The determinations of cloud presence, properties, and radiative influence are high priority requirements for both defense related operations and in the pursuit of research goals for the U.S. Global Change Research Program. A number of current programs address aspects of the remote sensing of clouds; this paper provides an overview of these programs, their objectives, requirements, and interrelationship. Requirements are divided into three categories: (1) impacts on defense operations and systems, (2) numerical weather prediction (NWP), and (3) climate study objectives, both prognostic and diagnostic. Additionally, current satellite sensor data resources and applicable cloud property retrieval algorithm approaches are described.

Global area coverage (GAC) pixel data from the polar-orbiter advanced very high resolution radiometer (AVHRR) and the cloud classifications for 2 X 2 pixel arrays, output from CLAVR Phase-I, constitute the input for the CLAVR Phase-II analysis. The total cloud amount product from Phase-I was based on the assignment of fractional amounts of 1.0 and 0.5 to every pixel labeled cloudy or mixed, respectively, but the latter assignment led to biased total cloud amounts near 50%. In Phase-II, radiometric cloud amounts are determined for the mixed pixels from 11 micrometer radiances within a mapped grid area. The Phase-II procedure was developed for the identification and stratification of overcast (OVCST) cloud types, according to the cloud classifications of Phase-I, in grid areas also containing subsets of mixed and clear pixels. All mixed pixels associated with an observed OVCST type are analyzed statistically to determine fractional amounts of each cloud type (layer) as well as the total cloud amount within the grid area. The bias of Phase-I estimates was eliminated.

One of the primary interests in digital image processing is the development of robust methods to perform feature detection, extraction, and classification. Until recently, classification methods for cloud discrimination were mainly based on the spectral information of the imagery. However, because of the spectral similarities of certain features (such as ice clouds and snow) and the effects of atmospheric attenuation, multi-spectral rule based classifications do not necessarily produce accurate feature discrimination. Spectral homogeneity of two different features within a scene can lead to misclassification. Furthermore, the opposite problem can occur when one feature exhibits different spectral signatures locally but is homogeneous in its cyclic spatial variation. The exploration of spatial information is often advantageous in these discrimination problems. A texture-based method for feature identification has been investigated. This method uses a set of localized spatial filters known as two dimensional Gabor functions. Gabor filters can be described as a sinusoidal plane wave within a two-dimensional Gaussian envelope. The frequency and orientation of the sine plane and the width of the Gaussian envelope are determined by the Gabor parameters. These tunable channels yield joint optimal information both in the spatial and the frequency domains. The new method has been applied to the thermal channels of the NOAA-advanced very high resolution radiometer (AVHRR) data for cloud type discrimination.

A multi-year research and development program is underway to develop an automated cloud model known as TACNEPH for use by the Air Force at tactical sites. Significant features of this model include the ability to analyze real-time DMSP/OLS and NOAA/AVHRR data using only the limited resources of transportable tactical ground stations and to automatically adapt to changes in the available data mix. No supporting data from a center are available (e.g., upper air temperature and moisture fields, surface reports). To satisfy these requirements it was necessary to develop separate algorithms for each sensor platform. An infrared algorithm developed for DMSP data relies on an estimate of the clear scene radiative brightness temperature based on a dynamic correction to a surface temperature climatology. A separate NOAA IR algorithm is an adaptation of the multispectral approach of Saunders and Kriebel. Both algorithms are designed to improve cloud detection capabilities over the current Air Force operational RTNEPH model, with particular emphasis on low cloud. A major aspect of the TACNEPH development program is the validation of the cloud algorithms over globally varying conditions. Since there is no universally accepted source of ground truth data for cloud, it was necessary to develop a validation procedure based on available data sources. Algorithm validation is based on subjective man/computer analysis of the input satellite sensor data using any available additional data sources as guidance.

Results are presented of the analyses of the 3D structure of the atmosphere with TOVS sounding data for two atmospheric situations over Europe. Both situations include the occurrence of thunderstorms. The retrieval methods used are the International TOVS Retrieval Package (ITPP-4) and the Improved Initialization Inversion (`3I-2') algorithm. These methods are applied to locally received real-time direct readout HRPT data. Both methods show meso scale phenomena in the temperature fields, represented by the thicknesses 500/1000. Radiosonde observations, as well as model analyses, do not show this phenomenon. The synoptic structure of the temperature field as well as the humidity field is comparable with all methods. Thus it can be concluded that there may be additional information in satellite derived atmospheric data, also on a scale at about meso-alpha/meso-beta. Statistics show that the structures of the fields are comparable at 98 - 99% correlation coefficient. This result can be improved by not using cloudy retrieval boxes. Collocations with radiosondes show a good agreement within the troposphere, thus a good simulation of stabilities can be expected by satellite data.

The influence of cloudiness on the shortwave radiation budget at the top of the atmosphere, at the surface, and, as a residual, for the atmosphere is investigated for an area of +/- 60 degree(s) longitude and latitude. The data used is exclusively taken from satellite measurements. Calculations for the top of the atmosphere are based entirely on measurements of the Earth Radiation Budget Experiment (ERBE). For the solar radiation budget at the surface, the incoming solar radiation at the surface is derived from Meteosat data and the surface albedo is calculated from ERBE clear sky planetary albedo measurements by applying an atmospheric correction scheme. As results, maps of absorbed solar radiation for the total earth-atmosphere system, the surface, and for the atmosphere are presented. To infer the contribution of clouds, the concept of cloud radiative forcing is applied to these different data sets.

In several instances during the International Cirrus Experiment (ICE) 1989, cloud types were detected by multisensor satellite data over the North Sea. The first cloud classification scheme is based on the maximum likelihood method using NOAA AVHRR and Meteosat data. The second is an algorithm using a combination of Meteosat and SSM/I data. By comparing these results together and with synoptical observations, good agreement can be achieved. The discrepancies can be explained either by time delay or different spatial resolution. Comparing the monthly mean cloudiness inferred from NOAA AVHRR-data with ISCCP C2 data, it seems that the ISCCP C2 results underestimate the real cloudiness for the North Sea area (approx. 55 degree(s) N latitude, 5 degree(s) E longitude). To determine the influence of clouds on the earth radiation budget and on the climate the cloud-climate efficiency was used. This index is similar to the cloud forcing, but is valid for an individual classified satellite image pixel. The cloud forcing is the sum of the cloud-climate efficiencies over an area, i.e., the heating or cooling of the earth/atmosphere system can be estimated. Using NOAA-AVHRR data the annual cycle of cloud forcing at top of atmosphere was calculated for the North Sea.

In order to get a better understanding of the influence of clouds on the Earth's energy budget, one needs a cloud classification taking into account cloud height, thickness, and cloud cover. The radiometer ScaRaB (scanner for radiation balance), launched in 1993, has in addition to the two broad-band channels (0.2 - 4 micrometers and 0.2 - 50 micrometers ) necessary for earth radiation budget (ERB) measurements, two narrow-band channels (0.5 - 0.7 micrometers and 10.5 - 12.5 micrometers ) in order to improve cloud detection. Most automatic cloud classifications have been developed with measurements of very good spatial resolution (200 m to 5 km). Earth radiation budget experiments, on the other hand, work at a spatial resolution of about 40 km (at nadir), and therefore we investigated a cloud classification algorithm adapted on this scale. The algorithm is based on the dynamic clustering method and uses co-located AVHRR-ERBE data, simulating the ScaRaB measurements. This cloud field classification is compared on one hand to results obtained by a well tested threshold algorithm using AVHRR (advanced very high resolution radiometer) measurements at reduced spatial resolution of 4 km and on the other hand to cloud parameters extracted from HIRS (high resolution infrared sounder)/MSU (microwave sounding unit) data. We find that classification of cloud fields is still possible at a resolution of 40 km, and by combining AVHRR, ERBE, and HIRS/MSU measurements one can undertake interesting studies on the influence of different cloud fields on the Earth radiation budget.

The scanner for radiation budget (ScaRaB), a satellite born radiometer for earth radiation budget measurements, is briefly described. The instrument has been calibrated versus the sun by means of a reference diffusor and a pyrheliometer. ScaRaB, the diffusor, and the pyrheliometer as well as a spectrometer are integrated in a newly developed ground calibration unit (GCU). During February of 1993, a successful solar ground calibration campaign took place in Inzana, Tenerife. First results are presented.

The Earth Radiation Budget Experiment (ERBE) consists of three satellites, each carrying a set of three scanning and four nonscanning Earth-viewing radiometric channels. These instruments are used to monitor the components of the Earth's radiative energy budget, which include reflected solar radiation and Earth-emitted radiation. The current paper deals with high-order modeling of the nonscanning channels. The nonscanning instruments are active cavity radiometers which view the Earth with two fields of view, medium and wide, and in two wavelength intervals, visible (0.2 to 5.0 micrometers ) and total (0.2 to 50.0 micrometers ). The visible channels are equipped with a hemispherical fused silica filter dome which absorbs long wave radiation. The optical front-end of the nonscanning instruments, including the field-of- view limiter and the filter dome for the visible channels, is modeled using the Monte-Carlo ray-trace method. Work in progress includes a finite difference thermal diffusion model of the filter dome. When completed, this model will be integrated with an existing dynamic electrothermal model of the total channel. This will permit studies of the contamination of the cavity signal due to emission and scattering of radiation from the field-of-view limiter and filter dome. Reported are results for the optical and thermal radiative behavior of the optical front-end of the ERBE WFOV total and visible channels.

The Earth Radiation Budget Experiment (ERBE) consists of a suite of three scanning and four nonscanning radiometric instruments on each of three satellites which monitor the solar- reflected and Earth-emitted components of the Earth's radiative energy budget. A numerical model has been formulated to study the dynamic behavior and equivalence of the ERBE scanning thermistor bolometer radiometers. The finite difference method is applied to the detector of the ERBE scanning radiometer to characterize its thermal and electrical dynamic behavior. The thermal analysis confirms the thermal time constant of the instrument claimed by the vendor. The model reveals that the instrument can be very sensitive to spatial variations of the incident thermal radiation. However, the analysis confirms that the hypothesis of equivalence is justified for viewing typical Earth scenes. The high-order numerical model described in this paper is a key component of the end-to-end dynamic electrothermal model under development for the ERBE and CERES scanning radiometric channels.

In recent yecas much attention has been devoted to the effect that clouds have on the radiative exchanges that take place in the Earth-atmosphere system. Solar arid thermal nonlinear radiation feedback processes that take place among layered clouds and the Earth's surface (water, land, snow, arid ice) play a central role in influencing and moderating global climatic change. The current consensus among cloud-climate researchers is that they have a net cooling effect on climate, but to how great a degree is as yet in question.

Clouds have significant effects on the radiative exchanges that take place between the Earth's suxface and its atmosphere. The role of clouds in influencing and moderating climatic change is central to the complex interaction of the sun's incoming radiation with the Earth-atmosphere system. In general, the current consensus among climate researchers is that clouds have a net cooling effect on the Earth's climate, but just how much of cn effect is still veiy much in question (Arking, 1991). Among the factors detennining the effect of clouds on climate are cloud fraction, altitude, and optical depth. These factors are considered essential to the cloud radiative feedback processes that take place in the Earth-atmosphere system. In order to properly assess the influence of clouds on climate, global measurements of cloud are needed.

Trends in global upper tropospheric semi-transparent cirrus cloud cover are beginning to emerge from a three year cloud climatology compiled using NOAA polar orbiting HIRS multispectral infrared data. Cloud occurrence, height, and amount have been determined with the CO₂ slicing technique on the three years of data (June 1989 - May 1992). Annual, seasonal, and geographical trends of cloudiness are presented. To date, semi-transparent clouds are found in more than one third of the observations. Large seasonal changes are found in areas dominated by the ITCZ, the sub-tropical high pressure systems, and the mid-latitude storm belts. Semi-transparent clouds associated with these features move latitudinally with the seasons. More thin clouds (effective emissivity less than .50) are found in the tropics than in midlatitudes. Global average of all clouds (semi-transparent and opaque) is about 75%; more clouds are found over the oceans than over land.

Cirrus clouds were the focus of an intensive field study in Kansas in November of 1991. During this period, measurements of the downwelling radiation from cirrus clouds in the water vapor rotational band (18 - 28 micrometers ) were made from the NCAR King Air Research Aircraft. The instrument, the SSH-2, is a 17 channel passive radiometer with seven channels spanning this spectral region. Measurements were made under a variety of atmospheric conditions ranging from clear to opaque multi-layered cloud systems. In this paper, two King Air flights were examined: one under clear conditions and one during a cirrus event. Calibration of the instrument was completed for each flight and the relative difference between measurements during the two cases are presented. The latter case included flight tracks below and through the cirrus clouds. It is clear from the measurements that a significant difference exists between the measurements during the cirrus event and under clear skies. Based upon model estimations these differences may be used to determine certain microphysical properties of the cirrus cloud.

We have developed a retrieval scheme for the inference of the emissivity and temperature of cirrus clouds using the data obtained from the advanced very high resolution radiometer (AVHRR) 3.7 and 10.9 micrometers channels. The scheme, which is applicable to the entire day, is based on the numerical solution of a set of nonlinear algebraic equations that are derived from the theory and parameterizations of infrared radiative transfer. The solution involves an effective extinction ratio for the two wavelengths, which is dependent on ice crystal size distribution. Based on radiative transfer calculations and cloud physics observations, the cloud optical depth and the ice crystal size distribution in terms of mean effective size can be determined from cloud emissivity and cloud temperature, respectively.

A review of the language and parameters that define subvisual cirrus is provided, including a literature review and classic examples of observations. Complementary measurement techniques are discussed, including lidar, solar aureole measurements, photographic techniques, and the use of radiometric observations. We address the microphysical and optical properties of these high, thin clouds, along with suggested dynamic sources. Five classes of subvisual cirrus clouds are suggested. Complementary subvisual cirrus observations recorded during the 1986 and 1991 FIRE (First ISSCP Regional Experiment) intensive field operations and other field programs support these classes. Detection of subvisual cirrus via satellite imagery is vital and a limiting case for discrimination or threshold methods. The lidar, satellite, rawinsonde, and meteorological conditions are used to derive a likely picture of the atmospheric conditions required to produce the subvisual cirrus class observed.

The accurate identification of clouds in meteorological satellite imagery by automated detection and classification algorithms is critical to environmental remote sensing studies, such as those related to Global Climate Change. Significant improvements in these algorithms were realized with the arrival of multispectral, meteorological satellite imagery, collected by NOAA's advanced very high resolution radiometer (AVHRR). However, deficiencies remained, especially with the positive identification of optically thin cirrus clouds due, in part, to the effects of atmospheric attenuation on cloud signatures caused primarily by variations in water vapor. Thus, the goal of this research was to enhance the accuracy of the automated classification of optically thin cirrus in nighttime, multispectral meteorological satellite imagery through an improved treatment of atmospheric attenuation caused by moisture.

Thin cirrus clouds are difficult to detect in visible and thermal infrared images, particularly over land. Using spectral imaging data measured with the airborne visible/infrared imaging spectrometer (AVIRIS) from an ER-2 aircraft, it has been found that narrow channels close to the center of the strong 1.38 micrometers water vapor band are very effective in separating thin cirrus clouds from clear surface areas. Due to the total absorption of solar radiation by atmospheric water vapor, pixels that do not contain cirrus clouds or stratospheric aerosols are black in images around 1.38 micrometers . Pixels containing cirrus clouds appear white in these images because of the scattering of solar radiation by cirrus clouds. We have selected a near- IR channel centered at 1.375 micrometers with a width of 30 nm for the moderate resolution imaging spectrometer (MODIS) for remote sensing of cirrus clouds from space. This channel may also be useful for remote sensing of stratospheric aerosols when cirrus clouds are absent.

The surface temperature retrieved from thermal infrared satellite data in the presence (either known or unknown) of cirrus clouds can be in error. This error results from the absorption and scattering by the cirrus cloud of the surface and intervening atmospheric radiation. Using parameterizations of standard and subvisual cirrus cloud micrometeorology (e.g., particle size distributions, density) based on recent observations, the effects of cirrus clouds have been modelled to determine the range of temperature errors. MODTRAN is the primary tool used in this evaluation to establish the temperature difference over the 3 - 14 micrometers spectral region as a function of: optical depth, asymmetry parameter, scattering albedo, cloud temperature, cloud base and thickness, and ground temperature. The advantage to computing temperature differences as a function of scattering albedo and asymmetry parameter is to permit these results to be applied when a more accurate model of scattering by cirrus crystals has been developed. Temperature errors for realistic cirrus clouds may then be derived from the albedo and asymmetry parameter implied by this model.

Extremely low column density ice clouds (subvisual cirrus) have significant impacts on passive IR remote sensing of both the atmosphere and the Earth. In this paper I review the physical properties of subvisual cirrus and discuss their detection and global distribution.

The Michelson interferometer for passive atmospheric sounding (MIPAS) is one of the ESA- developed instruments for the First European Polar Platform, which is planned to be launched in 1998. MIPAS is a limb sounding Fourier spectrometer measuring the thermal emissions of the earth's atmosphere from the upper troposphere to the lower thermosphere (5 - 150 km) in the 4.15 - 14.6 micrometers spectral range. This paper describes the optical modules of MIPAS where special care has been taken to reduce straylight and thermal self-emission. An important design driver for the material choice is the temperature difference between integration (at room temperature) and verification (at 170 K). Major design drivers for spectral and radiometric stability are the temperature variations over the orbit and the 4-year lifetime.

The Michelson interferometer for passive atmospheric sounding (MIPAS) is a high resolution Fourier-transform spectrometer to measure concentration profiles of stratospheric constituents on a global scale. It observes the atmospheric emissions from the earth horizon (limb) throughout the thermal infrared region (685 - 2410 cm^-1, 14.6 - 4.15 micrometers ), which allows the simultaneous measurement of more than 20 key trace gases, including the complete NO_y-family and several CFCs. MIPAS is selected as an ESA-payload for the Envisat-1 project of the Polar Orbiting Earth-observation Mission POEM. Launch is planned for 1998.

The next-generation Defense Meteorological Satellite Program (DMSP) satellites (Block 6) will include a minimum of five bands in the near- to long-wave infrared for passive remote sensing of clouds and the atmosphere, as well as the oceans and the land surface. Atmospheric temperature and humidity sounding are performed with microwave sounders, as with the current Block 5 system. Numerous hydrological variables are measured with a passive microwave imager, the data from which are synergistic with the infrared and visible bands. As a goal, the microwave imager provides contiguous swaths at the equator on successive orbits, to provide improved coverage of tropical storms. The present program and access to its data are also described, as is the close relationship of DMSP data to problems in global change.

Spectrally resolved radiometric measurements of middle infrared atmospheric emission can be used in conjunction with detailed radiative transfer calculations to retrieve cloud emissivity, and to estimate cloud liquid water path (LWP), optical depth, and equivalent radius of the droplet size distribution. Using a discrete-ordinates radiative transfer formulation, an algorithm has been developed to retrieve these cloud properties from FTIR data. The algorithm has been successfully applied to a four month Antarctic data set provided by the CalSpace FTIR Spectroradiometer. Radiative transfer calculations were performed to estimate spectral cloud emissivity for a range of cloud optical depth, liquid water content, and equivalent radius, sufficient to bracket values expected in the field. These calculations made use of bi-modal droplet size distributions actually observed in Antarctic clouds. Using a least- squares algorithm, a theoretical cloud emission spectrum is chosen which best reproduces a given measured brightness temperature spectrum. The results show marked differences in cloud emissivity between high and low overcast layers, and between clouds with and without precipitation. The results also suggest that the emissivity of a maritime Antarctic cloud deck should be smaller for a given LWP than the parameterization frequently used in general circulation models.

We attempt to bound microphysical properties of cirrus clouds via complementary satellite, rawinsonde, and lidar data analysis and radiative transfer modeling. Data acquired during the 1991 FIRE (First ISSCP Regional Experiment) Cirrus IFO (intensive field operation) includes AVHRR (advanced very high resolution radiometer) LAC (local area coverage), satellite imagery, high temporal resolution rawinsonde data, and lidar backscatter and depolarization information. We use the complementary rawinsonde and lidar data to generate profiles of the atmosphere, to place clouds at the correct height, and to verify the complex mixed phase nature of high altitude cirrus. Using the DISORT radiative transfer model (Stamnes et al., 1988), we generate brightness temperatures for a range of optical depths for single- and/or multi-level cloud systems composed of water or ice spheres or ice hexagons.

A method is presented to remove the thermal component of near-infrared radiances observed from geosynchronous satellites. The resulting solar component is then used to retrieve the effective radius of persistent marine stratiform clouds.

A bispectral approach for temperature determination of semitransparent clouds is applied to identify areas covered by Polar Stratospheric Clouds (PSCs). The method is based upon the information obtained from two satellite pixels and two spectral channels of the passive HIRS instrument onboard the NOAA satellites. Since PSCs form and emit in very cold stratospheric environments (< 195 K) solutions of the bispectral method yield low temperatures in a range that clearly differs from those derived for a cold but cloudfree stratosphere. Application of the method to NOAA satellite data gained during the major PSC event on January 31, 1989 demonstrates that large areas covered by dense PSCs are found with a horizontal distribution in very good agreement to independent airborne and satellite measurements.

Global scale satellite observations of cloud properties are a key to unraveling the physical processes that govern the role clouds play in the climate system. The retrieval of cloud particle size from satellite imagery data is of particular interest. Cloud emissivities and transmissivities depend substantially on the constituent single particle scattering and absorbing properties which in turn depend on particle size. Two clouds with the same ice water content but different particle sizes may have substantially different optical depths. This study presents a retrieval scheme for obtaining cloud droplet/particle size from satellite measurements at infrared wavelengths. The scheme employs Mie theory and the Eddington approximation and uses the advanced very high resolution radiometer (AVHRR) data at 3.7, 11, and 12 micrometers . It also relies on the nonlinear relationships between 3.7, 11, and 12 micrometers radiances observed for single-layered semitransparent cloud systems. Theory indicates that the nonlinear relationships disappear when the spherical particle radius reaches about 60 micrometers . An automated retrieval procedure has been developed. Some retrieved results are presented.

We have developed a simple analytical approximation for the infrared attenuation and emissivity of clouds and fogs composed of water spheres, based on a polynomial approximation to the Mie attenuation efficiencies. The attenuation coefficients can be obtained in simple analytic form as a function of the liquid water content and two parameters characterizing the size distribution of the water droplets, such as its effective radius and effective variance. The coefficients representing the approximation can be precalculated and stored in a form suitable for use in the inversion of remotely sensed cloud radiances, climate modeling, and numerical weather prediction. The resulting accuracies are within a few percent when compared with exact Mie calculations and integration over the size distribution, while the computational burden is reduced by several orders of magnitude. The radiances from clouds, as well as the cloud reflectance and emittances, calculated with radiative transfer theory using these approximations for the cloud scattering and absorption properties are compared with calculations using exact Mie scattering results. These comparisons show that the radiances, using the new approximations, generally are calculated with greater accuracy than might be expected on the basis of the accuracy of the approximations for the separate parameters.

A multiple-scattering model to simulate high-spectral resolution observations is discussed and examples presented.

Measurements of the IR optical thickness and radar reflectivity of cirrus clouds can be used to infer microphysical characteristics of these clouds. This paper presents a technique for estimating the optical thickness from spectral measurements of downwelling radiance of cirrus clouds in the infrared atmospheric transparency `window' near 10 micrometers . To illustrate, results of cirrus particle size retrieval from the FIRE-II experiment are given.

Traditionally the temperature dependent characteristics of the CO₂ band structures have been used by infrared atmospheric sounders to derive the temperature profiles of the troposphere and stratosphere. Since individual pressure/Doppler broadened CO₂ line structures are also temperature dependent, it is possible to derive the temperature profiles of the atmosphere by resolving single CO₂ spectral lines with high spectral resolution infrared sounder. In this paper, we first compare the techniques of deriving atmospheric temperature profiles from the CO₂ band structures and the pressure/Doppler broadened spectral line structures. Then, we discuss a new IR sounder concept based on the high resolution Fabry-Perot interferometer, which derives the temperature profiles of the atmosphere by resolving individual CO₂ spectral lines. Retrieval simulations using two different techniques are conducted to evaluate its performance relative to other IR sounders.

The vertical resolution capabilities of the VISSR atmospheric sounder (VAS) and the proposed GOES high-resolution interferometer sounder (GHIS) were considered with regard to temperatures at and near the ground surface. Simulated retrievals were performed, along with experiments on the sensitivity of radiances to profile perturbations. For a moderately moist atmosphere, ground surface temperature errors of about 1 degree(s)C and low-level air temperature errors of about 3 degree(s)C can occur in VAS retrievals due specifically to vertical resolution limitations and instrument noise. For a drier atmosphere the surface temperature errors tend to be smaller and the low-level air temperature errors tend to be larger. It appears that these vertical-resolution-related retrieval errors can be reduced by a factor of about 70 - 90% by going from VAS to an instrument with performance specifications such as those of GHIS. These results also imply that a high-spectral-resolution instrument can perform significantly better than VAS in the task of estimating cloud top heights and temperatures for low clouds.

From a pattern recognition technique applied on GOES-7 full disks, humidity profiles defined at six standard levels (300 - 1000 mb) are retrieved on 1 degree(s) X 1 degree(s) areas in all weather/cloud conditions. Using one of two retrievals in each direction, about 3000 profiles are assimilated every six hours from 60 degree(s) N to 60 degree(s) S, 50 degree(s) E to 170 degree(s) W in the Canadian global forecast model. Slightly improved forecasts of geopotential are obtained over North America. Significant differences are noted locally in the forecast instantaneous and accumulated precipitation, especially in the tropics. Both analyses and forecasts of low level stratocumulus which are extensive west of continents are clearly improved. The system became operational in April 1993 and will soon be extended to Meteosat-3 now at 75 degree(s) W; the two geostationary satellites provide humidity profiles on 40% of the globe four times daily.

Several methods exist for determining the presence of aerosol, and its optical thickness, from satellite visible and infrared observations. This paper suggests a new approach, physically based on the spectral variation of the index of refraction in the 8 - 12 micrometers region. The tri- spectral technique makes use of observations at wavelengths of 8, 11, and 12 micrometers . Observations from the AVHRR and HIRS/2 instruments demonstrate the potential of the technique.

During the last six years, extensive observations of atmospheric emitted radiance in the spectral region from 3.6 - 20 micrometers with resolving powers of 1000 - 4000 have been made, both from the ground and nadir viewing from NASA high altitude aircraft. Two recent field experiments in which both instruments participated are the FIRE II/SPECTRE experiment Nov. - Dec. 1991 in Coffeyville, KS and the STORMFEST experiment Feb. - Mar. 1992 in Seneca, KS. Experience with these instruments has led to instrument designs for advanced sounders on geostationary and polar orbiting satellites. Applications include remote sensing of atmospheric temperature and water vapor for improved weather forecasting, measurement of cloud radiative impact for improvement of global climate modelling, and trace gas retrieval for climate and air pollution monitoring.

Model calculations of upwelling spectral radiances at aircraft and satellite altitudes have been made to assess the capability of different current and planned sensors to extract information on the atmospheric aerosols. The visible and near infrared channels on the AVHRR, CZCS, and SeaWiFS satellite sensors were used, as well as hypothetical multichannel instruments covering 400 - 1000 nm with bandwidths of 100, 20, or 10 nm. The sensitivity to the aerosol and environmental properties increased as the bandwidth of the channel decreased.

Coincident lidar and satellite observations of thin and subvisual cirrus were collected to determine the probability of cirrus detection as a function of optical depth for several satellite systems. Satellite observations include those from DMSP (smooth), NOAA Polar Orbiter (GAC), GOES, GOES VAS and NOAA HIRS processed with the CO₂ slicing algorithm, and the RTNEPH. Different cirrus cloud detection techniques, namely, those of the RTNEPH, manual detection, Phillips Laboratory's (PL) multispectral image analysis scheme, and the CO₂ slicing algorithm were applied to the lidar-coincident satellite data. Each satellite image was examined for evidence of cirrus clouds at the lidar location. The binary (yes/no) results were then used in a nonlinear regression technique to determine the probability of detection as a function of optical depth. The results show that the VAS and HIRS data processed with the CO₂ slicing algorithm detected thin cirrus most of the time with probability of detection (POD) of 91% and 75%, respectively.