Multiplatforms and Multilevels - The "Multi" Approach to Applied Remote Sensing - Remote Sensing Tutorial -
Multiplatforms and Multilevels

Ground truth activities are an integral part of the "multi" ("more than one") approach. Thus, we should procure data whenever possible from different platforms ( multiplatform), at various altitudes (multistage; multilevel). (A "platform" is a synonym for any orbiting spacecraft, be it a satellite or a manned station, from which observations are made.) This gives rise to multiscaled images or classification maps. Ideally, we should aim to employ multisensor systems simultaneously to provide data, commonly at multiresolution, over various regions of the spectrum (multispectral). Often, we obtain data at different times (mutitemporal), whenever seasonal effects or illumination differences are factors or change detection is the objective. Supporting ground observations should come from many relevant, but not necessarily interrelated, sources (multisource). Some types of surface data may correlate with one another and with other types of remote sensing data (multiphase).

Many remote sensing investigations include several of the above "multi" categories but examination of both remote sensing textbooks and Internet sites generally does not highlight examples that include most of these together as applied to one study or application site. This proved the case in preparing this page. So, in order to illustrate the "multi" concept adequately, it is necessary to show images of different parts of the world that don't show the same piece of "real estate" sensed repeatedly by different systems. And, we mention now that most platforms in the last few years have multiple sensors on board operating simultaneously. The best example of that is Terra (page 16-9) which has five different but complementary sensors which often examine the same target.

Earlier in this Tutorial, there have already been individual and isolated examples of some of the "multi" types of images and photos. In the review erected on this page, we will develop one theme: agriculture, especially crop monitoring. We start with an expansion of the analysis of the Delaware/Chesapeake Bay classification shown on page 13-3, which was part of a NASA field study. Here we will follow the multi-level approach by looking successively at the farmlands around the Chesapeake and Delaware Canal, starting with a Landsat red band subscene and progressing to a low altitude photo.

Landsat MSS 5 subscene covering parts of Delaware and Maryland farmland; locate the Chesapeake & Delaware Canal.

Next is a high altitude U-2 photograph of part of the above area; locate yourself using the canal.

U-2 photograph of part of the Delaware-Maryland subscene.

Now, to zoom in further, consider this medium altitude aerial photo which contains part of the canal.

Aerial photo of the DE-MD subscene taken from an altitude of about 30,000 ft.

Finally, here reproduced is a paper print of a low altitude aerial photo that was actually taken into the Delaware test site. The Soil Conservation Service's field agent has made notations showing characteristics and yield for some of the crop acreage.

Low altitude (5000 ft) aerial photo of individual fields in the DE-MD study area.

The writer (NMS) was a participant in this field study. As part of the preparation for the Landsat phase, NASA flew an aircraft mission with a multispectral scanner over fields in the Delmarva Peninsula to the south of the study site. Here are four images designed to simulate the 4 Landsat MSS bands:

Aircraft multispectral scanner images simulating Landsat's 4 MSS bands, covering an area along the Choptank River in the Delmarva Peninsula.

From the data, an analog measure (using a photometer operating on a transparency) of the photo-density of selected fields in each of the MSS-equivalent bands led to this plot of relative darkness as a proxy for reflectance coming from the ground features and crops indicated:

Simplified plot of relative densities for each of 4 bands in the aircraft multispectral scanner images as applied to the labeled features.

Let us turn from this specific study to some more general examples. Many of the photographs taken from the Shuttle by the astronauts have agricultural areas as their subject matter. Often these photos are not particularly good owing mainly to atmospheric problems. But this one covering the land around Enid, Kansas is one of the better.

Farmlands around Enid, Kansas in a photo taken during the Shuttle STS-073 mission.

Other satellites produce excellent near-natural images of farmlands, such as this SPOT scene:

Fields in France in natural color, imaged by SPOT.

At higher resolution, here are fields in California's Great Valley near Fresno imaged by the IKONOS-2 satellite.

Field Crops near Fresno, California.

At the other extreme, the AVHRR, as demonstrated on page 3-4, is quite adept at providing small-scale, large area indications of crop and vegetation vigor, often expressed as NDVI. This next image is a black and white plot of the NDVI values (using channels 1 and 2) for the land in and around Dallas, TX. Light tones indicate high NDVIs.

AVHRR-generated NDVI image covering the area around Dallas, TX.

Crop stress results from insufficient soil and/or crop water (drought), improper nutrients; plant disease, insect infestations, and other factors. The next image is of cropland in Colorado's San Luis Valley. It was made by the AVIRIS sensor that will be described in detail on pages 13-9 and 13-10 of this Section. A classification of these Colorado crops is treated on page 13-10. Here we show AVIRIS hyperspectral data that use bands sensitive to crop moisture deficiency.


Soil moisture is one of the critical parameters a farmer needs to know in making decisions about planting conditions and need for irrigation. It is often the precursor indicator of potential or actual crop stress. This can be done through aerial photography, as shown here for some Indiana farms, but the cost of flying for specific water inventory is high.

Aerial photograph of Indiana fields showing higher moisture in darker tones, following irregular patterns controlled by drainage.

Thermal scanners are also good at detecting moisture, as indicated in this aerial image of a Wisconsin farm, taken around 9 PM at night shortly after the setting Sun. The bright spots in the upper left are a herd of (warm) cows. The black rectangle in the upper right is a sheet metallic roof on a farm outbuilding, which shows "cool" because of the very low emissivity of metal.

Thermal scanner image of a Wisconsin Farm.

AVHRR thermal bands can also provide useful agricultural information. And so did HCMM when it operated. Here is a HCMM Day Thermal image of much of California, taken in May. Note the farm patterns in the central Valley. Note also the very dark area in the High Sierras - this is spring snow that will eventually provide water for the crops during summer meltdown.

HCCM Day IR image of California.

Radar is a good means of imaging farmland, as seen in this low altitude aerial radar mission over the Maricopa area near Phoenix, AZ:

Fields near Phoenix imaged by radar.

Seasat radar imaged this next scene, in the Great Plains. Some fields are dark, others light, indicative of the stages of growth (light areas indicate crops that reflect more of the radar beam to the receiver). Of particular interest are the two dark patches which represent the effects of soil moisture (reduces returns) following two local thunderstorms passing over the plains.

Seasat image of Great Plains farmlands.

As was put forth in Section 8, radar images made from different bands disclose information in each not expressed in the same way as in the other(s). Below is a pair of images of the Medicine Lake area in Alaska west if Fairbanks that were fortuitously imaged 18 minutes apart by two different satellites. On the left is a ERS-1 radar C-band image; on the right is a JERS-1 L band image. Note that the ERS-1 image renders some bogs in bright tones; the JERS-1 image highlights creek beds.

Image pair - ERS and JERS - of the Medicine Lake area in east-central Alaska.

On the next two pages, we will finish this "multi" survey, starting with the information obtained when images from different sensors are merged and ending with multitemporal examples.