Analyses and Reanalyses for Climate

  

Kevin E. Trenberth  NCAR 

Workshop August 18-20, 2003, Boulder CO

Earth Observing Summit31 July 2003 Washington DC 

G8 Proposal:

Focussed on need for observations but really with a vision for an Earth Information System.Implies need for analysis and reanalysis: indeed a complete system 

 

Implementation to be developed over next year

Observations of climate and the assessment of the current state of the entire system, understanding of the past record, and probabilistic predictions of the future and climate variations on multiple time-scales have substantial value to local and national economies.

- hard to document- but widely believed

THE NEED FOR A SYSTEMS

APPROACH TO CLIMATE

OBSERVATIONSBY KEVIN E. TRENBERTH, THOMAS R. KARL, AND THOMAS W. SPENCE

Because climate is changing, we need to determine how and why. How do we best track and provide useful information of sufficient quality on climate?

Bulletin of the American Meteorological Society:

November 2002, 83, 1593-1602

• Global climate models that encompass all parts of the climate system and which are utilized in data assimilation and in making ensemble predictions.

• Global climate models that encompass all parts of the climate system and which are utilized in data assimilation and in making ensemble predictions.

Observing systemObserving system means a comprehensive approach, including

•Climate observations from both space-based and in situ platforms taken in ways that address climate needs and adhere to the ten principles outlined by the NRC (1999).

Observing systemObserving system means a comprehensive approach, including

•Climate observations from both space-based and in situ platforms taken in ways that address climate needs and adhere to the ten principles outlined by the NRC (1999). • A A global telecommunications networkglobal telecommunications network and satellite data telemetry and satellite data telemetry

capacity to enable data and products to be disseminated. capacity to enable data and products to be disseminated. • A A global telecommunications networkglobal telecommunications network and satellite data telemetry and satellite data telemetry capacity to enable data and products to be disseminated. capacity to enable data and products to be disseminated.

• A climate observations analysis capability that produces global and regional analyses of products for the atmosphere, oceans, land surface and hydrology, and the cryosphere.• Four dimensional data assimilation capabilities that process the multivariate data in a physically consistent framework to enable production of the analyses: for the atmosphere and oceans, land surface etc.

• A climate observations analysis capability that produces global and regional analyses of products for the atmosphere, oceans, land surface and hydrology, and the cryosphere.• Four dimensional data assimilation capabilities that process the multivariate data in a physically consistent framework to enable production of the analyses: for the atmosphere and oceans, land surface etc.

•A climate observations oversight and monitoring center that tracks the performance of observations, the gathering of the data, and the processing system. This center must have resources and influence to fix problems and be a prominent climate voice when observational systems are established, such as for weather purposes or in establishing requirements for instruments on satellites.

Trenberth, Karl and Spence 2002

•A climate observations oversight and monitoring center that tracks the performance of observations, the gathering of the data, and the processing system. This center must have resources and influence to fix problems and be a prominent climate voice when observational systems are established, such as for weather purposes or in establishing requirements for instruments on satellites.

Trenberth, Karl and Spence 2002

Many observations are made that are useful for climate. But they may not be suitable for climate without an evaluation of their quality and ability to detect climate-relevant signals. As the time horizon increases, the range of processes also increases.

Given the continuing improvement in Given the continuing improvement in our understanding of climate our understanding of climate observations and the need for long time observations and the need for long time series, reprocessing is a hallmark of series, reprocessing is a hallmark of every climate observing system.every climate observing system.

NOAA Climate and Global Change WG report, April 1-3 2003.

Overview 

 History of Reanalysis Accomplishments Problems  ERA-40 status ERA-40 Workshop results (Nov 2001) What have we learned?  

Where do we go from here? 

Data Assimilation merges observations & model predictions to provide a superior state estimate.

Observations of state and storage (temperature, wind, soil moisture, etc ) are blended with the state of the system as forecast by a model based on the previous set of observations. It provides a dynamically- consistent estimate of the state of the system using the best blend of past, current, and perhaps future observations.

Experience mainly in atmosphere; developing in ocean and land surface.

xt dynamics physics x

Obs Model4DDAImproved products,

predictions, understanding

1) Call for reanalysis:Trenberth, K. E., and J. G. Olson, 1988: An evaluation and intercomparison of

global analyses from NMC and ECMWF. BAMS, 69,1047-1057.Bengsston, L. and J. Shukla, 1988: Integration of space and in situ

observations to study global climate change. BAMS, 69, 1130-1143.

2) First generationSchubert, S.D., R. B. Rood and J. Pfaendtner, 1993: An Assimilated Data Set

for Earth Science Applications BAMS, 74, 2331-2342. Gibson, J.K., P. Kallberg, A. Nomura, and S. Uppala, 1994: The ECMWF re-

analysis (ERA) project- Plans and current status. 10th Int. Conf. Interactive Inf. Proc. Syst. Meteor., Oceanogr. and Hydrol., Nashville, TN, AMS, 164-167.

Kalnay,E.,et al., 1996: The NMC/NCAR 40-year Reanalysis Project. BAMS., 77,437-471.

Also NCEP/DOE. And limited time-period global re-analyses: 1) The stratosphere and the UARS period (S.Pawson), 2) TRMM period (A. Hou)  

3) Second generation ERA404) Next generation (this meeting): A US National

Analysis and Reanalysis Program to develop and distribute long-term consistent global climate datasets.

Prior to reanalyses, the analyzed climate record was beset with major discontinuities from changes in the data assimilation systems. It was difficult, if not impossible, to reliably infer anomalies and to analyze climate variability.

NWP Forecast skill scores continue to improve

and now SH skill is comparable to NH!

The use of a stable data assimilation system has produced fairly reliable records for monitoring, research and improved prediction that have enabled :

climatologies to be established anomalies to be reliably established time series, empirical studies, quantitative

diagnostics exploration of, improved understanding of processes model initialization and validation test bed for model improvement on all time scales,

especially seasonal-to-interannual forecasts Greatly improved basic observations and data bases.

Reanalysis:What have we gained and what are the benefits?

What have we learned?  Observing system changes affect variability

Trends and low frequencies unreliable Exacerbated by model bias

Budgets don’t balanceImpacts many diagnostic studies

Problems with hydrological cycleSensitivity to model physics (e.g., convection)Exacerbated by insertion of observationsProblems with warm season continental climates

precipitationdiurnal cycle

Unrealistic surface fluxes Ocean (radiative, freshwater)Land (precipitation, radiative)Limits usefulness for offline forcing; e.g. ocean modelingLimits ability to do coupled assimilation

Quantities/regions not a priority for weather centers SurfaceStratospherePolar regionsMany aspects of tropics

What have we learned?  Observing system changes affect variability

Trends and low frequencies unreliable Exacerbated by model bias

Budgets don’t balanceImpacts many diagnostic studies

Problems with hydrological cycleSensitivity to model physics (e.g., convection)Exacerbated by insertion of observationsProblems with warm season continental climates

precipitationdiurnal cycle

Unrealistic surface fluxes Ocean (radiative, freshwater)Land (precipitation, radiative)Limits usefulness for offline forcing; e.g. ocean modelingLimits ability to do coupled assimilation

Quantities/regions not a priority for weather centers SurfaceStratospherePolar regionsMany aspects of tropics

1. Goal of reanalysis (such as ERA-40) is to produce best analysis, given available data. This automatically makes the set of reanalyses inhomogeneous.

Outstanding issue is how to address this? e.g. T(1995) suggested overlapping 5 year periods with data-base held about fixed. Also OSEs.

2. 4DDA is founded in NWP: goal is to produce best forecast, not best analysis. Result, data are excluded (e.g. surface data) or altered (RH data).

3. NWP has developed true four-dimensional capabilities, but most analyses are 3.5 dimensional. Do not use look ahead capability optimally. 4DVAR at ECMWF uses a 12 hour window (-9 to +3 hours)

4. A major product is the input data-base and the output statistics on QC. Issue of how to keep track of and upgrade data bases with newly recovered data while not losing benefits from reanalyses?

5. New diagnostics, volumes of data, costs of getting data.

Some outstanding issues:

ERA-40Project at ECMWF under EU Fifth Framework Several partners, including NCAR. 3DVAR.Model 60L T159

• All streams completed.• Some times to be redone (esp. Pinatubo period)• Post processing: Many separate streams with

different starting points have to be brought together.

• Project to be completed by September 2003.• Data should become available beginning late

this summer through NCAR.

• Report by Adrian Simmonds

Major workshop held at ECMWF November 2001

• Reports on early evaluations of products.

• 5 WGs• Proceedings Published in Tech.

Rep.

Issue of datasets tracking. Should use ERA-40 input merge plus QC output as basis for

future dataset development. Metadata are invaluable.Other important products include: 1) blacklists; 2) the TEMP sonde database and bias correction; and 3) feedback statistics in gridded form.But what about improvements and updates?

Working Group 1: Observations for Reanalysis

Satellite ObservationsSatellite radiances go through pre-analysis thinning: cuts volume by two orders of magnitude (~1/80 ratio). Even with thinning the data volumes are enormous because of feedback blowup factor. EUMETSAT is reprocessing Meteosat winds for the period 1983-1988.US has full archives of GOES data from 1979-on, but need to recalculated winds 1979-88 owing to development of better algorithms.JMA is reprocessing GMS SATOB for 1987-1991.

Much improved stratospheric analyses, (sudden warmings, QBO, humidity and ozone (when TOMS/SBUV available))

Improved mid-latitude storm track precipitation ITCZ over Africa improved Soil water nudging over land has decreased High latitude frozen physics and water budget has

improved Severe cold bias in high-latitude near-surface temperature

has been replaced by a small warm bias Improved sea ice and associated 2-m temperatures Model clouds generally well distributed by the dynamics. Clear sky radiative budget is reasonable. Good detection of tropical cyclones Better coherence and location of mid-latitude cyclones Improved representation of ocean 10-m winds. Monthly mean 10-m winds and waves are OK. ERA-40 waves better than ERA-15, local and global

improvements. Improved distribution and variability of total water

column.

WG 2: Assessment of the quality of the ERA-40 analyses Summary of identified strengths Skill of 10-day forecasts from ERA-40 is excellent – compared to ERA-15 and the early 1980’s operations

Sudden shift (increase) in tropical precipitation, total water column and cloud albedo in mid-1991

P-E over oceans is erroneously positive Spin-up of the hydrological cycle Convective precipitation over land has maximum too early in the day (around noon)

Overestimation of OLR, possibly due to underestimated cloud ice

Tropical convective clouds too reflective, liquid water content overestimated.

Stratocumulus badly represented High ocean-wave heights underestimated, low heights overestimated.

Erroneous ERS-1 wave data used November 1992 to May 1993

Position errors in detected tropical cyclones. Cold bias in lower troposphere in the central Arctic and Antarctic

Too much precipitation in central Arctic in summer Before 1979 ozone is free-running may and not be useful. Trends before the satellite era may not be reliable

Summary of identified weaknesses Too large oceanic precipitation in the tropics.

ERA-40 E-P ERA-40 vs CMAP P

Conclusion: Tropical precipitation too high in model.

ERA40

CMAP

Diff.

model

moisture budget

diff.

Diurnal Cycle of Convective Precipitation for JJADiurnal Cycle of Convective Precipitation for JJA

Time of maximum Time of maximum

Modeled frequency occurs about 2 hours earlier than observed Trenberth et al. 2003

In ECMWF models the diurnal cycle in precipitation occurs with maximum ~local noon (about 3 hours early).00-03

03-06

09-12

12-15

15-18

18-21

21-00

ERA-40DataJun-Aug 1993

Courtesy:Per Kållberg

Working Group 3:Trends in Observations and Analyses

Issues for trends and low frequency variability:

The ERA-40 has the goal of producing the best set of analyses given the observations available.

The key advantage is the synthesis of all available observations.

However, this creates a dependency on theobserving system.

8 Nov 2002

Issues for trends and low frequency variability:Issues for trends and low frequency variability:

•Model bias• Analysis tends to revert to model climate in absence of data

•Real trends •SSTs and radiative gases;•But not total solar irradiance, aerosols, land use change

•Changes in observing systems•Perturbations (like Pinatubo)

While some trends may be captured by the observing system and can be reflected in other quantities through the dynamics, in general the null hypothesis should be that trends and low frequency variability are more likely tobe spurious unless proven otherwise.

Observing System Changes:

•In situ: SSTs (1982), Pibals, Aircraft, etc•Radiosondes

•Types, instruments, locations, times, coverage•Satellite data

•1972: VTPR•1973: some cloud tracked winds•1979: TOVS (HIRS, MSU, SSU); TOMS, SBUV (ozone)•1987: SSMI (sfc winds, column water vapour)•1992: ERS scatterometer•1998: ATOVS

Satellites vary in number,have finite lifetimes and are replaced every few years. There isorbital decay, change in times, platform heatingand instrument degradation

These require bias corrections

Problems occurred in ECMWF ERA-15 in assimilating satellite data• Spurious variability: cloud clearing problems e.g. 1980• Spurious changes: satellite transitions and bad data Only 1 satellite: 1986-87 2 Nov 1986: solar flare interference February 1989: change in satellites

From Trenberth, Hurrell, Fiorino, Stepaniak. 2001 J. Climate

Trends in Observations and Analyses

Possible example with CDAS?

Comparison of miscellaneous products: Puzzling changes at 100 mb in late 2000 in METO and NCEP reanalyses /CDAS:In UKMO: Change in assimilation system Nov 2000In NCEP??? TOVS ceased andATOVS NOAA-16 started 27 Feb 2001???

Bill Randel 2003

Temp anomalies 10N-10S

As well as puzzling changes at 100 mb in late 2000 in reanalyses /CDAS there are weird values at 850 mb: Too warm in 2002

Bill Randel 2003

TOVS/ATOVS for ERA-40

NOAA-16

SATELLITE

SUPPLIER

Example: Satellite based observations• Satellites typically last 3-5 years and have to be replaced• Orbits decay• Equator crossing times change• New satellite orbits differ• Instrument calibrations drift and can be changed by launch• Interference can occur from other instruments

• Need is for stable orbits• May require boosters• Need sufficient sampling of diurnal cycle• Launch on schedule, not on failure, to ensure overlap• Calibrations required• Ground truth validation required

Bias Corrections are NeededBut how good are they?Is there a baseline to establish real trends?

Bias corrections should be applied to satellite and radiosonde data.

Potential for unintended perturbations or bad data to be perpetuated.

There is evidence from alpine summit observationsthat spurious trends may exist.Most radiosonde stations do NOT have adequate recordsof changes

Need to document bias correction changes to allobserving systems.

Is there a baseline to establish real trends?

Closest to a baseline, but still unsatisfactory is theGUAN (Global Upper Air Network) of designated radiosonde stations

•More complete records•Independent analyses•Better metadata•Changes in instruments likely to be abrupt change

Track performance with “Feedback file”Validate ERA-40Improve GUAN

More generally: Extract subset of feedback filefor all radiosonde stations (emphasis on GUAN), and make available

64 23 34 2 20

Track system performance viz a viz trendsUse independent measures and constraints

Global mass of dry airSurface air temperature over land in selected regionsDobson ozone measurementsSAGE data (water vapor)Ocean wave measurementsAlpine summit station data

Time series of forecast performance measures“Satellite temperatures” MSU 2 and 4GEWEX and SPARC datasets and reports

e.g. ISCCP clouds, etcOther measurements:

Surface observations, River discharge, Glaciers (?)Time series of background and analysis fits to observations Recommendation:Document these issues and communicate them and the results of the studies to users.

Conservation of mass of dry air.Both total surface pressure ps and that due to moisture pw are independent measures of the water vapor contribution. Their difference pd should be constant.

Mean annual cycle is similar for ps and pw, except NCEP before 1966 and ERA-15 after 1989.ERA40 offset from global topography 5.5 m lower, mostly Antarctica.

No Sat VTPR TOVS

NoSat.

VTPR

TOVS

Problems

in subtro

pics

Problems

in tropics

Problems i

n

south

ern oce

ans

Recommendations:1. Carry out an ensemble of AMIP-type model simulations with the available forcings to establish the model climateand its natural variability.

2. Carry out selected OSEs with and without major new observing components (planned already)Such as 1973 (VTPR), 1979 (TOVS), 1987 (SSMI)

3. Carry out a series of OSEs to assess overall gradual changes in observing system by utilizing results froma recent year (1998-2000) and degrading the observingsystem to match that of:

Late 1950s (include simulated weather ship obs)Mid 1970s (include simulated VTPR from HIRS)Mid 1980s

For different seasons

A. Variability and trends due to changes in the observing systems With emphasis on the stratosphere, tropics and Southern Hemisphere, systematic errors

B. Physical consistency Pinatubo, 1997-98 El Niño, TRMM/TMI

C. Phenomenological studies Storm tracks, Easterly waves, Extreme events, Tropical cyclones

D. Validation against independent data Field programs, satellite/other data not used.

WG4: Observing System Experiments for ERA-40 The need for OSEs within ERA-40 The purpose of the Observing System Experiments is to be able to link features in the analysis to the characteristics in the observing systems.

WG 5: Distribution of ERA-40 DataFurther recommendations:•on time-series datasets, observations, metadata,•contents, formats, and costs of CD-Roms•basic and specialized datasets and costs.

There should be an online (web) interface to as much of ERA40 as is practicable (cf CDC NCEP/NCAR Reanalysis site).

ECMWF could consider collaboration with external data centers.

All dissemination of ERA40 including WEB should be accompanied by scientific metadata (including technical and scientific documentation) in an easily understandable and accessible form.

Where do we go from here?Limitations of Current Approach • Reliance on NWP-oriented development  Metrics must go beyond skill of weather forecasts

Need to have objective to produce best analysis, not best forecast

  • “Latest is the greatest” 

How do we integrate new observations into a consistent long term climate record?Need routine OSEs to assess ongoing observing system changes, and help design and optimize system

Need for continual ongoing program of analysis and reanalysis

From: Executive Summary of“The Second Report on the Adequacy of The Global Observing Systems for Climate in Support of the UNFCCC”.

Reanalysis has been applied to atmospheric data covering the past five decades. Although the resulting products have proven very useful, considerable effort is needed to ensure that reanalysis products are suitable for climate monitoring applications.

Conclusion:

Internationally-coordinated reanalysis activities need to be enhanced and sustained by the involved Parties to meet the requirements for monitoring climate trends, to establish ocean reanalysis for the recent satellite era, and to include variables related to atmospheric composition and other aspects of climate forcing.

From: Executive Summary of“The Second Report on the Adequacy of The Global Observing Systems for Climate in Support of the UNFCCC”.

A Proposal for a US National Program

Workshop on Ongoing Analysis of the Climate SystemBoulder CO, August 18-20, 2003NOAA, NSF, NASA

GOAL: Advance plans to establish ongoing capability to integrate and synthesize observations that have and will become available.

1. Steps needed to to ensure ongoing development of reanalysis efforts (atmosphere, ocean, land)

2. Identify near-term, high priority actions required1. Deal with changing observing system2. Improve description of atmospheric

interactions with land, ocean and cryosphere (sfc fluxes)

3. Improve description of the hydrological cycle

Workshop

 No Agency in US (or internationally) has responsibility for ongoing climate analysis and reanalysis

Organizing committee:Phillip ArkinSiegfried SchubertKevin TrenberthEugenia KalnayJim Laver

Open meeting

Welcome