IMPACT OF IASI DATA ON FORECASTING POLAR LOWS

Transcription

1 IMPACT OF IASI DATA ON FORECASTING POLAR LOWS Roger Randriamampianina rwegian Meteorological Institute, Pb. 43 Blindern, N-0313 Oslo, rway Abstract The rwegian THORPEX-IPY aims to significantly improve forecasts of high-impact weather events in the Arctic. This paper reports results of our study on using the IASI radiances to improve the forecasts of the polar lows over the rwegian Sea. Forty one active IASI channels, mainly peaking the stratosphere and the upper and middle troposphere, were used. The IASI data were assimilated together with the ATOVS/AMSU-A/AMSU-B/MHS and most of the conventional observations. For a better correction of the radiance biases, a new strategy for updating the coefficients was worked out. About one-month assimilation trail of the IASI radiances was performed, and showed a positive impact on analyses and forecasts of temperature, geopotential and humidity. An overall neutral impact on wind fields were also observed. Case studies showed a clear improvement on the analysis and forecast of polar lows when assimilating the IASI data in our analysis and forecast system. INTRODUCTION In the frame of the IPY-THORPEX/rway, aiming to improve the accuracy of the high-impact weather forecasts over the Arctic region, we decided to assimilate the Infrared Atmospheric Sounder Radiometer (IASI) data. After a successful implementation of the ALADIN/HARMONIE system on the rwegian domains of interest (Fig. 1), our next step was to look at the observations processing and the choice of the background error statistics to be used in the assimilation system. This paper reports our results on the assimilation of the IASI radiances in the HARMONIE assimilation system, with special focus on their impact on the analysis and forecast of the polar lows. Figure 1. The ALADIN-HARMONIE/rway domain, with 11 km horizontal resolution. The assimilation system consisted of i) updating the Sea Surface Temperature (SST) by using the ECMWF global SST analysis, ii) performing a surface Optimal Interpolation for updating soil moisture and skin temperature fields through a univariate analysis of 2 meters temperature and relative humidity using the synoptic stations network (SYNOP); iii) performing a spectral upper-air three dimensional variational data assimilation, which takes advantage of the SYNOP stations from ships and land for the surface pressure and for the 10 meters wind over sea only, the radiosonde network for the multi-layer observations of wind, temperature, humidity and geopotential, the wind profilers for the multi-layer observations of wind, the air-borne observations of temperature and wind, the surface pressure measurements from the oceanographic buoys, the wind vectors deducted from cloud-drift

2 satellite images (Atmospheric Motion Vectors, AMV), the microwave radiances from AMSU-A, AMSUB/MHS and IASI from the polar-orbiting satellites of NOAA and from MetOp (table 1). More details about the configuration of the HARMONIE/rway assimilation system can be found in Randriamampianina and Storto (2008). Type Parameter (Channel) Bias correction Thinning TEMP SYNOP PILOT (Europrof.) DRIBU AIREP AMV AMSU-A AMSU-B, MHS IASI GPS MSG/SEVIRI U, V, T, Q, Z Z U, V Z U, V, T U, V 5 to 13 3, 4, 5 41 channels available but not used in this experiment Only T using ECMWF tables -Use of quality flags Variational Variational Variational Static Variational Temporal and spatial Redundancy check against TEMP Temporal and spatial 25 km horizontal 25 km horizontal 80 km horizontal 80 km horizontal 80/120 km horizontal 60 km horizontal Table 1. Use of Observations in the ALADIN-HARMONIE/rway NEW STRATEGY TO UPDATE THE VARIATIONAL BIAS CORRECTION COEFFICIENTS The default strategy, designed originally for the global models for upgrading the variational bias correction (VarBC) coefficients, is so that at the given assimilation time, the coefficients estimated during the previous assimilation time is used, which will be used further during the next assimilation process. So in case of a global model, the coefficients are the averaged values taking into account statistical from both the winter and the summer hemispheres over the whole globe. For regional models, depending on the size of their domain, we have very specific situation: at a given assimilation time we deal with specific geographical region with specific season. Applying the default way of updating the VarBC coefficients, we observed signature of the diurnal temperature variation in the temporal bias, especially for the lower troposphere peaking channels (AMSU-A ch. 5, 6, and 7, for example). This resulted in a very slow convergence of the biases to their nominal values. So, we need longer time to estimate the bias coefficients. A separate daily update of the coefficients for each assimilation time permits a relatively fast computation of the nominal bias for the given assimilation time. The example described below shows that a period of about twenty (20) days was enough for the system to have convergence of the bias around its nominal value for certain IASI channels. Figure 2. Figures illustrating the functionality of the variational bias correction for the case of active (left) and passive (right) channels. Solid lines show the time series of the observation-minus-first-guess (omf) and bold lines stands for the observation-minus-analysis (oma). THE USE OF THE IASI DATA The IASI channels were used as described in the table 2, and their vertical distribution is shown in Fig. 3.

3 Over Sea 49, 51, 66, 70, 83, 109, 122, 125, 128, 131, 133, 135, 141, 144, 148, 151, 154, 159, 161, 165, 167, 180, 185, 189, 193, 201, 203, 207, 214, 217, 219, 222, 224, 226, 228, 230, 232, 236, 299, 301, 303 Over Land 70, 133, 154, 180, 214, 217, 219, 301, 303 Over Ice ne Table 2. Use of the chosen IASI channels. Figure 3. Weighting functions of the chosen 41 channels, showing the vertical distribution of channels. IMPACT OF THE IASI DATA OVERALL RESULTS FROM THE WINTER TRIALS During the winter 2008, a campaign measurements was organized from February 25 till March 17, Apart of additional surface measurements, airborne in-situ and remote-sensing observations of atmospheric wind, temperature and water vapor over the rwegian and Barents Seas was done from the DLR (Deutsche Zentrum für Luft- und Raumfahrt) Falcon 20 aircraft. The measurements were operated with special focus on Arctic fronts, polar lows and terrain-induced flow disturbances. As described in the Introduction, the IASI data were assimilated together with ATOVS microwave data and most of the conventional observations. We started our winter runs from February 22 to have a warming (spinup) period of four days before the start of the campaign measurements. Then, the experiments were stopped on March 17, the last day for the campaign. Skipping the spinup period, we have got more than 20 days for verification purposes. Analyses an forecasts were compared against observations (mainly surface and radiosonde observations) as well as against the ECMWF analyses. Figure 4. shows an example of such comparisons, where we can see the impact of the IASI data on the gopotential for the case when the additional campaign measurements were not taken into account. The impact of the IASI data on temperature and humidity was found significant in the lower troposphere. We found an overall neutral impact on wind fields of the IASI data. On Fig. 5, we can see the horizontal distribution of the positive impact over the HARMONIE domain. One can see that the positive impact is spread all over the domain with a bit better performance in the northern part. TESTING DIFFERENT THINNING DISTANCES WHEN ASSIMILATING THE IASI DATA First, the IASI data were assimilated with a thinning distance of about 120 km. Then, a thinning distance of 80 km, which is the screening distance for the ATOVS AMSU-A and AMSU-B/MHS in the HARMONIE, was tested. We observed a decrease of the error, so improvement on impact, when assimilating the IASI data in higher resolution (see Fig. 6), especially in lower troposphere.

4 Figure 4. Vertical cross-section of the impact of the IASI data in shown on the top graphs. Comparison against ECMWF analyses is shown on the left and against observations on the right, showing root mean square error (RMSE) difference between runs without minus that of with the IASI data. Red colour shows positive impact. The significance test of the impact at 500 hpa model pressure is shown (lower graphs). On the bottom graphs, RMSE of run with minus that of without IASI is shown, so negative values indicate positive impact. Figure 5. Impact of the IASI data on the 24-hour forecast of geopotential. Significance test showed that the impact was significant. One can see that the positive impact is observed over Sea like over Land. As expected the impact is more accentuated over the rthern part of the HARMONIE domain, where we have less conventional observations.

5 Figure 6. Difference of root mean square errors between runs with IASI data assimilated in 80 and 120km horizontal thinning distances. One can see that increasing the resolution of the assimilated data (lower thinning distance) reduces the errors especially in the middle and in the lower troposphere. CASE STUDIES IMPACT OF THE IASI DATA ON THE ANALYSES OF POLAR LOWS In this case study, we show the impact of the IASI data on the analysis of one mature polar low, situated over the rwegian Sea, situated north-west from Lofoten (see Fig 7.), at 00 UTC, the 4th March, Although the satellite picture was sensed at 01:25 UTC, March 4th, super-positioning the analysis of mean sea level pressure on top of it, gives an idea on the position and intensity of the polar low (Fig. 7.). On these pictures, one can see that the best analysis was described on the chart when assimilating the IASI data together with the additional dropsondes from the campaign measurements. Doing a vertical cross-section of different parameters form the model output, we can highlight the impact of the IASI on the analysis of the polar low. An example of the cross-section of the wind vcomponent is shown on Fig. 8, where we can see different intensities. For example, for the case with campaign measurements, IASI data is reducing the wind intensity in the frontal part of the low, comparing to the analysis without them. IMPACT OF THE IASI DATA ON FORECASTING THE POLAR LOWS To show the impact of the assimilated IASI data on forecast of polar low, we have chosen one case of very fast developing low. The low appeared almost at the and of the campaign period. This case was observed by dropsondes operated on 16th of March from 11:00 to 14:30 UTC and 17th of March from 05:00 to 08:30 UTC. The life of this polar low was roughly one and half - two days. Using the IASI data in both combinations (with or without additional campaign observations mainly dropsondes) the forecasts were less deep than in the cases without them. The best forecast was reach when the additional dropsondes were assimilated together with the IASI soundings. On Fig. 9, we show the position and the intensity of the developing low. We can see the apparition/formation of the low 16th of March at 00 UTC, then it reached its mature stage between somewhere between UTC, and start its dissipation after roughly one day at 12 UTC the 17th of March. This happened when reaching the land cost. On Fig. 10, one can see 12-hour forecasts issued from different scenarios (with and without campaign data, and with and without IASI soundings) superposed by analyses issued from run with campaign and IASI data, for the position at mature stage March 16 at 12 UTC. The best forecast is shown on left top graph on the result of run with both the campaign and IASI data. Comparing the top graphs, we can see that assimilating the IASI data in the HARMONIE system, we reduce the overestimation of the intensity of the low. Similar picture can be observed comparing the bottom graphs (runs without campaign data).

6 Figure 7. HARMONIE analyses for 4th of March 2008, at 00 UTC. Analyses from runs with (left) and without (right) the IPY-THORPEX/rway campaign observations, superposed with satellite picture sensed a bit later, at 01:25 UTC. One can see that analysis with campaign measurements and IASI (left, light green) comprises closed isobar, while the analysis without IASI (red) does not. So, better analysis was found in the run with IASI data. Figure 8. Satellite picture showing the polar low and the related front system (left graph), and a vertical cross-section through the low (direction of the cross-section shown by yellow line on the left graph) of the v-component of the wind vectors (right graph) for various combinations of assimilating campaign and IASI data.

7 Figure 9. Analyses at different times of the run with IASI and campaign observations. One can see how fast the low was developing. The arrows sow the positions of the polar low. Figure 10. Different scenarios of forecasts (red lines) valid for 16 March 2008 a 12 UTC, superposed with the analysis (blue lines).

8 CONCLUSIONS An ensemble of 41 active channels, peaking mainly the stratosphere, the upper and middle troposphere, was tested in trials of about one-month period. Most of the IASI channels were assimilated only over sea, while few of them were used over land. assimilation of the IASI data was prepared over sea ice. For a better correction of the radiances bias, a separate daily update of the coefficients was adopted. Comparing the analyses and forecasts issued from different runs with and without additional campaign measurements, we found positive impact of the IASI data. The impact on temperature (in the lower troposphere) and on geopotential (in the middle troposphere) was significantly positive. Also, significant positive impact on the humidity was observed around hpa. An overall neutral impact (comparison against analyses) on wind speed was observed, but comparison against radiosonde showed positive impact in lower troposphere. Highlighting on case studies the impact of the IASI data on the analysis and forecasts of polar lows, we found clear reduction in the overestimation of the intensity of the lows, and also better definition of their position. REFERENCES Randriamampianina, R. and A. Storto, 2008, ALADIN-HARMONIE/rway and its assimilation system - the implementation phase, HIRLAM Newsletter 54. ACKNOWLEDGEMENTS The author would like to thank the Met.no and Météo-France/GMAP staff for their help, in particular Andrea Storto and Vincent Guidard in the implementation of the HARMONIE system and pre- and processing of the IASI data in ALADIN/HARMONIE. The author acknowledge also help from Fiona Hilton for providing the Met-Office channel selection information, which helped a lot in choosing the active channels and Andrew Collard for his help in the pre-processing of the IASI channels and in the tuning of the cloud detection scheme. The present study was supported by the rwegian Research Council through its International Polar Year programme and the project IPY-THORPEX (grant number /S30)