Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R...

32
NN-DAR NN-QPPQ NN-SMS1R NN-SMS1L NN-SMS12R NN-SMS12L NN-SM1 NN-SM12 MC-DAR MC-QPPQ MC-SMS1R MC-SMS1L MC-SMS12R MC-SMS12L MC-SM1 MC-SM12 PRMS NN-AFINCH MC-AFINCH -1.0 -0.5 0 0.5 1.0 Efficiency Method of prediction in ungaged basins Figure C1. The distribution of the at-site Nash-Sutcliffe efficiencies of daily streamflow predictions for Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The Nash-Sutcliffe efficiency is along the vertical axis. Nash-Sutcliffe efficiency ranges from one to negative infinity; a value of one indicates a perfect fit, while a value of zero indicates that a mean value would have produced the same level of accuracy. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Transcript of Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R...

Page 1: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

−1.0

−0.5

0

0.5

1.0

Effic

ienc

y

Method of prediction in ungaged basins

Figure C1. The distribution of the at-site Nash-Sutcliffe efficiencies of daily streamflow predictions for

Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions

each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The Nash-Sutcliffe efficiency is along the vertical axis. Nash-Sutcliffe efficiency ranges from one to negative infinity; a value of one indicates a perfect fit, while a value of zero indicates that a mean value would have produced the same level of accuracy. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Page 2: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

−1.0

−0.5

0

0.5

1.0

Effic

ienc

y

Method of prediction in ungaged basins

Figure C2. The distribution of the at-site Nash-Sutcliffe efficiencies of the logarithms of daily streamflow

Nash-Sutcliffe Efficiencies of the Logarithms of Daily Streamflow Predictions

predictions for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The Nash-Sutcliffe efficiency of the logarithms is along the vertical axis. Nash-Sutcliffe efficiency ranges from one to negative infinity; a value of one indicates a perfect fit, while a value of zero indicates that a mean value would have produced the same level of accuracy. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Page 3: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0

500

1,000

1,500

2,000

Root

-mea

n-sq

uare

err

or

Method of prediction in ungaged basins

Figure C3. The distribution of the at-site root-mean-square errors of daily streamflow predictions for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis indicates the root-mean-square error in units of streamflow. Lower values represent less error, on average. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Root-Mean-Square Errors of Daily Streamflow Predictions

Page 4: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0

4

2

6

8

10

Root

-mea

n-sq

uare

-nor

mal

ized

erro

r

Method of prediction in ungaged basins

Figure C4. The distribution of the at-site root-mean-square-normalized errors of daily streamflow predictions for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis indicates the root-mean-square-normalized error. Lower values represent less error, on average. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Root-Mean-Square-Normalized Errors of Daily Streamflow Predictions

Page 5: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0

–100

100

200

Aver

age

perc

ent e

rror

Method of prediction in ungaged basins

Figure C5. The distribution of at-site average percent errors of daily streamflow predictions for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the average percent bias. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Average Percent Errors of Daily Streamflow Predictions

Page 6: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0

0.4

0.2

0.6

0.8

1.0

Pear

son

corr

elat

ion

Method of prediction in ungaged basins

Figure C6. The distribution of at-site Pearson correlations between simulated and observed daily streamflows for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the Pearson correlation, which ranges from negative to positive one. A perfect correspondence would exhibit a Pearson correlation of one. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Pearson Correlations Between Simulated and Observed Daily Streamflows

Page 7: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0.3

0.6

0.5

0.4

0.8

0.7

0.9

1.0

Spea

rman

cor

rela

tion

Method of prediction in ungaged basins

Figure C7. The distribution of at-site Spearman correlations between simulated and observed daily streamflows for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the Spearman correlation, which ranges fromnegative to positive one. A perfect correspondence between ranked streamflow would exhibit a Spearman correlation of one. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Spearman Correlations Between Simulated and Observed Daily Streamflows

Page 8: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0

0.6

0.4

0.2

0.8

1.0

Effic

ienc

y

Method of prediction in ungaged basins

Figure C8. The distribution of the at-site Nash-Sutcliffe efficiencies of the daily storage-yield curve (SYC) for each method of prediction in ungaged basins (PUB) is considered here. See text for a description of the methodology used to predict the SYC. The horizontal axis indicates each PUB method. The Nash-Sutcliffe efficiency is along the vertical axis. Nash-Sutcliffe efficiency ranges from one to negative infinity; a value of one indicates a perfect fit, while a value of zero indicates that a mean value would have produced the same level of accuracy. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribu-tion, have not been drawn.) Method abbreviations are in table 1.

Nash-Sutcliffe Efficiencies of the Daily Storage-Yield Curve

Page 9: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–1.0

0

–0.5

0.5

1.0

Effic

ienc

y

Method of prediction in ungaged basins

Figure C9. The distribution of the at-site Nash-Sutcliffe efficiencies of the logarithms of the daily storage-yield curve (SYC) for each method of prediction in ungaged basins (PUB) is considered here. See text for a description of the methodology used to predict the SYC. The horizontal axis indicates each PUB method. The Nash-Sutcliffe efficiency of the logarithms is along the vertical axis. Nash-Sutcliffe efficiency ranges from one to negative infinity; a value of one indicates a perfect fit, while a value of zero indicates that a mean value would have produced the same level of accuracy. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Nash-Sutcliffe Efficiencies of the Logarithms of Daily Storage-Yield Curve

Page 10: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0

150

200

50

100

250

300

Root

-mea

n-sq

uare

d er

ror

Method of prediction in ungaged basins

Figure C10. The distribution of the at-site root-mean-square errors of the daily storage-yield curve for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis indicates the root-mean-square error in units of storage fraction (time, storage volume divided by mean streamflow). Lower values represent less error, on average. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Root-Mean-Squared Errors of the Daily Storage-Yield Curve

Page 11: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0

6

2

4

8

10

Root

-mea

n-sq

uare

-nor

mal

ized

erro

r

Method of prediction in ungaged basins

Figure C11. The distribution of the at-site root-mean-square-normalized errors of the daily storage-yield curve for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis indicates the root-mean-square-normalized error. Lower values represent less error, on average. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Root-Mean-Square-Normalized Errors of the Daily Storage-Yield Curve

Page 12: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–100

100

0

200

300

Aver

age

perc

ent e

rror

Method of prediction in ungaged basins

Figure C12. The distribution of at-site average percent errors of the daily storage-yield curve for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the average percent bias. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Average Percent Errors of the Daily Storage-Yield Curve

Page 13: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0.95

0.98

0.96

0.97

0.99

1.00

Pear

son

corr

elat

ion

Method of prediction in ungaged basins

Figure C13. The distribution of at-site Pearson correlations between the simulated and observed daily storage-yield curves for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the Pearson correlation, which ranges from negative to positive one. A perfect correspondence would exhibit a Pearson correlation of one. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percen-tiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Pearson Correlations Between the Simulated and Observed Daily Storage-Yield Curves

Page 14: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0.990

0.996

0.992

0994

0.998

1.000

Spea

rman

cor

rela

tion

Method of prediction in ungaged basins

Figure C14. The distribution of at-site Spearman correlations between the simulated and observed daily storage-yield curves for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the Spearman correlation, which ranges from negative to positive one. A perfect correspondence between ranked values would exhibit a Spearman correlation of one. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Spearman Correlations Between the Simulated and Observed Daily Storage-Yield Curves

Page 15: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–100

–50

0

50

100

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C15. The distribution of at-site percent errors in the estimated coefficient of variation of annual streamflows for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated Coefficient of Variation of Annual Streamflows

Page 16: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–100

–50

0

50

100

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C16. The distribution of at-site percent errors in the estimated coefficient of variation of daily streamflows for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated Coefficient of Variation of Daily Streamflows

Page 17: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–200

–100

0

100

200

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C17. The distribution of at-site percent errors in the estimated 10th percentile of the distribution of 7-day average annual-minimum events for each method of prediction in ungaged basins (PUB) is considered here. This event is related to the 10-year, 7-day average annual event. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated 10th Percentile of the 7-Day Average Annual-Minimum Events

Page 18: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–200

–100

0

100

200

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C18. The distribution of at-site percent errors in the estimated 50th percentile of the distribution of 7-day average annual-minimum events for each method of prediction in ungaged basins (PUB) is considered here. This event is related to the 2-year, 7-day annual-minimum event. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated 50th Percentile of the 7-Day Average Annual-Minimum Events

Page 19: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–200

–100

0

100

200

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C19. The distribution of at-site percent errors in the estimated 90th percentile of the distribution of annual-maximum events for each method of prediction in ungaged basins (PUB) is considered here. This event is related to the 10-year annual-maximum event. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated 90th Percentile of the Annual-Maximum Events

Page 20: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–200

–100

0

100

200

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C20. The distribution of at-site percent errors in the estimated 90-percent-exceedance streamflow for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbrevi-ations are in table 1.

Percent Errors in the Estimated 90-Percent-Exceedance Streamflow

Page 21: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–200

–100

0

100

200

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C21. The distribution of at-site percent errors in the estimated 75-percent-exceedance streamflow for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated 75-Percent-Exceedance Streamflow

Page 22: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–150

–50

–100

0

100

50

150

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C22. The distribution of at-site percent errors in the estimated 50-percent-exceedance streamflow for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated 50-Percent-Exceedance Streamflow

Page 23: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–100

–50

0

50

100

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C23. The distribution of at-site percent errors in the estimated 25-percent-exceedance streamflow for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated 25-Percent-Exceedance Streamflow

Page 24: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–100

–50

0

50

100

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C24. The distribution of at-site percent errors in the estimated 10-percent-exceedance streamflow for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the 10-Percent-Exceedance Streamflow

Page 25: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–100

–50

0

50

100

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C25. The distribution of at-site percent errors in the estimated mean daily streamflow for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated Mean Daily Streamflow

Page 26: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–100

–50

0

50

100

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C26. The distribution of at-site percent errors in the estimated coefficient of variation (L-CV) of daily streamflow for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated Coefficient of Variation (L-CV) of Daily Streamflow

Page 27: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–50

0

50

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C27. The distribution of at-site percent errors in the estimated skewness (L-skew) of daily streamflows for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated Skewness (L-Skew) of Daily Streamflows

Page 28: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–50

0

50

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C28. The distribution of at-site percent errors in the estimated kurtosis (L-kurtosis) of daily streamflows for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated Kurtosis (L-kurtosis) of Daily Streamflows

Page 29: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–100

–50

0

50

100

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C29. The distribution of at-site percent errors in the estimated lag-1 autocorrelation of daily streamflows for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated Lag-1 Autocorrelation of Daily Streamflows

Page 30: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–100

–50

0

50

100

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C30. The distribution of at-site percent errors in the estimated amplitude of the sinusoidal seasonal trend of daily streamflows for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated Amplitude of the Seasonality of Daily Streamflows

Page 31: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

–40

–20

0

20

40

Perc

ent e

rror

Method of prediction in ungaged basins

Figure C31. The distribution of at-site percent errors in the estimated phase of the sinusoidal seasonal trend of daily streamflows for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis shows the percent error. Unbiased methods display a median near zero and minimum variability of at-site bias. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Percent Errors in the Estimated Phase of the Seasonality of Daily Streamflows

Page 32: Nash-Sutcliffe Efficiencies of Daily Streamflow Predictions...NN−DAR NN−QPPQ NN−SMS1R NN−SMS1L NN−SMS12R NN−SMS12L NN−SM1 NN−SM12 MC−DAR MC−QPPQ MC−SMS1R MC−SMS1L

NN

−DAR

NN

−QPP

Q

NN

−SM

S1R

NN

−SM

S1L

NN

−SM

S12R

NN

−SM

S12L

NN

−SM

1

NN

−SM

12

MC−

DAR

MC−

QPPQ

MC−

SMS1

R

MC−

SMS1

L

MC−

SMS1

2R

MC−

SMS1

2L

MC−

SM1

MC−

SM12

PRM

S

NN

−AFI

NCH

MC−

AFIN

CH

0

0.2

0.4

0.6

0.8

1.0

Root

-mea

n-sq

uare

-nor

mal

ized

erro

r

Method of prediction in ungaged basins

Figure C32. The distribution of the at-site root-mean-square-normalized errors of estimated Fundamental Daily Streamflow Statistics (FDSS) for each method of prediction in ungaged basins (PUB) is considered here. The horizontal axis indicates each PUB method. The vertical axis indicates theroot-mean-square-normalized error. Lower values represent less error, on average. (The dark line indicates the median of the distribution, the box outlines the 25th and 75th percentiles, and the whiskers extend to the data point a distance not more than 1.5 times the interquartile range away from the nearest quartile. Data points beyond this whisker length are defined as outliers and, to increase the visibility of the distribution, have not been drawn.) Method abbreviations are in table 1.

Root-Mean-Square-Normalized Errors of Estimated Fundamental Daily Streamflow Statistics