Kevin Donofrio AT 737 – Satellite Meteorology Literature Review March 28, 2005 Atmospheric humidity and water vapor has been a major source of interest for many satellite platforms. Because of this, there have been many types of satellite data used to retrieve water vapor profiles. Radiosonde data is currently being used by many numerical weather prediction models for moisture profiles. Radiosonde accuracy is equivalent to profiles with 1-K rms accuracy in 1-km thick layers and humidity profiles with 10% accuracy in the troposphere, and this is largely due to the fact that these are in situ measurements. Since the radiosonde network is rather sparse and costly, we seek global coverage of water vapor profiles, not only for forecasting, but for climate anaylsis. Since water vapor is the best greenhouse gas, it plays a large role in the radiation budget of Earth and the atmosphere. Satellite data is ideal for this task. Three main types of algorithms are used to retrieve water vapor profiles. Physical retrievals use first guess profiles, calculate weighting function and converge to a solution. Physical algorithms do not need large databases of radiosonde data and physical process are known during each phase of the retrieval, but they are computationally intensive. Statistical algorithms do not directly use radiative transfer equations. Regression methods are computationally less intensive, but you need a large set of coincident satellite and radiosonde data. Hybrid retrievals including optimal estimation methods which iterate a radiative transfer model subject to some constraint, and then minimize the cost function. Apriori knowledge is needed for these retrievals, and these retrievals are self diagnosing. Passive microwave satellite data has been popular for retrieving water vapor, due to the ability for global clear and cloudy sky water vapor retrieval ability. One example of satellite data used to retrieve water vapor profiles Special Sensor Microwave/Temperature-2(SSM/T-2) passive microwave brightness temperature measurements. A physically based retrieval algorithm has used this data alone and in conjunction with data from the Special Sensor Microwave/Imager (SSM/I). The SSM.T-2 is a microwave radiometer used to retrieve total integrated water vapor (TIWV) over ocean. This algorithm does not rely on radiosonde observations, and is said to be useful for measuring climate change on decadal scales and longer. Blakenship et al add that physical algorithms should perform better in areas where radiosonde data is not available. Combing SSM/T-2 data from three 183 GHz channels and SSM/I TIWV data showed an increased yield and decreased normalized brightness temperature error. Relative Humidity was found to be correlated more closely with ECMWF analyses and radiosonde measurements in lower layers (surface to 850mb, 850 to 700mb, and 700 to 500 mb). Failures of this retrieval where convergence did not occur happened in regions of increased precipitation and cloudiness. SSM/T-2 has been found (Engelen and Stephens) to perform well over ocean, but only works well in the mid-to upper troposphere with lower sensitivity than infrared measurements (I.E. TOVS, discussed later in this review). Another algorithm being used operationally is an algorithm from the National Environmental Satellite Data and Information Service (NESDIS), which includes adjustments for cloud liquid water and variable surface emissivity. This algorithm uses collocated radiosonde and satellite observations as its first guess in a regression retrieval. The NESDIS algorithm is derived from the radisonde data, which is the same data used for validation. And, an iterative maximum likelihood approach is taken to retrieve water vapor profiles from the Cross-track Infrared Sounder (CrIS). The CrIs sensor is flown on the National Polar-orbiting Operational Environmental satellite System (NPOESS) satellite. The resolution of this sensor is 15km at nadir for the moisture profile. High Resolution Infrared Radiation Sounder (HIRS) is an operational infrared sounder used to retrieve moisture and temperature profiles, as well as surface properties. HIRS is on the NOAA polar orbiting satellite and has 20 spectral channels (12 longwave, 7 shortwave and 1 visible). One technique used in the past involves a neural network technique. HIRS channels 2-8 and 11-12 are used on the HIRS sounder for the water vapor profile. The neural network dataset is constructed using profiles from the European Center for Medium-Range Weather Forecasts (ECMWF) system. Using a 3DVAR system, profiles are divided into seven groups differing by the total precipitable water vapor content of the profiles. The profiles are interpolated onto 35 presssure levels from 1000 to 0.1 hPA. A radiative transfer model simulates the HIRS channel brightness temperatures. Specifically for the water vapor profile, mixing ratio profile are derived from specific humidity and temperature profiles and collocated with the HIRS channel brightness temperatures from the radiative transfer model. The water vapor mixing ration values are from 1000 to 300hPa. Convergence tests are performed on the neural network and a mixing ratio retrieval occurs using the input test set. Retrieved values are compared to an output test set from the neural network to find the smallest rms. The output test set is the “truth”. Rms errors were found to be 2.1 g/kg at 1000mb, decreasing as you go up in the atmosphere to 1.2g/kg at 700mb, 0.7 g/kg at 600mb, and less than 0.4 g/kg above 500mb. The surface values ranged from rms error of 0.7g/kg to 3.5 g/kg. All values are consistent with ranges in past studies. Since data is available for all seasons, long term studies can be used on HIRS data and using neural network techniques. There have been physical retrievals and optimal estimation retrievals done using data from the Advanced Microwave Sounding Unit (AMSU). One algorithm used involved a forward model for calculating radiance and its Jacobian and a minimization process from the cost function. One of the key assumptions in these methods is that the errors in the observations and in the apriori information are neither biased or correlated and have Gaussian distributions, so that the best estimate of the state variable will minimize the cost function. (Liu and Weng) Liu and Weng use AMSU data from window channels at 23.8, 31.4, 89, and 150Ghz and sounding channels near 60 GHz and 183GHz and simulate the AMSU brightness temperatures at the top of the atmosphere. Results have found retrieved water vapor profiles agreeing with true profiles, with peaks occurring where clouds occur. Total precipitable water is found to be more accurate over oceans since microwave window channels over ocean are more sensitive to the lower tropospheric water. There are also surface emissivity problems over land, and therefore, microwave data needs to be coupled with other observations for better results over land. And Rosencranz uses a Bayesian approach and apriori statistics for parameters that characterize the state of the system. Moisture is obtained from the water-vapor and window channels. The modeling of the observation system is of critical importance, where again, the degree of freedom on the surface emissivity poses some errors for the remote sensing problem. Results from this approach showed fairly smooth curves, without the fine vertical structure of radiosonde profiles. Since the highest weighting function in this approach peaks between 300 and 500hPa, there are errors in dewpoint at higher altitudes. Improvements are needed for the surface model, and corrections for precipitation. Another algorithm uses the method of Engelen and Stephens (1999) to retrieve profiles of temperature and water vapor as well as cloud liquid water path and surface emissivity and is a 1DVAR system. AMSU data is compared with radiosonde data and matched up over the ocean. This algorithm uses also uses a physically based optimal estimation method similar to the one above, and uses a first guess on the water vapor profiles and surface emissivities, and minimizes the cost function to find the optimal solution. Results have found too much moisture at low levels (<850mb) and too little at high levels. And there is also the infrared TOVS water vapor retrieval from Engelen and Stephens, where an optimal estimation retrieval scheme is built, with the prime advantage being that a detailed description of the retrieval characteristics and errors can be obtained. In TOVS. water vapor profiles were limited to the upper and mid-troposphere (they also used SSM/T-2 for the lower troposphere), and the vertical resolution was limited. Their tests also worked in non-cloudy conditions. A lot of apriori information is retained in the retrievals, particularly in the lower troposphere. When they compared results to NVAP and Pathfinder, differences were found in the subsidence regions and in dry regions of the lower troposphere. Future work Any work that can be done to increase the yield is desired, particularly in cases where past retrievals have not converged. Also, retrievals of water vapor below clouds are desired, as well as in areas of precipitation, due to the fact that the retrieval problems are difficult. The problem of retrieving water vapor from the measurements is highly nonlinear even in clear atmospheres and the addition of clouds only makes it more so, particularly due to scattering. The World Meteorological organization defined that to significantly improve weather forecasting would require soundings of radiosonde accuracy. In my project, using the Atmospheric Infrared Sounder (AIRS) moisture granules and comparing with radiosonde data (which is typically supplied to the weather models), and we would expect the structure of the humidity profile to be more detailed. If we can improve our remote sensors to profile atmospheric variables with similar or better accuracy then in situ radiosondes, then we would no longer depend on the sparse radiosonde network that requires human intervention, and could have the advantage of worldwide coverage. We could then supply our models with satellite data and improve atmospheric humidity analysis and forecasting. The AIRS sensor has spectral coverage from 3.7 to 15.4 µm. The requirement of the AIRS sensor is to have the atmospheric humidity perform with accuracy of 20% in layers 2km thick in the troposphere. The goal is to have the humidity measurements perform with accuracy of 10%, which is the current accuracy rate of the radiosonde network. AIRS has 2378 channels, and has spectral resolution more than 100 times greater than previous IR sounders to provide more accurate information on the vertical profile of atmospheric moisture. Some day, coupling this data with microwave retrievals should aid in analysis of cloudy conditions. REFERENCES Atmospheric Infrared Sounder fact sheet. (n.d.). Retrieved March 10, 2005, from http://airs.jpl.nasa.gov/press/news_factsheets.html Blankenship, Clay B., Abdulrahman Al-Khalaf*, and Thomas T. Wilheit: Retrieval of water vapor Profiles Using SSM/T-2 and SSM/I Data. Journal of the Atmospheric Sciences, 57, No. 7, 939-955. Engelen R.J., and G.L. Stephens, 1999: Characterization of water vapour retrievals from TOVS/HIRS and SSM/T-2 measurements. Q.J.R. Meteorol. Soc., 125, 331-351. Forsythe, J., A. Jones and T. VonderHaar, 2004: Water Vapor Profile Retrievals from Satellite Microwave Sounding Instruments. 13th Conf. On Sat. Meteor. And Ocean. Liu, Q. and F. Weng. Physical Retrievals of Temperature, Water Vapor, and Cloud Profiles from AMSU. 13th Conf. On Sat. Meteor. And Ocean. Kidder, Stanley Q., and Thomas H. VonderHaar: Satellite Meteorology: An Introduction. 189-197. Rosencranz, P.W. Retrieval of Temperature and Moisture Profiles From AMSU-A and AMSU-B Measurements. IEEE Trans. Geosci. Remote Sensing, 39, No.11, 2429-2435. Shi, L. and J. Bates: Deriving Surface Skin Temperature and Atmospheric Profiles of Temperature and Water Vapor from HIRS Measurement Using a Neural Network Technique. 13th Conf. On Sat. Meteor. And Ocean. Wilheit, Thomas T., 1990: An Algorithm for Retrieving Water Vapor Profiles in Clear and Cloudy Atmospheres from 183 GHz Radiometric Measurements: Simulation Studies.J. App. Meteor., 209, 508-515.
Pages to are hidden for
"Literature Review"Please download to view full document