Adaptive observing strategies for active airborne remote sensing instruments

English


 

PANDOWAE-ADOBS (P7)
 
Adaptive observing strategies for active airborne
remote-sensing instruments
 
-This project has been finished in phase 1-
 
Project Team
Dr. Martin Weissmann (PI, funded by DFG)
Florian Harnisch (PhD student, funded by DLR)
 
Aim
 
The project intends to develop, apply and evaluate adaptive observing strategies with particular emphasis on the potential of airborne lidar instruments for improvements of numerical weather prediction (NWP) models.
 
Highlights
 
·    Successful Falcon operations during the T-PARC field campaign (cooperation P1), which contributed to the first
     coordinated four-airplane mission for typhoon targeting over the Western Pacific and to the first almost continuous
     observation of a typhoon from genesis in the tropics throughout the transition into an extra-tropical system over 13 days
·    Evaluation of the benefit of dropsonde observations for typhoon track and mid-latitude forecasts in 3 global models and
     one limited area model (Weissmann et al. 2010 )
·    Systematic evaluation of typhoon targeting strategies with T-PARC dropsondes (Harnisch and Weissmann 2010)
·    Assimilation of wind lidar observations in the global models of ECMWF and NRL, comparison of the analysis and forecast
     influence in both models
·    Assimilation of water vapor lidar observations in the ECMWF model (Harnisch et al. 2010)
 
 
T-PARC aircraft observations
 
The DLR Falcon aircraft was operated in Japan for a six week period during the THORPEX Pacific Asian Regional Campaign (T-PARC) in summer 2008. The coordination of research flights with two US aircraft (NRL P3 and US Airforce WC-130) and the Taiwanese Astra Jet of the DOTSTAR project enabled the first coordinated aircraft mission to observe a typhoon in the Western Pacific. Two major typhoons (Sinlaku and Jangmi) were observed from their early stages in the tropics throughout recurvature and extra-tropical transition for 13 and 9 days, respectively. An unprecedented set of more than 1000 dropsondes was collected, supplemented by airborne radar and wind and water vapor lidar observations.
 
 
Figure 1: T-PARC Falcon flights
 
 
Information about the field campaign and all observations can be found on:
 
 
Benefit of targeted observations
 
Several groups around the globe started to perform data denial experiments to evaluate the benefit of dropsondes after the T-PARC field campaign. The exchange of preliminary results emphasized the strong dependency of the benefit on the modeling system and consensus emerged among the groups that results from different models should be compared directly. Therefore, the original plan of evaluating the benefit of T-PARC dropsondes with the ECMWF system was expanded to an intercomparison of the influence in the global models of ECMWF, the Japan Meteorological Agency (JMA) and the National Centers for Environmental Prediction (NCEP) and the Weather Research and Forecasting (WRF) limited area model used at the National Institute of Meteorological Research (NIMR) in Korea (Weissmann et. al. 2010). To increase the significance of conclusions on the benefit of targeted observations, data denial experiments for historical typhoon events were conducted in addition to the original plan (Chou et al. 2010).
All models show an improving tendency of typhoon track forecasts (Fig. 2), but the degree of improvement varied from about 20-40% in NCEP and WRF to a comparably low influence in ECMWF and JMA. This is likely related to lower track forecast errors without dropsondes in the latter two models, presumably caused by a more extensive use of satellite data and 4D-Var assimilation at ECMWF and JMA compared to 3D-Var of NCEP and WRF.
 
Figure 2: Scatter diagrams of T-PARC typhoon track forecast errors for Sinlaku and Jangmi by (a) JMA, (b) WRF, (c) ECMWF and (d) NCEP. The linear fits are shown with dashed gray lines and their slopes and y-intersects are given in the lower right of the figures. Values beneath the diagonal indicate that the errors with dropsondes are lower than the ones without.
 
Due to the set-up of the data denial experiments, the downstream influence could only be investigated in detail for the ECMWF experiments. According to the track forecast improvement, these experiments revealed a reduction of the mean forecast error in mid-latitudes over the Pacific and on the Northern Hemisphere with the assimilation of dropsondes (Fig. 3). However, this improvement only occurred when the experiment was cycled, i.e. the first-guess field was modified in addition to the analysis, which leads to an accumulation of the impact over time. Interestingly, the improvement in mid-latitudes was the indirect result of typhoon track improvements, but not due to observations during extra-tropical transition although those flights were often guided by singular vector calculations optimized for forecasts over the mid-latitude Pacific.
 
Figure 3: (a) Mean error of 500 hPa geopotential in the ECMWF experiments with and without dropsondes in the period 9 September to 1 October averaged over the Pacific 30-65° N and 155° E to 130° W; (b) same as (a), but for experiment without cycling and only initial times where additional observations were available; (c) same as (a), but verified on the northern hemisphere north of 20° latitude. Empty (filled) triangles indicate times where the mean differences are significant at a 90% (95%) confidence level.
 
Given that there was no mean improvement without cycling and the limited mid-latitude improvement during individual T-PARC cases, it was decided not to investigate the mid-latitude effect of those observations with a case study approach. Instead, a data denial experiment for Hurricane Ike in the Atlantic was performed to investigate the influence of observations on error propagation and error growth as part of a joint paper with other members of the PANDOWAE Research Area C.
 
 
Evaluation of targeting strategies
 
A sensitivity study based on data denial experiments was carried out to investigate which dropsonde observations are most beneficial for typhoon forecasts of the ECMWF model and how to optimally observe tropical cyclones in the future. For this purpose, the dropsondes were separated into three different subsets (Fig. 4) depending on their location relative to the tropical cyclone (TC): core and centre region (CeObs), vicinity (ViObs) and remote sensitive regions (ReObs).
Figure 4: Idealized sketch illustrating the separation of the dropsondes into different subsets. Dropsondes are marked by triangles (CeObs), circles (ViObs) and asterisks (ReObs). Shading indicates a typical singular vector sensitivity pattern during T‑PARC with high (dark gray) and moderate (light gray) sensitivity.
 
 The largest forecast improvements are achieved due to observations in the vicinity of the storm (Fig. 5b). In contrast, observations in remote sensitive regions (Fig. 5c) seemed to have a relatively small influence which may be related to either insufficient resolution of the sensitive area calculations or the fact that the large scale synoptic flow is already well-represented in the analysis due to other observations. Observations in the TC core and centre (Fig. 5a) led to large analysis differences, but only very small mean forecasts improvements. Even modern high-resolution global models can not fully resolve the TC centre and thus the information provided by dropsondes within the TC centre can not be used optimally.
 
 
Figure 5:  Scatter plots of track forecast errors of experiments against the control run. The x-axis shows the track errors of the NoObs control run, and the y-axis the track errors of the (a) CeObs, (b) ViObs and (c) ReObs model runs.
 
Benefit of airborne wind lidar observations
 
Similarly to the investigation of dropsonde benefit, the original plan was expanded and the wind lidar impact is investigated in two different global models, the ECMWF Integrated Forecast System (IFS) and the Navy Operational Global Atmospheric Prediction System (NOGAPS) of the Naval Research Laboratory (NRL). This was enabled by the award of a three-month DLR Forschungssemester (sabbatical) at NRL. In particular, the calculation of adjoint observation impact, which was pioneered by NRL, is seen as a very valuable tool to further investigate the benefit of wind lidar observations. 
Several experiments with both models were conducted. Their analysis is still ongoing, but will be finished by the end of the project. Preliminary results show that the wind lidar data leads to forecast improvements in both models and a relatively high observation impact compared to other observation types.
 
Assimilation of airborne water vapour lidar observations
 
TheT-PARC data set of high-resolution humidity observations of the 4-wavelength DLR DIAL system was assimilated into the ECMWF model using the operational 4‑D variational data assimilation system. Different experiments were conducted to evaluate the influence of the observation resolution. The assimilation system successfully extracted the information content of the observations and the analysis including the humidity observations was improved compared to independent dropsonde observations. It turned out that observed water vapour structures with horizontal length scales smaller than the resolution of the assimilation system (T255; ~80 km) could not be corrected in the model analysis. Systematic differences between the water vapour observations and the model in the order of 5 to 10 % were identified which are presumably connected to errors of both, the model and the observations. The independent verification with dropsondes confirmed that the analysis with DIAL observations is more accurate. In terms of forecast impact, only a small but positive influence was found.
A large influence of the DIAL observations compared to other cases was found on 19 September 2008. The objectives of the flight were to examine the ridge building east of Japan that was triggered by the outflow of Typhoon Sinlaku and the interaction of the storm with the mid-latitude jet (Fig. 6). The assimilation of these DIAL observations lead to a better representation of the low-pressure system developing northeast of Japan and the forecast error in terms of a moist total energy was reduced compared to the control run without DIAL observations (Fig. 7).
 
 
Figure 6: Flight track (in clockwise direction) and height-distance transect of DIAL observations for the 00 UTC 19 Sept 2008 assimilation time. Observations are sampled between 2257 UTC 18 Sept and 0436 UTC 19 Sept 2008. Bold grey number label different sections of the flight track.
 
Figure 7: Geographical map of moist total energy forecast error reduction for the experiment with DIAL observations compared to the control run for (a) +24 h and (b) +48 h forecasts initialised at 00 UTC 19 Sept 2008. The forecasts are verified against the control analysis. The control analysis of the 500 hPa geopotential height is shown as solid line. Negative values represent an error reduction.
 
 
 
Publications
-  Harnisch, F., and M. Weissmann, 2010: Sensitivity of typhoon forecasts to different subsets of targeted dropsonde
    observations. Mon. Wea. Rev., 138, 2664-2680.
-  Harnisch, F., M. Weissmann, C. Cardinali, M. Wirth, 2010: Assimilation of DIAL watervapour observations in the
    ECMWF global model, submitted to Q. J. R. Meteorol. Soc. December 2010
-  Weissmann, M., F. Harnisch, C.-C. Wu, P.-H. Lin, Y. Ohta, K. Yamashita, Y. Kim, E.-H. Jeon, T. Nakazawa, S.
    Aberson, 2010: The influence of dropsondes on typhoon track and mid-latitude forecasts.Mon. Wea. Rev.,
-  Aberson, S., C.-C. Wu, M. Bell, J. Halverson, C. Fogarty, J. Cione, and M. Weissmann, 2010: Aircraft observations of
    tropical cyclones, in Global Perspectives on Tropical Cyclones: From Science to Mitigation, 2nd edition, World
    Scientific Publishing Company Ltd, J.C.L. Chan and J. Kepert, editors, pp 227-240.
-  Chou, K.-H., C.-C. Wu, P.-H. Lin, S. D. Aberson, M. Weissmann, F. Harnisch, T. Nakazawa, 2010: The Impact of
    Dropwindsonde Observations on Typhoon Track Forecasts in DOTSTAR and T-PARC. Mon. Wea. Rev.,