Towards a Definitive High- Resolution Climate Dataset for Ireland Promoting Climate Research in Ireland Jason Flanagan, Paul Nolan, Christopher Werner & Ray McGrath
Outline Project Background and Objectives Results Ongoing and Future Work Summary 1
Project Background and Main Objectives 2
High-Resolution Modelling of the Historical Irish Climate 2016: ICHEC completed 2 downscaled simulations of European and Irish Climate (1981-2016) using WRF (2km) and COSMO-CLM (1.5km). Note: 6km and 18km also. Met Éireann: MÉRA 1981-2016 reanalysis using HARMONIE NWP model at 2.5 km resolution with data assimilation All 3 datasets are archived at 1 hr intervals Estimates of observations and non-monitored ECVs to provide high-quality gridded datasets 3
Archived COSMO data Variable Units Variable Units Surface pressure Pa Surface net downward SW radiation W m-2 Mean sea level pressure Pa Average surf. net downward SW radiation W m-2 Surface temperature K Direct surface downward SW radiation W m-2 2m temperature K Averaged direct surface downward W m-2 SW radiation 2m dew point K Averaged surface diffuse downward W m-2 temperature SW radiation U-component of 10m m/s Averaged surface diffuse upward SW W m-2 wind radiation V-component of 10m m/s Averaged downward LW radiation at W m-2 wind the surface Surface roughness m Averaged upward LW radiation at the W m-2 length surface Maximum 10m wind m/s Averaged surface net downward LW W m-2 speed radiation Surface specific kg kg-1 Averaged surface photosynthetic W m-2 humidity active radiation 2m specific humidity kg kg-1 Surface albedo 0-1 2m relative humidity % Surface latent heat flux W m-2 Snow surface K Surface sensible heat flux W m-2 temperature Thickness of snow m Surface evaporation kg m-2 Height of freezing level m Total precipitation amount kg m-2 Variable Units Variable Units Precipitation rate kg m-2s-1 Soil temperature (8 levels) K Large scale rainfall kg m-2 Soil water content (8 levels) m Convective rainfall kg m-2 Large scale snowfall kg m-2 Daily average 2m temperature K Convective snowfall kg m-2 Daily maximum 2m temperature K Large scale graupel kg m-2 Daily minimum 2m temperature K Surface runoff kg m-2 Daily duration of sunshine s Subsurface runoff kg m-2 Daily relative duration of sunshine s Vertical integ. water kg m-2 vapour Vertical integrated cloud ice kg m-2 Below variables archived at 20,40,..200m Vertical integrated kg m-2 U-component of wind m s-1 cloud water Total cloud cover 0-1 V-component of wind m s-1 Low cloud cover 0-1 Air density kg m-3 Medium cloud cover 0-1 Wind speed m s-1 High cloud cover 0-1 Cube Wind speed m3s-3 CAPE 3km J/kg Wind Direction degree Surface lifted index K Showalter index Note: With the exception of the daily data, all variables are archived at one-hour intervals. K 4
Archived WRF Data Variable Surface temperature Surface pressure Sea level pressure Sea level pressure Units K Pa Pa Pa 2 m temperature C Total cloud fraction 0-1 Time varying roughness height Water vapour mixing ratio at 2m Relative humidity at 2m % U-component of wind at 10 m V-component of wind at 10m Maximum 10m windspeed previous output time Total precipitation Accumulated snowfall m kg kg-1 m/s m/s m/s mm mm Air density at lowest model level kg m-3 Variable Units u component of wind (Earth), at 40, 60, 80, 100, 120 hpa m/s v component of wind (Earth), at 40, 60, 80, 100, 120 hpa m/s Shortwave flux downward at surface instant W m-2 Shortwave flux downward at surface accumulated W m-2 Bucket Shortwave flux downward at surface accumulated Friction velocity W m-2 m/s Liquid Path Water kg m-2 Ice Path Water kg m-2 Ground Heat flux W m-2 Physical Snow Depth Water Evaporation flux at surface kg m-2 Soil Temperature, at 4 levels Soil Moisture, at 4 levels m3 m3-1 m K 5
MÉRA data Variable Units Variable Units Surface pressure Pa Cloud base m Mean sea level pressure Pa Cloud top m Surface temperature K Momentum flux, v-component N m-2 2m temperature K Momentum flux, u-component N m-2 2m relative humidity % Height of T w = 0 isotherm m u-comp of 10m wind m/s Height of 0 o isotherm m v-comp of 10m wind m/s Total precipitation kg m-2 Total cloud cover 0-1 Rain kg m-2 High cloud cover 0-1 Graupel kg m-2 Medium cloud cover 0-1 Snow kg m-2 Low cloud cover 0-1 Sensible heat flux J m-2 Mixed layer depth m Latent heat flux of evaporation J m-2 Direct SW irradiance W m-2 Latent heat flux of sublimation J kg-1 LW irradiance W m-2 Water evaporation kg m-2 Snow depth kg m-2 Snow sublimation kg m-2 Total cloud cover (fog) 0-1 Net shortwave irradiance J m-2 Visibility m Net longwave irradiance J m-2 Icing index - Direct shortwave irradiance J m-2 Precipitation type - Longwave irradiance J m-2 Variable Units Variable Units Global irradiance J m-2 U-component of wind m s-1 Direct normal irradiance J m-2 V-component of wind m s-1 Lightning m-3 Vertical velocity m s-1 Hail diagnostic kg m-2 Relative humidity % Maximum Temperature K Cloud ice kg m-2 Minimum Temperature K Cloud water kg m-2 Gust, u-component m s-1 Cloud top temperature (IR) K Gust, v-component m s-1 T b (WV) K Direct SW irradiance J m-2 T b (WV) + cld corr K Net SW irradiance J m-2 Cloud water reflectivity (VIS) - Net SW irradiance (Acc) J m-2 Precipitable water kg m-2 Net LW irradiance J m-2 Rain kg m-2 Net LW irradiance (Acc) J m-2 Snow kg m-2 Temperature K Graupel kg m-2 U-component of wind m s-1 Cloud ice kg m-2 V-component of wind m s-1 Cloud water kg m-2 Relative humidity % Soil temperature (2 levels) J m-2 Geopotential m2 s-2 Soil moisture content (3 levels) J m-2 Precipitation type K Surface soil ice (3 levels) J m-2 6
Objectives Uncertainty estimates (observational data) Evaluate relative skill of each model output Produce long-term (1981-2016) gridded and time-series datasets for priority ECVs User-defined. High-quality datasets that aid updates to / development of applications such as wind, solar and wave atlases Communicate and disseminate results to key stakeholders, policy makers and academia (reports, publications, website etc ) 7
Results to Date: A Selection 8
Uncertainty Estimates: Gridded Data Gridded datasets of 2m Temp and Precipitation from Met Éireann Step 1: Datasets are on Irish Grid and contain some negatives Step 2: Model precipitation is calculated over grid cell => calculate area averaged rainfall amounts from observations Step 3: Gather data over matching time periods Step 4: Interpolate observational data to model grid Step 5: Perform statistical analysis, e.g. 15 12-34 12-34 Bias =,34 mod,- obs,- / NxNy 15 MAE =,34 mod,- obs,- / NxNy Skill Scores 9
COSMO Precipitation mm mm Bias = 0.25% MAE = 10.02% 10
WRF Precipitation mm mm Bias = 5.98% MAE = 7.68% 11
MÉRA Precipitation mm mm Bias = 5.20% MAE = 11.61% 12
Skill Scores For a given weather parameter, which model performs best? What, if any, added value does an increase in resolution bring? Skill scores provide a measure of the value of model output Various measures available. Some examples Accuracy, Frequency Bias, Hit Rate, False Alarm Rate Hanssen-Kuiper Skill Score (KSS) Dahlgreen et al, 2016 Equitable Threat Score (ETS) Gandin and Murphy, 1992 Fractions Skill Score (FSS) - Roberts and Lean, 2008 13
Skill Scores 24-hour (0900-0900 UTC) accumulated precipitation Following Dahlgren et al (2016), zeros filtered out and thresholds chosen (0.1mm, 1mm, 5mm, 10mm, 20mm, 30mm, 50mm) Frequency distribution shows overall trend of observed rainfall is captured by the three models. 14
Skill Scores Appropriate contingency tables (shown opposite) are calculated for each threshold. Observed Modelled Yes No Total Yes Hits (H) False Alarms (FA) Modelled Yes (MY) No Misses (M) Correct Negatives (CN) Modelled No (MN) Total Observed Yes (OY) Observed No (ON) Total (T) Accuracy: 100 x (H+CN)/T % Freq Bias: (H + FA)/(H + M) Hit Rate: H /(H + M) 15
Skill Scores FA Rate: FA / (CN + FA) KSS: Hit Rate FA Rate ETS: (H - H rand )/(H + M + FA - H rand ) Fraction Skill Score (FSS) provides a measure of forecast skill against spatial scale for selected thresholds. Step 1: Convert to binary fields Suitable thresholds (q) are chosen which can be either characteristic values or percentiles I ; = < 1 O @ q 0 O @ < q and I E = < 1 M @ q 0 M @ < q 16
FSS continued Step 2: Generate fractions for neighborhoods of length n K N34 K M34 O n i, j = 4 K L I ; i + k 1 KQ4 R KQ4, j + l 1 R K N34 K M34 M n i, j = 4 K L I E i + k 1 KQ4 R KQ4, j + l 1 R Step 3: Compute fraction skill scores (FSS) MSE K = 4 1 V 1 W R O 1 V 1 K,,- M K,,- W,34-34 MSE K @XY = 4 1 V 1 W R 1 O 1 V 1 K,,- W,34 + V 1 W R M -34,34-34 K,,- FSS K = 1 E[\ ] E[\ ] ^_` 95 th %-ile COSMO MÉRA COSMO WRF 2km 1.5km 2.5km 18km WRF 18km Mean(FSS) 0.335 0.404 0.408 0.229 0.239 Std(FSS) 0.285 0.279 0.269 0.249 0.229 17
Station Observations Hourly observations of several parameters obtained from Met Éireann Hourly values of each model parameter interpolated to station locations Precipitation COSMO WRF MÉRA MAE (mm) 0.177 0.176 0.155 COSMO WRF MÉRA St.Dev. 0.326 0.327 0.294 18
2m Temperature: Station Observations Mn Err ( C ) COSMO WRF MÉRA -0.11-0.24-0.05 COSMO WRF MÉRA St.Dev. 1.93 1.85 1.31 10m Wind: Wind Roses, Bias, MAE, RMSE and StDE calculated In 68% of cases, MÉRA has the lowest MAE In all cases, COSMO has the largest MAE Some Examples. 19
Station Observations Dublin, Casement (01/01/1987 31/12/2015) WRF COSMO MÉRA Bias (m/s) -0.772 0.930 0.033 MAE (m/s) 1.649 1.911 1.055 RMSE (m/s) 2.256 2.516 1.408 StDE (m/s) 2.119 2.337 1.407 20
Station Observations Mayo, Claremorris (01/01/1987 31/12/2015) WRF COSMO MÉRA Bias (m/s) 0.050-0.014 0.887 MAE (m/s) 1.278 1.328 1.260 RMSE (m/s) 1.693 1.760 1.650 StDE (m/s) 1.693 1.760 1.391 21
Station Observations Westmeath, Mullingar (01/01/1987 31/12/2015) WRF COSMO MÉRA Bias (m/s) 0.726 1.686 1.380 MAE (m/s) 1.333 1.986 1.524 RMSE (m/s) 1.679 2.448 1.870 StDE (m/s) 1.514 1.774 1.262 22
Ongoing and Future Work Ongoing Further analysis of variables that can be validated against station and satellite data (e.g. upper-air winds and radiation) Work towards a definitive dataset for each weather parameter analysed Details and updates to appear on ICHEC website (https://www.ichec.ie) Future Determine and produce climate indices needed for climate research in Ireland Determine and produce those gridded datasets required to update the Agroclimatic Atlas of Ireland (Met Éireann) Continued collaboration with energy partners (e.g. ESBI, BNM) (wind, solar etc) Liaise with other agencies (e.g. Teagasc and Coillte) to determine their requirements of the data Selected datasets will be converted to various formats and made available through the ICHEC/EPA Climate Information Platform 23
Summary Outlined background to the project and the main objectives Overview of archived data variables A selection of results to date comparisons of precipitation, 2m temperature and winds (so far, MÉRA outperforms WRF and COSMO) Outlined ongoing and future work Questions? 24
Selected References Borsche M., A.K. Kaiser-Weiss, P. Unden and F. Kaspar, 2015, Methodologies to characterize uncertainties in regional reanalyses, Adv. Sci. Res., 12, 207-218 Dahlgreen P, T. Landelius, P. Kallberg and S. Gollvik, 2016, A high-resolution regional reanalysis for Europe. Part 1: Three-dimensional reanalysis with the regional High- Resolution Limited-Area Model (HIRLAM), Q.J.R. Meteorol. Soc. 142:2119-2131 Gandin L.S. and A.H. Murphy, 1992, Equitable Skill Scores for Categorical Forecasts, Mon. Weather Review, 120,361-370 Roberts N.M. and H.W. Lean, 2008, Scale-Selective Verification of Rainfall Accumulations from High-Resolution Forecasts of Convective Events, Mon. Weather Rev., 136, 78-97 Walsh, S., 2012. A Summary of Climate Averages for Ireland 1981 2010. Climatological Note No. 14. Met Éireann, Dublin. Eoin Whelan, John Hanley, Emily Gleeson, 'The MÉRA Data Archive', [report], Met Éireann, 2017-08-04, Met Éireann Technical Note, 65, 2017-08-04 25