Gridded observation data for Climate Services

Transcription

1 Gridded observation data for Climate Services Ole Einar Tveito, Inger Hanssen Bauer, Eirik J. Førland and Cristian Lussana Norwegian Meteorological Institute

2 Norwegian annual temperatures

3 Norwegian annual temperatures

4 Norwegian annual temperatures

5 Norwegian annual temperatures

6

7

8 Gridding of observation - aims to Provide fine scale climate information at any location or area of interest Provide data at relevant scales for various applications and purposes: - Hydrology - Agriculture and forestry - Infrastructure (roads, railways, electricity grids, municipalities) - Energy - Tourism -. Provide a observation basis for understanding and adapting to climate change Truth for calibrating post-processing (bias adjustment) of RCM / Dyn.Downscaling climate projection data. Given robust gridding methods provide space-time consistent and homogeneous information. (Local anomalies will often occur due to network changes though)

9 Gridded observation data sets at Met Norway. Monthly, seasonal and annual data - Period: 1900-present - Variables: Temperature, Precipitation - Methods: Spline interpolation of monthly anomalies. - Derived products: Degree day grids. Daily values (two parallell datasets) - Period: present - Variables: Temperature, Precipitation, Wind (in development) - Methods: Ver.1: Residual Kriging (temperature), Triangulation with terrain dependence (precipitation) Ver. 2: Bayesian spatial optimum interpolation (under operalization) Wind: Downcaling/post-processing of Hindcast data - Datasets covering Fennoscandia based on both ver.1 and 2 are in progress as a part of the FP7 UERRA project. Hourly values - Period: 2010-present - Variables: Temperature and precipitation - Bayesian spatial optimum interpolation (under operalization) Spatial resolution: 1x1 km

10 Precipitation

11 Days with precipitation

12 Number of days > p(0.99)

13 Trends in 1-day precipiation exceeding 10 mm A.V. Dyrrdal et al 2012 Dyrrdal, A.V., Isaksen, K., Hygen, H.O., and Meyer, N.K., 2012: Changes in meteorological variables that can trigger natural hazards in Norway. Climate Research, 55:

14 Heating degree days HDD = (17 TAM) TAM < 10

15 Heating degree days HDD = (17 TAM) TAM < 10

16 Growing degreedays in Oslo-Blindern

17 Change of the growing season, HadA GDD = (TAM 5) TAM > Norwegian Meteorological Institute

18 Change of the growing season, HadA GDD = (TAM 5) TAM > Norwegian Meteorological Institute

19 Days > 20 C

20 Days > 20 C

21 Days > 20 C

22 Days > 20 C, Western Norway

23 Days > 20 C, Western Norway

24 Days > 20 C

25 RCP 4.5 RCP 8.5

26 Shorter winters! Norwegian less dominant in cross-country skiing? - We will still have areas with winter conditions though..

27 Area averaged maps Municipality level MET is developing a area averaged data set on municipality level for temperature, precipitation and snow for the purpose of societal planning. Challenge: what is the most relevant area in a municipality? The entire municipality? The part with developed infrastructure?

28 Mean annual temperature ( ) Covering the whole area of the municipality

29 Mean annual temperature ( ) Covering the infrastructural area of the municipality

30 Observation gridding How we are doing it

31 Klimagrid v 2.0 Replace the current Klimagrid1.1 (SeNorge1.1) senorge: gridded datasets; Norwegian mainland +parts of Sweden and Finland; 1Km resolution source: Observations from MET Norway Climate database (manual + automatic weather station data) Precipitation 1h (RR_1) Precipitation 3h (based on hourly observations) Temperature hourly (TA) Precipitation daily (RR) Temperature daily mean (TAM) Periods: Hourly datasets onwards (hourly updated) Daily datasets onwards (updated daily)

32 Daily mean temperature 16.March 2015 Klimagrid 1.1 Temperature and precipitation Daily precipitation 29.October 2014 Period: present Resolution: 1x1 km Cover mainland Norway.

33 Klimagrid v1.1 Temperature Residual interpolation: ) u ( ) u ( ) u ( ) t(u ) u ( t m m n i i i ^ λ

34 Klimagrid v1.1 Temperature Residual interpolation: ) u ( ) u ( ) u ( ) t(u ) u ( t m m n i i i ^ λ Kriging (or any spatial interpolation method)

35 Klimagrid v1.1 Temperature Residual interpolation: ) u ( ) u ( ) u ( ) t(u ) u ( t m m n i i i ^ λ External trend/drift (linear regression) Kriging (or any spatial interpolation method)

36 Klimagrid v1.1 Temperature Residual interpolation: ) u ( ) u ( ) u ( ) t(u ) u ( t m m n i i i ^ λ External trend/drift (linear regression) Kriging (or any spatial interpolation method)

37 Correlation coefficients Coefficients terrain Coefficients position Klimagrid v1.1 Temperature Residual interpolation: ^ n t ( u0 ) λ t(ui ) ( u ) ( u ) m m( i u0 i1 ) Kriging (or any spatial interpolation method) External trend/drift (linear regression) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec LAT Month LONG ALT DEM_MEAN DEM_MIN Month ALT DEM_MEAN DEM_MIN LAT LONG -0.5

38 Trend climatological first guess

39 MET Norway TEMP1d - Evaluation Norwegian Meteorological Institute

40 Large(coarser) scale trend estimation Klimagrid 2.0 TEMP1d: OI + OI introduces the Local(finer) scale Norwegian Meteorological Institute

41 Temperature Spatial interpolation method is the same both for hourly and daily (in the future, it is possible that a post-processing of 24 hourly temperature fields will be used a first guess for the daily spatial interpolation) Spatial Interpolation method: hourly temperature Large scale after de-trending

42 Temperature: local scale

43 Old method New method New method reduce estimation error. (RMSE, independent cross-validation) Norwegian Meteorological Institute

44 Precipitation v1.1 Triangulation. Includes elevation dependence. Correction for wind-loss. - = x =

45 Precip v approach 1. Identification of Events a) Precipitation yes/no distinction (gridpoints) b) Contiguous Precipitation Areas (CPAs) identification and labeling 2. Iterative loop over Events - Multiscale Optimal Interpolation Event coarser scale At coarser spatial scales, precipitation events have a less Cascade of spatial scales: complex (smoother) from coarser to finer levels representation.... Event finer scale information At finer spatial scales, the measured precipitation shows a greater variability.... Prec Intensity strong weak

46 Spatial Interpolation Method based on Multi-scale Optimal Interpolation (Prec) Step 0: Identification of Precipitation Events (Observed Areas of Precipitation) (given the Station distribution) Event A Event B Event C Events D,... Norwegian Meteorological Institute

47 Multi-Scale Optimal Interpolation Step-by-step: from coarser to local scale Norwegian Meteorological Institute

48 Multi-Scale Optimal Interpolation Step-by-step: predicted field Norwegian Meteorological Institute

49 Challenges & limitations (i) Observation gridding is basically based on statistical relations - Quality depends on station density and representativity of the station network. The choice of external predictor should be based on a good understanding of the physical processes of the predictand (on a local scale).

50 Challenges & limitations (ii) Combine datasources? - Observations - Radar - Model fields (NWP, RA) Ensemble gridding? - Conditional stochastic simulations - Bayesian statistical analysis (OI) in «an ensemble» mode - Analogue techniques - Atm. circulation as conditional input? - Observations itself set the conditions?

51 Challenges & limitations (iii) High resolution timeseries are more «noisy» than long-term data series spatial covariance is lower (estimates are more uncertain) and the «signal» is weaker. Physical concistency between elements (precip, temperature) will not be retained. For long-term data with «low» (daily, monthly.) resolution observation gridding provide useful and precise information. For historical dataseries (> 50 years) it is probably the best (only) reference.

52 Future NWP, RCM and RRA models are «chasing» the spatial resolution of observation gridding. - Provide physical consistent data - Spatial scale is an issue. 4-5 x the resolution. The gap between observation gridding and atmospheric model approaches for providing climatological data are ready be closed. - Apply gridding concepts in postprocessing of NWP/RA - Apply NWP/RA fields as first guess - Independent validation

53 Thank you Norwegian Meteorological Institute