| 英文摘要: | Wu et al. reply —
Macias-Fauria et al.1 highlight deficiencies in the high-resolution gridded climate database2, 3 prepared by the Climate Research Unit (CRU). In our analysis4, yearly averaged land surface air temperature (SAT) at each grid from this database was decomposed using the multidimensional ensemble empirical mode decomposition5, 6, 7, 8 (MEEMD) and these nonlinear secular trends from all grids were then pieced together into the spatiotemporal evolution of land SAT trends. Land SAT was independently decomposed grid by grid. The spatial and temporal biases of land SAT in the equatorial and Arctic coastal regions therefore do not impact the results derived for other regions.
Replacing unobserved data at an isolated temporal location with climatological values does not affect the determination of the trend (see methodology papers5, 6, 7, 8 and Supplementary Information accompanying Ji et al.4). Therefore, only those regions with an extended period of climatological values need extra attention. Because the trend derived for each spatial grid is local in time, the land SAT trends of later decades are not sensitive to earlier data bias as previously demonstrated8. This spatiotemporal locality of the MEEMD-determined trends can be verified using the MEEMD code and detailed computational information (Supplementary Information for Ji et al.4).
To assess whether our conclusions3 are affected by the data bias, we have repeated our analysis after removing regions lacking sufficient coverage before 1950. We identified these regions by plotting the standard deviation of the yearly mean SAT for the time periods 1901–1950 and 1951–2009 (Supplementary Fig. 1) and using a standard deviation of <0.05 K as the threshold. This threshold is a good indicator for identifying the non-interpolatable locations as they are filled with some spatially local and temporally optimized mean monthly climatology in the CRU database and therefore the land SAT values have a standard deviation close to zero for an extended period dominated by climatological filling. Supplementary Fig. 1 is in agreement with Fig. 1 of Macias-Fauria et al. Our new figure (Fig. 1b) does not differ much from the original Fig. 3 of Ji et al.4 (Fig. 1a) and our conclusions remain valid. |