Extended Data Fig. 2: Asymmetric impacts of boreal forest cover change on resilience derived from alternative datasets and methods.
From: Asymmetric sensitivity of boreal forest resilience to forest gain and loss
Each row corresponds to one alternative data or method: VOD as the state variable for the forest ecosystem to construct TAC (a, b), ESA CCI land cover product as an alternative dataset to derive forest cover fraction (c, d), variance as an alternative resilience indicator (e, f), 3-order harmonic deseasoner with rolling mean detrender as complementary pre-processing approaches (g, h). Bars in panels a, c, e and g indicate the mean ΔTAC weighted by pixel areas, with error bars representing the 95% confidence interval of the mean ΔTAC. ΔFC values in panel a were divided into 40 bins with 0.5% intervals, while in panels c, e and g, they were divided into 25 bins with 2% intervals. Bars in panels b, d, f and h indicate the linear regression coefficient, with error bars representing the 95% confidence interval. Brown and magenta numbers in panels indicate the sample size of forest loss pixels, while green numbers indicate the sample size of forest gain pixels.
