Extended Data Fig. 7: Drought-related dieback from NDVI.
From: Forest vulnerability to drought controlled by bedrock composition
NDVI data from 2014 to 2016 generally plot below 1:1 line, indicative of drought-related declines at all 3 sites. Like the ET-ET plot in the main text (Fig. 3c), the data are consistent with a three parameter regression, with a slope, an intercept, and an offset between Bald Mountain and the other two sites (see Extended Data Table 3 for regression parameters, uncertainties, and p values). The statistically significant offset is consistent with a species-related control on forest dieback, as discussed in the main text. In addition, the slope of the relationship is not ≥1 (p < 0.0001), so we can reject the null hypothesis that greater water storage capacity at sites with higher pre-dieback forest productivity compensates for the higher water demand. Hence, dieback from NDVI increases disproportionately with increasing pre-dieback NDVI. Moreover, our results are consistent with bedrock control of dieback, irrespective of whether dieback is quantified using NDVI or ET derived from NDVI.
