Expanded use of the information content of infrared hyperspectral radiance data has resulted in an improvement in the beneficial impact of these data on numerical weather prediction. Experiments which have shown the benefit of improved spatial coverage, spectral coverage and the use of moisture channel data, have been briefly summarised in this paper. In addition, an experiment which has recorded the benefit of using hyperspectral radiance data from fields of view containing clouds is also described. Again it is demonstrated that a more complete use of the information content in the observations available from hyperspectral sounders has resulted in improved benefits to numerical weather prediction. This conclusion is also supported by early experiments reported using IASI data.

The Atmospheric Infrared Sounder (AIRS) (Aumann et al. 2003; Chahine et al. 2006) was launched in 2002 on AQUA, the second of the EOS polar-orbiting satellites. The AIRS was the first of a new generation of meteorological advanced sounders able to provide hyperspectral data for operational and research use. Initially, we briefly review the first assimilation trials to use full spatial resolution and higher spectral resolution hyperspectral radiance data, available in real time from the AIRS. The result from these assimilation trials was significant improvement in forecast skill in the National Centers for Environmental Prediction (NCEP) Global Data Assimilation System (GDAS), compared to the global system without AIRS data over both the northern and southern hemispheres. A second trial was an experiment which showed the advantage of using all AIRS fields of view (fov) in the analysis as opposed to the use of sampled fields of view (typically one-in-eighteen) often used by numerical weather prediction (NWP) centres. Another trial showed the benefit of using hyperspectral data with expanded spectral coverage. We then describe recent experiments where radiances, derived from cloudy AIRS fovs and which represent the radiance emanating from the clear part of the cloudy fov, have been assimilated for global NWP. The beneficial impact of these data in the GDAS is recorded. The impact is an initial indication of the potential benefit of using cloudy hyperspectral radiances routinely in global NWP. Background

The Atmospheric InfraRed Sounder (AIRS), flying aboard NASA's Aqua satellite with the Advanced Microwave Sounding Unit‐A (AMSU‐A) and four other instruments, has been providing data for use in numerical weather prediction and data assimilation systems for over three years. The full AIRS data set is currently not transmitted in near‐real‐time to the prediction/assimilation centres. Instead, data sets with reduced spatial and spectral information are produced and made available within three hours of the observation time. In this paper, we evaluate the use of different channel selections and error specifications. We achieve significant positive impact from the Aqua AIRS/AMSU‐A combination during our experimental time period of January 2003. The best results are obtained using a set of 156 channels that do not include any in the H2O band between 1080 and 2100 cm−1. The H2O band channels have a large influence on both temperature and humidity analyses. If observation and background errors are not properly specified, the partitioning of temperature and humidity information from these channels will not be correct, and this can lead to a degradation in forecast skill. Therefore, we suggest that it is important to focus on background error specification in order to maximize the impact from AIRS and similar instruments. In addition, we find that changing the specified channel errors has a significant effect on the amount of data that enters the analysis as a result of quality control thresholds that are related to the errors. However, moderate changes to the channel errors do not significantly impact forecast skill with the 156 channel set. We also examine the effects of different types of spatial data reduction on assimilated data sets and NWP forecast skill. Whether we pick the centre or the warmest AIRS pixel in a 3 × 3 array affects the amount of data ingested by the analysis but does not have a statistically significant impact on the forecast skill. Copyright © Published in 2007 by John Wiley & Sons, Ltd.

(Aumann et al. 2003) was launched on AQUA the second of the EOS polar-orbiting satellites. The AIRS was the first of a new generation of meteorological advanced sounders able to provide hyper-spectral data for operational and research use. A large investment has been made internationally to upgrade the meteorological satellite systems to carry these advanced instruments. The US Cross-track Infrared Sounder (CrIS) and Geosynchronous Imaging Fourier Transform Spectrometer (GIFTS), as well as the European Infrared Atmospheric Sounding Interferometer (IASI) represent significant investment in these systems. As a result, demonstration of the benefit of hyperspectral data on numerical weather prediction (NWP) has been a high priority. Observing system experiments designed to examine effective methods to use AIRS hyperspectral radi-ances are summarised here. The first experiments to use full spatial resolution hyperspectral radiance data, available in real time from the AIRS instrument are reviewed. The result of these assimilation trials was significant improvements in forecast skill, compared to the global system without AIRS data over both the northern and southern hemispheres. In addition , an experiment is described which showed the advantage of using all AIRS fields of view in analysis as opposed to the use of sampled fields of view (typically one-in-eighteen) often used for NWP. Experiments showing the impact of using hyperspec-tral data of different spectral coverage are also described, and show the importance of careful selection of instrument channels for assimilation. Overall, the results indicate the significant benefits to be derived from AIRS data assimilation and the benefits to be gained from an enhanced use of the information content contained in the AIRS radiance observations.

NASA, NOAA, and U.S. Department of Defense Joint Center for Satellite Data Assimilation, Camp Springs, Maryland;NASA Jet Propulsion Laboratory, Pasadena, CaliforniaCORRESPONDING AUTHOR: John Le Marshall, Joint Center for Satellite Data Assimilation, NOAA Science Center, 5200 Auth Road, Camp Springs, MD 20746 E-mail: John.Lemarshall@noaa.gov

Full spatial resolution AIRS hyperspectral data have been used in a data assimilation study over the globe utilizing the operational NCEP Global Forecast System (GFS). The result of this assimilation trial has been significant improvements in forecast skill over the Southern Hemisphere and improvement over the Northern Hemisphere.

Experimental weather forecasts at the Joint Center for Satellite Data Assimilation (JCSDA) using Atmospheric Infrared Sounder (AIRS) radiance observations indicate significant improvements in global forecast skill compared with the operational system without AIRS data. The improvement in forecast skill at six days is equivalent to gaining an extension of forecast capability of several hours.This magnitude of improvement is quite significant when compared with the rate of general forecast improvement over the last decade. A several hour increase in forecast range at five or six days normally takes several years to achieve at operational weather centers.

Since the publication of the Optical Path Transmittance (OPTRAN) algorithm [Appl. Opt. 34, 8396 (1995)], much of the code and implementation has been refined and improved. The predictor set has been expanded, an objective method to select optimal predictors has been established, and the two-interpolation method has been discarded for a single-interpolation method. The OPTRAN coefficients have been generated for a wide range of satellites and instruments. The most significant new development is the Jacobian-K-matrix version of OPTRAN, which is currently used for operational direct radiance assimilation in both the Global Data Analysis System and the ETA Data Analysis System at the National Oceanographic and Atmospheric Administration, National Weather Service, National Centers for Environmental Prediction Environmental Modeling Center. This paper documents these improvements and serves as a record of the current status of the operational OPTRAN code.

The year 2000 marks the 40th anniversary of the launch of the first weather satellite. The images of cloud systems from the early satellites enabled forecasters to locate and monitor the movements of storms. Today's satellites provide a wealth of quantitative information about the constantly changing state of the Earth's atmosphere, ocean, and land surface. Significant strides are being made by operational centers around the world to effectively use these remotely-sensed observations in forecast models. The satellite measurements are used to initialize, provide boundary conditions for, and verify predictions of models. As an example of the state of the art, this paper reviews how satellite observations are used in the numerical weather and climate prediction models of the U.S. National Weather Service. The National Weather Service, National Centers for Environmental Prediction (NCEP), Environmental Modeling Center (EMC) develops regional and global weather prediction models, coupled ocean-atmosphere models for seasonal to interannual climate predictions, and a coastal ocean forecast model. A three dimensional variational data assimilation system is used to specify the initial conditions for the forecast models. Data from the following satellite instruments are currently used in one or more of these models: High Resolution Infrared Sounder (HIRS), Microwave Sounding Unit (MSU), Advanced Microwave Sounding Unit-A (AMSU-A), Geostationary Operational Environmental Satellites (GOES) sounder, GOES, METEOSAT, and Geostationary Meteorology Satellite (GMS) imagers, Advanced Very High Resolution Radiometer (AVHRR), Special Sensor Microwave/Imager (SSM/I), ESA Remote-sensing Satellite-2 (ERS-2) scatterometer, Solar Backscatter Ultraviolet Spectrometer/2 (SBUV/2), and Oceanic Topography Experiment (TOPEX) and ERS-2 altimeters.

The formulation of the National Centers for Environmental Prediction four‐dimensional variational data‐assimilation (4D‐Var) system is described. Results of applying 4D‐Var over a one‐week assimilation period, with a full set of physical parametrizations, are presented and compared with those of 3D‐Var. The linearization has been performed without simplifications and, therefore, the tangent‐linear and adjoint codes are consistent with the nonlinear physical parametrizations. The 4D‐Var assimilation is similar in formulation to the 3D‐Var analysis, except that observations are used at the appropriate time in 4D‐Var. Compared with the 3D‐Var runs, the 4D‐Var results showed good convergences, smaller analysis increments, and a comparable fit of analyses and short‐range forecasts to observations. A consistent improvement with the 4D‐Var system is observed in short‐range (six‐hour) forecasts of all model variables except the specific humidity. The temperature analyses from 4D‐Var were found to be better in most of the areas where the analysis errors from 3D‐Var were largest, although the globally averaged root‐mean‐square difference in the 4D‐Var temperature analysis was larger due to a very small degradation in some parts of the globe that include data‐rich areas. The globally averaged root‐mean‐square difference in the 4D‐Var specific‐humidity analysis, compared with that of 3D‐Var, was larger and was found to result from slightly increased analysis‐error maxima in the 4D‐Var results over data‐sparse tropical regions. The 3–4 day forecasts from 4D‐Var analyses compared more favourably than forecasts from the 3D‐Var analyses with the targeted mid‐Pacific dropwindsonde observations available from the 1998 North Pacific Experiment. Compared with conventional observations, a consistent improvement in the 1‐5 day forecasts of wind and temperature was shown in the tropics and the southern hemisphere.

The background error covariance plays an important role in modern data assimilation andanalysis systems by determining the distribution of the information in the data in space andbetween variables. A new formulation has been developed for use in the ECMWF system. Thenon-separable structure functions depend on the horizontal and vertical scales and a generalizedlinear balance operator to imply multivariate structure functions. The balance operator isincorporated into the definition of the analysis variables to ensure good preconditioning of theproblem. The formulation and structure of the background error covariance are presented, andthe implications for the analysis increments are examined. This reformulation became the operational ECMWF formulation in 3D-Var in May 1997 and in 4D-Var in November 1997.

The second workshop on the development and application of adjoint models in the field of dynamic meteorology was held in Visegrad, Hungary, from 2 to 6 May 1994. Topics included sensitivity analysis, data assimilation, stability analysis, optimal evaluation of model parameters, Kalman filtering, and limitations of the adjoint technique.

A dynamical model-based ocean analysis system has been implemented at the National Meteorological Center (NMC). This is used to provide retrospective and routine weekly analyses for the Pacific and Atlantic Oceans. Retrospective analyses have been performed for the period mid-1982 to mid-1993. The analyses are used for diagnostics of past climatic variability, real-time climate monitoring, and as initial conditions for coupled multiseason forecasts. The assimilation system is based on optimal interpolation objective analysis solved using an equivalent variational formulation. Analysis errors are estimated by comparisons to independent datasets such as temperature data from moorings and sea level information from tide gauges. In the near equatorial zone rms errors in thermocline depth are of order of 6-15 m. Comparisons of sea level estimates from the reanalyses with the records from tide gauges indicate that the rms sea level errors for monthly analysis are of the order of 0.04-0.09 m. For the weekly analyses, which potentially have more accurate forcing fields, the rms sea level errors are about 0.02-0.06 m.The analysis system can be used to infer the net heat flux at the air-sea interface on mean annual and interannual timescales. Examination of the dominant components to the oceanic heat budget shows that advection, storage changes, and the net surface heat Bur can all be of the same order of magnitude; however, frequently the net surface heat flux is much smaller than the other components. The annual variations in the components are as large or larger than the interannual variability. In the equatorial region interannual changes are of the order of 50-100 W m(-2) and act as a negative feedback to the anomalous SSTs. In the subtropics the interannual variability is only in the order of 5-10 W m(-2).Principal component analysis of the monthly analyzed ocean fields revealed an interannual sea level and SST empirical orthogonal function that has an intradecadal timescale. This mode is characterized by meridional adjustments of the thermal field. It is probably forced by the changes in the curl of the stress caused by changes in the intensity and location of the trade winds associated with the ENSO.

The use of adjoint equations is proving to be invaluable in many areas of meteorological research. Unlike a forecast model which describes the evolution of meteorological Fields forward in time, the adjoint equations describe the evolution of sensitivity (to initial, boundary and parametric conditions) backward in time. Essentially, by utilizing this sensitivity information, many types of problems can be solved more efficiently than in the past, including variational data assimilation, parameter fitting, optimal instability and sensitivity analysis in general. For this reason, the adjoints of various models and their applications have been appearing more and more frequently in meteorological research. This paper is a bibliography in chronological order of published works in meteorology dealing with adjoints which have appeared prior to this issue of Tellus. Also included are meteorological works regarding variational methods (even without adjoints) and Kalman filtering in data assimilation, plus some references outside meteorology. These additional works are included here because the main thrust for adjoint application within meteorology is currently concentrated in the development of next-generation data assimilation systems.