A great deal of attention is being paid to understanding the terrestrial carbon budget in high northern latitudes because of the role ecosystems in this region play in intra- and interannual variations in the atmospheric concentration of carbon dioxide. Randerson et al. (1997, 1999) showed that the amplitude of the seasonal CO₂ signature at latitudes above 50° N has increased since 1980. The increase in amplitude has been attributed to increases in net primary production (CO₂ uptake) early in the growing season, and a possible increase in respiration (CO₂ release) late in the growing season (Randerson 1999). These conclusions are consistent with satellite observations of an increased vegetation index (normalized difference vegetation index or NDVI) in this region over the same time period (Myneni et al. 1997).

While it has been argued that the increase in early season primary production/ NDVI may be the result of recent climate warming in this region (Myneni et al. 1997), other factors can also result in these differences. Variations in primary production and NDVI in the boreal forest over the past three decades are the result of complex interaction between temperature, fire activity, and changes in forest cover after disturbance. In particular, the role of disturbance by fire has yet to be factored into the estimates of NDVI variability and primary production. Fire is ubiquitous throughout the boreal forest region, burning large areas during years when drought conditions prevail. Analysis of fire records (Murphy et al. 2000) and satellite imagery (Cahoon et al. 1994; Kasischke et al. 1999) shows that at sub-continental scales between 3 and 6% of the entire land surface can burn in a single fire season. During the decade prior to 1981, nearly 7% of the North American boreal forest burned in fires.

Fires cause increases in NDVI and primary productivity in two ways. First, it has been shown that after severe fires (that consume a large portion of the organic mat found in boreal forests), deciduous forests replace coniferous forests (Landhauesser and Wein 1995; Zimov et al. 1999; Kasischke et al. 2000a,b). Because deciduous forests have higher NDVI and productivity than coniferous forests, a shift from mature coniferous forests to young deciduous forests would result in an increase in both of these variables. Second, studies by Kasischke and French (1997) show a dramatic rise in NDVI in burned Alaskan forests (both deciduous and coniferous) as vegetation recovers during the first 15 to 20 years after a fire. These two factors result in increases in both NDVI and primary production regardless of changes in temperature. However, it should be noted that fires that occurred during the 1981 to 1991 would cause significant decreases in NDVI and loss of primary production during the year they occur.

Recent advances in the development of large scale-models (many of them driven by satellite observations) have improved our capabilities to understand and quantify the sources of variation in the atmospheric carbon dioxide record (Potter et al. 1993; Running and Nemani 1988; Haxeltine and Prentice 1996; McGuire et al. 1997). For the boreal forest region, significant advances were made in the development and validation of satellite-based models of NPP during the BOREAS program sponsored by NASA and other U.S. and Canadian agencies (see, e.g., Goetz et al. 1999).

In spite of these advancements, there are two shortcomings in the approaches used to estimate changes in net carbon flux in the boreal forest. First, most terrestrial carbon models produce estimates of net ecosystem production (NEP), not net biome production. Net biome production models account not only for variations in NPP and heterotrophic respiration, but also direct emissions from fire. Recent studies by French et al. (2000) show that on average fire emissions are 9.7 g C m^-2 for the entire North American boreal forest, but are as high as 22 g C m^-2 in certain regions, such as the boreal forest region of Alaska. A recently approved NASA IDS study by the principal investigator for this proposal will further refine estimates of carbon emissions from fires throughout the boreal forest.

Of critical importance to carbon cycling in the boreal forest is a more complete understanding of the processes that influence carbon storage in the deep organic soils. Increasing temperatures in the boreal region may result in a significant reduction of this soil carbon pool in two ways:

1. Studies by Kasischke (2000) have shown that in the deep soiled black spruce forests of Alaska, on average >40% of the pre-fire organic mat carbon (which includes litter, moss, and organic soil) is burned during fires. These studies clearly showed that fires occurring later in the growing season resulted in a higher level of organic mat consumption than did early season fires. This results indicates that if climate warming results in longer growing seasons and warmer, drier soil moisture, then higher levels of direct carbon release from fires can be expected.

2. Studies by O'Neill show that a further 10 to 20% of pre-fire organic soil carbon is lost in the first several years after a fire due to increased soil respiration that result from the warming soils resulting from the fire.

This latter result is particularly important because it addresses a fundamental limitation of the current generation of net ecosystem production models. These models are not presently configured to recognize that a significant degree of warming of the upper one meter of soil (between 4 and 10° C during the first two years after a fire) occurs in boreal forests, depending on fire severity (e.g., level of consumption of organic soil).

For example, a recent exercising of the TEMs model for Alaska shows there to be a significant increase in NEP after fires due to increased NPP associated with vegetation that re-establishes on burned sites (Dave McGuire, personal communication). However, site specific studies by O'Neill (2000) show a decrease in NEP for 2 to 7 years after a fire because of the effects of enhanced soil respiration.

The overall goal of the research proposed herein is to develop improved approaches to estimate patterns of net primary and ecosystem production in fire-disturbed boreal forests located in interior Alaska. To achieve this goal, we will address the following objectives:

1. Modify a satellite-based net primary production (NPP) model so that it can use MODIS observations of seasonal variations in vegetation index and surface temperature variations, imaging radar data to assess soil moisture conditions, finer-scale Landsat-7 data to account for small-scale variations in vegetation cover, and ASTER and Landsat-7 thermal IR data to account for finer-scale surface temperature variations;

2. Based on an extensive set of surface measurements collected at the study sites, develop a satellite-based model of NPP and soil-surface carbon dioxide emissions based on variations in vegetation cover, surface temperature and soil moisture measurements derived from satellite observations, and combine the outputs of these models to produce estimates of NEP.

3. Validate the model estimates against surface observations obtained through chamber and eddy covariance techniques through the appropriate up-scaling of surface measurements to the same spatial scales as the outputs of NPP models.