Forest ecosystems can store large amounts of carbon (C) but those stores can change rapidly due to harvest, natural disturbance, or growth. While regional and national trends in forest carbon flux are well-known from extensive Forest Inventory and Analysis (FIA) assessments, substantial uncertainty remains about the effects of community composition, ecosystem productivity, climatic effects (e.g. drought), and disturbance severity. Advances in our ability to describe ecosystem change from remote sensing using a Landsat time-series-based approach called Timesync have opened the possibility of assessing the severity of disturbance events and providing current estimates of forest carbon flux. We studied C flux in a diverse landscape that includes National Vegetation Classification System macrogroups Vancouverian Lowland & Montane Forest, Vancouverian Subalpine Forest, Southern Vancouverian Dry Foothill Forest & Woodland, and Central Rocky Mountain Mesic Lower Montane Forest. Sixty-two percent of the forested lands are managed by federal agencies for a range of uses, while 38% are managed by private landowners primarily for timber revenue. We examined changes in C pools on 676 forested plots with a mean remeasurement period of 10 years and used measures of magnitude and duration of Landsat change to predict C flux on plots.
Results/Conclusions
Plot-based and Timesync-based assessments agreed that 69% of the plots were not disturbed, and that 19% of the plots were disturbed (overall agreement of 88%). Nine percent of the forested plots were disturbed but not detected by Timesync; most were in vigorous stands where 33% of the initial live tree C was cut or died, primarily in smaller tree sizes, yet the net change in C over the measurement period was comparable to undisturbed plots (11.1 vs. 11.9 Mg ha-1). The 3% of plots that FIA did not code as cut or burned but Timesync coded as disturbed did experience elevated mortality rates compared to plots that both called undisturbed, with a mean mortality of 31.2 Mg C ha-1 compared to 7.0. The best disturbance model explained 85% of the variation in live tree C at the second measurement, based on measured C at the first measurement, Tassel Cap Angle (TCA) at the time of the first measurement, the magnitude of TCA change, and the time since disturbance. Such strong results in such a diverse landscape suggest the utility of Landsat time series to update forest inventory estimates.