Understanding Vegetation from a SAR Image

Synthetic Aperture Radar (SAR) is a high-resolution radio wave imaging sensor that transmits microwave signals to the Earth and detects the reflected signals for the observation. Unlike the optical sensors counterpart, SAR can operate both day and night, and does not get affected by clouds or rainfall.

Microwave used by SAR has the longer wavelength than the visible light, resulting in lesser resolution than the optical sensors. Depends on the parameters, SARs have the resolution of 10 meters or so.

Though many people wonders, it is not so surprising that forests can be observed thorough radio waves. Since the radio waves transmitted by SAR reflect off tree crowns, branches, trunks and the ground, many researchers have found that the amount of radio waves reflected has a correlation with the forest biomass. Much attention has been attracted to the forest observation with SAR recently.

As shown in Fig. 1, radio waves transmitted from SAR reflect off trees, and the satellite antenna receives the part of it. Some waves reflect off the tree crowns, and others penetrate the tree crowns but reflect off trunks and the ground. Generally, the radio waves with longer wavelength and more horizontal polarization tends to penetrate the woods further, and eventually reaches to the ground, before getting reflected back to the satellite.

Forest biomass, which is related to the absorption of carbon dioxide (CO2) by forests, may be expressed with the volume of tree trunks per area. Therefore, obtaining data on the density of tree trunks in an area will lead to an understanding of forest biomass. As radio waves penetrate tree crowns and branches deeper, more waves reflect off tree trunks. In addition, considering that the scattering of microwaves is proportional to relative permittivity of tree trunks (relative permittivity is related to the parameters such as the water content and the mass), we can establish a relationship between the intensity of the reflected waves and forest biomass.

Fig. 2 shows a relationship that many researchers have found. With the horizontal axis representing forest biomass and the vertical axis representing the backscattering cross-section, the diagram shows that forest biomass up to 60 t/ha have a one-to-one relationship to the backscattering cross-section. Although it is not shown in the diagram, received P-band waves have a similar relationship with forest biomass up to 150t/ha. By utilizing the SAR characteristics, we expect to learn the role of forestry resources in the land CO2 cycle, which is a determinant factor in global warming.

Overlaying Multiple Microwave Images

Tropical rain forests and boreal forests expands over vast regions, and a full SAR scene of about 7575 km is not sufficient to cover them. Therefore, several adjacent images need to be put together to form a big mosaic.

First, SAR data obtained in a flight is processed to compose a path image (JERS-1/SAR has obtained as large as 400075 km in the past). Creating multiple path images makes the subsequent data processing easy because the satellite travels on almost the same orbital planes (orbital inclination).

Next, the coordinate values of these images are converted to the coordinate system of the final output image (e.g. Mercator Chart). If there was no errors in the orbital elements, a mosaic image could be created by simply patching the path images together. However, these errors cannot be ignored. So, a registration process is performed using a correlation in the overlapping portions of immediate neighbors to align one image with another for an accurate overlay.

Another problem is the brightness of the image. When the satellite observes the land surface with a swath of 75 kilometers, the incidence angle varies within a range of about six degrees. The incidence angle influences the backscatter cross-section of forests, resulting in different levels of brightness for each path image (see Figs. 3a and 3c). Variation in the backscattering cross-section of a target is a natural and unavoidable phenomenon, and it is information in its right. However, in order to see a big picture in an aesthetically pleasing manner, images are given uniform brightness by correcting the gray level (see Figs. 3b and 3d).

Correction of this kind is easier to perform if the data were obtained during the same period of time. The creation of path images not only simplifies the mosaic forming process but also expands the coverage to include remote islands. A mosaic image of Southeast Asia that created by the process described in this section is shown in. Fig.4.

Findings from SAR Images

Fig. 4 is a mosaic image that covers eleven nations in most of the Indonesian region (lat. 32ºN - 1ºS, long. 90ºE - 112ºE). The nations include Vietnam, Cambodia, southern China, Tibet, Laos, Thailand, Malaysia, Singapore, Myanmar, Bangladesh and India. The mosaic image was generated using a total of thirty-nine path images. Eleven hundred JERS-1/SAR scenes were used just for the land area. The image represents an area of 3.5 million km² where 8% of the world’s population resides (470 million people).

Most data used for the mosaic was obtained in the winter of 1997 (January and February). However, a portion of Myanmar was obtained in August 1998, causing the slight difference in the brightness.

Geographically, the Indonesian region was formed over three different time eras: the Palezoic era (central Vietnam, Malaysia, Myanmar and Tibet), the Mesozoic era (Cambodia, Laos, northern Vietnam and Thailand) and the Cenozoic era (southern Thailand, Vietnam, western Myanmar and Bangladesh).

Eight million years ago the Indian Plate collided with the Eurasian Plate to form the Himalayan mountains and the Tibet plateau (elev. 4,000 meters). This event not only pushed Indochina to the East but also formed the Red River Fault (northern Vietnam) and faults along the Arakan Yoma (Myanmar). Precipitous variations in elevation and the faults running across a portion of Tibet and the Brahmaputra River like snakes can be seen in Fig. 4. This kind of images provides useful information for geographical research in this region.

Located at an elevation of less than 300 meters above sea level, southern Cambodia, Thailand and Vietnam has a large amount of forestland. Most of the forests are found in mountainous areas. With the exception of the Korat Plateau in southern Thailand -low biomass region- and Myanmar and southern Vietnam -areas where biomass changes are deeply related to monsoon-, this area is covered with dense tropical forests.

The mosaic image of forests in Southeast Asia appears darker than its counterpart in the Amazon and Africa, which may be related to the fact that these forests stands on the inclined surface, since most of the forest in this region are in the mountainous area.

Annamitic Cordillera, which marks the border of Laos and northern Vietnam, has an elevation of 2,400 meters that decreases northward. We have data showing a backscatter coefficient of –10 db for forests in Vietnam, Laos and southern China.

In the oceans in the image, three interesting features can be found. The shadowy pattern in the Gulfs of Siam and Martaban is probably the product of the wind. A dark strip offshore of Yangon seems to be a spillage from an oil tanker. Offshore of Phuket in the Andaman Sea features internal waves with the wavelength of 10 km for two days. This indicates that the state of the ocean remains unchanged for a period of two days.

With a geometric accuracy of 406 meters and a spatial resolution of 200 meters, the mosaic image of Fig. 4 proves itself to be a useful intermediate-resolution image.