Towards a Convective Scale Reanalysis with a - PowerPoint PPT Presentation

Towards a Convective Scale Reanalysis with a Soil-Vegetation-Atmosphere- Transfer-Model International Symposium on Regional Reanalyses Clarissa Figura, Insa Thiele-Eich, Jan D. Keller, Wolfgang Kurtz, Clemens Simmer, Andreas Hense 19.07.2018

Motivation Reanalyses offer spatially and temporally consistent data sets for global or regional grids and a certain vertical exent within the land surface and atmosphere. Applications: • Renewable Energy Sector • Regional climate analyses • Hydrology • Agricultural economics

Motivation Reanalyses offer spatially and temporally consistent data sets for global or regional grids and a certain vertical exent within the land surface and atmosphere. Applications: • Renewable Energy Sector • Regional climate analyses • Hydrology • Agricultural economics Demand for regional reanalyses is growing!

Challenges 1 Misrepresentation of hydrology (soil moisture-evaporation- precipitation feedback) in atmospheric reanalyses, especially at catchment scale • Betts et al. 1998 and 2003: Investigations of water and energy budget in ECMWFs reanalysis for Mississippi and subbasins → Coupled reanalysis approach

Challenges 2 There exists a coupling between convection triggering and soil moisture (Cioni and Hohenegger, 2017) • idealized LES-Simulations show a strong coupling between soil moisture and diurnal precipitation cycle  Convective scale setup

Challenges 3 Models are not perfect due to parametrizations and the chaotic nature of the systems they represent (Lorenz, 1969).  Ensemble approach with perturbed realizations

Experimental setup TerrSysMP With kind permissions of P. Shrestha and M. Sulis

Experimental setup KENDA (Kilometre Scale ENsemble Data Assimilation) is a Local Ensemble Transform Kalman Filter

Experimental setup KENDA

Experimental setup KENDA

Experimental setup KENDA → Basic Cycling System (BaCy) Observations Observations COSMO COSMO COSMO COSMO … LETKF LETKF Ensemble Ensemble Ensemble Ensemble Since March 2017 operational at Deutscher Wetterdienst

Experimental setup KENDA with TerrSysMP ...work in progress Observations Observations TerrSysMP TerrSysMP TerrSysMP TerrSysMP … LETKF LETKF Ensemble Ensemble Ensemble Ensemble

Experimental setup EMVORADO (Efficient Modular VOlume RADar Operator) → Radar forward operator QR H EMVORADO QS … Model Simulation of = x b = y b output radar beams

Experimental setup LETKF LETKF TerrSysMP TerrSysMP TerrSysMP TerrSysMP ?

Experimental setup

Preliminary work COSMO-REA6 COSMO 4.25/TERRA • ∆xy= 6 km • Atmospheric boundary: ERA-Interim • Data assimilation: • Nudging of conventional observations (e.g., buoys, radio soundings, aircraft) Soil Moisture / Snow / SST Analysis

Preliminary work Long-term deterministic TerrSysMP simulation (Mauro Sulis) TerrSysMP fully coupled (COSMO 4.x, CLM 3.5) • ∆xy=1000 m/ 500 m • Atmospheric boundary: COSMO-DE-Analysis • Downscaling (no data assimilation) •

First results Ensemble TerrSysMP downscaling TerrSysMP fully coupled (COSMO 5.1, CLM 3.5) • ∆xy=1000 m/ 500 m • Atmospheric boundary: COSMO-DE-KENDA-Analysis • Time period: 16.05.-13.06.2014 •

First results Case study: Front 12 h accumulated rain [mm] for 21.05.2014 00 UTC - 21.05.2014 12 UTC

First results Case study: Convection 12 h accumulated rain [mm] for 22.05.2014 12 UTC - 23.05.2014 00 UTC

First results Feedback precipitation – soil moisture 28.05.2014 IQR Stdev Soil moisture (pressure) Precipitation (mm/24h)

First results Feedback precipitation – soil moisture 26.05.2014 IQR Stdev Soil moisture (pressure) Precipitation (mm/24h)

First results Contingency table Observed yes no false hit yes alarm Model correct no miss negatives Observations: www.bremerhaven-wetter.de 68 DWD rain gauges stations in NRW domain

First results Contingency table 16.-22.05.2014 Threshold: 0.1 mm/h Model BIAS ETS Log Odds COSMO-REA6 1.71 -0.03 -2.36 TSMP det 1.13 -0.04 -2.09 TSMP ens 1.04 -0.05 -1.78 Threshold: 0.2 mm/h Model BIAS ETS Log Odds COSMO-REA6 2.95 -0.01 -∞ TSMP det 1.06 -0.03 -∞ TSMP ens 0.95 -0.03 -2.62

First results Contingency table 16.-22.05.2014 Threshold: 0.1 mm/h Model BIAS ETS Log Odds COSMO-REA6 1.71 -0.03 -2.36 TSMP det 1.13 -0.04 -2.09 TSMP ens 1.04 -0.05 -1.78 Threshold: 0.2 mm/h Model BIAS ETS Log Odds COSMO-REA6 2.95 -0.01 -∞ TSMP det 1.06 -0.03 -∞ TSMP ens 0.95 -0.03 -2.62

First results Shortcomings in representation of precipitation: → Data assimilation (LETKF) is expected to enhance the results

Conclusion • Qualitatively TerrSysMP with dynamic downscaling of analyses is able to better reproduce small scale precipitation events in comparison to COSMO-REA6 • Extended simulation time period necessary to evaluate the quantitative accuracy of precipitation with verification scores

Outlook • Extended time period with TerrSysMP-KENDA with EMVORADO planned • Quantitative model comparison and verification in terms of precipitation and soil moisture with independent observations • Evaluation of feedback of soil moisture on precipitation • Evaluate impact of initial conditions

Towards a Convective Scale Reanalysis with a Soil-Vegetation-Atmosphere- Transfer-Model International Symposium on Regional Reanalyses Clarissa Figura, Insa Thiele-Eich, Jan D. Keller, Wolfgang Kurtz, Clemens Simmer, Andreas Hense 19.07.2018

First results Contingency table Frequency Bias (BIAS): • BIAS=(hits+false alarms)/(hits+misses) • ratio of the frequency of modeled events to the frequency of observed events → BIAS <1: underforecast, BIAS >1: overforecast, perfect: 1 Equitable Threat Score (ETS): ETS=(hits+hits random )/(hits+misses+false alarms-hits random ) • • fraction of observed and/or modeled events that were correctly predicted, adjusted for hits associated with random chance → Range: -1/3 to 1, no skill:0, perfect: 1 Log Odds Ratio (LOR): • LOR=log((hits+correct negatives)/(misses+false alarms)) • ratio of the odds of making a hit to the odds of making a false alarm → Range: -∞ - +∞ no skill:0, perfect: +∞