Woods Hole Research Center

Monitoring Post-fire Regrowth

Two Landsat images displaying the pre- (l) and post-burn (r) conditions of the 1999 Donnelly Flats fire, outlined in blue.

Woods Hole Research Center scientists examined the regrowth rates of boreal forest vegetation after fire using a time series of satellite data derived from the NOAA Advanced Very High Resolution Radiometer (AVHRR) series of satellites. Data from the Global Inventory Modeling and Mapping Studies Group (GIMMS) Normalized Difference Vegetation Index (NDVI) data sets, a surrogate for vegetation photosynthetic activity, was used.

Graph showing 1995 departure from the mean FPAR in the burned areas. Notice the large dip in 1995.

WHRC results showed a noticeable pattern of post-fire regrowth with 5 or more years passing before returning to pre-fire conditions. For example, for three episodic fire years (1981, 1989, and 1995) in the Canadian boreal forest, all three years show a significant drop from average productivity relative to the period before the fire. By coupling this information with similar statistics aggregated from unburned areas in the same eco-region of the fires for each of the high fire years, Center researchers removed changes in the mean productivity over time that occurred due to other environmental factors (such as precipitation).

Click image to view animation of 1995 FPAR

Continent-wide monitoring of vegetation response to fire over time can be an effective ingredient for climate and carbon system models. Although the AVHRR NDVI data lack the spatial detail needed to trace small fires, the efficacy of this method on continental and global scale studies is apparent.

Examining Regrowth Trends using MODIS

In order to understand how fire and regrowth affect tree cover, WHRC scientists examined the MODIS continuous vegetation fields, or percent tree cover (MOD44B), over burn scars in Alaska ranging in age from 4 to 50 years. Using a database of Alaska fire boundaries to identify areas which burned since 1950, an average percent tree cover of the total area burned throughout that year was calculated. The plot demonstrates that MODIS percent tree cover exhibits a large amount of variability both between burns of the same age and burns from different years. As the linear regression suggests, however, recently burned areas was found to have significantly less tree cover than areas which burned years before.

The large amount of variability in tree cover observed between areas which burned during the same year is most likely driven by differences in burn severity and landscape type. In severely burned areas both the understory and canopy are completely combusted, killing shrubs, grasses, herbaceous species, and trees. Because deciduous species tend to grow faster than coniferous species, deciduous regrowth typically dominates in severely burned areas. In less intense burns, however, fire remains predominantly in the understory, and most of the coniferous tree species survive. In these areas we tend to observe a coniferous dominated succession. These fire-related changes in forest cover lead to obvious differences in percent tree cover and this is observed in the MODIS data.

Maps of the relative cover of deciduous and evergreen vegetation in Central Alaska. Dark red outlines in the smaller map show areas burned since 1950. Click for larger image.

Examining Burn Severity using MODIS

Burn severity has significant effects on both the instantaneous flux of carbon to the atmosphere during fire, as well as the rate and timing of the successional recovery following fire. When examining the effects that fire has on net carbon exchange and forest regrowth, it is important to be able to characterize the severity of individual burns.

Individual pixels of MODIS NDVI, located across a gradient of burn severities, were plotted against time for the entire MODIS collection period. These yearly profiles of MODIS NDVI, seen below, display the seasonal variability associated with leaf-out and leaf-off. Differences between sites of varying burn severity are distinguished by the magnitude of the NDVI value. MODIS 250m NDVI is, therefore, clearly able to distinguish between these sites, despite fine scale landscape heterogeneity in the area.