the contribution of soil moisture information to forecast skill: two studies randal koster and...

19
The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office, NASA/GSFC Ben Livneh Dept. of Civil and Env. Engineering, U. Washington With contributions from the GLACE-2 team, Dennis Lettenmaier, Rolf Reichle, and Qing Liu Direct questions to: [email protected]

Upload: juliana-joseph

Post on 22-Dec-2015

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

The Contribution of Soil Moisture Information

to Forecast Skill: Two Studies

Randal Koster and Sarith MahanamaGlobal Modeling and Assimilation Office, NASA/GSFC

Ben LivnehDept. of Civil and Env. Engineering, U. Washington

With contributions from the GLACE-2 team, Dennis Lettenmaier, Rolf Reichle, and Qing Liu

Direct questions to: [email protected]

Page 2: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

Long-term question:

To what extent might hydrological prediction benefit from satellite-based soil moisture data (e.g., from SMAP or SMOS)?

This talk:

Describe two recent studies quantifying the benefits to prediction of model-based soil moisture data (produced with observed met. data); make inferences regarding satellite data.

Page 3: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

Study #1Subseasonal air temperature and

precipitation forecasts

GLACE-2: A quantification of the impact of realistic soil moisture initialization on the prediction of precipitation and air temperature at subseasonal timescales.

Page 4: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

GLACE-2:Experiment Overview

Perform ensembles of retrospective

seasonal forecasts

Initialize land stateswith “observations”,

using GSWP approach

Prescribed, observed SSTs or the use of a coupled ocean

model

Initialize atmosphere with “observations”, via

reanalysis

Evaluate P, T forecasts against

observations

Series 1: In a subseasonal forecast system (GCM),

4

Page 5: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

GLACE-2:Experiment Overview

Perform ensembles of retrospective

seasonal forecasts

Initialize land stateswith “observations”,

using GSWP approach

Prescribed, observed SSTs or the use of a coupled ocean

model

Initialize atmosphere with “observations”, via

reanalysis

Evaluate P, T forecasts against

observations

Series 2: In a subseasonal forecast system (GCM),

“Randomize” land

initialization!

5

Page 6: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

GLACE-2:Experiment Overview

Step 3: Compare skill in two sets of forecasts; isolate contribution of realistic land initialization.

Forecast skill,Series 1

Forecast skill, Series 2

Forecast skill due to land initialization

6

Examine 60 independent subseasonal forecasts during JJA (10 ensemble members each) 600 2-month simulations.

Page 7: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

Participant List

Group/Model Points of Contact

1. NASA/GSFC (USA): GMAO seasonal forecast system (old and new)

2. COLA (USA): COLA GCM, NCAR/CAM GCM

3. Princeton (USA): NCEP GCM

4. IACS (Switzerland): ECHAM GCM

5. KNMI (Netherlands): ECMWF

6. ECMWF

7. GFDL (USA): GFDL system

8. U. Gothenburg (Sweden): NCAR

9. CCSR/NIES/FRCGC (Japan): CCSR GCM

10. FSU/COAPS

11. CCCma

# models

S. Seneviratne, E. Davin

E. Wood, L. Luo

P. Dirmeyer, Z. Guo

R. Koster, S. Mahanama2

B. van den Hurk

T. Gordon

J.-H. Jeong

T. Yamada

2

1

1

1

1

1

1

13 models

1 G. Balsamo, F. Doblas-Reyes

M. Boisserie1

1 B. Merryfield

7

Page 8: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

Temperature forecasts: Increase in skill (r2) during JJA due to land initialization(Multi-model results, conditioned on strength of local initial soil moisture anomaly)

Extreme tercilesall dates

Extreme quintiles

Extreme deciles

16-30 days

31-45 days

46-60 days

8

Forecast skill: r2 with land ICs vs r2 w/o land ICs

Dates for conditioning vary w/location

Page 9: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

Study #2Seasonal streamflow prediction

A quantification, using multiple land models, of the degree to which soil moisture and snow initialization contribute to streamflow forecast skill at seasonal timescales.

Page 10: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

10

Experiment:

1. Perform multi-decadal offline simulation covering CONUS, using observations-based meteorological data. Determine streamflows in various basins for MAMJJ and compare against (naturalized) streamflow observations.

2. Repeat, but doing forecasts: Simulate MAMJJ streamflow knowing only soil moisture and snow conditions on January 1. (Use climatological met forcing for January – July.) Compare forecasts to observations. (Not a synthetic study!)

3. Repeat, knowing only snow conditions on January 1.

4. Repeat, knowing only soil moisture conditions on January 1.

Page 11: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

MAMJJ Streamflow Forecast Skill (r2)

a. CTRL: Forcings, initial snow, initial SM known (not true forecasts) b. Exp1: Initial snow, initial SM known

c. Exp2: Initial snow known d. Exp3: Initial SM known

belongs to 5 ( also 1&2) belongs to 2 (also 1)

Skill of model simulation of MAMJJ streamflow given: -- Realistic January 1 initial

conditions -- “Perfect” prediction of

forcing during forecast period

Skill (r2)

Skill of model simulation of MAMJJ streamflow given: -- Realistic January 1 initial

conditions -- No skill in prediction of

forcing during forecast period (use

climatology)

Page 12: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

MAMJJ Streamflow Forecast Skill (r2)

a. CTRL: Forcings, initial snow, initial SM known (not true forecasts) b. Exp1: Initial snow, initial SM known

c. Exp2: Initial snow known d. Exp3: Initial SM known

belongs to 5 ( also 1&2) belongs to 2 (also 1)

Skill (r2)

“Snow initialization only” test: snow important toward northwest of study area.

“Soil moisture initialization only” test: SM more important toward southeast of study area.

Page 13: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

Oct. 1 initialization

Jan. 1 initialization

Apr. 1 initialization

July 1 initialization

Synthetic study results: Lead, in months, for which some significant (95% confidence level) streamflow forecast skill is obtained from soil moisture initialization.

0 1 22 3 4 5 6 87 109Number of Lead Months

Soil moisture initialized

Snow initialized

Page 14: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

These two studies show that accurate soil moisture initialization can lead to improvements in:

subseasonal air temperature forecastsseasonal streamflow forecasts

What are the implications for satellite-derived soil moisture data?

Page 15: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

15

In the global GLACE-2 analysis, the skill levels obtained are clearly connected to the accuracy of the soil moisture initialization, indicating that improved soil moisture estimates can lead to improved forecasts.

Land-Derived Skill (r2) for Air Temperature Forecasts

Rain-gauge density(# gauges /

2ox2.5o grid cell)

Measure of Underlying Model “Predictability”Surrogate for soil moisture accuracy

Page 16: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

16

In the global GLACE-2 analysis, the skill levels obtained are clearly connected to the accuracy of the soil moisture initialization, indicating that improved soil moisture estimates can lead to improved forecasts.

Land-Derived Skill (r2) for Air Temperature Forecasts

Rain-gauge density(# gauges /

2ox2.5o grid cell)

Measure of Underlying Model “Predictability”

SMAP or SMOS would effectively increase the ordinates of the dots…

… suggesting that we’d get more skill for these locations with SMAP or SMOS data

Page 17: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

SMAP data coverage

Rain gauge density: a reasonable surrogate metric for the accuracy of soil moisture initial conditions in a forecast

SMAP data will be available over most of the world (the white areas), allowing first-order increases in soil moisture accuracy in many regions. (Of course, GPM will help, as well…)

Rain gauge density

GLACE-2 cutoff

Page 18: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

Rain gauge density

Even in regions of high rain gauge density, data assimilation studies* with AMSR-based and SMMR-based soil moisture data show that satellite-based products improve the estimation of soil moisture over that obtained with the rainfall forcing alone. SMAP and SMOS will improve over AMSR and SMMR, providing even higher accuracy than indicated here.

*Reichle et al., JGR, 112, 2007 Reichle and Koster, GRL, 32, 2005 Liu et al., Journal Hydromet., submitted.

From Liu et al. (submitted)

Skill (r)

Page 19: The Contribution of Soil Moisture Information to Forecast Skill: Two Studies Randal Koster and Sarith Mahanama Global Modeling and Assimilation Office,

19

Summary

Accurate soil moisture initialization (as derived from met forcing, particularly precipitation) does contribute significantly to skill in temperature and streamflow forecasts.(Koster et al., GRL, 2010; Koster et al., Nature Geosci., 2010)

Forecasts are indeed found to improve with the accuracy of the soil moisture initialization.

Given AMSR & SMMR experience, SMAP & SMOS should contribute significantly to the accuracy of initialization.

Inference: SMAP & SMOS should contribute to forecast skill.

Obvious next challenge: quantify these contributions!