Reanalysis: past, present and future
Adrian Simmons Climate Change Consultant, Copernicus Climate Change Service European Centre for Medium-Range Weather Forecasts Objectives of this presentation
Climate Change To provide brief basic introductions to
– the early evolution of the atmospheric observing system
– the origin and evolution of reanalysis
– the further development of the climate observing system
– some of the issues to be addressed in preparing, assessing and using reanalyses
and thus to set the scene for the talks and posters to be presented this week Reanalysis and data assimilation
Climate Change Reanalysis entails applying a fixed modern data assimilation system to observations that typically have been made over several decades or more.
The data assimilation blends information from: ̶ many types of observation ̶ a short “background” model forecast (or forecasts) ̶ estimates of observational and background errors (including biases) ̶ dynamical relationships used in the representation of background errors. The model carries information from earlier observations forward in time. The model and background-error relationships spread information in space and from one variable to another. Development of atmospheric observation up to 1979
Climate Change Early Growth of network of surface measurements years Development of measurements from balloons
Ship 1940s Establishment of network of radiosonde Alpha measurements from North Atlantic and North Pacific Weather Ships
1957 Radiosonde network enhanced in southern hemisphere for the International Geophysical Year NOAA-2 15 Oct 1972 Operational sounding of temperature and 1972 humidity from polar-orbiting satellite Some data from commercial aircraft
1979 Improved sounding from polar orbiters Winds from geostationary orbit Much more data from commercial aircraft Drifting buoys Development of global atmospheric reanalysis
Climate Change Global modelling became established for climate in the 1960s, and global systems for numerical weather prediction were introduced at NMC (NCEP) and ECMWF in the 1970s
ECMWF and GFDL produced analyses for 1979 from Global Weather Experiment data
Analysis datasets were quite extensively used, but soon supplemented by global analyses from operational weather forecasting for multi-year studies
Frequent operational changes clouded the picture, leading to a call for reanalysis (Trenberth & Olson, 1988)
Bengtsson & Shukla (1988) made a more specific proposal for atmospheric reanalysis of the period 1979-1988 …
… and stated that the concept is equally applicable to the ocean and biosphere, and that reanalyses would be “quite useful for studying global climate change” Comprehensive atmospheric reanalyses
Climate Change The first multi-year reanalyses were produced in the early to mid 1990s − ERA-15 (1979 - 93), NASA/DAO (1980 - 93) and NCEP/NCAR (1948 - …)
A second round of production followed − ERA-40 (1958 - 2001), JRA-25 (1979 - 2014) and NCEP/DOE (1979 - …)
And a third − ERA-Interim (1979 - …), JRA-55 (1958 - …), NASA MERRA (1979 - 2016) and NOAA CFSR/CFSv2 (1979 - 2010/2011 - …)
A fourth round is proceeding − MERRA-2 (1979 - …) is now up-to-date and continued close to real time − ERA5 production is well advanced − JRA-3Q is planned to enter production in Japanese Fiscal Year 2018 − CRA-40 will be produced by CMA using NOAA and NASA systems, with completion by 2020 Diversification: atmospheric reanalysis Climate Reanalyses using subsets of the observing system Change − JRA-55C using no satellite data; NCEP CORe using no radiance data − NOAA/CIRES 20CR and ERA-20C assimilating only surface data over periods of more than a century − JRA-55AMIP, AMIP20C, ERA-20CM using observations only through specified SST and sea-ice boundary conditions and time-varying atmospheric composition
Reanalyses including trace species additional to ozone − aerosols in MERRA-2 − reactive chemical species, aerosols and greenhouse gases from 2003 in the GEMS/MACC/CAMS series
Reanalyses exploring the use of early upper-air data
Regional reanalyses − NARR, ASR, EURO4M/UERRA, …, MÉRA, … Diversification: land, ocean and coupled reanalysis Climate Land reanalyses Change − coupled with atmosphere, possibly using observed not model precipitation − or stand-alone and possibly downscaled
Ocean-wave reanalyses − coupled with atmosphere
Ocean-circulation reanalyses − possibly including sea-ice and/or biogeochemistry
Coupled atmosphere/ocean/land reanalyses − moving us closer to a climate-system (or Earth-system) reanalysis The climate observing system changes over time
Climate WCRP’s younger sibling GCOS reviews status and needs Change for observing-system improvements The needs are real, but the changing observing system poses challenges or questions for reanalysis producers − whether to assimilate observations available for only a limited period of time − how to deal with long-term variations in observation and background errors − how to ensure consistent observational bias correction over long periods of time − how to balance requirements for ◦ high quality data over a decade or less ◦ data well-suited for studying multi-decadal variations in climate − how to support appropriate use of the data Development of the oceanic observing system
Climate Change Sea surface temperature data evolve from buckets to engine intakes, drifting buoys and (from the early 1980s) satellites Sea ice data from microwave imagery began in 1978 Temperature and salinity profiles from ship transects (XBTs and CDTs) increased from the 1960s to the 1990s Tropical moored array was built up 1984-1994 under the WCRP TOGA programme Sea level has been sensed from space since late 1992 Argo profiling-float network was established between 2003 and 2007, and expanded thereafter … and more … Developments in mapping the land surface and its changes
Climate Change Developments in mapping the land surface and its changes
Climate Change
Challenges and opportunities are for validation, refinement of classification, use of improved imagery, e.g. from Sentinel-2, reprocessing and updating. And for improved land-cover datasets for earlier periods, such as the 1984-2015 record of inland and coastal surface water produced by Pekel et al. (2016). New atmospheric observing system components since 1979
Climate Change Rain-sensitive microwave imagery data in substantial numbers since 1992
Surface wind information from scatterometry since 1992
ATOVS (AMSU/MHS and improved HIRS) sounding starts in 1998; MSU & SSU end in 2006
Hyperspectral infrared sounding since 2002
Microwave limb sounding (Aura MLS) from 2004 to 20??
GNSS (GPS) radio occultation data in substantial numbers since 2006
and more … Tropical tropopause temperature
Climate Change
Significant amounts of GPSRO data assimilated in ERA-Interim but not MERRA Tropical tropopause temperature
Climate Change
Significant amounts of GPSRO data assimilated in ERA-Interim and JRA-55 Tropical tropopause temperature
Climate Change
Significant amounts of GPSRO data assimilated in all except MERRA Assimilating GPSRO data brings reanalyses together from 2006, and gives confidence in the earlier values from ERA5 and MERRA-2 Global stratospheric temperature bias
Climate Change
Significant amounts of GPSRO data assimilated Global stratospheric temperature bias
Climate Change
Significant amounts of GPSRO data assimilated Global stratospheric temperature bias
Climate Change
ERA5 data assimilation background Significant amounts of error structure functions have GPSRO data assimilated been re-derived using the observing system as it was in 1979
Assimilating GPSRO data brings reanalyses together from 2006, and but gives low confidence in the earlier values from ERA5 and MERRA-2 Enhancements of observing- system components since 1979
Climate Change Data numbers and quality from in situ observing systems have in general increased since 1979.
Data in new BUFR code (providing higher resolution radiosonde ascents) are used in ERA5 but not ERA-Interim. Enhancements of observing- system components since 1979
Climate Change Data numbers and quality from in situ observing systems have in general increased since 1979.
Data in new BUFR code (providing higher resolution radiosonde ascents) are used in ERA5 but not ERA-Interim.
Databases need to flag incorrect Drifter numbers reach target, but early BUFR data and duplicated surface pressure is measured data that were provided in both from only about half the buoys alphanumeric and BUFR codes.
Scope to recover data not used Data from moored TAO Drop in data numbers during changeover to BUFR codes array were first reported due to longevity using DRIBU code problem for US buoys Recovery, rehabilitation and reprocessing of observations Climate The needs include Change ̶ discovery and digitisation of older data held on paper or in other non-digital forms ̶ further development of bias and other error corrections for data that will not be adjusted within reanalysis systems ̶ merging of existing and new holdings of in situ data into modern fit-for-purpose archives, including metadata related to quality control and estimated biases, some of which may be provided by existing reanalyses ̶ tidying up after the change from alphanumeric to BUFR codes is complete ̶ gap-filling and recalibration to construct improved “Fundamental Climate Data Records” for data to be assimilated as radiances or in other basic forms ̶ reprocessing of data to be assimilated or otherwise used in the form of retrieved geophysical variables
Progress has been good in some areas, and slow to start in others A substantial and expanding number of users
Climate Change
Users of the public data server for ERA-Interim, 2016
ERA-Interim data are served to tens of thousands of users per year More than 2000 articles published to date in 2017 cite ERA-Interim The user survey by the EU FP7 CORE- CLIMAX project
Climate Change More than 2500 respondents
“Climate” and “weather” top the list of respondents’ work fields; “oceans and seas” come third
Characterization of uncertainty, provision of training material and information on observational input were the topics where users on the whole were least satisfied
Spatial resolution, timeliness of data delivery and biases were of particular concern for users from some specific sectors But one can hear from “authoritative” sources that:
Climate Change Multiple reanalysis products yield substantially different and even opposing trends … Consequently, reanalysis products are still considered to be unsuitable … Newer reanalyses have to be different from the older reanalyses that they improve on
Reanalyses, being “model based”, might be deemed to lack independence from the model projections Not for variables for which the differences between reanalyses are very much smaller than the differences between models; or for which a reanalysis fits the observations much more closely than a free model does
When both are available, priority is given to “observational datasets” over reanalyses Why? Neither will be perfect, and the reanalysis may not be the more imperfect Evaluation is key to establishing confidence in reanalyses
Climate Change What data are assimilated? How well do the background and analysis fit these data? How does this change over time? How large are the changes made each assimilation cycle? How do the changes vary in space and time? How do the reanalyses compare with one another and with any alternative observationally-based datasets for a particular variable? How well do reanalyses fit independent unassimilated data? How well do they perform locally? How consistent are their global budgets? How small are any adjustments made to ensure balance? How are applications improved by including reanalysis data? To conclude
Climate Change Reanalysis has achieved much since production of the initial multi-decadal datasets began some 25 years ago
Looking back to the time of the Fourth International Conference in May 2012 ̶ there have been many welcome developments since then in terms of the scope of reanalyses, who is producing, assessing or using them, and their improved quality ̶ but challenges remain, some of which are only now beginning to be addressed
But we should not dwell on conclusions so early in this conference ̶ for now we should learn from the talks and poster presentations ̶ and return later in the week to take stock of where we are, where we are going, and what we need to do to get there in good shape