M odels David Bromwich Marco Tedesco Mike Bevis and Steve Businger GNET Workshop 25 Jan 2017 Thoughts from David Bromwich on the GPS Network GNET BENEFITS High temporal resolution 1015 min intervals ID: 715397
Download Presentation The PPT/PDF document "GNET in Regional A tmospheric" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
GNET in Regional Atmospheric Models
David Bromwich, Marco Tedesco, Mike Bevis, and Steve
Businger
GNET Workshop: 25 Jan 2017Slide2
Thoughts from David Bromwich on the GPS Network (GNET)
BENEFITS
High temporal resolution (10-15 min. intervals)
European studies for
convection in summer
show assimilating GPS
ZTD improves
the prediction of large (extreme) precipitation events
How do these results translate to the Greenland environment?
Greenland events of interest that may profit from GPS
ZTD assimilation
: synoptic-scale cyclones, fronts, hurricane remnants in late summer, lee cyclones east of Cape Farewell, cold air outbreaks…Slide3
The Greenland EnvironmentSlide4
ISSUES
Reliability:
Precipitable Water (PW) values are very low (a few mm in cold season in N. Greenland); not much larger than the accuracy of PW from
GPS ZTD.
Robust Siting:
Set
up for
geodetic purposes, not necessarily optimal for meteorology
Complex flow: A challenge for this vertically integrated quantity - PW (valleys vs peaks; sheltering effects; multilevel flows)Harsh Conditions: Inaccurate during icing conditions – needs a robust quality control procedureNeed Additional Information: Atmospheric motion fields of comparable detail (e.g., GC-Net) and accuracy to GNET. The dry delay is related to surface pressure thus greatly assisting the accurate definition of the motion fields. Thus, GNET is an untapped opportunity for research by graduate students – data need comprehensive evaluation.
Greenland GPS Network (GNET) Slide5
Intense Barrier Wind in Denmark Strait @ 15
km –
Hi. Res. is Key.
03UTC Mar 03, 2007
ASR 30km
ASR 15km
Wind Speed
m/s
Iceland
Greenland
Greenland
Iceland
North
NorthSlide6
Arctic System Reanalysis (ASR)
Regional reanalysis of the Greater Arctic (
2000-2012, currently)
Includes major Arctic rivers and NH storm tracks
Will be updated through
2016 later in 2017 for v2.
Uses
Polar WRF with WRFDA (3D-VAR)
Two Versions
ASRv1-30km & ASRv2-15 km
71 Vertical Levels (1
st
level – 4m)
3h output
ASRv1 30 km available online at the NCAR CISL Research Data ArchiveASRv2 – 15 km Coming Very Soon!
Bromwich et al. 2016 QJRMSSlide7
ASRv3: The Next Generation
Same domain and horizontal resolution (15 km)
Vertical levels: up to 100 total
Period: 1979-2020Nested within ERA5 global reanalysisAtmospheric assimilation: Hybrid-ensemble 3D-VAR possibly 4D-Var New land surface model optimized for GreenlandAssimilate GNET and GC-Net data for Greenland to improve surface mass balanceOur research motivation: Pan-Arctic extreme weather and climate; Greenland SMB. Proposed to NSF late 2016Slide8
Thoughts from Marco Tedesco
Surface mass balance (SMB) plays a crucial role in modulating the contribution of the Greenland ice sheet (GrIS) to sea level rise, and recent work suggests that the role of SMB is becoming increasingly important.
a
NASA-sponsored workshop was held at the Lamont Doherty Earth Observatory (LDEO) of Columbia University on 7-9 September 2016 to provide guidance to the scientific community and funding agencies on actions to be undertaken for reducing uncertainties of SMB estimates of the
GrIS
The workshop participants engaged in discussions to address key questions such as:
Which parameters most affect SMB and how well can we model their current and
historical
evolution? Which measurements are currently available to constrain these parameters? What are the uncertainties associated with estimates of the parameters identified above and how are they spatially and temporally distributed? Which measurements are most needed and where? Slide9
Tedesco et al. (2016)
Modeling Ice Sheet Change
Four major themes
F
our priorities emerged to be addressed to reduce uncertainty of GrIS SMBSlide10
A highlight from the workshop
"
Interest was also expressed in formally collaborating with atmospheric modelers or cloud physics specialists to better understand and represent the processes associated with snow accumulation and albedo." Slide11
Improvements in accumulation
Example: annual
snowfall in 1996 (units mmWE/
yr
) simulated by MAR at a resolution of 10 km
vs
MAR at 20km, 30km,
35km vs. ice core
higher resolution, this bias is a lot reduced.Higher ResolutionSlide12
Thoughts from Mike Bevis on Assimilating ZTD using 4DVar
f
rom
Bennitt
and
Jupp
(2012) Mon.
Weath
. Rev.The observational constraints are applied where (3DVar) and also when (4DVar) each measurements is made.4DVar is now widely used in meteorology, also in oceanography, and is beginning to be used with ice sheet models. Slide13
Peter Bauer, Alan Thorpe and Gilbert Brunet (2015) Nature
The quiet revolution of numerical weather prediction
Schematic of the ensemble analysis of forecast cycle.
Ensemble 4DVar (En4DVar)
Operational implementation of 4D
variational
data assimilation techniques (based on optimal control theory) represents a major milestone in global NWP. The R&D preceding operational deployment was often >10 years. 4DVar was first developed in Europe:
ECMWF 1997
Météo France 2000 UK Met Office 2004 Météo France 2000 Japan 2005 Envir. Canada 2005 US NRL 2009Slide14
Nested Grids
Meteorologists use multi-scale nested grids for the modeling of weather and climate
In Greenland and the Arctic, regional atmospheric models (e.g. RACMO2, MAR and HIRLAM) are embedded in an ECMWF
reanalysis/global model (e.g., ERA-Interim or the operational model at ~9km horizontal grid spacing)Slide15
Example of a regional model
High Resolution Limited Area Model (HIRLAM)
HIRLAM is a synoptic scale (5-15 km horizontal resolution) hydrostatic grid scale model with a 4DVar data assimilation scheme (Huang et al., 2002.)
HIRLAM can assimilate all conventional observations, a wide range of satellite observations, radar, wind profile and
scatterometer
data and GPS zenith total delay (GPS ZTD).
Initial and boundary conditions are normally taken from the ECMWF model or from a larger scale HIRLAM model.
The HIRLAM consortium:
Météo France is an associate member of this consortium.Météo France and the UK Met Office have their own high resolution, regional atmospheric models, and they too have highly developed 4DVar modulesSlide16
Demonstrated impacts of assimilating GPS delay data (usually ZTD)
While GPS ZTD has long been used with regional atmospheric models for Cal/Val purposes, nearly all ZTD assimilation studies have documented a positive impact on predictive skill.
Nearly all of these studies have been performed in Europe and Japan.
Typically GPS ZTD data comprises only a small fraction of the data being assimilated,
but adding the ZTD data has been shown to improve the regional atmospheric models’ representations of humidity and precipitation, and it often improves in addition the representation of clouds, surface radiation and surface temperature.Slide17
Steve Businger’s Agenda
Steve
Businger
was a co-founder of ground-based GPS meteorology (along with M. Bevis and others) in the 1990s.His group runs an operational NWP program in Hawaii.Steve proposes to acquire a copy of HIRLAM from Europe, to set up a reanalysis system largely focused on Greenland from 2000-present.This will assimilate GNET ZTD time series computed by Bevis at OSU, along with all met data assimilated operationally at ECMWF that lies within the HIRLAM model domain.This will assess the possible impact of GNET ZTD data on regional atmospheric models such as MAR and RACMO2.