The proposed research project will be carried out in the Upper Tanana River Valley region of Alaska located between Delta Junction and Tok, Alaska. The uniqueness of this region lies in the fact that it contains a set of large areas that were affected by forest fires in 1956, 1987, 1990, 1994, and 1999 adjacent to undisturbed mature forests. The majority of these fires occurred in black spruce forests, which is also the dominant forest type found in Interior Alaska (Viereck et al. 1983; 1986). Aspen, white spruce, mixed aspen/spruce, and spruce/willow forests occur to a lesser extent in this area. All the burn sites in this area are accessible via the Alaska Highway located adjacent to the Tanana River. The existence of this burn chronosequence has made it possible to study patterns of biomass burning (Kasischke 2000; Kasischke et al. 2000c,d) and a wide range of ecosystem processes as a function of time since disturbance as well as the severity of burning (Chambers et al. 1999; Chambers and Chapin 1999; Kasischke et al. 2000a,e; O'Neill 2000; O'Neill et al. 1997; Richter et al. 2000).

The 1990's saw the deployment of 6 spaceborne synthetic aperture radar (SAR) systems (ERS-1 and 2, JERS-1, SIR-C, X-SAR, and Radarsat) which allowed scientists to begin to develop a number of unique techniques for monitoring terrestrial ecosystems (Kasischke et al. 1997). In terms of mapping specific surface characteristics using a single overpass, one frequency/ polarization systems (such as ERS-1/2, JERS-1, and Radarsat), however, are considerably less useful than multiple channel systems such as the SIR-C/X-SAR (Kasischke et al. 1997). The single frequency/polarization systems that have been in continuous operation for most of the past decade offer a distinct advantage over the multi-channel systems (that operated for only a few weeks). Studies have shown that ERS and JERS have the ability to monitor variations in aboveground woody biomass up to levels between 3 and 6 kg ha^-1 (Harrell et al. 1995; Kasischke et al. 1994a).

The signatures they record over vegetated surfaces vary to a considerable degree when changes in the moisture or temperature of the surface occur. It has been shown that single-channel, orbiting radars can be used to: (1) detect and map fires and monitor changes in relative soil moisture in the disturbed forests (Kasischke et al. 1992; 1994b; 1995; 1997; French et al. 1996; 1999; Bourgeau-Chavez et al. 1997; 2000; Wang et al. 2000); (2) detect variations in soil moisture and flooding in wetland ecosystems (Morrissey et al. 1996; Kasischke and Bourgeau-Chavez 1997; Kasischke et al. 1995; 1997), and monitor changes in surface backscattering associated with temperature variations (Way et al. 1991; Rignot et al. 1994a,b). Because of these seasonal variations, care must be taken in selection of the data sets used to estimate biomass. This is because variations in soil moisture in these low biomass stands result in changes in backscatter (2 to 3 dB) that are comparable to those resulting from biomass differences (Wang et al. 1994; Harrell et al. 1995; Pullianen et al. 1996). It has been found that using radar imagery collected in early spring prior to snow melt minimizes the effects of soil moisture variations when estimating biomass (Harrell et al. 1995).

Several of the investigators associated with this proposal have been studying the use of spaceborne imaging radars to monitor fire-disturbed boreal forests since 1991, and have focused a great deal of their research on areas within the burn chronosequence discussed above. These studies have shown:

1. At incidence angles less than 25°, there are characteristic signatures that appear in both ERS and Radarsat imagery of forested areas disturbed by fires in the boreal forest (Kasischke et al. 1992, 1994, 1995, 1997; Bourgeau-Chavez et al. 1997; French et al. 1999). These characteristic signatures have been found in the boreal forests of Alaska, Canada, and Russia.

2. At times during the growing season when soil moistures are high (e.g., immediately after a snow melt or after rainfall), fire-disturbed forests are 3 to 6 dB brighter than adjacent unburned forests (Kasischke et al. 1994; Bourgeau-Chavez et al. 1997; French et al. 1996).

3. As soils dry out during times with low precipitation, the radar image intensity in both burned and unburned areas diminishes (French et al. 1996; Bourgeau-Chavez et al. 2000; Harrell et al. 1995).

4. Within the same fire scar, radar image intensity is linearly proportional to the moisture in the upper 5 cm of the soil surface (French et al. 1996; Kasischke et al. 1995).

5. Vegetation recovery after a fire will begin to significantly attenuate the soil moisture signature within 7 to 10 years after a fire, although fires > 10 years in age have been detected through an increased image intensity (Bourgeau-Chavez et al. 1997; Wang et al. 2000).

6. A composite signature derived from multiple-radar images collected during the same growing season can be used to assess relative seasonal moisture conditions within the same fire scar.

It is our belief that being able to monitor variations in relative moisture conditions in boreal forests (both burned and unburned) represents a unique application for the both the Radarsat and ERS imaging radar systems. Direct measurement of moisture conditions is important to models that estimate net primary productivity and heterotrophic respiration in boreal forests (Bonan 1991a,b; Bonan and Van Cleve 1992; Schlentner and Van Cleve 1985). Available soil moisture is a limiting factor in both plant photosynthesis and respiration and microbial respiration, with these processes also being regulated by air and ground temperature.

Different modeling approaches have been used to estimate rates of net primary production (NPP) and total soil respiration. For NPP, production efficiency models (PEM) have been developed that physiologically restrict the conversion efficiency of absorbed photosynthetically active radiation (APAR) to primary productivity over the short term based on temperature and moisture conditions. Models have been developed (GLO-PEM and GLOPEM2) that use satellite observations as a basis for providing model input parameters (Prince and Goward 1995; Goetz et al. 1999). GLO-PEM2 currently consists of linked models of canopy radiative transfer, canopy utilization of APAR, and physical environmental parameters that reduce potential production. Seasonal vegetation indices and GOES PAR data are combined with air temperature to estimate potential production. A set of physical parameters (stress terms) are determined via satellite observations to restrict production, including vapor pressure deficit, a soil moisture cumulative stress index and air temperature. To estimate surface and air temperature, as well as vegetation temperature, a contextual, split-window approach that uses satellite measured surface temperature and spectral vegetation indices, known as TVX is used (Goward et al. 1994). Similar techniques have been developed to estimate soil moisture and atmospheric humidity.

Recent field and laboratory studies carried out as part of a NASA Global Change Fellowship (O'Neill 2000) have resulted in a model of net soil respiration (CO₂ emission) from burned and unburned spruce forests in interior Alaska. In this model, total CO₂ emission from the soil (E_t) is defined as

where H_r is microbial respiration, R_r is root respiration, M_np is net primary production from the moss layer. A model was developed by O'Neill based on laboratory measurements of variations in both H_r and M_np from samples collected in actual stands. Algorithms were produced to estimate both H_r and M_np based on variations in moisture and temperature. Root respiration models were developed based on percent vegetation cover constrained by the model estimates of H_r and M_np and field measures of total respiration. Field observations by O'Neill (2000) showed that the seasonal temperature patterns at various depths in the organic soil profiles in unburned stands followed air temperature patterns, while temperatures in the burned stands were also strongly controlled by the amount of organic mat material remaining after the fire. Algorithms were produced to estimate rate of thaw/soil warming in the burned stands based on depth of organic soil.

We believe that the models developed by O'Neill (2000) can be driven by parameters provided by satellite observations, particularly imaging radars. Specifically, many of the same parameters used for the GLO-PEM2 model (e.g., air and surface temperature, variations in vegetation cover, soil moisture) can also be used for estimating net soil respiration. One additional parameter on fire severity (which can be directly converted to depth of organic soil remaining) can be provided through analysis of Landsat TM imagery (Michalek et al. 2000).