This presentation describes recent progress in the following two strategic approaches for improving operational climate monitoring and seasonal hydrologic prediction undertaken by EMC and its CPPA partners: A) Improving the land model and land data

1
CAHMDA III International Workshop, 9-11 January
2008, Melbourne, Australia
Contact information Kenneth.Mitchell_at_noaa.gov,
Youlong.Xia_at_noaa.gov
ABSTRACT
This presentation describes recent progress in
the following two strategic approaches for
improving operational climate monitoring and
seasonal hydrologic prediction undertaken by EMC
and its CPPA partners A) Improving the land
model and land data assimilation component of
NCEPs coupled climate forecast model and its
coupled global data assimilation, together known
as the NCEP Climate Forecast System (CFS) B)
Multiple uncoupled high-resolution hydrologic
models, driven by bias-corrected and downscaled
ensemble seasonal predictions from the CFS. The
presentation will consist of two parts, with one
part on each of the aforementioned approaches.
Part 1 will present results from summer-season
hindcast experiments with the NCEP coupled global
CFS. The CFS experiments use either the old OSU
(Oregon State University) or newly developed Noah
land surface model (LSM) as the land component,
and it utilizes either NCEP-DOE Global Reanalysis
2 (GR2/OSU), which uses the OSU LSM, or Noah-LSM
based Global Land Data Assimilation System
(GLDAS/Noah) for initial land states. To
demonstrate the extent to which upgrades to the
land model and land data assimilation component
can improve CFS summer season predictions over
the continental U.S. (CONUS), four basic CFS
configurations were tested, including 1) CFS
using the old OSU LSM and initial land states
from GR2/OSU, 2) CFS with the new Noah LSM and
initial land states from GR2/OSU, 3) CFS with the
new Noah LSM and initial land states from
GLDAS/Noah, and 4) CFS with the new Noah LSM and
climatology land States from GLDAS/Noah. For each
of the four configurations, the NCEP EMC CPPA
project executed 10-members of 6-month CFS
forecasts extending from mid-to-late April
initial conditions through the subsequent
Jun-July-August (JJA) summer season for each of
25 years 1980-2004 altogether representing a
total of 4 x 250 CFS 6-month forecasts. The
performance of the four CFS experiments is
assessed against analyses of precipitation, 2-m
air temperature, Nino-3.4 SST and 500 hPa height.
The results demonstrate the pivotal role of
providing self-consistent and compatible initial
land states to an upgraded land model component
of a global model. Hence it is naïve to merely
upgrade the land component of a global climate
model for seasonal forecasting without
simultaneously upgrading to the same new LSM in
the companion global data assimilation system.
Part 2 presents the progress by EMC and
its CPPA partners in constructing and executing
an uncoupled high-resolution land surface
monitoring and seasonal prediction system over
CONUS using five land models (Noah, VIC, Mosaic,
SAC, CLM). This uncoupled approach includes both
an analysis/monitoring mode and an ensemble
seasonal prediction mode. The analysis/monitoring
mode consists of a retrospective 29-year
execution (1979-2007) and realtime daily update
execution of the aforementioned five land models
on a common 1/8th degree CONUS grid using common
hourly land surface forcing. The
non-precipitation forcing is derived from the
NCEP N. American Regional Reanalysis System
(NARR). The NARR downward solar radiation is
bias-corrected by using GOES satellite data, and
the precipitation forcing is anchored to daily
gauge-only precipitation analysis over CONUS,
which is temporally disaggregated to hourly
precipitation amounts and is topographic-bias-corr
ected. This analysis/monitoring mode provide both
the initial land states for the uncoupled
ensemble seasonal prediction mode and basic
products such as soil moisture, snow water
equivalent and runoff for NCEP CPC drought
monitoring. The prediction mode utilizes 1)
Ensemble Streamflow Prediction (ESP), 2) ensemble
forecasts downscaled and bias-corrected from an
ensemble seasonal predictions of several climate
models (i.e. CFS, GFDL climate model , NASA
climate model, NCAR CCM), via downscaling methods
developed by Princeton University (PU), and 3)
statistical downscaling of the official seasonal
climate outlooks of NCEP CPC, via downscaling
developed by University of Washington (UW). The
presentation below provides examples of ensemble
seasonal prediction from all three of these
uncoupled prediction methods.
3. Uncoupled Hydrologic Prediction System
Skill of seasonal precipitation forecasts (left
panel) and 2-m air temperature (right panel) as
measured by the correlation coefficient between
the forecasted quantities and observations
achieved with different combinations of forecast
systems and soil moisture initial conditions.
Top left CFS forecast system using Noah land
model, initialized with land states from an
offline analysis with the Noah land model. Top
right Same CFS forecast system but using the OSU
land model, initialized with land states from the
NCEP/DOE Global Reanalysis that used the OSU land
model. Bottom left Same CFS forecast system
using Noah land model, but initialized with land
states from the NCEP/DOE Global Reanalysis that
used the OSU land model. Bottom right spatial
averages of the correlations produced over the
continental U.S., for each of the three
combinations.
1. Coupled Climate Prediction System
U. Washington CPC and ESP Approach
Princeton University Approach
CFS shows the better performance when Noah model
and GLDAS initial states are used
3 month Forecast
3 month Forecast
Application of Princeton U. and U. Washington
Approaches for Drought Prediction Soil Moisture
and Runoff
JJA Mean 200 mb Geopotential Height Correlation
Skill with Different LSM/ICs
6-month forecast
3-month forecast
Princeton U.
VIC
VIC
Illinois
Illinois
JJA Mean SST Correlation Skill with Different
LSM/ICs
U. Washington
Illinois
Illinois
3 month Forecast
3 month forecast
2. Future Work for Coupled System

• For a new Climate Forecast System (CFS)
  implementation, two essential components are
  included
• A new Reanalysis of the atmosphere, ocean, sea
  ice and land over the 31-year period (1979-2009)
  is required to provide consistent initial
  conditions for
• A complete Reforecast of the new CFS over the
  28-year period (1982-2009), in order to provide
  stable calibration and skill estimates of the new
  system, for operational seasonal prediction at
  NCEP

VIC
CLM
4. Future Work for Uncoupled System

• Under CPPA, NCEP and CPPA PIs are collaborating
  improving and applying the above new suites of
  uncoupled hydrometeorological monitoring
  prediction systems
• Downscaling Focus plus bias-correction multi
  land models, (2) NLDAS monitoring mode (analysis
  and reanalysis), (3) NLDAS prediction mode (3a.
  Dynamical prediction forced with dynamical
  coupled global models and 3b. Empirical
  prediction ENSO composites, CPC outlooks)