Is The Boreal Forest Greening or Browning? An Explanation of Differing Results.
Alcaraz-Segura, D., Chuvieco, E., Epstein, H.E., Kasischke, E.S., and Trishchenko, A. 2010. Debating the greening vs. browning of the North American boreal forest: differences between the satellite datasets. Global Change Biology 16: 760-770.
The authors argue that a significant factor not considered in these past studies is fire history. If points in time are compared before and after a fire, for example, NDVI will be seen to have decreased, but not for the reason for which the data were sought. Similarly, trends that begin right after a fire will show increasing NDVI that is unrelated to climate factors, as the vegetation recovers from fire. Hence, they use this latter phenomenon of vegetation recovery after fire as a ground-truth test for the two different satellite datasets.
Working with GIMMS satellite data that represent 64-km2 cells and newly available CCRS data that represent 1-km2 cells, the five researchers used an algorithm that has been shown to detect recent fires to correctly classify pixels as burned or not-burned, comparing the performance of the two datasets for detecting trends. In doing so, they found that the GIMMS data were unable to properly detect increases in NDVI over time in burned areas compared to the CCRS data, and that GIMMS data are thus a poor choice for this type of study. The CCRS data, on the other hand, detected strong greening in burned areas (as expected) and a weaker but consistent greening in unburned forest areas over 1996 to 2006. As a result, Alcaraz-Segura et al. suggest that (1) the widely-used GIMMS data may have produced false results in other studies and should be used with caution, (2) that satellite data need to be better calibrated with ground data before use, and that (3) the greening of the Canadian boreal forest is probably real for the most recent decades.