6/17/2016. Content. My credentials. Who am I. Inhomogeneities. Global temperature changes

Transcription

1 Content Is the global mean temperature trend too low? Victor Venema Ralf variable-variability.blogspot.com About me Global mean temperature change Independent lines of research 1. Statistical homogenization 2. Physical understanding (parallel measurements) 3. Modelling (UHI, irrigation, radiation screens) Other changes in the climate system Who am I Studied physics in Groningen PhD on radar measurements of clouds in Delft Postdoc on radiative transfer through clouds in Bonn Stochastic modelling of clouds Stochastic modelling of climate data to validate homogenization methods COST Action HOME: Since 2011: PI on homogenization daily data Since 2014: development of a robust homogenization method for global collections My credentials Executed the validation study of homogenization methods of the COST Action HOME Member of the benchmarking group of the International Surface Temperature Initiative (ISTI) Chair of the ISTI Parallel Observations Science Team (POST) Chair of the Task Team on Homogenization Steering Committee of MEDARE, MEditerranean climate DAta REscue Global temperature changes Inhomogeneities Land Surface Temperature Sea Surface Temperature Figures: Zeke Hausfather Figure: IPCC (2013) 1

2 Global land surface trend Homogeneous NOAA-GHCNv3 Trend: 0.8 C per century since 1880 Raw data: 0.6 C HadISH Breaks: 0.12 C bias One break every 15 years Period C per century Station by station basis Homogeneous: means of the same nature (comes from the Greek!!) A homogeneous climate time series is defined as one where variations are caused only by variations in climate Conversely, an inhomogeneous climate time series is one which contains variations (biases) caused by factors other than climate: inhomogeneities Homogenisation: WHY? Example of PAU-UZEIN temperature Minimum temperature Reno, USA 1912 PAU-LESCAR (EN) 2005 PAU-UZEIN (AERO) Slide: Olivier Mestre Physical causes of inhomogeneities Shelter type, exposure Radiation & wetting protection Natural or forced ventilation Snow cover Plastic screen: insolation on hot days Change surrounding Siting quality, Urbanization, growing vegetation, Irrigation Relocation of station City-> airport, suburbs, Village centre to outside lower heights Deurbanisation of network Instrument Response, integration time Zero drift, shrinking glass initial years Calibration errors Temperature out of range Quicksilver thermometers: T < -39 C Definitions Computation daily means Measurement procedures Reading times Maintenance procedures AWS: Icing, damage detection Painting & cleaning schedule Digitisation & database Minus sign forgotten Station names mixed up Pre-homogenised data 2

3 Relative homogenization approaches Pairwise Mathematically tractable Inhomogeneous reference Candidate & reference: noise Composite Better signal to noise ratio Reference ideally regional climate signal Half the noise, noise of candidate Reference assumed to be homogeneous Difficult programming task Also due to gaps in the data Advances in relative homogenization Detection Single breakpoint + splitting series Multiple breakpoint (number of breaks) If SNR low high position error of detected breaks Correction Single breakpoint: One after another Accumulation of errors Multiple breakpoint (so-called ANOVA) Decomposition Regional climate signal Inhomogeneities per station (step function) Noise to be minimized On average: under-correct trend biases Conclusions from HOME validation homogenization algorithms Relative homogenisation improves temperature records Best algorithms Function with an inhomogeneous reference Multiple breakpoint methods (ACMANT), Craddock, MASH, PRODIGE, PHA/USHCN, (HOMER) Automatic algorithms among the best (no metadata) Moderate correlation between error metrics Break detection scores not good predictor of skill 3

4 Trend uncertainties Trend uncertainties Benchmarking gives qualitative idea of uncertainties HOME: 10%, NOAA: 90% Both high density networks (EU and USA) Several caveats of unknown importance Size of the breaks in HOME too large Network density like Europe Much of the world SNR will be less Length of the series not varied Benchmark more comparable to real data ISTI benchmarking cycle These points will be fixed in the benchmarking of the International Surface Temperature Initiative Not much work on computing uncertainties from remain in homogeneities Multiple sources of uncertainty: Not all breaks in the candidate station can be detected Uncertainty in the estimation of correction parameters due to insufficient data Uncertainties in the corrections due to remaining inhomogeneities in the references Especially network-wide transitions The date of the break may be imprecise (see Lindau & Venema, 2013b) Inhomogeneities Well-homogenized national datasets Land Surface Temperature Sea Surface Temperature Figures: Zeke Hausfather Compared global collection Annual mean averaged over same countries Berkeley Earth Surface Temperature (BEST) GHCNv3 GISS CRUCY (& CRUTEM4) National datasets are expected to be better More data: better correlated references More metadata: station history More care and better methods Trend since 1901 Trend since 1961 Trend since 1901 Region Regional GHCNv3 Diff GHCNv3 BEST Diff BEST Austria 1,43 1,24 0,20 1,16 0,27 Italy 1,41 0,80 0,61 1,04 0,37 Switzerland 1,57 1,43 0,14 1,18 0,39 Average 0.31 (p=0.17) 0.34 (p=0.01) Region Regional CRUCY Diff CRUCY GISS Diff GISS Austria 1,43 1,36 0,07 0,99 0,45 Italy 1,41 0,89 0,52 1,02 0,39 Switzerland 1,57 1,48 0,08 1,19 0,38 Average 0.23 (p=0.27) 0.41 (p=0.002) Trend since 1961 Region Regional GHCNv3 Diff GHCNv3 BEST Diff BEST Australia 1,60 1,50 0,10 1,62-0,02 Austria 3,69 3,19 0,51 3,14 0,55 France 3,47 3,45 0,02 3,10 0,37 Hungary 2,88 2,81 0,07 2,73 0,15 Italy 3,28 2,83 0,45 2,77 0,51 Slovenia 3,41 3,12 0,28 3,28 0,13 Switzerland 3,95 3,70 0,25 3,11 0,84 Average 0.24 (p=0.02) 0.36 (p=0.02) Region Regional CRUCY Diff CRUCY GISS Diff GISS Australia 1,60 1,23 0,38 1,66-0,06 Austria 3,69 3,31 0,38 3,10 0,60 France 3,47 3,38 0,09 3,15 0,32 Hungary 2,88 2,96-0,08 2,80 0,07 Italy 3,28 2,67 0,61 2,81 0,47 Slovenia 3,41 3,72-0,32 3,13 0,28 Switzerland 3,95 3,91 0,04 3,09 0,86 Average 0.16 (p=0.24) 0.36 (p=0.02) 4

5 Trend since 1981 Physical reasons: Parallel data Trend since 1981 Region Regional GHCNv3 Diff GHCNv3 BEST Diff BEST Australia 1,95 1,57 0,37 1,55 0,40 Austria 4,74 4,06 0,68 4,02 0,72 France 4,15 3,98 0,18 3,33 0,82 Hungary 4,80 4,71 0,08 4,46 0,33 Italy 4,53 3,83 0,70 3,61 0,92 Slovenia 5,32 5,37-0,05 4,49 0,82 Switzerland 4,35 4,85-0,51 3,80 0,54 Average 0.21 (p=0.24) 0.65 (p=3e-04) Region Regional CRUCY Diff CRUCY GISS Diff GISS Australia 1,95 0,82 1,13 1,74 0,20 Austria 4,74 3,96 0,78 4,13 0,61 France 4,15 3,77 0,38 3,51 0,65 Hungary 4,80 4,51 0,29 4,51 0,28 Italy 4,53 3,27 1,26 3,88 0,65 Slovenia 5,32 5,48-0,17 4,47 0,85 Switzerland 4,35 4,37-0,02 3,71 0,63 Average 0.52 (p=0.05) 0.55 (p=7e-04) ISTI Parallel Observations Science Team Produce an open database Initially data is restricted to contributors Incentive to contribute Until first joint paper(s) by contributors are written First actions: Inventory of parallel datasets (WMO RA Europe) Data processing for database More information Radiation error Radiation error Climates largest radiation errors: * Strong insolation * Low wind * Dry ground * High specific humidity Tropical and continental climates Parallel measurements Transition to Stevenson screens North-West Europe: < 0.2 C (Various, Parker) Basel, Switzerland: 0 C (Wild screen) Kremsmünster, Austria: 0.2 C (North-wall) Adelaide, South Australia: 0.2 C (Glaisher stand) Spain: 0.35 C (French screen) Sri Lanka: 0.37 C (Tropical screen) India: 0.42 (Tropical screen) 5

6 Sources of global temperature trend bias Details: Transition to Stevenson screens Transition to Automatic Weather Stations Urbanization Urban Heat Island and relocations Relocations to airports Station siting quality Centre of villages to current location outside Problems for more warming hypothesis Tropospheric temperatures Radiosonde Passive microwave satellites (AMSU) Sea surface temperature Irrigation & watering Ratio warming over land to warming sea Sutton et al. Land/sea warming ratio in response to climate change: IPCC AR4 model results and comparison with observations. GRL, Ratio warming over land to warming sea Sutton et al. Land/sea warming ratio in response to climate change: IPCC AR4 model results and comparison with observations. GRL, ~2 HadCRT2v ( ) Erring in the side of least drama? Temperature reconstructions Melting of Arctic sea ice Relocations of storm tracks Slowdown Atlantic Meridional Overturning Circulation Permafrost? Sea level rise Ocean heat content Trend in mean precipitation Severe precipitation Climate sensitivity & aerosols in GCMs Trend in down welling infra-red flux Trend in lake temperatures, ice season shorter Snow extent Phenology? Temperature reconstructions The Copenhagen Diagnosis Figure 19: Northern Hemisphere reconstructed temperature change since 200AD 6

7 Global temperature changes Lake and rive freeze and breakup times Figure: IPCC (2013) Magnuson et al., Science 2000 Sea level rise Church & White, 2011 Summary Global mean temperature similar for all global collections Considerable corrections needed Small bias per break is important Relative homogenization has improved a lot Understand that removing large-scale bias is hard Trend difference between well-homogenized dataset and global collections Historical changes allow for a trend bias Cooling biases much too little studied Many other changes in climate system fast Conclusions & outlook Need better mathematical understanding of how well trend biases can be removed Need better homogenization methods and apply them to global datasets Need to exchange more data & metadata Understanding of cooling biases is poor Large global parallel dataset can help ISTI-POST Need to study other climatic changes in the light of possible temperature trend biases If trend in other parameters do not fit with models, temperature trend bias often not considered Questions? Victor.Venema@uni-bonn.de 7