Synthetic aperture radar (SAR) can be used to study carbon flux in boreal forests by coupling SAR observations to ecosystem carbon flux models, thus allowing improved quantification of canopy transpiration and C0₂flux. The seasonal C0₂flux in boreal forests is characterized by a large increase at the start of the growing season in the Spring and a collapse into senescence in the Fall. The research thrust for using SAR is to:

1. quantify the landscape in terms of functional groups of species and their relative magnitude of flux,

2. define growing season length (thus bounding the period during which flux occurs), and

3. quantify short-term variations observed during the growing season.

The goal of this research is to use the backscatter signature to assist with landscape classification, determination of growing season length, and changes in vegetation water status as related to canopy C0₂flux in boreal forests. Benefits of using radar include the fact that it penetrates clouds, can be used at night, and is directly related to surface structure and dielectric properties.

Radar imagery can provide inputs to process models and biosphere/atmosphere models by supplying ecological parameters such as classification of vegetation functional groups. Each functional group is characterized by a unique behavior in C0₂flux; this is the reason for partitioning the landscape into different classes. Radar is also useful in determining the timing of deciduous leaf onset and leaf drop, as well as the timing of soil and vegetation thaw and freeze.

The radar backscatter signature of a forest is determined by a combination of the geometric and dielectric properties of the vegetation and soil. Vegetation dielectric properties are related to fluid chemistry (sugars, electrolytes) and vegetation water status (water content, precipitation, dew, temperature, freeze/thaw state, xylem water potential, xylem flow). Soil dielectric properties are related to soil composition (textural components, organic materials) and soil water status (water content, freeze/thaw state, snow). The vegetation structural class is related to geometric properties of the canopy. Canopy phenology, variations of which define growing season, is determined by a combination of geometric and dielectric properties. Vegetation hydrologic status is related to the dielectric properties alone.

A C0₂flux estimate model is used to determine how the carbon flux varies for various species or functional groups within a landscape. It provides an estimate of C0₂flux in a region. Figure 8.1 shows the major controls on seasonal C0₂flux in boreal forests. Generally, growing season length is bound by vegetation freeze/thaw state in coniferous species and by leaf on/off in deciduous species. For coniferous species, radar data are used to bound the start of the growing season, when the vegetation thaws, and the end of the growing season, when the vegetation freezes. These estimates can be used to complement CO 2 flux models and improve their estimates of the total C0₂flux for a boreal landscape.

At the Bonanza Creek experimental forest near Fairbanks, Alaska, landscape classification by radar successfully identified areas of Balsam Poplar, White Spruce and Black Spruce. Large backscatter decreases (6 to 8 dB at L-band) were observed when the landscape underwent a transition from a thawed to a frozen state. Marked changes were also observed when conditions changed from flooded to unflooded. In this area of discontinuous permafrost, the timing of the freeze and thaw as detected by the radar help bound the growing season, and thus the C0₂flux.

Quantification of C0₂flux with flux models revealed radical differences over just 3 days in Spring. Ground truth measurements verified this rapid shift. Temperature measurements made with thermisters in the soil and trees revealed that the thaw occurred in the soil, tree stems, and needles and twigs at distinctly different times. With the soil thaw begins the period of respiration the start of carbon exchange.

In a regional extension of this approach, ERS multi-temporal transects collected by the ERS -1 spaceborne SAR from August through November across Alaska were used to determine when the landscape froze. This was accomplished by use of a change detection algorithm based upon the difference in backscatter between unfrozen and frozen conditions.

As part of a canopy water status monitoring study carried out in California, dielectric probes were placed in trees to monitor dielectric properties of different parts of the trunk to characterize temporal variance of the dielectric constant. Order of magnitude changes in trunk dielectric constant were found over 24 hour periods. At mid-day, during the period of maximum evapotranspirative demand, the trunk dielectric constant was very low, resembling that of frozen trees. At night, levels recovered, and approached the dielectric constant of pure water. These large diurnal changes in dielectric constant were also reflected as observable changes in radar backscatter.

In situ  observations from the Boreal Ecosystem-Atmosphere Study (BOREAS) sites in Manitoba and Saskatchewan were recorded during long-term seasonal and short-term focused field studies to monitor dielectric properties and other variables. As an example, Figure 8.2 shows seasonal xylem water flux, vapor pressure deficit, air temperature, soil temperature profile, bole temperature, and bole dielectric constant, over an entire growing season for an old black spruce site in northern Manitoba.

There is normally a low point in the dielectric constant during the midday high evapotranspiration period, followed by recovery during the pre-dawn hours. However, it has also been observed that during nighttime periods when no mass transport of water occurs, there may still be a marked decrease in dielectric constant. This may be related to some chemical diffusion or mass diffusion laterally in the trunk related to electrolyte exchange and is reflected in the dielectric constant. However, the variation in dielectric constant from tree to tree may be significant, and this highlights the need to study a number of trees in a given stand. These variations could be related to the insertion depth of the probes in each tree or to differences in sizes between trees.