This discussion is confined to the process of deducing temperature of the atmosphere from observations of thermal radiation emitted by gases of the atmosphere; presently the most productive method of remote measurement of atmospheric temperature. This technique has evolved through experiments on three satellites of the Nimbus series (Wark, 1970; Hanel and Conrath, 1970; Smith et al., 1974) and is currently in operational use aboard the NOAA series of satellites. There are other methods of remotely sensing temperature, such as Raman backscatter (Strauch, Derr and Cupp, 1971) or molecular density measurements from. Rayleigh Scattering (Elterman, 1953), but they are not considered here.

For many years the strong contrast between radiation received from the edge of the atmosphere and from space has been used to orient spacecraft near the earth. Recently, observations and calculations of outgoing radiation have shown sufficient agreement to give confidence that such measurements could be inverted to give information on the state and composition of the atmosphere. (Much of this work is cited in Ref. 1.)

The success of vertical inversion measurements and processes from low altitude satellites has led us to speculate on the possibilities and returns of performing inversion from synchronous altitudes. We postu3ated the question as to whether one could accomplish temperature and composition inversion with a lateral resolution of about one mile square over a total field of about one hundred miles, and repeat this measurement about every hour. This would provide a continual picture of the temperature, water vapor and ozone content for specific mesoscale analysis. It would give a representation of "thermal winds" on a meaningful meteorological time scale.

Since much of the information presented in this paper is my interpretation of earth observation developments during the past twenty years, it is well that I offer some qualifications. I am an engineer of the "old school" with degrees in aeronautical and electrical engineering. I say old school because it was hammered into my head at Georgia Tech that engineers do useful things or they starve. I have been concerned during my twenty-five years of R&D experience in airborne instrumentation developments that the output of results of practical use to society has been minimal. I have become more enthused in the past nine years from my participation in the development of multispectral scanning and discrimination techniques. I believe we have finally developed a useful new remote sensing tool. As you might imagine I am biased toward aircraft as a widely useful earth observation platform for optical scanners. However, I see the merit of other platforms and I am an advocate of using appropriate tools for a particular job.

While Mr. Hasell's argument that optimal earth resource inventory and monitoring systems will necessarily require data from several sources is a sound one, a comment on the use of the phrase "multistage approach" is perhaps in order. The basic problem is to differentiate between two real and significantly different, though complementary, approaches to quantity estimation. The first method involves the use of several levels of information--satellite imagery, low and high resolution aircraft data in resource areal extent classification inventories calibrated to known ground data models. The second estimation technique ultilizes all data levels in an integrated, randomized sample design. The former approach is that described by Mr. Hasell while the latter embodies sample plans such as introduced in William G. Cochran's Sampling Techniques (Second Edition, John Wiley & Sons, Inc.).

An important aspect in designing scanners for earth resources users is the specification of signal to noise and resolution size which is compatible with the cost and space-craft burden that can be supported by the program. There are two areas of conflict between these performance factors. First, either resolution or noise can be improved only at the expense of the other within each cost/ burden ceiling. Second, emphasis on the importance of the two factors differs according to whether the user depends on visual examination of pictures or on automated signature analysis of the multispectral data.

Optical observation of the earth from a satellite has many useful applications, in fact, far too many to be satisfied by any particular embodiment of a scanning system. Each potential user -the cartographer, the urban development analyst, the agronomist, the water pollution sentry, the forester - has his own specific requirements for spectral bands, spatial resolution, sampling frequency, response time, and mapping fidelity. Therefore, instead of seeking a general purpose scanner, we must look for versatility at the design level. The particular approach described here, the roof wheel scanning technique*, leads to a rather large family of instruments applicable to satellite, airborne and other operations. (Ref. 1)

Earth resource scientists are demanding greater classification capability from remote sensing instruments. The trend is for more spectral bands with finer spectral resolution and improved spatial resolution. These requirements greatly impact instrument design, particularly since data bandwidth is increased proportionally. An instrument concept is described which, from a spacecraft platform, provides seven band coverage, 30 meter ground resolution and 100 nautical miles swath width. Its 80% scan efficiency yields a near minimum data transmission band-width.

During operation, space-borne sensor system calibrations may drift away from preflight measured values. Present sensor internal radiometric calibration systems are also subject to error and drift. Natural targets can be used to determine temporal radiometric response changes. In addition, use of common targets having uniform reflectance and thermal emittance permits cross-checking the responses of different sensors.

The corrections are presented for the visible and near infrared spectrum. The specifications of earth-atmosphere models are given. Herman's and Dave's methods of computing the four Stokes parameters are presented. The relative differences between the two sets of values are one percent. Such precision is insufficient for many remote sensing applications, since satellite measurements of radiance will soon have SNR exceeding 100. The absolute accuracy of the computations can be established only by comparisons with measured data. Suitable observations do not yet exist. Nevertheless, comparisons are made between computed and aircraft and satellite measured radiances.

Variations in sun angle and haze level change the spectral signatures collected by multispectral scanners (MSS). This paper describes methods and computer programs that have been developed to simulate the effect of such variations and to correct for them. Simulation and correction are really the same process since correction consists of simulating a different haze level or sun angle than is actually present when the data was taken.

Variations in the relative value of the blue and green reflectances of a lake can be correlated with important optical and biological parameters measured from surface vessels. Measurement of the relative reflectance values from color film imagery requires removal of atmospheric effects. Data processing is particularly crucial because: (1) lakes are the darkest objects in a scene; (2) minor reflec-tance changes can correspond to important physical changes; (3) lake systems extend over broad areas in which atmospheric conditions may fluctuate; (4) seasonal changes are of im-portance; and, (5) effects of weather are important, precluding flights under only ideal weather conditions. Data processing can be accomplished through microdensitometry of scene shadow areas. Measurements of reflectance ratios can be made to an accuracy of ±12%, sufficient to permit monitoring of important eutrophication indices.

During the Skylab missions detailed ground measurements of atmospheric effects and target characteristics, at selected targets, were made in order to quantitatively evaluate the performance of the Earth Resource Experiment Package (EREP) remote sensors. The ground measurements consisted of absolute measurements of: (1) incident total (global, 180 F.O.V.) solar irradiance 400 to 1300 nm; (2) diffuse (scattered) sky irradiance (180° F.O.V.) - 400 to1300 nm; (3) normal incident direct solar beam irradiance - 400 to 1300 nm; (4) reflected solar target radiance - ERTS bands*; (5) target reflectivity - 400 to 1300 nm and ERTS bands; (6) target thermal (8 to 14 μm) brightness temperature; and (7) atmospheric temperature and humidity - profiles from ground level to 15,000 ft. A.S.L.

Since its launch in July 1972, ERTS has generated multispectral data and images of a sizable portion of the world. While much analysis has been done by photointerpretation of the imagery, computer-implemented analysis of the digital tape data is increasing. At the same time, because of the large area coverage and synoptic view of the ERTS system, investigators are processing data from larger areas--in some cases up to several ERTS frames. The combination of detailed computer analysis and large area coverage has focused investigators on all factors which cause the data to vary--sensor system stability, atmospheric and solar illumination effects, and variations. in the reflectance of scenes, e.g., because .of phenological differences. To successfully conduct large area surveys which will demonstrate the ultimate utility of the ERTS system, all these factors need to be accounted for.

For more than a decade, satellite-borne instruments, in programs such as TIROS, Nimbus, and ATS, have been providing information about the earth and its atmosphere. Without satellites, much of this information could not be derived as data must be obtained from remote and inaccessible areas, large area coverage, and/or through timely and rapid assessment of dynamic events.

Solid-state is a popular phrase today for its use indicates improvements in reliability and system efficacy. The purpose of this paper is to identify the systems engineering aspects of applying solid-state technology to earth observation applications that have traditionally been performed by point (or multiple-point) .detector line scanned mechanisms. The paper shows that the translation from a basically serial data flow point (or multiple-point) detector mechanically-scanned sensor to a solid-state highly parallel linear-array pushbroom sensor results in the minimization of mechanical complexity, but it maximizes electronics complexity, and for some applications, it also places increased demands upon optical performance. The paper discusses technical aspects indigenous to highly parallel photodiode linear array pushbroom applications. Examples of systems engineering applications are included. The applications are for a high resolution (10-m ground sample distance) narrow swathwidth (50 km), offset pointing sensor and a coarse resolution (80-m ground sample distance), wide swathwidth (185 km) sensor both operated in four bands from 0. 5 μm to 1.1 μm.

As will be pointed out many times in the next few months, the Earth Resources Technology Satellite (ERTS-1) has now completed two years of successful operation in orbit as of July 23, 1974. This satel-lite provides two sensor systems capable of imaging the Earth's surface with nominally 100m ground resolution across a 185 Km swathwidth. In addition, accurate relative radiometry is provided in four spectral bands by one of these ERTS systems known as the Multispectral Scanner (MSS). The four spectral bands cover the visible to near-IR (.5 -.6; .6 -.7; .7 -.8; and .8 -1.1 micrometers). The high quality of the ERTS-1 MSS imagery provides a known standard to which future systems can be compared.

The Fiscal Year 1973 phase of the Man-ned Earth Observatory (MEO): Mission Requirements and Preliminary Design Study was recently completed for NASA/MSFC under Contract NAS-28013. The study was performed by TRW Systems Group and Aerojet Electro Systems Company. This phase of the study was concerned with the use of the Space Shuttle in earth observation sensor development. The MEO was considered to be a secondary or "piggyback" Shuttle/Spacelab payload. The specific objectives of the study were to: Gather detailed information on two sensor class development programs and present these as case histories Assume the Shuttle existed at the beginning of these programs and discuss how sensor development would have been conducted utilizing the Shuttle/Spacelab Estimate the feasibility of using the Shuttle/Spacelab for sensor development and the potential savings in schedule and development cost that could be realized.