Overcoming limitations of using radar to estimate vegetation variables: Part 2 of 2
Because radar is an active sensor, it presents enormous advantages to quantify vegetation attributes in areas constantly covered with clouds. However, limitations described in Part 1 present some challenges. Here, I describe a few alternatives to mitigate these limitations.
Typical available radar bands are:
- X (wavelength 3 cm)
- C (wavelength 4–8 cm)
- L (wavelength 15–30 cm)
- P (wavelength ~75 cm)
The longer the wavelength, the deeper radar bands penetrate dense vegetation. While the X-band, for example, having roughly the size of a leaf, will reflect back straight from the canopy, longer wavelength bands — such as L and P-bands — will pass through the canopy and can even reach the ground underneath the canopy.
The longer the wavelength, the deeper radar bands penetrate dense vegetation.
For example, X-band amplitudes provide poor data for quantifying the structure of the vegetation. The P or L-bands will provide more information because, after passing through the canopy, it interacts with the vegetation structure. This interaction occurs at several levels. The microwave can simply bounce back from tree trunks. But it can also alter its polarization, (i.e., once emitted horizontally, with respect to the radar, the microwave might interact with the landscape and return to the radar vertically or horizontally). The same occurs if it is emitted vertically. As a result, a single band may have four different polarizations (HH, HV, VH, and VV), and these polarizations carry information about the landscape.
Interferometry is a radar technique that uses another kind of information from microwaves: its phase, and phase difference — somewhat similar to stereo pair imagery. One of the main products of this technique is the production of Digital Elevation Models (DEMs), which, depending on the band choice, will provide a variety of information about the landscape.
A DEM using the X-band is a model of the landscape surface, including the vegetation canopy; a Digital Surface Model (DSM). A DEM using the P-band is a model of the terrain even under the vegetation; a Digital Terrain Model (DTM). The difference between DTM and DSM is a good estimate of the vegetation height. While a DTM will likely not change over time (except in rare circumstances, e.g., an earthquake), changes due to vegetation growth or removal will result in a different DSM, and vegetation heights. Of course, DEMs using bands that partially penetrate the vegetation canopy are harder to interpret and are not as suitable for this kind of analyses as X- and P-bands are.
Interferometry is a radar technique that uses another kind of information from microwaves: its phase, and phase difference — somewhat similar to stereo pair imagery.
Other techniques, such as PolInSAR, can employ C or L-bands and quantitatively assess the vegetation profile/height without the need of producing a DEM by combining interferometry and polarimetry. Estimating the vegetation height either through DSM/DTM differences or PolInSAR promises to overcome barriers resulting from the saturation problem because vegetation height is an important allometric quantity used in conventional forest inventories (see also Forest Inventory: Can we do without it? by Ian Moss).
SAR data are becoming increasingly available due the expansion in the number of satellite and airborne sensors as well as advances in radar data processing. The development of radar applications is still a field open for exploration, and will rely on a variety of techniques that, when wisely combined with the right sensor, enables the assessment of different aspects of the landscape and address very specific questions.
Learn more about advantages and limitations to using radar to estimate vegetation attributes in Part 1.
Read more about Tesera’s approach to high resolution forest inventory.
Dr. Julianno Sambatti is a Data Scientist working on high resolution inventory solutions at Tesera.com
