Center for Weather Forecast and Climate Studies Brazilian
Center for Weather Forecast and Climate Studies Brazilian National Institute for Space Researches SIXTH WMO WORKSHOP ON THE IMPACT OF VARIOUS OBSERVING SYSTEMS ON NWP Shanghai, China 10 -13 May 2016 Assessing observation impacts on the INPE/CPTEC global data assimilation system over South America by LUIZ SAPUCCI Data Assimilation Group Modeling and Development Division luiz. sapucci@cptec. inpe. br
Motivation • CPTEC/INPE’s mission: "To provide to country with weather and climate predictions, for the benefit of the Brazilian people. “ • The proposal of this center is to make global modeling to offer the best forecast over Brazilian territory and neighboring region; • The focus of this presentation is the observation impact in the improvement of the forecast quality over South America. • The results from several studies recently developed (or in development) at CPTEC are showed here and the respective collaborators are mentioned.
NWP’s challenger over South America Figure from BRAMS website Andes mountains Amazonian forest Larger portion in the tropical region; Specific adjustments in the model are necessary, consequently the observation impact presents some particularity, for example larger dependency of satellite observations, which are explored in this presentation. Located between the two largest oceans: larger regions with poor network of observation at surface.
Content of the presentation • Model used: – CPTEC global model T 213 L 42 (42 km) • Data assimilation systems available: – LETKF (Local Ensemble Transform Kalman Filter); • Research mode; – GSI (Gridpoint Statistical Interpolation); • Operational mode; • Studies about impact of observation at CPTEC: – – – FSO using LETKF at CPTEC/INPE (state-space); FSO using GSI at CPTEC/INPE (observation-space); Impact of Land Data Assimilation at CPTEC model on precipitation; Impact of RO-GNSS refractivity data using LETKF/CPTEC; Impact of Radar Data Assimilation on precipitation.
FSO using LETKF (state-space) In Liu and Kalnay [2008, see also Li et al (2010)] is introduced an ensemble-based approach to assess the impact of observations on the forecasts formulated using a state-space aspect where Ck is a positive semidefinite suitable weighting matrix, and the forecast error is calculated for a forecast started at time tl < tk. Considering this measure the ensemble-based forecast error reduction can be written as When m=5 represents the case with verification at 24 -h forecast (assuming 6 -h assimilation cycle) Figure: Schematic representation of time line and the relevant forecast error definitions. J. Liu, E. Kalnay: Estimating observation impact without adjoint model in an ensemble Kalman filter. 2008. In QJRMS, 134, 1327 -1335. H. Li, J. Liu, E. Kalnay: Correction of ‘Estimating observation impact without adjoint model in an ensemble Kalman filter’. 2010. In QJRMS, 136, 1652 -1654.
FSO using LETKF at CPTEC/INPE (state-space) Global observation usage summary Figure: Bar plots of fractional observation impact (top left), observation count (top right), impact per observation (bottom left) and positive impact (bottom right) on 24 -h forecasts (m=5) during February 2004. Diniz, F. L. R. et al. 2016: An Observation Impact Tool for CPTEC LETKF. Weather and Forecasting (AMS). In preparation.
FSO using LETKF at CPTEC/INPE (state-space) Global observation usage summary Figure: Bar plots of fractional observation impact (top left), observation count (top right), impact per observation (bottom left) and positive impact (bottom right) on 24 -h forecasts (m=5) during February 2004. Diniz, F. L. R. et al. 2016: An Observation Impact Tool for CPTEC LETKF. Weather and Forecasting (AMS). In preparation.
Content of the presentation • Model used: – CPTEC global model T 213 L 42 (42 km) • Data assimilation systems available: – LETKF (Local Ensemble Transform Kalman Filter); • Research mode; – GSI (Gridpoint Statistical Interpolation); • Operational mode; • Studies about impact of observation at CPTEC: – – – FSO using LETKF at CPTEC/INPE (state-space); FSO using GSI at CPTEC/INPE (observation-space); Impact of Land Data Assimilation at CPTEC model on precipitation; Impact of RO-GNSS refractivity data using LETKF/CPTEC; Impact of Radar Data Assimilation on precipitation.
FSO using GSI (observation-space) In Todling (2013) is introduced an observation-space approach to assess the impact of observations on the forecasts formulated using a residual Where Ck is a positive semidefinite suitable weighting matrix, and the forecast error is calculated for a forecast started at time tl < tk. Considering this measure the observation-space forecast error reduction can be written as When m=1 represents the case with verification at analysis time and choosing the particularly convenient norm Ck=Rk-1, results Todling, R. Comparing two approaches for assessing observation impact. 2013. In Monthly Weather Review, 141, 1484 -1505.
FSO using GSI at CPTEC/INPE (observation-space) Global observation usage summary Figure: Bar plots of fractional observation impact (left), observation count (right), on all analysis for control experiment (blue) and for the experiment adding mesonet data over Brazil (red) during January 2013. Diniz, F. L. R. et al. 2016: Assessing observation impacts using CPTEC Global GSI. Meteorological Applications (RMet. S). In preparation.
FSO using GSI at CPTEC/INPE (observation-space) Global observation usage summary Figure: Bar plots of impact per observation (left) and positive impact percentage (right) on all analysis for control experiment (blue) and for the experiment adding mesonet data over Brazil (red) during January 2013. Diniz, F. L. R. et al. 2016: Assessing observation impacts using CPTEC Global GSI. Meteorological Applications (RMet. S). In preparation.
FSO using GSI at CPTEC/INPE (observation-space); Global observation usage summary There is the indication that in this region more dense surface station network associated with satellite data can be useful for reduce the radiosonde data dependence. This access method to observation impact is relatively simple and can be used in operational mode, which will be implemented at CPTEC for monitoring of the assimilated data base. Figure: Bar plots of fractional observation impact (left), observation count (right), on all analysis for control experiment (blue) and for the experiment adding mesonet data over Brazil (pink) during January 2013. Diniz, F. L. R. et al. 2016: Assessing observation impacts using CPTEC Global GSI. Meteorological Applications (RMet. S). In preparation.
Content of the presentation • Model used: – CPTEC global model T 213 L 42 (42 km) • Data assimilation systems available: – LETKF (Local Ensemble Transform Kalman Filter); • Research mode; – GSI (Gridpoint Statistical Interpolation); • Operational mode; • Studies about Impact of observation: – – – FSO using LETKF at CPTEC/INPE (state-space); FSO using GSI at CPTEC/INPE (observation-space); Impact of Land Data Assimilation at CPTEC model on precipitation; Impact of RO-GNSS refractivity data using LETKF/CPTEC; Impact of Radar Data Assimilation on precipitation.
Land Data Assimilation at CPTEC-AGCM Soil Moisture Increment Screen-Level (Open. Loop – LDAS) Diference of Latent Heat flux (Qle) Data assimilation method: OI Period of study: 1998 to 2014. de Mattos, J G Z et al. 2016: A screen-level Data Assimilation at CPTEC AGCM, Journal of Hydrometeorology, in submission
Impact land data assimilation on simulated precipitation Pearson correlation without data assimilation = 0. 44 with data assimilation = 0. 80 The figure above shows the average difference of the precipitation over Bias complete period (1998 to 2014) removed → (Open. Loop - CMAP) fix annual cycle Global Medium Mattos, J G Z et al. 2016: A screen-level Data Assimilation at CPTEC AGCM, Journal of Hydrometeorology, in submission
Content of the presentation • Model used: – CPTEC global model T 213 L 42 (42 km) • Data assimilation systems available: – LETKF (Local Ensemble Transform Kalman Filter); • Research mode; – GSI (Gridpoint Statistical Interpolation); • Operational mode; • Studies about Impact of observation: – – – FSO using LETKF at CPTEC/INPE (state-space); FSO using GSI at CPTEC/INPE (observation-space); Impact of Land Data Assimilation at CPTEC model on precipitation; Impact of RO-GNSS refractivity data using LETKF/CPTEC; Impact of Radar Data Assimilation on precipitation.
Impact of RO-GNSS refractivity data at LETKF CPTEC using ROPP A specific operator of the RO-GNSS refractivity data into the Local Ensemble Transform Kalman Filter (LETKF) system is being developed applying the Radio Occultation Processing Package (ROPP) from GRAS-SAF/Eumetsat (GNSS Receiver for Atmospheric Sounding- Satellite Applications Facilities/European Organization for the Exploration of Meteorological Satellites) Schematic figure showing the LETKF assimilation cycle, in which involve CPTEC global model, observations operator and data pre-processing. Flow chart of the operator of the RO-GNSS observations coupled in the LETKF system calling the ROPP modules in parallel processing.
Experiment setup • Experiment to evaluate the impact: • Running with and without refractivity data; • Period September/2011.
Spatial field of gain percentage (green) and losses (red) of the RMSE with data assimilation of GNSS-RO. • Anomaly correlation in geopotential hight at 500 Hpa Results
Impact over South American • Score card for RMSE gain percentage Sapucci, L. F. et al. (2016): Inclusion of GNSS-RO data into CPTEC LETKF using the ROPP as an observation operator. Met. Apps, 23: 328– 338. doi: 10. 1002/met. 1559
Outline • Model used: – CPTEC global model T 213 L 42 (42 km) • Data assimilation systems available: – LETKF (Local Ensemble Transform Kalman Filter); • Research mode; – GSI (Gridpoint Statistical Interpolation); • Operational mode; • Studies about Impact of observation: – – – FSO using LETKF at CPTEC/INPE (state-space); FSO using GSI at CPTEC/INPE (observation-space); Impact of Land Data Assimilation at CPTEC model on precipitation; Impact of RO-GNSS refractivity data using LETKF/CPTEC; Impact of Radar Data Assimilation on precipitation.
Radar Data Assimilation l. S-band Doppler Radar l. Observations: l. Refletictivity l. Radial Velocity l. Radar Data Assimilation using the WRF Data Assimilation System (WRFDA). l. Radar data assimilation will cycle hourly and restart each 6 hour using initial condition from domain d 02 (from regional cycle) d 03 Pico do Couto Radar (Rio de Janeiro) Sao Roque Radar (Sao Paulo)
Impact of radar data assimilation • Radar Data = huge amount of data. • 3 D-Var + Radar Data = lack of proper balance in the final analysis. Was added a constraining to Control Noise in High-Resolution and obtain Analysis with Multi-scale Balance Vendrasco, E. P. et al. (2016): Constraining a 3 DVAR Radar Data Assimilation System to Improve Short-Range Precipitation Forecasts. JAMC. DOI: http: //dx. doi. org/10. 1175/JAMC-D-15 -0010. 1
Center for Weather Forecast and Climate Studies Brazilian National Institute for Space Researches Thank you for your attention. Assessing observation impacts on the INPE/CPTEC global data assimilation system over South America
- Slides: 24