Advanced Search
Article Contents

Ensemble Transform Sensitivity Method for Adaptive Observations


doi: 10.1007/s00376-015-5031-9

  • The Ensemble Transform (ET) method has been shown to be useful in providing guidance for adaptive observation deployment. It predicts forecast error variance reduction for each possible deployment using its corresponding transformation matrix in an ensemble subspace. In this paper, a new ET-based sensitivity (ETS) method, which calculates the gradient of forecast error variance reduction in terms of analysis error variance reduction, is proposed to specify regions for possible adaptive observations. ETS is a first order approximation of the ET; it requires just one calculation of a transformation matrix, increasing computational efficiency (60%-80% reduction in computational cost). An explicit mathematical formulation of the ETS gradient is derived and described. Both the ET and ETS methods are applied to the Hurricane Irene (2011) case and a heavy rainfall case for comparison. The numerical results imply that the sensitive areas estimated by the ETS and ET are similar. However, ETS is much more efficient, particularly when the resolution is higher and the number of ensemble members is larger.
  • 加载中
  • Aberson S. D., 2003: Targeted observations to improve operational tropical cyclone track forecast guidance. Mon. Wea. Rev., 131, 1613- 1628.10.1175//2550.1ad2f7ecd-5392-4c75-b42e-66cf73a57be37c1adda31bef5224881266b355c1b7eahttp%3A%2F%2Fwww.researchgate.net%2Fpublication%2F249621307_Targeted_Observations_to_Improve_Operational_Tropical_Cyclone_Track_Forecast_Guidancerefpaperuri:(dffa4eededa6f5f6a7b7e101a35070c1)http://www.researchgate.net/publication/249621307_Targeted_Observations_to_Improve_Operational_Tropical_Cyclone_Track_Forecast_GuidanceSince 1997, the Tropical Prediction Center and the Hurricane Research Division have conducted operational synoptic surveillance missions with a Gulfstream IV-SP jet aircraft to improve numerical forecast guidance. Due to limited aircraft resources, optimal observing strategies for these missions must be developed. In the current study, the most rapidly growing modes are represented by areas of large forecast spread in the NCEP bred-vector ensemble forecasting system. The sampling strategy requires sampling of the entire target region with regularly spaced dropwindsonde observations. Three dynamical models were employed in testing the targeting and sampling strategies. With the assimilation into the numerical guidance of all the observations gathered during the surveillance missions, only the 12-h Geophysical Fluid Dynamics Laboratory Hurricane Model forecast showed statistically significant improvement. Assimilation of only the subset of data from the subjectively found fully sampled target regions produced a statistically significant reduction of the track forecast errors of up to 25% within the critical first 2 days of the forecast. This is comparable with the cumulative business-as-usual improvement expected over 18 yr.
    Aberson S. D., S. J. Majumdar, C. A. Reynolds, and B. J. Etherton, 2011: An observing system experiment for tropical cyclone targeting techniques using the global forecast system. Mon. Wea. Rev., 139, 895- 907.10.1175/2010MWR3397.1b6ad9d96-43ec-400a-b37d-babb280015d991fc63b52e16fe25f8c42493e5b2162fhttp%3A%2F%2Fwww.researchgate.net%2Fpublication%2F253363084_An_Observing_System_Experiment_for_Tropical_Cyclone_Targeting_Techniques_Using_the_Global_Forecast_System%3Fev%3Dauth_pubrefpaperuri:(86bf87434622384e14b80027c7474e14)http://www.researchgate.net/publication/253363084_An_Observing_System_Experiment_for_Tropical_Cyclone_Targeting_Techniques_Using_the_Global_Forecast_System?ev=auth_pubIn 1997, the National Oceanic and Atmospheric Administration''s National Hurricane Center and the Hurricane Research Division began operational synoptic surveillance missions with the Gulfstream IV-SP jet aircraft to improve the numerical guidance for hurricanes that threaten the continental United States, Puerto Rico, the U.S. Virgin Islands, and Hawaii. The dropwindsonde observations from these missions were processed and formatted aboard the aircraft and sent to the National Centers for Environmental Prediction and the Global Telecommunications System to be ingested into the Global Forecasting System, which serves as initial and boundary conditions for regional numerical models that also forecast tropical cyclone track and intensity. As a result of limited aircraft resources, optimal observing strategies for these missions are investigated. An Observing System Experiment in which different configurations of the dropwindsonde data based on three targeting techniques (ensemble variance, ensemble transform Kalman filter, and total energy singular vectors) are assimilated into the model system was conducted. All three techniques show some promise in obtaining maximal forecast improvements while limiting flight time and expendables. The data taken within and around the regions specified by the total energy singular vectors provide the largest forecast improvements, though the sample size is too small to make any operational recommendations. Case studies show that the impact of dropwindsonde data obtained either outside of fully sampled, or within nonfully sampled target regions is generally, though not always, small; this suggests that the techniques are able to discern in which regions extra observations will impact the particular forecast.
    Ancell B., G. J. Hakim, 2007: Comparing adjoint- and ensemble-sensitivity analysis with applications to observation targeting. Mon. Wea. Rev., 135, 4117- 4134.10.1175/2007MWR1904.1ca259e2e-115d-40f9-af6b-99737ecb9d827255f9a3835d66e4ae0c776fce5d94f5http://www.researchgate.net/publication/249621559_Comparing_Adjoint_and_Ensemble-Sensitivity_Analysis_with_Applications_to_Observation_Targetinghttp://www.researchgate.net/publication/249621559_Comparing_Adjoint_and_Ensemble-Sensitivity_Analysis_with_Applications_to_Observation_TargetingAbstract The sensitivity of numerical weather forecasts to small changes in initial conditions is estimated using ensemble samples of analysis and forecast errors. Ensemble sensitivity is defined here by linear regression of analysis errors onto a given forecast metric. It is shown that ensemble sensitivity is proportional to the projection of the analysis-error covariance onto the adjoint-sensitivity field. Furthermore, the ensemble-sensitivity approach proposed here involves a small calculation that is easy to implement. Ensemble- and adjoint-based sensitivity fields are compared for a representative wintertime flow pattern near the west coast of North America for a 90-member ensemble of independent initial conditions derived from an ensemble Kalman filter. The forecast metric is taken for simplicity to be the 24-h forecast of sea level pressure at a single point in western Washington State. Results show that adjoint and ensemble sensitivities are very different in terms of location, scale, and magnitude. Adjoint-sensitivity fields reveal mesoscale lower-tropospheric structures that tilt strongly upshear, whereas ensemble-sensitivity fields emphasize synoptic-scale features that tilt modestly throughout the troposphere and are associated with significant weather features at the initial time. Optimal locations for targeting can easily be determined from ensemble sensitivity, and results indicate that the primary targeting locations are located away from regions of greatest adjoint and ensemble sensitivity. It is shown that this method of targeting is similar to previous ensemble-based methods that estimate forecast-error variance reduction, but easily allows for the application of statistical confidence measures to deal with sampling error.
    Anderson J. L., 1997: The impact of dynamical constraints on the selection of initial conditions for ensemble predictions: Low order perfect model results. Mon. Wea. Rev., 125, 2969- 2983.10.1175/1520-0493(1997)1252.0.CO;2035a0e79-e31a-4183-a9fc-d0ca291b25fcc05e5f8ae8355b83e3b71a52479bfd0dhttp://www.researchgate.net/publication/241565040_The_Impact_of_Dynamical_Constraints_on_the_Selection_of_Initial_Conditions_for_Ensemble_Predictions_Low-Order_Perfect_Model_Resultshttp://www.researchgate.net/publication/241565040_The_Impact_of_Dynamical_Constraints_on_the_Selection_of_Initial_Conditions_for_Ensemble_Predictions_Low-Order_Perfect_Model_ResultsA number of operational atmospheric prediction centers now produce ensemble forecasts of the atmosphere. Because of the high-dimensional phase spaces associated with operational forecast models, many centers use constraints derived from the dynamics of the forecast model to define a greatly reduced subspace from which ensemble initial conditions are chosen. For instance, the European Centre for Medium-Range Weather Forecasts uses singular vectors of the forecast model and the National Centers for Environmental Prediction use the "breeding cycle" to determine a limited set of directions in phase space that are sampled by the ensemble forecast. The use of dynamical constraints on the selection of initial conditions for ensemble forecasts is examined in a perfect model study using a pair of three-variable dynamical systems and a prescribed observational error distribution. For these systems, one can establish that the direct use of dynamical constraints has no impact on the error of the ensemble mean forecast and a negative impact on forecasts of higher-moment quantities such as forecast spread. Simple examples are presented to show that this is not a result of the use of low-order dynamical systems but is instead related to the fundamental nature of the dynamics of these particular low-order systems themselves. Unless operational prediction models have fundamentally different dynamics, this study suggests that the use of dynamically constrained ensembles may not be justified. Further studies with more realistic prediction models are needed to evaluate this possibility.
    Avila L. A., S. Stewart, 2012: Atlantic hurricanes 2011: All about Irene and Lee. Weatherwise, 65, 34- 41.10.1080/00431672.2012.670078ab2f2bc3-fdf6-4375-baf2-2fc6b19d867598767489f3b58bc6c61f1510218f37f6http%3A%2F%2Fwww.tandfonline.com%2Fdoi%2Fabs%2F10.1080%2F00431672.2012.670078http://www.tandfonline.com/doi/abs/10.1080/00431672.2012.670078STACY STEWART senior hurricane specialists at the National Hurricane Center in Miami, Florida. The cyclone summaries are based on Tropical Cyclone Reports prepared by the authors and Jack Beven, Robbie Berg, Eric Blake, Michael Brennan, Daniel Brown, John Cangialosi, Todd Kimberlain, and Richard Pasch.
    Bauer P., R. Buizza, C. Cardinali, and J.-N. Th\'epaut, 2011: Impact of singular vector based satellite data thinning on NWP. Quart. J. Roy. Meteor. Soc., 137, 286- 302.10.1002/qj.733b07a562c-73bb-42ed-84c8-dfb042e611857288668dafd8839513d302282ca1f522http://onlinelibrary.wiley.com/doi/10.1002/qj.733/abstracthttp://onlinelibrary.wiley.com/doi/10.1002/qj.733/abstractAbstract Singular-vector(SV)-based selective satellite data thinning is applied to the Southern Hemisphere (SH) extratropics to reduce analysis uncertainty and forecast error. For two seasons, the European Centre for Medium-Range Weather Forecasts (ECMWF) four-dimensional variational data assimilation system has been run in five different configurations with different satellite data coverage: two reference experiments used low-density and high-density coverage over the globe; in the SH two SV-based selective thinning experiments used low-density data everywhere apart from targeted regions; and one random-based thinning experiment used low-density data everywhere apart from randomly defined regions. The SV-based target regions have been defined either by daily operational SVs computed for the ECMWF Ensemble Prediction System, or by the previous year's mean seasonal distribution. Results indicate that the impact of the additional data largely depends on the season. Overall, forecast errors grow faster in the SH cold season than in the warm season. In the SH cold season, the general impact of adding data is smaller and the relative difference between the impact of the individual targeting experiments is smaller as well. In the cold season, the data assimilation system failed to extract the meteorological signal carried by the extra satellite data in sensitive regions. In the SH warm season, all experiments with more data produce a statistically more significant and longer-lasting positive impact on forecast skill. In this season, the SV-based targeting experiment performs best and as well as the reference experiment in which the data density is increased globally. Copyright 2011 Royal Meteorological Society
    Berger H., R. Langland , C. S. Velden, C. A. Reynolds, and P. M. Pauley, 2011: Impact of enhanced satellite-derived atmospheric motion vector observations on numerical tropical cyclone track forecasts in the western North Pacific during TPARC/TCS-08. J. Appl. Meteor. Climatol., 50, 2309- 2318.10.1175/JAMC-D-11-019.15be4c2f6-87a2-435c-9674-1ac5fa64b003a8b0030d930a866be1ade5d8eabef8e9http%3A%2F%2Fwww.researchgate.net%2Fpublication%2F258495652_Impact_of_Enhanced_Satellite-Derived_Atmospheric_Motion_Vector_Observations_on_Numerical_Tropical_Cyclone_Track_Forecasts_in_the_Western_North_Pacific_during_TPARCTCS-08refpaperuri:(70fcad3b5ba2179bf75403e73d0fbcce)http://www.researchgate.net/publication/258495652_Impact_of_Enhanced_Satellite-Derived_Atmospheric_Motion_Vector_Observations_on_Numerical_Tropical_Cyclone_Track_Forecasts_in_the_Western_North_Pacific_during_TPARCTCS-08ABSTRACT Enhanced atmospheric motion vectors (AMVs) produced from the geostationary Multifunctional Transport Satellite (MTSAT) are assimilated into the U.S. Navy Operational Global Atmospheric Prediction System (NOGAPS) to evaluate the impact of these observations on tropical cyclone track forecasts during the simultaneous western North Pacific Ocean Observing System Research and Predictability Experiment (THORPEX) Pacific Asian Regional Campaign (TPARC) and the Tropical Cyclone Structure-2008 (TCS-08) field experiments. Four-dimensional data assimilation is employed to take advantage of experimental high-resolution (space and time) AMVs produced for the field campaigns by the Cooperative Institute for Meteorological Satellite Studies. Two enhanced AMV datasets are considered: 1) extended periods produced at hourly intervals over a large western North Pacific domain using routinely available MTSAT imagery and 2) limited periods over a smaller storm-centered domain produced using special MTSAT rapid-scan imagery. Most of the locally impacted forecast cases involve Typhoons Sinlaku and Hagupit, although other storms are also examined. On average, the continuous assimilation of the hourly AMVs reduces the NOGAPS tropical cyclone track forecast errors in particular, for forecasts longer than 72 h. It is shown that the AMVs can improve the environmental flow analyses that may be influencing the tropical cyclone tracks. Adding rapid-scan AMY observations further reduces the NOGAPS forecast errors. In addition to their benefit in traditional data assimilation, the enhanced AMVs show promise as a potential resource for advanced objective data-targeting methods.
    Bishop C. H., Z. Toth, 1999: Ensemble transformation and adaptive observations. J. Atmos. Sci., 56, 1748- 1765.10.1175/1520-0469(1999)056<1748:ETAAO>2.0.CO;221bd8cc0-1848-4fcb-94e0-884754d7a876766d4dfae61ff755e9f7e006911a2b70http%3A%2F%2Fwww.researchgate.net%2Fpublication%2F259226957_Ensemble_Transformation_and_Adaptive_Observationsrefpaperuri:(c2396610ba8d4f06e9f7cfc45dfe07f0)http://www.researchgate.net/publication/259226957_Ensemble_Transformation_and_Adaptive_ObservationsDeals with a study which described the theoretical basis of ensemble transform technique and illustrate it with an example from the Fronts and Atlantic Storm Tracks Experiment (FASTEX). Details on the transformation of matrices C...; Production of rms prediction error maps during FASTEX; Conclusions.
    Bishop C. H., B. J. Etherton, and S. J. Majumdar, 2001: Adaptive sampling with the ensemble transform Kalman filter. Part I: Theoretical aspects. Mon. Wea. Rev., 129, 420- 436.10.1175/1520-0493(2001)129<0420:ASWTET>2.0.CO;294ad37df-1cea-4cdf-a53a-67db05d15b5e24610b9bb142a4812e0dd6e3b446b34chttp%3A%2F%2Fwww.researchgate.net%2Fpublication%2F228867225_Adaptive_sampling_with_the_ensemble_transform_Kalman_filter._Part_I_Theoretical_aspectsrefpaperuri:(4e9fe37b1086b9615d7af1ea2e445fc2)http://www.researchgate.net/publication/228867225_Adaptive_sampling_with_the_ensemble_transform_Kalman_filter._Part_I_Theoretical_aspectsAbstract A suboptimal Kalman filter called the ensemble transform Kalman filter (ET KF) is introduced. Like other Kalman filters, it provides a framework for assimilating observations and also for estimating the effect of observations on forecast error covariance. It differs from other ensemble Kalman filters in that it uses ensemble transformation and a normalization to rapidly obtain the prediction error covariance matrix associated with a particular deployment of observational resources. This rapidity enables it to quickly assess the ability of a large number of future feasible sequences of observational networks to reduce forecast error variance. The ET KF was used by the National Centers for Environmental Prediction in the Winter Storm Reconnaissance missions of 1999 and 2000 to determine where aircraft should deploy dropwindsondes in order to improve 24-72-h forecasts over the continental United States. The ET KF may be applied to any well-constructed set of ensemble perturbations. The ET KF technique supercedes the ensemble transform (ET) targeting technique of Bishop and Toth. In the ET targeting formulation, the means by which observations reduced forecast error variance was not expressed mathematically. The mathematical representation of this process provided by the ET KF enables such things as the evaluation of the reduction in forecast error variance associated with individual flight tracks and assessments of the value of targeted observations that are distributed over significant time intervals. It also enables a serial targeting methodology whereby one can identify optimal observing sites given the location and error statistics of other observations. This allows the network designer to nonredundantly position targeted observations. Serial targeting can also be used to greatly reduce the computations required to identify optimal target sites. For these theoretical and practical reasons, the ET KF technique is more useful than the ET technique. The methodology is illustrated with observation system simulation experiments involving a barotropic numerical model of tropical cyclonelike vortices. These include preliminary empirical tests of ET KF predictions using ET KF, 3DVAR, and hybrid data assimilation schemes he results of which look promising. To concisely describe the future feasible sequences of observations considered in adaptive sampling problems, an extension to Ide et al. unified notation for data assimilation is suggested.
    Buizza R., A. Montani, 1999: Targeted observations using singular vectors. J. Atmos. Sci., 56, 2965- 2985.10.1175/1520-0469(1999)0562.0.CO;27f69c9d6-6a06-4c5d-932b-33dcdc5b4d9fdccfe90c6e41c861395cf64f95df2bb2http://www.researchgate.net/publication/247698475_Targeting_Observations_Using_Singular_Vectorshttp://www.researchgate.net/publication/247698475_Targeting_Observations_Using_Singular_VectorsAbstract Singular vectors with maximum energy at final time inside a verification area are used to identify the target area where extra observations should be taken, at an initial time, to reduce the forecast error inside the verification area itself. This technique is applied to five cases of cyclone development in the Atlantic Ocean, with cyclones reaching the British Isles at the final time. Three verification areas centered around this region are considered. First, the sensitivity of the target area to the choice of the forecast trajectory along which the singular vectors are evolved, to the choice of the verification area where singular vector energy is maximized, and to the number of singular vectors used to define the target area is investigated. Results show little sensitivity to the choice of the verification area, but high sensitivity to the choice of the trajectory. Regarding the number of singular vectors used, results based on the first 4 or the first 10 singular vectors are shown to be very similar. Second, the potential forecast error reduction that could be achieved by taking extra observations inside the target area is estimated by contrasting the error of a forecast started from the unperturbed analysis with the error of a forecast started by subtracting so-called pseudo-inverse perturbations (estimated using the leading singular vectors) to the unperturbed analysis. Results indicate that root-mean-square errors in the verification region could be reduced by up to 13% by adding targeted observations. Overall, results suggest that linear models can be used to define the target area where adaptive observations should be taken.
    Chang E. K. M., M. H. Zheng, and K. Raeder, 2013: Medium-range ensemble sensitivity analysis of two extreme pacific extratropical cyclones. Mon. Wea. Rev., 141, 211- 231.10.1175/MWR-D-11-00304.116791b5b-1476-4a1c-960e-aa422f1d6415e38e63459b28d2604bbde76278bfe457http%3A%2F%2Fwww.researchgate.net%2Fpublication%2F258805685_Medium-Range_Ensemble_Sensitivity_Analysis_of_Two_Extreme_Pacific_Extratropical_Cyclonesrefpaperuri:(b4f7c9ee3ea455b4de6a95b000a607c5)http://www.researchgate.net/publication/258805685_Medium-Range_Ensemble_Sensitivity_Analysis_of_Two_Extreme_Pacific_Extratropical_CyclonesAbstract In this study, ensemble sensitivity analysis has been applied to examine the evolution of two extreme extratropical cyclones over the Pacific. Sensitivity using, as forecast metrics, forecast cyclone minimum pressure and location, as well as principal components (PCs) of the leading EOFs in forecast SLP variations near the cyclone center, has been computed for medium-range forecasts of up to 7.5 days. Results presented here show that coherent sensitivity patterns can be tracked from the forecast validation time back in time to at least day 6, with the sensitivity signal exhibiting downstream development characteristics in most cases. Comparing the different forecast metrics, sensitivity patterns derived from the PCs of the leading EOFs in forecast SLP variations are apparently more coherent than those derived from cyclone parameters. To test whether the linear sensitivity analyses provide quantitatively accurate guidance under the highly nonlinear evolution of the atmospheric flow, perturbed initial condition experiments have been conducted using initial condition perturbations derived based on ensemble sensitivity. Results of this study suggest that in the medium range, perturbations derived from cyclone parameters are quite effective in modifying the evolution of the cyclones out to 5.5 days, but are largely ineffective for 7.5-day forecasts. On the other hand, perturbations derived based on the PCs of the leading EOFs are still quite effective in modifying forecast cyclone location out to 7.5 days. These results suggest that EOF-based sensitivities perform better than cyclone parameter-based sensitivities in the medium range.
    Chou K.-H., C.-C. Wu, P.-H. Lin, S. D. Aberson, M. Weissmann, F. Harnisch, and T. Nakazawa, 2011: The impact of dropwindsonde observations on typhoon track forecasts in DOTSTAR and T-PARC. Mon. Wea. Rev., 139, 1728- 1743.10.1175/2010MWR3582.17a7b821b-089a-40b2-9880-de84e43764e7c1faee216c64280eb55cf5719d33a188http%3A%2F%2Fwww.researchgate.net%2Fpublication%2F225021009_The_Impact_of_Dropwindsonde_Observations_on_Typhoon_Track_Forecasts_in_DOTSTAR_and_T-PARCrefpaperuri:(4d76fb01f7e7ed16109e75624792cc5b)http://www.researchgate.net/publication/225021009_The_Impact_of_Dropwindsonde_Observations_on_Typhoon_Track_Forecasts_in_DOTSTAR_and_T-PARCThe typhoon surveillance program Dropwindsonde Observations for Typhoon Surveillance near the Taiwan Region (DOTSTAR) has been conducted since 2003 to obtain dropwindsonde observations around tropical cyclones near Taiwan. In addition, an international field project The Observing System Research and Predictability Experiment (THORPEX) Pacific Asian Regional Campaign (T-PARC) in which dropwindsonde observations were obtained by both surveillance and reconnaissance flights was conducted in summer 2008 in the same region. In this study, the impact of the dropwindsonde data on track forecasts is investigated for DOTSTAR (2003--09) and T-PARC (2008) experiments. Two operational global models from NCEP and ECMWF are used to evaluate the impact of dropwindsonde data. In addition, the impact on the two-model mean is assessed. The impact of dropwindsonde data on track forecasts is different in the NCEP and ECMWF model systems. Using the NCEP system, the assimilation of dropwindsonde data leads to improvements in 1- to 5-day track forecasts in about 60%% of the cases. The differences between track forecasts with and without the dropwindsonde data are generally larger for cases in which the data improved the forecasts than in cases in which the forecasts were degraded. Overall, the mean 1- to 5-day track forecast error is reduced by about 10%%--20%% for both DOTSTAR and T-PARC cases in the NCEP system. In the ECMWF system, the impact is not as beneficial as in the NCEP system, likely because of more extensive use of satellite data and more complex data assimilation used in the former, leading to better performance even without dropwindsonde data. The stronger impacts of the dropwindsonde data are revealed for the 3- to 5-day forecast in the two-model mean of the NCEP and ECMWF systems than for each individual model.
    Ehrendorfer M., R. M. Errico, and K. D. Raeder, 1999: Singular-vector perturbation growth in a primitive equation model with moist physics. J. Atmos. Sci., 56, 1627- 1648.10.1175/1520-0469(1999)056<1627:SVPGIA>2.0.CO;273e44d83-74ae-43be-abba-166f9135e7f0061fdc5ad53d7841cd5e42e0d697f6e7http%3A%2F%2Fwww.researchgate.net%2Fpublication%2F236268983_Singular-vector_perturbation_growth_in_a_primitive_equation_model_with_moist_physics%3Fev%3Dauth_pubrefpaperuri:(86cff832a1a789843922e77950601a43)http://www.researchgate.net/publication/236268983_Singular-vector_perturbation_growth_in_a_primitive_equation_model_with_moist_physics?ev=auth_pubABSTRACT Finite-time growth of perturbations in the presence of moist physics (specifically, precipitation) is investigated using singular vectors (SVs) in the context of a primitive equation regional model. Two difficulties appear in the explicit consideration of the effect of moist physics when studying such optimal growth. First, the tangent- linear description of moist physics may not be as straightforward and accurate as for dry-adiabatic processes; second, because of the consideration of moisture, the design of an appropriate measure of growth (i.e., norm) is subject to even more ambiguity than in the dry situation. In this study both of these problems are addressed in the context of the moist version of the National Center for Atmospheric Research Mesoscale Adjoint Modeling System, version 2, with emphasis on the second problem. Leading SVs are computed in an iterative fashion, using a Lanczos algorithm, for three norms over an optimization interval of 24 h; these norms are based on an expression related to (total) perturbation energy. The properties of these SVs are studied for a case of explosive cyclogenesis and a case of summer convection. The consideration of moisture leads to faster growth of perturbations than in the dry situation, as well as to the appearance of new growing structures. Apparently, moist processes provide for new mechanisms of error growth and do not simply lead to a modulation of SVs obtained with the dry version of the model. Consequently, consideration of the linearized moist processes is essential for revealing all structures that might potentially grow in a moist primitive equation model. In the context of this investigation growth rates depend more on the choice of the basic state and linearized model (moist vs dry) than on the choice of the norm (moist vs dry total energy norm). A reference is cited that supports the validity of the moist tangent-linear SV perturbation growth studied here in the nonlinear regime.
    Hamill T. M., J. S. Whitaker, and C. Snyder, 2001: Distance-dependent filtering of background error covariance estimates in an ensemble Kalman filter. Mon. Wea. Rev., 129, 2776- 2790.10.1175/1520-0493(2001)1292.0.CO;21522b1b3-a0e6-4dc3-98af-8d160276d4139ba88ba1c3d55b861d1cac96eb6d6d8chttp%3A%2F%2Fci.nii.ac.jp%2Fnaid%2F10013126187%2Frefpaperuri:(616cc3a07f8cd71204778c8a384b8339)http://ci.nii.ac.jp/naid/10013126187/Abstract The usefulness of a distance-dependent reduction of background error covariance estimates in an ensemble Kalman filter is demonstrated. Covariances are reduced by performing an elementwise multiplication of the background error covariance matrix with a correlation function with local support. This reduces noisiness and results in an improved background error covariance estimate, which generates a reduced-error ensemble of model initial conditions. The benefits of applying the correlation function can be understood in part from examining the characteristics of simple 2-2 covariance matrices generated from random sample vectors with known variances and covariance. These show that noisiness in covariance estimates tends to overwhelm the signal when the ensemble size is small and/or the true covariance between the sample elements is small. Since the true covariance of forecast errors is generally related to the distance between grid points, covariance estimates generally have a higher ratio of noise to signal with increasing distance between grid points. This property is also demonstrated using a two-layer hemispheric primitive equation model and comparing covariance estimates generated by small and large ensembles. Covariances from the large ensemble are assumed to be accurate and are used a reference for measuring errors from covariances estimated from a small ensemble. The benefits of including distance-dependent reduction of covariance estimates are demonstrated with an ensemble Kalman filter data assimilation scheme. The optimal correlation length scale of the filter function depends on ensemble size; larger correlation lengths are preferable for larger ensembles. The effects of inflating background error covariance estimates are examined as a way of stabilizing the filter. It was found that more inflation was necessary for smaller ensembles than for larger ensembles.
    Ito K., C.-C. Wu, 2013: Typhoon-position-oriented sensitivity analysis. Part I: Theory and verification. J. Atmos. Sci., 70, 2525- 2546.10.1175/JAS-D-12-0301.18ae3f718-1d75-46a3-8136-058efd7e59a51cc110fb5b0f3b582f983c5790c07863http%3A%2F%2Fwww.researchgate.net%2Fpublication%2F260617763_Typhoon-Position-Oriented_Sensitivity_Analysis._Part_I_Theory_and_Verificationrefpaperuri:(9b29a8f29d2649c71fcef8cf937d4a10)http://www.researchgate.net/publication/260617763_Typhoon-Position-Oriented_Sensitivity_Analysis._Part_I_Theory_and_VerificationAbstract A new sensitivity analysis method is proposed for the ensemble prediction system in which a tropical cyclone (TC) position is taken as a metric. Sensitivity is defined as a slope of linear regression (or its approximation) between state variable and a scalar representing the TC position based on ensemble simulation. The experiment results illustrate important regions for ensemble TC track forecast. The typhoon-position-oriented sensitivity analysis (TyPOS) is applied to Typhoon Shanshan (2006) for the verification time of up to 48 h. The sensitivity field of the TC central latitude with respect to the vorticity field obtained from large-scale random initial perturbation is characterized by a horizontally tilted pattern centered at the initial TC position. These sensitivity signals are generally maximized in the middle troposphere and are far more significant than those with respect to the divergence field. The results are consistent with the sensitivity signals obtained from existing methods. The verification experiments indicate that the signals from TyPOS quantitatively reflect an ensemble-mean position change as a response to the initial perturbation. Another experiment with Typhoon Dolphin (2008) demonstrates the long-term analysis of forecast sensitivity up to 96 h. Several additional tests have also been carried out to investigate the dependency among ensemble members, the impacts of using different horizontal grid spacing, and the effectiveness of ensemble-Kalman-filter-based perturbations.
    Joly, A., Coauthors, 1997: The fronts and Atlantic storm-track experiment (FASTEX): Scientific objectives and experimental design. Bull. Amer. Meteor. Soc., 78, 1917- 1940.10.1175/1520-0477(1997)0782.0.CO;2e0e04279-176c-435a-83c8-c349edd4ffb613a745313a5b3e73b2667495da80557ahttp%3A%2F%2Fwww.researchgate.net%2Fpublication%2F260730779_The_Fronts_and_Atlantic_Storm-Track_Experiment_%28FASTEX%29_Scientific_Objectives_and_Experimental_Design%3Fev%3Dprf_citrefpaperuri:(0c3384ba71043dfcb285c3c6290c184c)http://www.researchgate.net/publication/260730779_The_Fronts_and_Atlantic_Storm-Track_Experiment_(FASTEX)_Scientific_Objectives_and_Experimental_Design?ev=prf_citAbstract The Fronts and Atlantic Storm-Track Experiment (FASTEX) will address the life cycle of cyclones evolving over the North Atlantic Ocean in January and February 1997. The objectives of FASTEX are to improve the forecasts of end-of-storm-track cyclogenesis (primarily in the eastern Atlantic but with applicability to the Pacific) in the range 24 to 72 h, to enable the testing of theoretical ideas on cyclone formation and development, and to document the vertical and the mesoscale structure of cloud systems in mature cyclones and their relation to the dynamics. The observing system includes ships that will remain in the vicinity of the main baroclinic zone in the central Atlantic Ocean, jet aircraft that will fly and drop sondes off the east coast of North America or over the central Atlantic Ocean, turboprop aircraft that will survey mature cyclones off Ireland with dropsondes, and airborne Doppler radars, including ASTRAIA/ELDORA. Radiosounding frequency around the North Atlantic basin will be increased, as well as the number of drifting buoys. These facilities will be activated during multiple-day intensive observing periods in order to observe the same meteorological systems at several stages of their life cycle. A central archive will be developed in quasi-real time in Toulouse, France, thus allowing data to be made widely available to the scientific community.
    Joly, A., Coauthors, 1999: Overview of the field phase of the fronts and Atlantic Storm-Track EXperiment (FASTEX) project. Quart. J. Roy. Meteor. Soc., 125, 3131- 3163.10.1002/qj.49712556103fd7896e5-cd89-4fc3-9cad-13fe4c8e0c6be1bf48a6c586579bd949e626dd0dd7f5http://onlinelibrary.wiley.com/doi/10.1002/qj.49712556103/fullhttp://onlinelibrary.wiley.com/doi/10.1002/qj.49712556103/fullThese objectives were successfully achieved. Intensive Observation Periods were conducted on 19 occasions. High-resolution vertical profiles through the same cyclones at three different stages of their life cycle have been obtained on more than 10 occasions. the calculation of critical areas where further observations were needed to limit the growth of forecast error, was undertaken using different techniques, and flights were planned and executed in these areas in time to achieve this. Combined dropsonde and Doppler radar observations of cloud systems are available for 10 cases. A unique air-sea turbulent exchange dataset has been obtained.
    Langland, R. H., R. Gelaro, G. D. Rohaly, M. A. Shapiro, 1999a: Targeted observations in FASTEX: Adjoint-based targeting procedures and data impact experiments in IOP17 and IOP18. Quart. J. Roy. Meteor. Soc., 125, 3241- 3270.10.1002/qj.497125561070078f5f9-5b43-4bab-82a1-6c0380610b62c82b8845327663bd5380005ccf26f2f2http%3A%2F%2Fonlinelibrary.wiley.com%2Fdoi%2F10.1002%2Fqj.49712556107%2Ffullrefpaperuri:(d4f839f5368396dcd114ea4013a1be09)http://onlinelibrary.wiley.com/doi/10.1002/qj.49712556107/fullAbstract The Fronts and Atlantic Storm-Track EXperiment (FASTEX) provided an opportunity for testing targeted-observing procedures in a real-time framework during January and February 1997. This study describes the use of singular vectors (SVs) for objective targeting during FASTEX, and the evaluation of the impact obtained from targeted dropsonde data, satellite wind data, and other observations on 1-2 day forecast skill in intensive observation periods (IOPs) 17 and 18. In IOP17, targeted dropsondes improve a 42 h forecast of L41 (Low 41; cyclones were numbered in sequence throughout FASTEX) in terms of sea-level pressure, but the forecast skill is degraded in the upper troposphere. It is suggested that the degraded forecast may be caused by an incomplete survey of the SV target area, that improved the analysis in one region, but made the analysis less accurate in an adjacent part of the target area where no dropsonde data were provided. In a series of experiments, the best 42 h forecast of L41 is obtained by the addition of a few radiosonde profiles provided specially for FASTEX at off-times, that provide observational data in the most sensitive part of the SV target area. the analysis differences introduced by the radiosonde profiles are much smaller in magnitude than those from the dropsonde data, but have a larger forecast impact, because they occur in an area that has larger error growth rates in this forecast. In a series of experiments for IOP18, the best 24 h forecast of L44 is obtained using a combination of targeted-dropsonde data and satellite wind data. Both data types can also be used separately to improve this forecast. the assimilation of satellite wind data and ship-based soundings in areas of weak initial-condition sensitivity (&lsquo;null&rsquo; areas) is shown to have minimal impact on the forecast error. the target areas identified by SVs in these two IOPs occur in strongly baroclinic regions, tending to favour the right-entrance and left-exit regions of the upper-level jet, but with greatest sensitivity near 600 hPa.
    Langland, R. H., Coauthors, 1999b: The North Pacific experiment (NORPEX-98): Targeted observations for improved North American weather forecasts. Bull. Amer. Meteor. Soc., 80, 1363- 1384.10.1175/1520-0477(1999)0802.0.CO;21ef83f07-a678-4207-9aaa-02190e5e450d44bb43d95c76f1759fd54999d99c7b94http://www.researchgate.net/publication/247933839_The_North_Pacific_Experiment_(NORPEX-98)_Targeted_Observations_for_Improved_North_American_Weather_Forecastshttp://www.researchgate.net/publication/247933839_The_North_Pacific_Experiment_(NORPEX-98)_Targeted_Observations_for_Improved_North_American_Weather_ForecastsAbstract The objectives and preliminary results of an interagency field program, the North Pacific Experiment (NORPEX), which took place between 14 January and 27 February 1998, are described. NORPEX represents an effort to directly address the issue of observational sparsity over the North Pacific basin, which is a major contributing factor in short-range (less than 4 days) forecast failures for land-falling Pacific winter-season storms that affect the United States, Canada, and Mexico. The special observations collected in NORPEX include approximately 700 targeted tropospheric soundings of temperature, wind, and moisture from Global Positioning System (GPS) dropsondes obtained in 38 storm reconnaissance missions using aircraft based primarily in Hawaii and Alaska. In addition, wind data were provided every 6 h over the entire North Pacific during NORPEX, using advanced and experimental techniques to extract information from multispectral geostationary satellite imagery. Preliminary results of NORPEX data impact studies using the U.S. Navy and National Weather Service forecast models include reductions of approximately 10% in mean 2-day forecast error over western North America (30-60N, 100-130W) from assimilation of targeted dropsonde and satellite wind data (when measured against control forecasts that contain no special NORPEX observations). There are local reductions of up to 50% in 2-day forecast error for individual cases, although some forecasts are degraded by the addition of the special dropsonde or satellite wind data. In most cases, the positive impact of the targeted dropsonde data on short-range forecast skill is reduced when the full set of advanced satellite wind data is already included in the model analyses. The NORPEX dataset is being used in research to improve objective methods for targeting observations, to study the "mix" of in situ and space-based observations, and to understand the structure and dynamics of fast-growing errors that limit our ability to provide more accurate forecasts of Pacific winter storms.
    Majumdar S. J., C. H. Bishop, B. J. Etherton, I. Szunyogh, and Z. Toth, 2001: Can an ensemble transform Kalman filter predict the reduction in forecast-error variance produced by targeted observations? Quart. J. Roy. Meteor. Soc., 127, 2803- 2820.10.1002/qj.49712757815d8dd8243-c5e4-49ee-93f1-e86b878c638bd05dd01cfc4f7fe8d4887667d5781f89http%3A%2F%2Fonlinelibrary.wiley.com%2Fdoi%2F10.1002%2Fqj.49712757815%2Fpdfrefpaperuri:(e74dfc2d05079ee7fc2163a705bad830)http://onlinelibrary.wiley.com/doi/10.1002/qj.49712757815/pdfABSTRACT The ensemble transform Kalman filter (ET KF) is currently used at the National Centers for Environmental Prediction (NCEP) to identify deployments of aircraft-borne dropwindsondes that are likely to significantly improve 1-3 day forecasts of winter storms over the continental United States. It is unique among existing targeted observing strategies in that it attempts to predict the reduction in forecast-error variance associated with each deployment of targeted observations. To achieve this, the ET KF predicts the variance of &lsquo;signals&rsquo; for each feasible deployment, where a signal represents the difference between two forecasts, initialized with and without the targeted observations. For linear forecast-error evolution, the signal variance is equal to the reduction in forecast-error variance, provided that observation- and background-error covariances are accurately specified and identical to those produced by the operational data-assimilation scheme. However, background-error covariances assumed by the ET KF are both imperfect and different from the imperfect error covariances used in NCEP's 3D-Var data-assimilation scheme, and hence their signal statistics are likely to differ.In spite of these differences, a linear relationship of positive gradient is found to exist between the ET KF signal variance and the sample variance of NCEP signal realizations at both the targeted analysis and forecast verification times, for 30 forecasts from the 2000 Winter Storm Reconnaissance Program. This relationship enables the NCEP signal variance to be predicted by the ET KF, via a statistical rescaling that corrects the ET KF's current over-prediction of signal variance magnitude. A monotonically increasing relationship is also found to exist between the NCEP signal variance and the reduction in NCEP forecast-error variance. The ET KF signal variance predictions can be used to make quantitative estimates of the forecast-error-variance reducing effect of targeted observations. Potential benefits include (i) making rapid decisions on when and where to deploy targeted observations, (ii) warning operational data quality-control schemes against the rejection of observational data if the signal variance is large, and (iii) estimating the likelihood of economic benefit due to any future deployment of observations.
    Majumdar S. J., C. H. Bishop, B. J. Etherton, and Z. Toth, 2002: Adaptive sampling with the ensemble transform Kalman filter. Part II: Field program implementation. Mon. Wea. Rev., 130, 1356- 1369.10.1175/1520-0493(2002)130<1356:ASWTET>2.0.CO;23ff96b29-eadd-4596-a77b-5b037aba91a09d4bae588597458fc42fb9aac64c3c4ehttp%3A%2F%2Fwww.researchgate.net%2Fpublication%2F228980604_Adaptive_sampling_with_the_ensemble_transform_Kalman_filter._Part_II_Field_program_implementationrefpaperuri:(f0e4effd00f6c1265ef705c173cd2c7c)http://www.researchgate.net/publication/228980604_Adaptive_sampling_with_the_ensemble_transform_Kalman_filter._Part_II_Field_program_implementationThe practical application of the ensemble transform Kalman filter (ET KF), used in recent Winter Storm Reconnaissance (WSR) programs by the National Centers for Environmental Prediction (NCEP), is described. The ET KF assesses the value of targeted observations taken at future times in improving forecasts for preselected critical events. It is based on a serial assimilation framework that makes it an order of magnitude faster than its predecessor, the ensemble transform technique. The speed of the ET KF enabled several different forecast scenarios to be assessed for targeting during recent WSR programs. Each potential observational network is broken down into idealized routine and adaptive components. The adaptive component represents a predesigned flight track along which GPS dropwindsondes are released. For a large number of flight tracks, the ET KF estimates the forecast error reducing effects of these observations (via the 070705signal variance070705). The track that maximizes the average forecast signal variance within a selected verification region is deemed optimal for targeting. Secondary flight tracks can also be chosen using serial assimilation, by calculating the signal variance for each flight track given that the primary track had already been selected. For the second consecutive year the ET KF was able to estimate, via a statistical rescaling, the variance of NCEP signal realizations produced by the dropwindsonde data. A monotonic increasing relationship between the ET KF signal variance and the reduction in NCEP forecast error variance due to the targeted observations was then deduced for the operational 2001 WSR program.
    Majumdar, S. J., Coauthors, 2011: Targeted observations for improving numerical weather prediction: An overview. WWRP/THORPEX No. 15.10.1002/j.1477-8696.1966.tb02810.x545a5777982b70e624680490453ba51dhttp%3A%2F%2Fwww.researchgate.net%2Fpublication%2F267715561_Targeted_Observations_for_Improving_Numerical_Weather_Prediction_An_Overviewhttp://www.researchgate.net/publication/267715561_Targeted_Observations_for_Improving_Numerical_Weather_Prediction_An_OverviewABSTRACT argeted observations- refers to the augmentation of the regular observing network with additional, specially chosen observations to be assimilated into operational numerical weather prediction models. Observation locations are chosen in order to improve forecasts of high-impact weather events of importance to society. Examples include dropwindsondes launched from aircraft or balloons, additional rawinsonde ascents, remotely sensed observations, and the inclusion of enhanced regular satellite observations (such as radiances or winds) that may normally be excluded from data assimilation due to routine thinning or quality control procedures. As a consequence of many field campaigns worldwide during the past decade, advancements have been made in the development of objective strategies for targeting observations, and in quantitative evaluations of the impact of assimilating these extra observations on numerical weather predictions. The successes and shortcomings of these efforts are reviewed here and recommendations are made to the community for the use of targeted observations in the future to maximize the impact on forecasts.
    Mu M., F. F. Zhou, and H. L. Wang, 2009: A method for identifying the sensitive areas in targeted observations for tropical cyclone prediction: Conditional nonlinear optimal perturbation. Mon. Wea. Rev., 137, 1623- 1639.10.1175/2008MWR2640.11ff05c96-c6d3-4672-ae1c-f5f7b5a663e132149bf9c8850652f8990e7ea9cba731http://www.researchgate.net/publication/236324200_A_Method_for_Identifying_the_Sensitive_Areas_in_Targeted_Observations_for_Tropical_Cyclone_Prediction_Conditional_Nonlinear_Optimal_Perturbationhttp://www.researchgate.net/publication/236324200_A_Method_for_Identifying_the_Sensitive_Areas_in_Targeted_Observations_for_Tropical_Cyclone_Prediction_Conditional_Nonlinear_Optimal_PerturbationConditional nonlinear optimal perturbation (CNOP), which is a natural extension of the linear singular vector into the nonlinear regime, is proposed in this study for the determination of sensitive areas in adaptive observations for tropical cyclone prediction. Three tropical cyclone cases, Mindulle (2004), Meari (2004), and Matsa (2005), are investigated. Using the metrics of kinetic and dry energies, CNOPs and the first singular vectors (FSVs) are obtained over a 24-h optimization interval. Their spatial structures, their energies, and their nonlinear evolutions as well as the induced humidity changes are compared. A series of sensitivity experiments are designed to find out what benefit can be obtained by reductions of CNOP-type errors versus FSV-type errors. It is found that the structures of CNOPs may differ much from those of FSVs depending on the constraint, metric, and the basic state. The CNOP-type errors have larger impact on the forecasts in the verification area as well as the tropical cyclones than the FSV-types errors. The results of sensitivity experiments indicate that reductions of CNOP-type errors in the initial states provide more benefits than reductions of FSV-type errors. These results suggest that it is worthwhile to use CNOP as a method to identify the sensitive areas in adaptive observation for tropical cyclone prediction.
    Palmer T. N., R. Gelaro, J. Barkmeijer, and R. Buizza, 1998: Singular vectors, metrics, and adaptive observations. J. Atmos. Sci., 55, 633- 653.10.1175/1520-0469(1998)055<0633:SVMAAO>2.0.CO;22cd34d32-a504-4133-b3b8-d99d3a039ed0beba7df480bfe51079b3cba0eb60fcfbhttp%3A%2F%2Fwww.researchgate.net%2Fpublication%2F236268980_Singular_Vectors_Metrics_and_Adaptive_Observationsrefpaperuri:(56028d93b39b0d4af77e2768cd0923cb)http://www.researchgate.net/publication/236268980_Singular_Vectors_Metrics_and_Adaptive_ObservationsProvides information on the use of singular vectors of the linearized equations of motion to study the instability properties of the atmosphere-ocean systems and its related predictability. Information on singular vector formulation; Detailed information on the covariance metric; Conclusions reached.
    Szunyogh I., Z. Toth, R. E. Morss, S. J. Majumdar, and C. H. Bishop, 2000: The effect of targeted dropsonde observations during the 1999 winter storm reconnaissance program. Mon. Wea. Rev., 128, 3520- 3537.10.1175/1520-0493(2000)128<3520:TEOTDO>2.0.CO;279ab6050-0c70-4e11-8643-358e6a7a4749da12f6d6053cbc9526d865f020a0223dhttp%3A%2F%2Fwww.researchgate.net%2Fpublication%2F252656173_The_Effect_of_Targeted_Dropsonde_Observations_during_the_1999_Winter_Storm_Reconnaissance_Programrefpaperuri:(d87786ac48c4989bb0847d2cbc15eed8)http://www.researchgate.net/publication/252656173_The_Effect_of_Targeted_Dropsonde_Observations_during_the_1999_Winter_Storm_Reconnaissance_ProgramAbstract In this paper, the effects of targeted dropsonde observations on operational global numerical weather analyses and forecasts made at the National Centers for Environmental Prediction (NCEP) are evaluated. The data were collected during the 1999 Winter Storm Reconnaissance field program at locations that were found optimal by the ensemble transform technique for reducing specific forecast errors over the continental United States and Alaska. Two parallel analysis orecast cycles are compared; one assimilates all operationally available data including those from the targeted dropsondes, whereas the other is identical except that it excludes all dropsonde data collected during the program. It was found that large analysis errors appear in areas of intense baroclinic energy conversion over the northeast Pacific and are strongly associated with errors in the first-guess field. The ignal,- defined by the difference between analysisorecast cycles with and without the dropsonde data, propagates at an average speed of 30 per day along the storm track to the east. Hovm02ller diagrams and eddy statistics suggest that downstream development plays a significant role in spreading out the effect of the dropsondes in space and time. On average, the largest rms surface pressure errors are reduced by 10%-20% associated with the eastward-propagating leading edge of the signal. The dropsonde data seem to be more effective in reducing forecast errors when zonal flow prevails over the eastern Pacific. Results from combined verification statistics (based on surface pressure, tropospheric winds, and precipitation amount) indicate that the dropsonde data improved the forecasts in 18 of the 25 targeted cases, while the impact was negative (neutral) in only 5 (2) cases.
    Szunyogh I., Z. Toth, A. V. Zimin, S. J. Majumdar, and A. Persson, 2002: Propagation of the effect of targeted observations: The 2000 winter storm reconnaissance program. Mon. Wea. Rev., 130, 1144- 1165.10.1175/1520-0493(2002)1302.0.CO;207af1587-272f-40e6-a86e-801c605959532dd412d2c99f3721f236d6b05546daefhttp%3A%2F%2Fwww.researchgate.net%2Fpublication%2F249621530_Propagation_of_the_Effect_of_Targeted_Observations_The_2000_Winter_Storm_Reconnaissance_Programrefpaperuri:(654714f7afc64ded354817e927cb38bf)http://www.researchgate.net/publication/249621530_Propagation_of_the_Effect_of_Targeted_Observations_The_2000_Winter_Storm_Reconnaissance_ProgramAbstract The propagation of the effect of targeted observations in numerical weather forecasts is investigated, based on results from the 2000 Winter Storm Reconnaissance (WSR00) program. In this field program, nearly 300 dropsondes were released adaptively at selected locations over the northeast Pacific on 12 separate flight days with the aim of reducing the risk of major failures in severe winter storm forecasts over the United States. The data impact was assessed by analysis orecast experiments carried out with the T62 horizontal resolution, 28-level version of the operational global Medium Range Forecast system of the National Centers for Environmental Prediction. In some cases, storms that reached the West Coast or Alaska were observed in an earlier phase of their development, while at other times the goal was to improve the prediction of storms that formed far downstream of the targeted region. Changes in the forecasts were the largest when landfalling systems were targeted and the baroclinic energy conversion was strong in the targeted region. As expected from the experience accumulated during the 1999 Winter Storm Reconnaissance (WSR99) program, downstream baroclinic development played a major role in propagating the influence of the targeted data over North America. The results also show, however, that predicting the location of significant changes due to the targeted data in the forecasts can be difficult in the presence of a nonzonal large-scale flow. The strong zonal variations in the large-scale flow over the northeast Pacific during WSR00 did not reduce the positive forecast effects of the targeted data. On the contrary, the overall impact of the dropsonde data was more positive than during WSR99, when the large-scale flow was dominantly zonal on the flight days. This can be attributed to the improved prediction of the large-scale flow that led to additional improvements in the prediction of the synoptic-scale waves.
    Wang H. L., M. Mu, X. Y. Huang, 2011: Application of conditional non-linear optimal perturbations to tropical cyclone adaptive observation using the weather research forecasting (WRF) model. Tellus A, 63, 939- 957.10.1111/j.1600-0870.2011.00536.x3af78b06-f0d0-422e-a0c5-b242fa15e14355400776b167c97099b7237d82f8e9dchttp://onlinelibrary.wiley.com/doi/10.1111/j.1600-0870.2011.00536.x/abstracthttp://onlinelibrary.wiley.com/doi/10.1111/j.1600-0870.2011.00536.x/abstractConditional non-linear optimal perturbation (CNOP), which is a natural extension of the linear singular vector into the non-linear regime, has been suggested to identify data-sensitive regions in the adaptive observation strategy. CNOP is the global maximum of a cost function, whereas, local CNOP is the local maximum of the cost function if the local maximum exists. The potential application of CNOPs to tropical cyclone adaptive observation is researched. The CNOPs and the first singular vector (FSV) are numerically obtained by a spectral projected gradient algorithm with the Weather Research Forecasting (WRF) model. This paper examines two tropical cyclone cases, a fast straight moving typhoon Matsa (2005) and a slow moving recurving typhoon Shanshan (2006). The CNOPs and FSVs are obtained using the norms of background error at initial time and total dry energy at final time with a 36-h optimization time interval. The spatial structures of CNOPs, their energies, non-linear evolutions and impacts on track simulations are compared with those of the FSVs. The results show that both the CNOPs and the FSVs are localized, and evolve into the verification area at the final time with the upscale growth of perturbations. However, the CNOPs are different from the FSVs in spatial patterns, wind maximum distribution, growth rate of energy and impact on track simulation. Compared to FSV, CNOP and local CNOP have greater impact on the forecast in the verification region at the final time in terms of total energy, and have larger, at least similar impact on track simulation too. This indicates the CNOP method with constraint of the norm of background error at initial time and total energy norm at final time is a reasonable candidate in tropical cyclone adaptive observation. Therefore, both CNOP and local CNOP are suggested to be considered in tropical cyclone adaptive observation.
    Wu C.-C., K.-H. Chou, P.-H. Lin, S. D. Aberson, M. S. Peng, and T. Nakazawa, 2007a: The impact of dropwindsonde data on typhoon track forecasts in DOTSTAR. Wea. Forecasting, 22, 1157- 1176.1ce5196b-f539-4923-a9d4-68ce2ad3dbca9536005b28a3c31d94cba4faee2a1f03http://www.researchgate.net/publication/254562105_Impact_of_dropwindsonde_data_on_typhoon_track_forecasts_in_DOTSTAR/s?wd=paperuri%3A%2884349b5c39887024bcc0d97e22ac25e6%29&filter=sc_long_sign&sc_ks_para=q%3DThe%20Impact%20of%20Dropwindsonde%20Data%20on%20Typhoon%20Track%20Forecasts%20in%20DOTSTAR&tn=SE_baiduxueshu_c1gjeupa&ie=utf-8
    Wu C.-C., J.-H. Chen, P.-H. Lin, and K.-H. Chou, 2007b: Targeted observations of tropical cyclone movement based on the adjoint-derived sensitivity steering vector. J. Atmos. Sci., 64, 2611- 2626.
    Wu C.-C., S.-G. Chen, J.-H. Chen, K.-H. Chou, and P.-H. Lin, 2009: Interaction of Typhoon Shanshan (2006) with the mid-latitude trough from both adjoint-derived sensitivity steering vector and potential vorticity perspectives. Mon. Wea. Rev., 137, 852- 862.
    Xie B. G., F. Q. Zhang, Q. H. Zhang, J. Poterjoy, and Y. H. Weng, 2013: Observing strategy and observation targeting for tropical cyclones using ensemble-based sensitivity analysis and data assimilation. Mon. Wea. Rev., 141, 1437- 1453.10.1175/MWR-D-12-00188.105eb8aa5-8b57-489b-a023-d51fb0b10437e115bf6ecf23560940f8a42da974f0e5http%3A%2F%2Fwww.researchgate.net%2Fpublication%2F258805701_Observing_Strategy_and_Observation_Targeting_for_Tropical_Cyclones_Using_Ensemble-Based_Sensitivity_Analysis_and_Data_Assimilationrefpaperuri:(e1da7d6e3f6428108d854a5b688d84e0)http://www.researchgate.net/publication/258805701_Observing_Strategy_and_Observation_Targeting_for_Tropical_Cyclones_Using_Ensemble-Based_Sensitivity_Analysis_and_Data_AssimilationAn ensemble Kalman filter data assimilation system for the Weather Research and Forecasting Model is used with ensemble-based sensitivity analysis to explore observing strategies and observation targeting for tropical cyclones. The case selected for this study is Typhoon Morakot (2009), a western Pacific storm that brought record-breaking rainfall to Taiwan. Forty-eight hours prior to making landfall, ensemble sensitivity analysis using a 50-member convection-permitting ensemble predicts that dropsonde observations located in the southwest quadrant of the typhoon will have the highest impact on reducing the forecast uncertainty of the track, intensity, and rainfall of Morakot. A series of observing system simulation experiments (OSSEs) demonstrate that assimilating synthetic dropsonde observations located in regions with higher predicted observation impacts will, on average, lead to a better rainfall forecast than in regions with smaller predicted impacts. However, these OSSEs also suggest that the effectiveness of the current-generation ensemble-based tropical cyclone targeting strategies may be limited. The limitations may be due to strong nonlinearity in the governing dynamics of the typhoon (e.g., moist convection), the accuracy of the ensemble background covariance, and the projection of individual dropsonde observations to the complicated targeted sensitivity vectors from the ensemble.
  • [1] QIN Xiaohao, MU Mu, 2014: Can Adaptive Observations Improve Tropical Cyclone Intensity Forecasts?, ADVANCES IN ATMOSPHERIC SCIENCES, 31, 252-262.  doi: 10.1007/s00376-013-3008-0
    [2] Zhiyong MENG, Eugene E. CLOTHIAUX, 2022: Contributions of Fuqing ZHANG to Predictability, Data Assimilation, and Dynamics of High Impact Weather: A Tribute, ADVANCES IN ATMOSPHERIC SCIENCES, 39, 676-683.  doi: 10.1007/s00376-021-1362-x
    [3] ZHONG Ke, DONG Peiming, ZHAO Sixiong, CAI Qifa, LAN Weiren, 2007: State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics \mbox{\rm (LASG)}, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, ADVANCES IN ATMOSPHERIC SCIENCES, 24, 435-448.  doi: 10.1007/s00376-007-0435-9
    [4] Jie FENG, Ruiqiang DING, Jianping LI, Deqiang LIU, 2016: Comparison of Nonlinear Local Lyapunov Vectors with Bred Vectors, Random Perturbations and Ensemble Transform Kalman Filter Strategies in a Barotropic Model, ADVANCES IN ATMOSPHERIC SCIENCES, 33, 1036-1046.  doi: 10.1007/s00376-016-6003-4
    [5] Zhemin TAN, Qinghong ZHANG, Xudong LIANG, Kun ZHAO, Xin XU, Lili LEI, 2023: Preface to the Special Issue on the 14th International Conference on Mesoscale Convective Systems and High-Impact Weather, ADVANCES IN ATMOSPHERIC SCIENCES, 40, 745-746.  doi: 10.1007/s00376-023-2022-0
    [6] Zhiyong MENG, 2022: Preface to the Special Issue: Predictability, Data Assimilation, and Dynamics of High Impact Weather—In Memory of Dr. Fuqing ZHANG, ADVANCES IN ATMOSPHERIC SCIENCES, 39, 673-675.  doi: 10.1007/s00376-022-2002-9
    [7] Xuanming ZHAO, Jiang ZHU, Lijing CHENG, Yubao LIU, Yuewei LIU, 2020: An Observing System Simulation Experiment to Assess the Potential Impact of a Virtual Mobile Communication Tower–based Observation Network on Weather Forecasting Accuracy in China. Part 1: Weather Stations with a Typical Mobile Tower Height of 40 m, ADVANCES IN ATMOSPHERIC SCIENCES, 37, 617-633.  doi: 10.1007/s00376-020-9058-1
    [8] Xuanming ZHAO, Jiang ZHU, Lijing CHENG, Yubao LIU, Yuewei LIU, 2020: An Observing System Simulation Experiment to Assess the Potential Impact of a Virtual Mobile Communication Tower–based Observation Network on Weather Forecasting Accuracy in China. Part 1: Weather Stations with a Typical Mobile Tower Height of 40 m, ADVANCES IN ATMOSPHERIC SCIENCES.  doi: 10.1007/s00376-020-9058-1-bug
    [9] Jorge A. REVELLI, Miguel A. RODR, Horacio S. WIO, 2010: The Use of Rank Histograms and MVL Diagrams to Characterize Ensemble Evolution in Weather Forecasting, ADVANCES IN ATMOSPHERIC SCIENCES, 27, 1425-1437.  doi: 10.1007/s00376-009-9153-6
    [10] Jiangshan ZHU, Fanyou KONG, Xiao-Ming HU, Yan GUO, Lingkun RAN, Hengchi LEI, 2018: Impact of Soil Moisture Uncertainty on Summertime Short-range Ensemble Forecasts, ADVANCES IN ATMOSPHERIC SCIENCES, 35, 839-852.  doi: 10.1007/s00376-017-7107-1
    [11] Qian ZOU, Quanjia ZHONG, Jiangyu MAO, Ruiqiang DING, Deyu LU, Jianping LI, Xuan LI, 2023: Impact of Perturbation Schemes on the Ensemble Prediction in a Coupled Lorenz Model, ADVANCES IN ATMOSPHERIC SCIENCES, 40, 501-513.  doi: 10.1007/s00376-022-1376-z
    [12] XUE Feng, JIANG Dabang, LANG Xianmei, WANG Huijun, 2003: Influence of the Mascarene High and Australian High on the Summer Monsoon in East Asia: Ensemble Simulation, ADVANCES IN ATMOSPHERIC SCIENCES, 20, 799-809.  doi: 10.1007/BF02915405
    [13] Yanni Qu, Mitchell D. Goldberg, Murty Divakarla, 2001: Ozone Profile Retrieval from Satellite Observation Using High Spectral Resolution Infrared Sounding Instrument, ADVANCES IN ATMOSPHERIC SCIENCES, 18, 959-971.
    [14] Mengmeng SONG, Dazhi YANG, Sebastian LERCH, Xiang’ao XIA, Gokhan Mert YAGLI, Jamie M. BRIGHT, Yanbo SHEN, Bai LIU, Xingli LIU, Martin János MAYER, 2024: Non-crossing Quantile Regression Neural Network as a Calibration Tool for Ensemble Weather Forecasts, ADVANCES IN ATMOSPHERIC SCIENCES.  doi: 10.1007/s00376-023-3184-5
    [15] Fuzhong WENG, Xinwen YU, Yihong DUAN, Jun YANG, Jianjie WANG, 2020: Advanced Radiative Transfer Modeling System (ARMS): A New-Generation Satellite Observation Operator Developed for Numerical Weather Prediction and Remote Sensing Applications, ADVANCES IN ATMOSPHERIC SCIENCES, 37, 131-136.  doi: 10.1007/s00376-019-9170-2
    [16] Xiaoqiong ZHEN, Shuqing MA, Hongbin CHEN, Guorong WANG, Xiaoping XU, Siteng LI, 2022: A New X-band Weather Radar System with Distributed Phased-Array Front-ends: Development and Preliminary Observation Results, ADVANCES IN ATMOSPHERIC SCIENCES, 39, 386-402.  doi: 10.1007/s00376-021-1114-y
    [17] Xun LI, Noel E. DAVIDSON, Yihong DUAN, Kevin J. TORY, Zhian SUN, Qinbo CAI, 2020: Analysis of an Ensemble of High-Resolution WRF Simulations for the Rapid Intensification of Super Typhoon Rammasun (2014), ADVANCES IN ATMOSPHERIC SCIENCES, 37, 187-210.  doi: 10.1007/s00376-019-8274-z
    [18] Seung-Woo LEE, Dong-Kyou LEE, 2011: Improvement in Background Error Covariances Using Ensemble Forecasts for Assimilation of High-Resolution Satellite Data, ADVANCES IN ATMOSPHERIC SCIENCES, 28, 758-774.  doi: 10.1007/s00376-010-0145-6
    [19] X.L. Wang, H.L. Rui, A. Leetmaa, 1996: The Dynamics of ENSO Anomaly as Revealed in Ensemble Climate Simulations-Impact of Mean Stationary Wave, ADVANCES IN ATMOSPHERIC SCIENCES, 13, 425-444.  doi: 10.1007/BF03342035
    [20] Kang Hongwen, Gu Xiangqian, Liu Chongjian, Wang Pengyun, 2002: Adaptive Grid Technique Based on the Variational Principle and Its Weight Functions, ADVANCES IN ATMOSPHERIC SCIENCES, 19, 705-718.  doi: 10.1007/s00376-002-0010-3

Get Citation+

Export:  

Share Article

Manuscript History

Manuscript received: 23 January 2015
Manuscript revised: 11 May 2015
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Ensemble Transform Sensitivity Method for Adaptive Observations

  • 1. Nanjing University of Information Science and Technology, Nanjing 210044
  • 2. Chinese Academy of Meteorological Science, Beijing 100081
  • 3. Global Systems Division, Earth System Research Laboratory, NOAA, Boulder, CO 80305, USA
  • 4. Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, CO 80523, USA
  • 5. National Meteorological Centre, Beijing 100081

Abstract: The Ensemble Transform (ET) method has been shown to be useful in providing guidance for adaptive observation deployment. It predicts forecast error variance reduction for each possible deployment using its corresponding transformation matrix in an ensemble subspace. In this paper, a new ET-based sensitivity (ETS) method, which calculates the gradient of forecast error variance reduction in terms of analysis error variance reduction, is proposed to specify regions for possible adaptive observations. ETS is a first order approximation of the ET; it requires just one calculation of a transformation matrix, increasing computational efficiency (60%-80% reduction in computational cost). An explicit mathematical formulation of the ETS gradient is derived and described. Both the ET and ETS methods are applied to the Hurricane Irene (2011) case and a heavy rainfall case for comparison. The numerical results imply that the sensitive areas estimated by the ETS and ET are similar. However, ETS is much more efficient, particularly when the resolution is higher and the number of ensemble members is larger.

1. Introduction
  • For high-impact weather (HIW) events, adaptive mobile observation instruments or vehicles can be deployed to improve analysis quality and forecast accuracy. Several field campaigns have shown that observations sampled in dynamically sensitive areas have positive impacts on numerical weather prediction (Majumdar et al., 2001; Majumdar et al., 2011). For example, 1-2 day forecast skill was increased by assimilating targeted data deployed in the Fronts and Atlantic Storm Track Experiment and the North Pacific Experiments (Joly et al., 1997; Joly et al., 1999; Langland et al., 1999a; Langland et al., 1999b). Assimilation of adaptive observations significantly reduced typhoon track forecast errors over the western North Pacific and the Atlantic (Aberson, 2003; Wu et al., 2007a; Aberson et al., 2011; Chou et al., 2011). A major challenge is to identify sensitive areas for deploying the adaptive observations in the hours or days ahead of HIW events. There are several approaches that have been developed to estimate sensitive areas, such as the singular vector method (Palmer et al., 1998; Buizza and Montani, 1999), the conditional nonlinear optimal perturbation method (Mu et al., 2009; Wang et al., 2011), and the adjoint sensitivity method (Wu et al., 2007b, 2009). In general, an adjoint model is usually required in the above three approaches. In addition, ensemble-based methods, such as the ensemble transformation (ET) method (Bishop and Toth, 1999), the ensemble transform Kalman filter (ETKF) method (Bishop et al., 2001), and ensemble sensitivity (Ancell and Hakim, 2007) are widely used in field campaigns (Chang et al., 2013; Xie et al., 2013).

    The ensemble-based methods are less demanding computationally and have been extensively employed in practical applications (Ancell and Hakim, 2007; Ito and Wu, 2013). These methods consider sensitivity in the subspace spanned by the ensemble forecasts and are computationally inexpensive in operational centers where ensemble forecasts are routinely produced. Among these methods, ET (Bishop and Toth, 1999; hereafter BT1999) provides a practical method for adaptive observations. It has been used for targeted dropsonde deployments in winter storm reconnaissance (WSR) (Szunyogh et al., 2000). Later, ETKF was used to identify the sensitive region in WSR (Szunyogh et al., 2002). The dropsonde data collected over these sensitive areas improved the weather forecasts over the continental United States and Alaska (Szunyogh et al., 2000). However, the impact of dropsonde data may be limited in global forecasts (Hamill et al., 2001), and high-resolution observation datasets are suggested for HIW (Bauer et al., 2011, Berger et al., 2011).

    It is noted that ET is still expensive for high-resolution applications or those applications with large numbers of ensemble members. ET has been used at relatively coarse resolutions and a few vertical levels, e.g., usually three vertical levels at the National Centers for Environmental Prediction (NCEP), and a relatively small number of ensemble members (30-60). As resolutions increase for HIW applications, the computational cost grows exponentially. This is because ET, as well as the ETKF, has been implemented to exhaust all possible observation deployments. For example, it currently estimates sensitive areas by adding an observation at every analysis grid location, horizontally and vertically, and calculating the ensemble transformation and the reduction of forecast variance for each observation. Because of the use of a matrix decomposition of ensemble covariance, the computational cost also increases as a cubic function of the number of ensemble members. For high-resolution adaptive observation applications, or those with large numbers of ensemble members, the computational cost could be significant. ETKF can also be computationally expensive, the same as the ET method (Bishop et al., 2001). In order to further improve these methods for fine scale HIW applications, efficiency is an important factor to investigate.

    A new ET-based sensitivity (ETS) method is proposed in this paper to specify sensitive regions for adaptive observations. The proposed method calculates the sensitivity (gradient) of forecast error variance reduction in terms of analysis error variance reduction. The newly proposed ETS method is the first order approximation of the original perturbation ET method and reduces computational cost because only a single transformation matrix calculation is required.

    This paper is organized as follows. In section 2, we review the ET method. In section 3, we describe how ETS calculates the sensitivity with a single transformation matrix calculation. We compare sensitive regions using ET and ETS for a hurricane case and a heavy rainfall case in section 4. Conclusions and discussions are presented in section 5.

2. Review of the ET method
  • First, we review the ET method (BT1999), with some matrix notations used to simplify the discussion (see Table 1). This review in matrix forms helps illustrate the ETS derivation.

  • Let E(t) denote a set of ensemble forecasts at a given forecast time t. This E(t) is a matrix with M rows and K columns, where M is the number of gridded values of all the state variables and K is the number of ensemble members. Then, the ensemble perturbation matrix X e is also an M× K matrix, \begin{equation} {X}_{ e}(t)={E}(t)-\overline{{E}} , (1)\end{equation} where \(\overline E\) is the ensemble mean and also an M× K matrix, but all columns are the same ensemble mean vector. Here is the adaptive observation strategy in ET: Use a set of ensemble forecasts E(t) to determine which possible deployments of observational resources at a future analysis time t fa will minimize the expected prediction error of forecasts for the verification time t v, which are initialized with, inter alia, the supplemental data taken at t fa (BT1999).

    Let X e(t fa) denote the perturbations at t fa, X e(t v) the perturbations at the t v, and Y(t fa) the ensemble perturbations after assimilating a set of data from a possible adaptive observation deployment. ET finds a transformation of the ensemble perturbation X e(t fa) to Y(t fa). Such a transformation can be uniquely determined if the number of ensemble forecasts is very large and unconstrained (Anderson, 1997). Assume such a transformation exists, and denote it as a matrix C, such that \begin{equation} {Y}(t_{ fa})={X}_{ e}(t_{ fa}){C} . (2)\end{equation} This C is a K× K matrix.

    Mathematically, the ensemble-based error covariance at the t a approximates the truth analysis error covariance of A e(t fa), \begin{equation} {A}_{ e}(t_{ fa})\approx\dfrac{1}{K}{X}_{ e}(t_{ fa}){C}{C}^{ T}{X}_{ e}(t_{ fa})^{ T} . (3)\end{equation} The ensemble-based forecast error covariance at t v, P e(t v), can be approximated by \begin{equation} {P}_{ e}(t_{ v})=\dfrac{1}{K}{X}_{ e}(v){C}{C}^{ T}{X}_{ e}(t_{ v})^{ T} . (4)\end{equation}

    In reality, the true analysis error covariance of A e(t fa) is unknown, but an approximation or guess can be estimated by a given data assimilation system (BT1999). Let A g(t fa) denote the approximation of A e(t fa) and Eq. (3) is approximately satisfied. Thus, forcing transformed ensemble-based error covariance to be equal to the guessed adaptive analysis error covariance, \begin{equation} {A}_{ g}(t_{ fa})=\dfrac{1}{K}{X}_{ e}(t_{ fa}){C}{C}^{ T}{X}_{ e}(t_{ fa})^{ T}.(5) \end{equation} The ET method finds a solution of CC T satisfying Eq. (5). Note that there is really no need to explicitly calculate the transformation matrix C in the ET method but CC T, a product of the transformation. For the sake of simplicity, here we assume that A g is a full rank matrix (BT1999) and Eq. (5) can be rewritten as \begin{equation*} {A}_{ g}^{-1}(t_{ fa})=K^{-1}{A}_{ g}^{-1}(t_{ fa}){X}_{ e}(t_{ fa}) {C}{C}^{ T}{X}_{ e}(t_{ fa})^{ T}{A}_{ g}^{-1}(t_{ fa}) .(5') \end{equation*} Then if X e(t fa) is a full rank (e.g., a set of independent ensemble members), the matrix X e(t fa) TA g-1X e(t fa) is invertible. By multiplying X e(t fa) T and X e(t fa) from the left and right of Eq. (5'), the solution of the product of the ET transformation matrix (BT1999) is, \begin{equation} {C}{C}^{ T}=K({X}_{ e}(t_{ fa})^{ T}{A}_{ g}^{-1}{X}_{ e}(t_{ fa}))^{-1} . (6)\end{equation} Equation (6) is an equivalent variation of the Equation (8) in BT1999. This matrix derivation of the product simplifies the derivation of BT1999.

  • Let β be a parameter measuring the percentage reductions in the analysis error after a set of adaptive observations are assimilated. For example, β=1 means zero percent reduction, β=0.5 means 50% reduction etc. Let A g(β) denote the best guess of the analysis error covariance with all possible adaptive observation reduction by β, and αi(i=1,…,M) denote the i-th diagonal element of A g. \begin{equation} {A}_{ g}({\beta})={ diag}(\alpha_1\beta_1,\ldots,\alpha_M\beta_M). (7)\end{equation} Note that β=(β12,…) is a vector of all possible adaptive observation locations. Notice that we used analysis error variance only as BT1999 did. The components of β may or may not be equal to 1 corresponding to the given adaptive observation datasets. A value of 1 indicates no error variance reduction at this location, while a value of <1 indicates an observation at this location is assimilated. The corresponding transformation matrix C(β) can be calculated by (6). One can use (4) to estimate the forecast error covariance P e(t v) associated with the adaptive observation scheme.

    In order to calculate the observation sensitivity for an adaptive observation scheme, one has to define an output scalar measuring the sensitivity. A measurement is usually defined by an energy norm using forecast variance information from (4). For example, a total dry energy norm (Ehrendorfer et al., 1999) is expressed as, \begin{equation} \dfrac{1}{2}\dfrac{1}{D}\int_D\int_0^1\left[u'^{2}+v'^{2}+\dfrac{c_p}{T_{ r}}T'^2+RT_{ r}\left(\dfrac{{ pre}_{ s}}{{ pre}_{ r}}'\right)^2\right]{ d}\delta{ d}D , (8)\end{equation} where (u',v',T', pre' s) are the forecast error variance of Eq. (4) corresponding to two wind components, temperature, and surface pressure. cp, and R are specific heat at constant pressure and the gas constant of dry air, respectively (with numerical values of 1005.7\;J kg-1 K, and 287.04\;J kg-1 K). The integration extends over the full horizontal domain D and vertical directions δ. The T r and pre r are the reference temperature and pressure. In this study, zonal and meridional horizontal wind, and temperature are used to estimate the reduction of forecast error variance, since the contribution from the pressure term is very small, and thus ignored. The reference temperature is 270 K, the same as in Martin et al. (1999). In an adaptive observation measurement, it is common to use this norm to sum of the diagonal elements of P e(t v) corresponding to u,v and T. Using Eqs. (4) and (8), the sum can be calculated as follows.

    • A projection matrix \(\wp= diag(P_i),i=1\ldots M\), where Pi=1 if the i-th position is either the u or v state variables, and \(P_i=\sqrt c_p/T_ r\) if the i-th position is the T state variable, otherwise the values are zeros.

    • The norm of forecast error variances is the sum of the diagonal elements of \begin{equation*} \dfrac{1}{K}\wp{X}_e(t_{ v}){C}({\beta}){C}^{ T}({\beta}){X}_{ e}(t_{ v})^{ T}\wp .\end{equation*}

    • The measurement of the adaptive data impact is calculated as follows. Let \(Z=(Z,\ldots,Z_M)=X_ e(t_ v)^ T\wp\), where Zi is the i-th column of matrix Z. The forecast error J is \begin{equation} J[{\beta}]=\dfrac{1}{K}{\sum}_{i=1}^M{Z}_i^{ T}{C}[{\beta}]{C}^{ T}[{\beta}]{Z}_i . (9)\end{equation}

    In general, the forecast error variance reduction in ET at a possible adaptive deployment at a location l is \begin{equation} S_l=J[\beta_l=1]-J[1-\Delta \beta_l] . (10)\end{equation} The forecast error reduction estimations are obtained by repeating the above process for all possible adaptive deployments. In BT1999, it was assumed the analysis error variance is reduced by 0.5 (∆βl=0.5).

3. ETS method
  • By perturbing all possible adaptive observation data, the ET method may yield high order information about the sensitivity regions but it could be costly for high-resolution applications with large ensemble members. In this section, we consider a first order approximation of the ET method, ET sensitivity.

  • The basic idea of the ETS in this paper is to use the sensitivity (gradient) of forecast error variance over the verification region in terms of analysis error variance to determine data sensitive regions for adaptive observations. It is the first order approximation of Eq. (10), but only a single transformation matrix computation will be needed, thus improving computation efficiency when compared to ET. The main objective is to derive a mathematical formulation of ∂ J/∂βl in this paper.

    Following Eq. (8), the gradient of J to the analysis error variance reduction ratio β is \begin{align} \nabla J&=\left(\dfrac{\partial J}{\partial \beta_1},\ldots,\dfrac{\partial J}{\partial\beta_l},\ldots,\dfrac{\partial J}{\partial\beta_M}\right)^{ T}\nonumber\\ &=\bigg(\dfrac{1}{K}{\sum}_{i=1}^M{Z}_i^{ T}\dfrac{\partial{C}{C}^{ T}}{\partial\beta_1}{Z}_i,\ldots,\dfrac{1}{K}{\sum}_{i=1}^M {Z}_i^{ T}\dfrac{\partial{C}{C}^{ T}}{\partial\beta_l}{Z}_i,\ldots,\nonumber\\ &\quad\ \dfrac{1}{K}{\sum}_{i=1}^M{Z}_i^{ T}\dfrac{\partial{C}{C}^{ T}}{\partial\beta_M}{Z}_i\bigg)^{ T} . (11a)\end{align}

    The estimated forecast error variance reduction is \begin{align} dJ&=\bigg(\dfrac{1}{K}{\sum}_{i=1}^M {Z}_i^{T}\dfrac{\partial{C}{C}^{T}} {\partial\beta_1} {Z}_i,\ldots,\dfrac{1}{K}{\sum}_{i=1}^M{Z}_i^{T}\dfrac{\partial{C}{C}^{T}}{\partial\beta_l}{Z}_i,\ldots,\nonumber\\ &\quad\ \dfrac{1}{K}{\sum}_{i=1}^M {Z}_i^{ T}\dfrac{\partial{C}{C}^{T}}{\partial\beta_M}{Z}_i\bigg)^{T}d{\beta}.(11b)\end{align}

    In the BT1999 implementation, dβ is set to a constant with a value of 0.5. The ET method is approximated by ETS derivatives. In operational applications, β can be set to different values at different locations that can really take advantage of an analysis error covariance, e.g., a large reduction (β) occurs over a large analysis error variance.

  • The main contribution of the ETS method is the derivation of an analytic gradient formulation of CC T/∂β in Eq. (11a) in terms of the error reduction coefficient βl.

    For the ET transformation matrix C, let us introduce a matrix \begin{equation} {\Psi}=K({C}{C}^{ T})^{-1}={X}_{ e}(t_{ fa})^{ T}{A}_{ g}^{-1}{X}_{ e}(t_{ fa}) . (12)\end{equation} and then the product of ET transformation matrix CC T=KΨ-1 as given in Eq. (6). Using an inverse matrix derivative formulation, the ET transformation product derivative is, \begin{equation} \dfrac{\partial{C}{C}^{ T}}{\partial\beta_l}=-K{\Psi}^{-1}\dfrac{\partial{\Psi}}{\partial\beta_l}{\Psi}^{-1} , (13)\end{equation} where, \begin{eqnarray} \dfrac{\partial{\Psi}}{\partial\beta_l}&=&{X}_{ e}(t_{ fa})^{ T}\dfrac{\partial{A}_{ g}^{-1}}{\partial\beta_l}{X}_{ e}(t_{ fa})\nonumber\\ &=&-{X}_{ e}(t_{ fa})^{ T}{A}_{ g}^{-1}\dfrac{\partial{A}_{ g}}{\partial\beta_l}{A}_{ g}^{-1}{X}_{ e}(t_{ fa}) . (14)\end{eqnarray} Thus, ET sensitivity can be obtained using Eqs. (11-14).

    \begin{equation} \dfrac{\partial J}{\partial\beta_l}={\sum}_{i=1}^M{Z}_i^{ T}{\Psi}^{-1}{X}_{ e}(t_{ fa})^{ T}{A}_{ g}^{-1} \dfrac{\partial{A}_{ g}}{\partial\beta_l}{A}_{ g}^{-1}{X}_{ e}(t_{ fa}){\Psi}^{-1}{Z}_i. (15)\end{equation} Note the ?A g/?βl is usually a constant matrix. For an example of a diagonal matrix of A g= diag(α1β1,…,αMβM) (BT1999), ?A g/?βl|β=1 is equal to a diagonal matrix of diag(0,…,0,αl,0,…,0). For a given guessed analysis error variance A g, the ET transformation matrix CC T is determined. So the ET sensitivity from Eq. (15) can be obtained after a single computation of a transformation matrix instead of calculating ensemble transformations (\(C[\beta]C^ T[\beta]\)) for all possible perturbations in ET using Eq. (10).

  • When applying ETS in practice, Eq. (15) is not solved directly. Here are the implementation procedures:

    Step 1: Compute the perturbation fields at the t fa:X e(t fa), and t v:X e(t v).

    Step 2: Initial the projection matrix \(\wp\) and the guessed analysis error variance A g.

    Step 3: Compute the inverse of Eq. (12): X e(t fa) TA g-1X e(t fa)

    Step 4: Compute the matrix \(Z:Z=X_ e(t_ v)^ T\wp\)

    Step 5: Obtain all the signals [Eq. (11a)]: ? J/?βl,l=1… M. Decomposition of the K× K symmetric matrix (CC T) takes K3/6 computing operations (Step 3 costs ∼ K3/6). In order to obtain all of the sensitivity, ETS needs to estimate the sensitivity at all the elements in the state vector (M), with an Eq. (11a) cost of ∼ M2 (Step 5 costs ∼ M2). So, the computation count of ETS is about M2+K3/6. However, ET needs to decompose the CC T matrix at each element in the state vector [Eqs. (9) and (10) with a cost of ∼ M2+K3/6] to obtain all the signals (forecast error reduction). The magnitude of K is about 102. The M is 103 in a very coarse resolution. It could rise to 108 in the high-resolution case. Table 2 shows the estimation of the computation counts of ET and ETS. ETS gains greatly in efficiency as it only needs to decompose the CC T once. When M and K are large, the difference is significant.

4.Numerical experiments
  • In this section, we apply ET and ETS for a hurricane case and a rainfall case. The first case, Hurricane Irene (2011), formed on 21 August, and became a hurricane on 22 August 2011. It then passed Exuma and Cat Islands. It made landfall near Cape Lookout, North Carolina at 1200 UTC on 27 August. It continued tracking north northeastward, and moved over Manhattan, New York on 28 August. The heavy rainfall and strong wind caused severe damage (Avila and Stewart, 2012). We also apply ET and ETS for a heavy rainfall case. The heavy precipitation in this case is associated with a low level vortex that developed over western China during 3-5 August 2013. The hourly accumulate precipitation was >30\;mm over the Beijing areas at 1200 UTC 4 August 2013.

  • The European Centre for Medium-range Weather Forecasts ensemble forecasts are used in this study, which can be downloaded from the THORPEX Interactive Grand Global Ensemble (TIGGE) portal (http://apps.ecmwf.int/datasets/data/tigge/). The initial time of the ensemble forecasts are at 0000 UTC 24 August 2011 and 1200 UTC 3 August 2013 for the hurricane and heavy rainfall case, respectively. The length of prediction time is 72-h with a 6-h interval for the ensemble prediction outputs. The variables selected for the ET dry energy norm are the temperature and horizontal wind components at the 850, 500 and 200 hPa pressure levels. The diagonal values of guessed analysis error covariance A g used are the same as ETKF (Majumdar et al., 2002): the guessed analysis error covariance of wind at the 850, 500 and 200 hPa pressure levels is 2.72, 3.16 and 4.66 m s-1 separately; the guessed analysis error vovariance of temperature at the 850, 500 and 200 hPa pressure levels is 1.22, 0.92°C and 1.82°C separately.

    For the hurricane case, the verification area (26°-40°N, 86°-70°W) is marked by the inner rectangle showed in Fig. 1a). The estimation or potential targeting observation area is the whole domain (10°-50°N, 100°-60°W). The ensemble mean indicated that the hurricane was moving towards the east coast of the U.S. at 0000 UTC 27 August, and it is selected as the t v in this case. There are seven t fa for the adaptive observations, -0 h, -12 h, -24 h, -36 h, -48 h, -60 h, and -72 h. The negative hours t fa indicate the number of hours ahead of the t v, correspondingly. For the heavy rainfall case, the verification area is over the Beijing area (38°-42°N, 114°-120°E), (Fig. 1b). The estimation potential targeting area was covered from 100°E to 124°E and 34°N to 50°N. The t v is the heavy rainfall time, 1200 UTC 4 August 2013. The t fa are set to 6 h and 12 h ahead of the t v. As claimed in section 3, ETS is more efficient when the number of ensemble members (K) and the elements in state vectors (M) are mathematically large. We set up six experiments (Table 3) with different resolutions and ensemble members to demonstrate this claim numerically. The number of ensemble members vary from 10 to 30 to 50. Two resolutions are used, 1°× 1° and 2°× 2°. As an example, K30R1 means the number of ensemble members is 30 and the resolution is 1°× 1°.

    Figure 1.  The domain and verification areas for the (a) Hurricane Irene (2011) and (b) Beijing rainfall cases. Contours are the ensemble mean geopotential height at 500 hPa (gpm). The inner rectangle is the verification areas.

    Figure 2.  The signals (color filled areas) identified by (a, c, e) ETS and (b, d, f) ET at a different $t_ a$ in K30R2 for the Hurricane Irene case. The $t_ a$ are (a, b) 0000 UTC 25 August; (c, d) 0000 UTC 26 August, (e, f) 0000 UTC 27 August. The contours are the 500 hPa geopotential height of the ensemble mean forecast at each $t_ a$. The inner rectangle is the verification area. The $t_ v$ is 0000 UTC 27 August.

    Figure 3.  The signals (color-filled areas) identified by (a, c) ETS and (b, d) ET in the Hurricane Irene case: (a, b) are Experiment K50R1; (c, d) experiment K10R2. The $t_ a$ is 0000 UTC 26 August 2011. The contours are the 500 hPa geopotential height of ensemble mean forecasts.

    Figure 4.  (a) Computational cost of different experiments. (b) Relative computation time reduction.

    Figure 5.  The signals (color filled areas) identified by (a, c) ETS and (b, d) ET in K30R2 for the Beijing rainfall case. The $t_ v$ is 1200 UTC 4 August 2013. The $t_ a$ are (a, b) is 0000 UTC 4 August; (c, d) 0600 UTC 4 August. The wind barbs are the horizontal wind component at 850 hPa (units: m s$^-1$). The contours are the 500 hPa geopotential height of the ensemble mean forecast at each $t_ a$. The inner rectangle is the verification area.

    Figure 6.  The signals for the Hurricane Irene (2011) case. Contours are the signals from (a-c) ET and (d) ETS. The experiment is K30R2. The $t_ a$ is 0000 UTC 26 August 2011. Color filled areas show the signal differences between ETS and ET. The analysis error reduction for ET is (a) 0.2, (b), 0.4 and (c) 0.6.

    Figure 7.  As in Fig. 6 but for the rainfall case. The $t_ a$ is 0000 UTC 4 August 2013.

  • A summary map——the signals of sensitivity identified by ET or ETS over the whole calculation domain——shows the sensitive area. ET considered each grid point as a hypothetical adaptive site and identified sensitive areas by perturbing the analysis error variance at each observation site. The signal——the reduction of forecast error covariance associated with this grid point——can be obtained. The summary map can be plotted after perturbing the variance and calculating the reduction over all the grid points.

    In contrast to these perturbations of ET, the ETS method can obtain these values by a single computation of the derivative, using the same amount of computation as would be needed for each individual ET perturbation. The derivatives are shown, as well as the reduction of analysis error variance. They represent the sensitivity signals over the calculation domain. In order to compare the signals from ETS and ET at different t fa, the signals from Eqs. (10) and (15, 11b) are normalized in this study, \begin{equation} \overline{{S}}=({S}-S_{\min})/(S_{\max}-S_{\min}) ,(16)\end{equation} where \(S_\max/S_\min\) are the maximum/minimum values over the whole domain. Thus, the summary maps show the relative sensitivity of the ET or ETS methods.

  • The color filled areas of Fig. 2 show the normalized signals identified by ET and ETS at different t fa for experiment K30R2 in the hurricane case. It is seen that ETS and ET give similar signal patterns and evaluation. As the t fa approached the t v, the signal (color filled contour areas) approached the verification areas. The sensitivity areas are distributed around the hurricane itself, and evolved into the verification areas at the t v. Figure 3 shows the normalized signals identified from ET and ETS from K50R2 and K10R1. The results are very close to K30R2. The signals are both located at the hurricane's eastern center. It shows that the data sensitive region identified by ETS is very close to ET even when the results are from a different number of ensemble members and resolutions. Generally, the ETS can obtain the same sensitive areas as ET without high consumption.

    ETS is much faster than ET because ET needs to loop over all the possible elements in the state vector (M), especially when the number of ensemble prediction members (K) is large. Figure 4 shows the computational costs and relative computation time reduction with ETS and ET using a different number of ensemble prediction members (K). The cost is less than 60 seconds with a fine resolution and few ensemble prediction members for ETS and ET. This is acceptable for the adaptive observations. However, the computational cost rises to about 1200 seconds with a 1°× 1° resolution in the horizontal direction, with three vertical levels and 50 ensemble prediction members. ETS only costs about 200 seconds. Overall, the computation time saved by ET was 60%-80% (Fig. 4b). If the computational domain is larger (particularly for a global model) with higher resolution in the horizontal and vertical directions (here the computations were conducted in three vertical levels only), the reduction in computational costs would be much more significant with ETS compared to ET.

    The signals from the heavy rainfall case are shown in Fig. 5. It can be seen that the signals are similar between ET and ETS. The sensitive areas were distributed around the wind divergence (850 hPa) and the trough (500 hPa) at 12 hours ahead of the t v (Figs. 5a and b). The signals are located in the west of the verification areas at 6 hours ahead of the t v (Figs. 5c and d). Although the sensitive areas from the ET covered a slightly larger area compared to ETS, signals with maximum values are located at almost the same position in ET and ETS. The following section provides more discussion on the differences between ETS and ET.

  • In BT1999, the signals are calculated by Eq. (10). ETS used Eq. (15) to calculate the signals. So ETS is a first order approximation of the ET. This also means the results from ETS should get closer to those of ET when β approaches zero. Here we set up three more numerical experiments using different β values for ET (β=0.2,0.4,0.6). When β=0.2 this means the ET signals are estimated by: sl=J[βl=1]-J[βl=0.8] It is noted that ETS signals [Eqs. (10a), (14)] do not vary with different β. Figures 6 and 7 show the ETS and ET signals for the Hurricane Irene (2011) and Beijing rainfall cases, respectively. The differences between ETS and ET are presented by the color shading colors. It is seen that for the smallest β, the two methods produce almost the same data sensitive region (Figs. 6 and 7). And for larger values of β, the ETS signal distribution is still close to the ET signal; in particular, the centers of the signals from the two methods are almost the same even with a large β. Overall, for the Hurricane Irene (2011) case, both ET and ETS identify one sensitive region (Fig. 6); for the rainfall case, one region with global maximum signals and two local regions with local maximum signals (Fig. 7) are identified. The differences between ET and ETS are acceptable, since the targeting observation focuses on the sensitive areas with a maximum (the center of the signals). Generally the signals from ET and ETS are similar.

5. Conclusion and discussion
  • Adaptive observations have the potential to improve weather forecasts. Among existing methods of identifying observation sensitivity regions, ET is attractive because of its use of analysis error covariance information and its efficiency compared to other more complex methods. In this study, a newly proposed ETS approach for adaptive observations is derived and demonstrated. The ETS method only uses a single computation of a transformation matrix to yield a sensitivity summary map, instead of calculating ensemble transformations for all possible perturbations, as in the ET method. Thus, it further increase the computational efficiency. If the computational domain is larger (even global in the horizontal direction), with higher resolution in the horizontal and vertical directions, the reduction in computational cost would be far greater with ETS compared to ET. Numerical experiments with Hurricane Irene (2011) and a heavy rainfall case in Beijing showed that ETS reduced the computation cost by 60%-80%.

    The summary maps from the two cases show that the ETS method produces a similar data sensitive region as the ET method, especially for the region with large signal values. Thus, the new method gains computational efficiency without losing the positive characteristics of the ET method. It is noted that, in general, the more realistic the analysis covariance is, the better the targeting region is that can be identified under the assumption of ET. As the main aim of this paper is mainly to introduce the ETS method, the best guess of analysis covariance, which can be provided by a data assimilation system (e.g., ETKF), will be further studied in future work. Our plan is to implement ETS at the NCEP Environmental Modeling Center for WSR, and compare it to the existing ET adaptive observation method in future works. With its improved efficiency, ETS can be applied to severe weather events with high spatial resolution and a large number of ensemble members.

Reference

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return