1 / 23

ERA-CLIM - European Re-Analysis of Global Climate Observations

ERA-CLIM - European Re-Analysis of Global Climate Observations WP 4 : Quantifying and reducing uncertainties Workshop on observations errors 19-20 April 2012, Vienna, Austria FFCUL - Fundação da Faculdade de Ciências da Universidade de Lisboa IDL – Instituto Dom Luiz

neylan
Download Presentation

ERA-CLIM - European Re-Analysis of Global Climate Observations

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. ERA-CLIM - European Re-Analysis of Global Climate Observations WP4: Quantifying and reducing uncertainties Workshop on observations errors 19-20 April 2012, Vienna, Austria FFCUL - Fundação da Faculdade de Ciências da Universidade de Lisboa IDL – Instituto Dom Luiz Rua da Escola Politécnica, 58 - Lisboa, Portugal – Meteorological station Edifício C8, Campo Grande, Faculdade de Ciências da Universidade de Lisboa, Portugal

  2. WP4: Quantifying and reducing uncertainties Errors in surface observations and first steps in homogenising Portuguese records People involved in WP4 for FFCUL: Researchers: Maria Antónia Valente – Coordination IDL Researcher and Meteorological Station Manager Ricardo Trigo IDL Researcher and Universidade Lusófona Associate Professor Pedro Gomes Research scholarship and trainee meteorological observer IDL Julia Bethke Research Scholarship WP4 (from 1st May 2012)

  3. WP4: Quantifying and reducing uncertainties Task 4.1: Bias assessments, homogenisation, and variational bias correction FFCUL/IDL will participate in this task essentially by performing homogeneity tests and homogenisation of data resulting from pilot reanalysis. Meanwhile, the first steps for testing the long time series of partially QC’d observations homogeneity are being taken. WP4: Quantifying and reducing uncertainties Coimbra Tmax and Tmin for 1864-2007 has been subjected to the SNHT, Buishand and Pettitt tests. The breakpoints for the annual and monthly series of DTR=Tmax-Tmin are shown in the figure below Figure courtesy of Javier Sigró, 2007 Using the SNHT with several Spanish reference series, Javier Sigró has homogenised the Lisbon/Geofísico surface pressure series. The procedure took into account the metadata for the station, which included several barometer replacements. Figure courtesy of Anna Morozova, 2011

  4. WP4: Quantifying and reducing uncertainties • Error and inhomogeneity sources in surface data • Pressure: • Gravity corrections • Change of barometers • Change of station location • Mixing gravity corrections with units conversion (mm to hPa) • Considering gravity correction with only minutes of latitute or including seconds • For coastal stations mixing slp with mslp. • Bias in observations, depending on the observer (where to position the rule, at top of mercury column or more in the middle). • Change of observations hours (change in local time) • Case study Lisbon Pressure • Long time series of Lisbon surface pressure were digitised separately for 1864-2006 (at 9am, 12pm, 3pm, 6pm and 9pm). • These series were subjected to homogeneity tests and a homogenisation procedure.

  5. WP4: Quantifying and reducing uncertainties • Case study Lisbon Pressure • Before homogeneity tests were applied it was verified that: • Pressure read in mmHg from 1856 to 14 December 1993, afterwards read in hPa. • Pressure published in hPa since 1938 -> 1938-1993 conversion from mm to hPa • Metadata – Barometer change dates 1896, 1918, 14/12/1993, 1995. • Correction for dates before 1896 +0.25mm (or +0.15mm? As indicated in 1904?) • Gravity correction not applied before 1938. After 1938 we don’t know exactly when they started to be applied • After 1947 observations done at GMT, before 1947 at local time (+37 min) • But legally in 1911 the local time changed to GMT. • After 1938, gravity correction done with g=9.8009m/s2 for latitude 38º 42’ 59.4’’ N, altitude 95.4m. Normal gravity at 45ºN is 9.80665m/s2. • After 15 December 1993 gravity correction done with 9.7997m/s2 meaning that a latitude of 38º 42’ N is being used.

  6. WP4: Quantifying and reducing uncertainties Case study Lisbon Pressure Manuscript pressure avaliable Pressure published in hPa, gravity correction introduced? Observation hours changed to GMT Barometer changed, gravity correction changed slightly Barometer changed After 1941 there are 2 pressure series, IM and IGIDL Between 1970 and 1993 many problems were detected: IM(GC), IGIDL(no GC) IM=IGIDL(no GC) IM wrong, IGIDL(GC) IM(no GC), IGIDL(GC) IM=IGIDL (GC)

  7. WP4: Quantifying and reducing uncertainties • Case study Lisbon Pressure • IGIDL published annales with surface station pressure (1856-1999) • IM gave us digitised data from 1941-2006. • We have manuscript data from 1970 onwards and verified that: • Jan/1970 to Mar/1971, IM and IGIDL series are equal and include both gravity correction; • Apr/1971 to 15/Dec/1971, IM series has no grav. correction, IGIDL has gravity correction; • 16/Dec/1971 to 31/Dec/1971, IM has non identifiable wrong values, IGIDL series has grav. correction; • Jan/1972 to Dec/1974, IM and IGIDL series are equal and include both gravity correction; • Jan/1975 to Jun/1985, IM series does have grav. correction and IGIDL doesn’t; • Jul/1985 to 14/Dec/1993, IM and IGIDL series are equal and don’t include grav. correction; • 15/Dec/1993 to 2006, IM and IGIDL series are equal and include both gravity correction. • Between 1941 and 1969 IM and IGIDL series are equal and probably include grav. correction (verification was possible with some scattered manuscript data). • - From 1938-1940 (only IGIDL series) we don’t exactly know if gravity correction is included, but supposed that it is

  8. WP4: Quantifying and reducing uncertainties Case study Lisbon Pressure How on earth can we work with series like these? What we did: Took off all supposed gravity corrections. We got a series with no gravity corrections (which was sent to ISPD) Wanted to test its homogeneity (Javier Sigró – C3SNHT) Figure courtesy of Javier Sigró, 2007 First attempt to catch breakpoints detected too many abrupt changes, so homogenisation had to be performed using metadata information. Breakpoints detected and corrected First correction – 1993 ; Second correction - 1938 Third – 1917; Fourth – 1895

  9. WP4: Quantifying and reducing uncertainties Case study Lisbon Pressure C3SNHT Javier Sigró, Enric Aguilar After doing QC and using the C3SNHT with several Spanish reference series (A Coruña, Cadiz-San Fernando, Madrid and Barcelona), Javier Sigró has homogenised the Lisbon/Geofísico surface pressure series. The procedure took into account the metadata for the station, which included several barometer replacements, imposing breakpoints. PRODIGE? Olivier Mestre Figure courtesy of Olivier Mestre, 2008 Olivier Mestre has also looked for breakpoints on the Lisbon yearly pressure data. He found 1895, 1994 and 1972!!??

  10. WP4: Quantifying and reducing uncertainties • Error and inhomogeneity sources in surface data • Temperature: • Change of station location • Introduction of shelters • Many shelters were taken from higher points to the lower ground • Change of thermometers • Change in local time observations • Changes in the enveloping area • Volcanic eruptions have to be taken into account when detecting temperature homogeneity breaks • Case study Lisbon, Porto and Coimbra Temperatures • (Anna Morozova) • Long term series series of daily tmax and tmin for Lisbon, Porto and Coimbra were digitised by IGIDL, IGUC and IM • Homogeneity tests programmed and used: • Von Neumann ratio test – non-parametric; Buishand test – parametric • SNHT – likelihood ratio; Pettit test – non parametric rank

  11. WP4: Quantifying and reducing uncertainties Metadata and need for correction

  12. WP4: Quantifying and reducing uncertainties Case study Lisbon, Porto and Coimbra Temperatures Tmin, Coimbra 1864-2007 DTR=Tmax-Tmin, Porto 1888-2002 DTR=Tmax-Tmin, Lisbon 1854-2010 Annual data Gray lines – metadata changes Each month was treated separately Means calculated before and after breakpoint with an appropriate time interval dependent on the homogeneous periods

  13. WP4: Quantifying and reducing uncertainties Case study Lisbon, Porto and Coimbra Temperatures Buishand, SNHT and Pettit test results for Lisbon annual DTR series Cyan lines – volcanic eruptions Gray lines – metadata changes

  14. WP4: Quantifying and reducing uncertainties Case study Lisbon, Porto and Coimbra Temperatures Tmin, Coimbra 1864-2007 DTR=Tmax-Tmin, Porto 1888-2002 DTR=Tmax-Tmin, Lisbon 1854-2010 Temp. homogenisation (Anna Morozova) – method uses no reference stations However corrections are done only when metadata supports the detection of breakpoints We’re looking for reference stations (Spanish series that have been homogenised, Manola Brunet), SST series (UKMetOffice?) Applying methodology for series resulting from doing difference between stations (stationA –stationB, etc)

  15. WP4: Quantifying and reducing uncertainties • Error and inhomogeneity sources in surface data • Wind: • Change of station location • Change in the enveloping area • Change of anemometers • Change in local time observations • The units of wind speed can be misleading, some appear in km/h, others in an old scale • Lisbon wind speed data have to be corrected up to 1938 because of type of anemometer • Relative humidity, water vapor pressure: • Changes in station location and enveloping area • Change in the methodology of calculating rel. humidity and water v.p. • Changes in thermometers • Introduction of shelters • Wet bulb thermometer maintenance • For Lisbon, and other mainland Portuguese stations, an increase in relative humidity has been detected since the 1940’s. • Difference for the automatic station in Lisbon is too big at the moment. Automatic stations tend to get drier values after a year.

  16. WP4: Quantifying and reducing uncertainties • Error and inhomogeneity sources in surface data • Precipitation: • Change in the station location and enveloping area • Change in rain gauge location, many changed from higher places to parks at lower altitude • Using the wrong rain colector • Distribution of precipitation along the day depends on the rain gauge register and for large values, the tipping after 10 mm can be delayed, leading to less precipitation registered. • Automatic stations (after 1994 in Portugal) introduce many errors, underestimating precipitation values. • Could cover: • The definition of cloud cover has changed with time. • Depends on the observer, but is not the most likely value to get wrong. • Evaporation: • Location of station • Introduction of shelters • Maintenance of the filters on the Piche evaporimeter (can increase evaporation) • Filters type.

  17. WP4: Quantifying and reducing uncertainties • Error and inhomogeneity sources in surface data • Grass temperatures: • Location of station • Minimum grass temperature thermometers tend to be faulty after being exposed to sun (alcohol column breaks and decreases Tmingrass), it is substituted by EMA values and they are substancially different. • Soil temperatures: • Change in thermometers location • How correctly positioned on the soil they are • Depths have varied much along the last 150 years. • Visibility: • Depends on how well positioned the observation point is • Depends on the observer, lower values are easy to get wrong • Cloud types and height of cloud base: • Varies wildly with observer. • Sunshine duration: • Dependent on instrument used and strips used.

  18. WP4: Quantifying and reducing uncertainties • Surface data to be subjected to homogeneity tests and homogenisation procedures • Lisbon, Porto and Coimbra long time series 1800’s to 2000’s • Angra, Ponta Delgada (Azores), Funchal (Madeira) 1800’s to 1946 • Former Portuguese colonies: Angola, Cape Verde, São Tomé, Macau (China) • 1800’s to 1946 • Mozambique, Guiné-Bissau, Goa (India) • 1915 to 1946 • Other Portuguese mainland stations: 1800’s to 1946. • Data series are being digitised, many are complete and are undergoing Quality Control procedures (specially the Azores, Madeira and former Portuguese colonies) • Some Lisbon, Porto and Coimbra series are being used to test homogeneity procedures. • We’ll suply the data QC’d to WP3 and expect to receive feedback from OFA (Pilot Reanalyses). The bias calculated by the Reanalyses will guide our homogeneity tests and homogenisation procedures, together with metadata.

  19. WP4: Quantifying and reducing uncertainties Cape Verde – Ilha S. Vicente – Mindelo February 1915 Pressure Colonies Annales Two different sources give different values for pressure in Mindelo – Cape Verde during February and March 1915. For the remaining months the values are the same in both sources As the IGIDL Annales have surface pressure, the challenge is to know what pressure values were printed for February and March in the Colonies Annales IGIDL Annales

  20. WP4: Quantifying and reducing uncertainties Homogenisation procedures used and to be used: Used for Pressure: C3SNHT (Javier Sigró, Enric Aguilar, with reference series and metadata) Olivier Mestre (PRODIGE?) Used for Temperature: SNHT, Buishand, Pettit, Von Neuman Software programmed by Anna Morozova (no reference series used; we need SSTs as reference series near Portugal and other territories) Other tests to be used: RhtestV3 (Xiaolan Wang) (learning to use it with pressure with and without reference series) HOME-R (HOME Homogenization Cost Action) AnClim (Petr Stepanek) ACMANT (Peter Domonkos temperatures – recommended by Manola Brunet) MASH(?), Climatol (?) (Guijarro)

  21. WP4: Quantifying and reducing uncertainties RhtestV3 Lisbon annual pressure series 9am (1864-2005), no reference series used Finds the same breakpoints as Olivier Mestre, 1895, 1972 and 1994

  22. WP4: Quantifying and reducing uncertainties RhtestV3 Porto – Serra do Pilar annual pressure series 9am (1893-2001), no reference series used Finds the breakpoints 1923 and 1954, plus 3 periods with trends.

  23. WP4: Quantifying and reducing uncertainties • Questions raised so far: • We’re not sure if colonies pressure data have gravity correction, probably not. • For colonies data, do we apply homogenisation procedures with series that go only up to 1946? (Yes, methinks, as long as we have the Reanalysis feedback). • Other than temperature and pressure we want to test the homogeneity of other variables (wind, speed and direction, precipitation, humidity, etc) • Published metadata is not always reliable. In the printed logbooks there is a tendency to repeat on the following year the text used before, so changes are not always given at the right times. • Monthly data per se is more often wrong than monthly data obtained from daily data. We detected many wrong monthly means in old logbooks. • Mixing of classical series, synop series and automatic ones is very common and not advisable. • (Freak upper air data question) In pilot balloon data why do the altitudes published vary so frequently from year to year?

More Related