceeds 44 80 44 0

75% exceeds 49 89 49 0

90% exceeds 50 91 50 0

95% exceeds 50 91 50 0

99% exceeds 50 91 50 0

Low 90-day 44 80 44 0

Low 60-day 46 84 46 0

Low 30-day 53 96 53 0

Low 14-day 55 100 55 0

Low 7-day 55 100 55 0

Low 3-day 52 95 52 0

Low 1-day 51 93 51 0

Trend analysis results for high flows for

1984-2013 record for 55 streamgages

Streamflow

No. of significant

trends (p<0.05)

Percent with

significant trends

Number with

positive trends

Number with

negative trends

Peak flow 1 2 1 0

High 1-day 0 0 0 0

High 3-day 0 0 0 0

High 7-day 0 0 0 0

High 15-day 0 0 0 0

High 30-day 0 0 0 0

1% exceeds 0 0 0 0

2% exceeds 0 0 0 0

5% exceeds 0 0 0 0

10% exceeds 0 0 0 0

25% exceeds 0 0 0 0

Trend analysis results for low flows for

1984-2013 record for 55 streamgagesStreamflow

No. of significant

trends (p<0.05)

Percent with

significant trends

Number with

positive trends

Number with

negative trends

50% exceeds 0 0 0 0

75% exceeds 0 0 0 0

90% exceeds 0 0 0 0

95% exceeds 0 0 0 0

99% exceeds 2 4 2 0

Low 90-day 0 0 0 0

Low 60-day 1 2 0 1

Low 30-day 0 0 0 0

Low 14-day 0 0 0 0

Low 7-day 0 0 0 0

Low 3-day 0 0 0 0

Low 1-day 0 0 0 0

Trend analysis results for high flows

for years of record for 53 streamgages Percent of significant positive trends for

Streamflow

50-59 yrs,

n = 8 (1955-63)

60-69 yrs,

n = 14 (1946-54)

70-79 yrs,

n = 17 (1935-42)

80-89 yrs,

n =9 (1915-34)

90-99 yrs,

n = 5 (1903-15)

Peak flow 12.5 0 12 44 20

High 1-day 12.5 0 6 0 20

High 3-day 12.5 0 12 0 20

High 7-day 25 0 18 0 40

High 15-day 37.5 36 47 11 40

High 30-day 50 36 59 11 60

1% exceeds 25 7 24 0 40

2% exceeds 25 36 59 0 40

5% exceeds 87.5 50 76 22 80

10% exceeds 100 86 88 78 100

Percent of streamgages, out of total of 55, with

statistically significant positive trends for the POR

Per

cen

t o

f st

ream

gag

es

Streamflow statistic

p<0.05

Seasonal low-flow trends

Winter Spring

Summer Fall

Per

cen

t o

f st

ream

gag

es

1-day7 14 30

1 7 14 30

POR

1984-2013

Summary of trend analysis results

Sig trends are positive, few negative trends

Long-term (period of record) analyses: high flows

indicate sig trends for 16-45% of streamgages and low

flows indicate sig trends for 80-100% of streamgages,

strong pattern of increasing number of trends from

high-flow to low-flow statistics – notable increase

around 5-10% exceedance flows, maybe some trend

differences in high flows due to record lengths

Recent-term (1984-2013) analyses: no trends (2-4%)

for entire range of streamflow statistics

Possible causes for period of record

streamflow trends in Iowa

High flows: Proportion of total precipitation

contributed by extreme, one-day events has

increased significantly during the last century;

although increases in precipitation are modest, they

are concentrated in the high-flow statistics

Low flows: Increases in row-crop production are

related to increases in base flows; increased base

flows could also be caused by improved conservation

practices, added artificial drainage, and channel

incision (Schilling, 2005; Schilling and Libra, 2003)

Climate the Main Cause of Increasing

Streamflow in Eastern South Dakota

Increased precipitation is the primary cause of rising

streamflow in many eastern SD streams.

USGS analyzed streamflow and precipitation trends

from 1945 to 2013 for 10 South Dakota streamgages.

Land-use changes likely are minor factors in rising

streamflow, with the main factors probably being

changes in the timing and frequency of large

precipitation events and persistently wetter preceding

conditions.

http://dx.doi.org/10.3133/sir20155146

Flow-duration curves for Cedar

River at Cedar Rapids 054645001,000,000

300,000

100,000

100

300

1,000

3,000

10,000

30,000

Dis

char

ge,

in c

ub

ic f

eet

per

sec

on

d

Percentage of time indicated discharge is equaled or exceeded

0 10010 20 30 40 50 60 70 80 90

Mean daily mean discharges for Cedar

River at Cedar Rapids 05464500

DayOct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep

1,000

2,000

3,000

4,000

5,000

6,000

7,0008,0009,000

10,000

20,000D

isch

arg

e, in

cu

bic

fee

t p

er s

eco

nd

Annual mean discharges for Cedar

River at Cedar Rapids 05464500

Water Year1900 1920 1940 1960 1980 2000 20200

2,000

4,000

6,000

8,000

10,000

12,000

14,000

16,000

18,000

19,000D

isch

arg

e, in

cu

bic

fee

t p

er s

eco

nd

Annual mean discharge tests

Wilcoxon rank-sum test computed for each of the 55

streamgages, a nonparametric test - to test whether two

independent, paired records have the same medians

Annual mean discharges for each streamgage from the

beginning of record to 1980 and from 1984-2013 were

tested (1st data set ranged from 18-78 yrs of record (ave.

43 yrs of record), 2nd data set had 30 yrs of record)

39 of 55 streamgages (71%) had statistically significant

differences between the median of the annual mean

discharges measured prior to 1981 compared to the

median of annual mean discharges measured after 1983

Conclusions

Sig trends in Iowa are positive, few negative trends

Long-term (period of record) analyses: Few

streamgages (16-45%) with trends for high flows, many

streamgages (80-100%) with trends for low flows

Recent-term (1984-2013) analyses: No trends (2-4%) for

entire range of streamflow statistics

Assumption of stationarity is a concern, to minimize the

bias of significant trends, streamflow statistics, other

than peak flow, can be computed using a common or a

variable record length without a significant trend

QUESTIONS