so it’s a land-falling atmospheric river, can that help the...

So it’s a Land-falling Atmospheric River, Can That Help the Forecaster Make a Better

QPF?

David W. Reynolds Cooperative Institute for Research in Environmental Sciences

Boulder, CO

Brian KawzenukCenter for Western Weather and Water Extremes

August 8, 2016

International Workshop on Atmospheric Rivers

Outline• Review historical CNRFC (California Nevada River Forecast Center)

derived 24-hr Mean Areal Precipitation (MAP) and IVT correlations for Lake Mendocino Watershed– Is IVT a good proxy for 24-hr rainfall?

• Review CNRFC and GEFS forecast to observed MAPs– Are model and model derived QPFs better related to observed rainfall than IVT?

• Review a land-falling AR case from 2016• Looked at some 10 cases but will only show early March 2016

• Summary and Recommendations

Russian River Watershed

Potter Valley Dam

Healdsburg

Guerneville

Hopland

40N 125W

BBY

IVT calculated here

37.5N 125W

Landfalling AR depicted by IWV and IVT00Z 6 March 2016

http://cw3e.ucsd.edu/?page_id=491#IVT

Not an AR

Minimal strength AR

Moderate strength AR

Normal-duration AR landfall (12-24 hours)

Days from 10 AM PT Thursday 3 March 2016

AR summary for Pt Reyes, CA area, including Russian River

Precipitation forecast from NOAA/NWS California/Nevada River Forecast Center

6-9 inches of rain in favored mountain locations over 3 days (4 AM Fri – 4 AM Mon)

1000

600

800

400

200

0

0 2 3 4 5 61

AR

Str

en

gth

(wat

er

vap

or

tran

spo

rt; k

g/m

/s)

Summary by F.M. Ralph 8 AM PT Fri 4 March 2016

AR Plume Diagram by J. Cordeira/Plymouth St.Univ

What is a more reliable predictor of AR impacts?

IVT strength or QPF derived from models?

IVT QPF Methodology

• Utilize Rutz NCEP/NCAR* AR catalog from1949-2015 and NASA MERRA data** from 1980-2012. • Use 1000-200 mb IVT – show max 6-hr IVT 12z-12z as this showed best results

• Correlate to 24-hr Mean Areal Precipitation (MAP) available from 1949-2015 • Recalculate using upslope wind direction, precipitation threshold of >=.1 “/day and for matching sample

years using Rutz catalog.• Lavers et al. (2016) – showed IVT more predictable than rainfall but is IVT well correlated with

subsequent rainfall?

• Correlate 16-yrs of CNRFC 24-hr forecast MAPs issued daily at 12z to observed MAPs

• Compare 1985-2012 11 member GEFS (V9.1 ) Ensemble Mean MAPs utilized by CNRFC for probabilistic inflows to Lake Mendocino• Rutz and Alcott, 2014 did a similar study using 20 yr GEFS data and CPC QPE for CONUS

* http://rda.ucar.edu/datasets/ds090.0/ ** http://disc.sci.gsfc.nasa.gov/mdisc/data-holdings

Case-study Approach

• Focus on Lake Mendocino Watershed• Use MAP

• Utilize 12z model runs to coordinate with CNRFC 5-Day 6-hr QPFs

• Utilize NWS GFE software• Calculate MAPs for each model available

• Using GFS40 analysis compute layer average vapor transport for each case at 40N 125W at 6-hr intervals• 550-700, 700-850, 850-1000, 200-1000 MB

QPF verification using NWS GFE Software

Lake Mendocino Watershed

Derive the Mean Areal Precipitation (MAP)for each model available

Compare to MAP provided by CNRFC

Compare to IVT at landfall

Max 6-hr IVT NCEP/NCAR from 12z-12z at 40n 125W and 12-12z 24-hr QPE Lake Mendocino 1949-2015

0

200

400

600

800

1000

1200

1400

0 1 2 3 4 5 6 7 8

Max

6-h

r IV

T 1

2z-

12

z

24-hr sum

R2 = .24

Max 6-hr IVT MERRA 12z-12z at 40n 125W and 12-12z 24-hr QPE Lake Mendocino 1980-2012

0

200

400

600

800

1000

1200

1400

0 1 2 3 4 5 6

Max

6-h

r IV

T 1

2z-

12

z

24-hr sum

R2 = .24

CNRFC Day 1 24 hr QPF Correlation to Observed Lake Mendocino Watershed (2000-2016)

y = 1.0306x + 0.0168R² = 0.8252

0.00

1.00

2.00

3.00

4.00

5.00

6.00

0.00 1.00 2.00 3.00 4.00 5.00 6.00

Fcst

Day

1 2

4 h

r (i

n)

Obs 24 hr (in)

Day 1 Fcst vs Obs

Fcst vs Obs Linear (Fcst vs Obs)

R2 Values for CNRFC-GEFS Mean and IVT to24 hr MAP Lake Mendocino

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

24 48 72 96 120 144 168 192

Co

effi

cie

nt

of

Det

erm

inat

ion

R2

Forecast Lead Time (Hours)

CNRFC vs GEFS Mean 24 hr QPF

CNRFC 24 hr QPF R2 GEFS Mean 24 hr QPF R2

Range of 6-Hr Max IVTRutz, Merraor match GEFS or CNRFC yearsAlso SW flow only or >.1” day

CNRFC - 2000-2016

GEFS v9.1 11 member mean 1985-2012

R2 values fall in a very narrow range For various correlations listed- .22-.24

NCEP/NCAR 3-Day Time-Average 6-hr IVT vs 3-day MAP Lake Mendocino 40N and 37.5N

0

200

400

600

800

1000

1200

0 2 4 6 8 10 12 14

3d

ay T

ime

Avg

IVT

kg/m

/s

3-Day Precip

Regression of 3day Time Avg Ivt 40N by 3-Day Rainfall (R²=0.297)

Active Model

Conf. interval (Mean 95%) Conf. interval (Obs. 95%)

0

200

400

600

800

1000

1200

0 2 4 6 8 10 12 14

3-D

ay IV

T 3

7.5

N

3 Day precip

Regression of 3-Day Time Avg IVT 37.5N by 3 Day Rainfall (R²=0.364)

Active Model

Conf. interval (Mean 95%) Conf. interval (Obs. 95%)

3 Day CNRFC Total QPF vs Observations 2000-2016

y = 0.9483x + 0.0344R² = 0.8175

0

1

2

3

4

5

6

7

8

9

0 1 2 3 4 5 6 7 8

3-D

ay F

cst

(in

che

s)

Observed 3 day (inches)

3 Day CNRFC ForecastLake Mendocino

y = 0.0426x - 1.0919

-4

-3

-2

-1

0

1

2

2/1

5/2

000

1/2

/20

02

3/1

5/2

003

2/1

8/2

004

2/2

7/2

004

12

/9/2

004

12

/29

/20

04

3/2

/20

05

12

/21

/20

05

12

/31

/20

05

2/2

8/2

006

3/2

/20

06

2/1

1/2

007

1/6

/20

08

1/2

7/2

008

2/3

/20

08

2/1

7/2

011

3/2

1/2

011

3/2

6/2

011

1/2

2/2

012

11

/30

/20

12

12

/3/2

012

12

/24

/20

12

2/1

0/2

014

12

/5/2

014

12

/12

/20

14

2/8

/20

15

12

/23

/20

15

1/7

/20

16

1/1

8/2

016

3/7

/20

16

3/1

4/2

016

Me

an E

rro

r (i

nch

es)

Year

CNRFC Mean Error 3-Day Obs >= 3"

Mean Error Obs >= 3" Linear (Mean Error Obs >= 3")

Trends in 3-Day Forecast Mean Error (Bias)

IVT Plume Diagrams GEFS

00Z 6 March 00Z 11 March

GFS control forecast (black line), the ensemble-mean (green line), and the maximum (red line) and minimum (blue line) ensemble value at each forecast time. White shading represents +/- 1 standard deviation forecast from the ensemble mean

138 hrs > 250 kg/m/s42 hrs > 500 kg/m/s

5-7 Mar 2016 Layer IVT and CNRFC 6-hr Fcst and Obs MAPs

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0

100

200

300

400

500

600

700

800

900

0 6 12 18 0 6 12 18 0 6

Rai

nfa

ll (i

nch

es)

IVT

pe

r la

yer

kg m

-1s-1

hr in Z day 5-7 Mar

Mar 5-7 2016 Layer IVT (40n 125W) and Lake Mendocino Watershed 6-hr Precip

6-Hr Precip COY Res Day 1 Fcst Day 2 Fcst Day 3 Fcst Day 4 Fcst Day 5 Fcst 550-700 700-850 850-1000 200-1000

00Z 6 March

Mar 5-6, 2016 MAP QPF Verification

0

0.5

1

1.5

2

2.5

3

3.5

0-24 24-48 48-72 72-96 96-120

24

-hr

rain

fall

(in

che

s)

Forecast Lead Time

MAP QPF valid 12z-12z 5-6 Mar 2016Lake Mendocino Watershed

CNRFC

GFS40

WestWrf

NAM12

HiResNmm

OBS

9-11 Mar 2016 Layer IVT and CNRFC 6-hr Fcst and Obs MAPs

0

0.2

0.4

0.6

0.8

1

1.2

0

100

200

300

400

500

600

700

800

900

0 6 12 18 0 6 12 18 0 6 12 18 0

Rai

nfa

ll (i

nch

es)

IVT

per

laye

r kg

m-1

s-1

hr in Z day 9-11 Mar

Mar 9-11 2016 Layer IVT (40n 125W) and Lake Mendocino Watershed 6-hr Precip

6-Hr Precip COY Res Day 1 Fcst Day 2 Fcst Day 3 Fcst Day 4 Fcst

Day 5 Fcst 550-700 700-850 850-1000 200-1000

11 March

Mar 10-11, 2016 MAP QPF Verification

0

0.5

1

1.5

2

2.5

3

0-24 24-48 48-72 72-96 96-120

24

-hr

rain

fall

(in

che

s)

Forecast Lead Time

MAP QPF valid 00z-00z 10-11 Mar 2016Lake Mendocino Watershed

CNRFC

GFS40

WestWrf

NAM12

HIRESNMM

OBS

Based on Correlations QPF better Predictor of Impacts than IVT

Lake Mendocino

CNRFC 6 Day QPF from 1 Mar 2016

GEFS Model Probability of >= 6”March 1-March 12, 2016

Plume Diagrams from GEFS

Conclusions and Recommendations

• IVT (200-1000mb) not as good an indicator of potential rainfall impacts of land-falling ARs as is model or manual QPFs• Supports Rutz-Alcott 2014 study for CONUS• IVT may be more predictable (Lavers) but IVT not well correlated to

subsequent rainfall for this study area over long term

• Utilize model QPFs (3 day totals) to define strength of land-falling ARs?• R-Cats to define AR strengthas defined by Ralph and Dettinger,2012-BAMS

• Forecaster situational awareness better served by focusing on model QPF- especially looking at 3-5-7 day totals given timing and exact locations can be off utilizing 6-12-24 hr model QPFs

International AR Workshop

Thank Youdavid.reynolds@noaa.gov

EXTRAS

Feb 5-10, 2015 Layer IVTand CNRFC 6-hr Fcst and Obs MAPs

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0

200

400

600

800

1000

1200

12 18 0 6 12 18 0 6 12 18 0 6 12 18 0 6 12

Rai

nfa

ll (i

nch

es)

IVT

per

laye

r kg

m-1

s-1

hr in Z day 5-10 Feb

Feb 5-10, 2015 Layer IVT and Lake Mendocino Watershed 6-hr Precip

LMAC1 Precip Day 1 Fcst Day 2 Fcst Day 3 Fcst Day 4 Fcst Day 5 Fcst 550-700 700-850 850-1000 200-1000

Feb 6-7, 2015 24-hr MAP QPF Verification

0

1

2

3

4

5

6

0-24 24-48 48-72 72-96 96-120

24

-hr

rain

fall

(in

ches

)

Forecast Lead Time

MAP QPF valid 12z-12z 6-7 Feb 2015Lake Mendocino Watershed

CNRFC

GFS40

WestWrf

NAM12

HIRESNMM

OBS

Dec 10-12, 2014 IVT vs CNRFC MAP 6-hr QPE250* and 500** kg m-1 s-1 IVT annotated with respect to 200-1000 mb IVT

0

0.4

0.8

1.2

1.6

2

2.4

2.8

0

200

400

600

800

1000

1200

0 6 12 18 0 6 12 18 0

Rai

nfa

ll (i

nch

es)

IVT

per

laye

r kg

m-1

s-1

hr in Z day 10-12 Dec

Dec 10-12, 2014 Layer IVT and Lake Mendocino Watershed 6-hr Precip

LMAC1 Precip Day 1 Fcst Day 2 Fcst Day 3 Fcst Day 4 Fcst Day 5 Fcst 550-700 700-850 850-1000 200-1000

* After Rutz et al, 2014 Mon. Wea. Rev. **Cordeira - http://cw3e.ucsd.edu/?page_id=491#LFT

Comparison of MAPs CNRFC and NWP Lake Mendocino Watershed 24 hr Amounts

0

1

2

3

4

5

6

7

0-24 24-48 48-72 72-96 96-120

24

-hr

rain

fall

(in

che

s)

Forecast Lead Time

MAP QPF valid 00z-00z 11-12 Dec 2014Lake Mendocino Watershed

CNRFC

GFS40

WestWrf

NAM12

HIRESNMM

OBS

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0

100

200

300

400

500

600

700

800

900

0 6 12 18 0 6 12 18 0 6 12 18 0 6

Rai

nfa

ll (i

nch

es)

IVT

pe

r la

yer

kg m

-1s-1

hr in Z day 19-22 Dec

Dec 19-22, 2015 Layer IVT and Lake Mendocino Watershed 6-hr Precip

LAMC1 6-hr Precip Day 1 Fcst Day 3 Fcst Day 4 Fcst Day 5 Fcst 550-700 700-850 850-1000 200-1000

Dec 19-22, 2015 Layer IVT and 6-hr Fcst and Obs CNRFC MAPs

21-22 Dec 2015 MAP QPF Verification

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

0-24 24-48 48-72 72-96 96-120

24

-hr

rain

fall

(in

che

s)

Forecast Lead Time

MAP QPF valid 12z-12z 21-22 Dec 2015Lake Mendocino Watershed

CNRFC

GFS40

WestWrf

NAM12

HIRESNMM

OBS

Lake Mendocino Operations 2015-2016

01

23

45

60

20000

40000

60000

80000

100000

120000

1-O

ct

6-O

ct

11

-Oct

16

-Oct

21

-Oct

26

-Oct

31

-Oct

5-N

ov

10

-No

v

15

-No

v

20

-No

v

25

-No

v

30

-No

v

5-D

ec

10

-Dec

15

-Dec

20

-Dec

25

-Dec

30

-Dec

4-J

an

9-J

an

14

-Jan

19

-Jan

24

-Jan

29

-Jan

3-F

eb

8-F

eb

13

-Fe

b

18

-Fe

b

23

-Fe

b

28

-Fe

b

4-M

ar

9-M

ar

14

-Mar

19

-Mar

24

-Mar

29

-Mar

3-A

pr

8-A

pr

13

-Ap

r

18

-Ap

r

23

-Ap

r

28

-Ap

r

3-M

ay

8-M

ay

13

-May

18

-May

23

-May

28

-May

Lake

Me

nd

oci

no

Wat

ers

he

d R

ain

fall

(in

che

s)

Sto

rage

(ac

re-f

eet

)

Date

Lake Mendocino Guide Curve

COY MAP MAP 2 Baseline 2007-2014 Deviation STO 1 STO 2

Forecasts Flows and Stage at HoplandDay 5 Forecast

Forecasts Flows and Stage at HoplandDay 4 Forecast

Forecasts Flows and Stage at HoplandDay 2 Forecast

Forecasts Flows and Stage at HoplandDay 1 Forecast