Elements of Change 1995: Session I Emery

Remote sensing of the Earth began with the advent of aerial photography and in a relatively short time, advanced to the mounting of cameras in rockets. In 1957, the Russian satellite Sputnik demonstrated the viability of operating systems in low Earth orbit (LEO). This began a series of meteorological satellites designed to monitor cloud conditions on Earth. Known as the Television Infrared Observing Satellite (TIROS), the first U. S. space craft in the early 1960s was spin stabilized with its camera pointing towards Earth only for the latitudes of North America. A change in the spin axis led to the series of "wheel" satellites with cameras pointing outwards giving daily global coverage.

AVHRR data lends itself to the study of many land surface processes, including monitoring snow pack for river runoff, detecting/monitoring forest fires and assessing the extent of vegetation damage, and monitoring the status of large areas of terrestrial vegetation.

In the early 1970s, ballistic missile guidance technology was incorporated into LEO satellites making it possible to operate a 3-axis stabilized spacecraft. This was accomplished through a considerable increase in weight (from ~300 to over 700 lbs. (136 to over 317 kg). Today, LEO environmental satellites typically weigh over 3,000 lbs. (1,360 kg) and carry a multitude of instruments including imagers, radiometric profilers, space environment monitors, and data collection systems. A summary of the presently operating Earth observing spacecraft and instruments is shown in Table 3.1 .

Agribusinesses are beginning to use this technology to increase their knowledge of crop productivity.

One of the biggest stumbling blocks in the routine analysis of satellite data is the need for considerable processing to correct for image distortions and sensor calibration. At present, individual groups perform most of these functions. In addition, there is considerable difficulty in gaining access to digital satellite data. NASA is planning a new data system called the Earth Observing System Data and Information System (EOSDIS) that will ingest, process and distribute all NASA (and some non-NASA) satellite data. There will be a number of Distributed Active Archive Centers (DAACs) in the U. S. responsible for the processing, archiving and distribution of these data.

The regular daily coverage of the Advanced Very High Resolution Radiometer (AVHRR) lends itself to the study of many land surface processes. These include: monitoring snow pack for river runoff, detecting/monitoring forest fires and assessing the extent of vegetation damage, and monitoring the status of large areas of terrestrial vegetation. As an example, Figure 3.1 shows AVHRR images converted into the Normalized Difference Vegetation Index (NDVI) before and after a severe freeze in Florida on Christmas Day 1989. The reduced vegetation indicated by the brown colors in the second image indicates the damage to the orange groves brought about by this freeze. Similar analyses can be used to infer the conditions of the corn crop in Iowa or the wheat crop in Kansas. Agribusinesses are beginning to use this technology to increase their knowledge of crop productivity.

It is possible to use the fact that repeat coverage by the AVHRR instrument never covers the same exact spot. This oversampling at a scale smaller than the 1 km resolution of the AVHRR images can be used with a Bayesian reconstruction technique to produce 180 m (and possibly 120 m) images from 18 AVHRR images of the same area. Applied to Death Valley, California (Figure 3.2 ) this technique resolved spatial features that could only be seen clearly in the higher resolution (80 m) of Multi Spectral Scanner (MSS) data for the same region. A statistical test of this procedure using two-dimensional Fourier transforms (Figure 3.3 ) demonstrates that the finer-scale features of the MSS image are mostly resolved by the 180 m AVHRR surface reconstruction.

This surface reconstruction technique has the potential for generating higher spatial resolution satellite imagery from repeat images from lower resolution images without the need for more complex or dedicated imaging systems. In addition using a reflectance model provides a technique for routine subpixel image navigation.

A statistical test of this procedure using two-dimensional Fourier transforms (Figure 3.3) demonstrates that the finer-scale features of the MSS image are mostly resolved by the 180m AVHRR surface reconstruction.