Technical Comments

Comment on “Changes in Climatic Water Balance Drive Downhill Shifts in Plant Species’ Optimum Elevations”

See allHide authors and affiliations

Science  14 Oct 2011:
Vol. 334, Issue 6053, pp. 177
DOI: 10.1126/science.1203791


Crimmins et al. (Reports, 21 January 2011, p. 324) reported that plant species moved downhill between 1935 and 2005. They compared plot data for two time periods, ignoring that the modern plots were farther north than the historical plots. I contend that there is no support for a general downhill shift after correcting for this geographic bias.

Crimmins et al. (1) compared California survey plot data from two time periods (around 1935 and around 2005) and concluded that plant species moved downhill (88 m on average), tracking increased precipitation, not temperature. The historical plots were mostly located in the coastal ranges south of San Francisco and in the central Sierra Nevada. The modern plots were predominantly from places much farther to the north—the Klamath Mountains and the Cascade Range—and from the entire Sierra Nevada [see figure S2 in (1)]. These more northern regions are colder and generally wetter than the more southern regions. In California, the average temperature at a given elevation decreases with about 0.4°C per 50 km when going from south to north, and the temperature decreases by about 0.4°C for each 100-m increase in elevation (2). Therefore, a downhill shift of 100 m would be expected when comparing sites that are 50 km apart in latitude.

Crimmins et al. state that they adjusted their data for “geographic and/or environmental bias between our samples” [supporting online material for (1)]. They subsampled the two data sets to obtain comparable distributions in terms of water deficit for both time periods. However, they did not rigorously define bias, nor did they explain why their bias correction is warranted or provide information about its effect.

I computed latitudinal bias between the two sets of plots as the difference in the median distance south of 42°N (the latitude of the California-Oregon border) (3). The entire data set had a bias of 311 km for the plots and 161 km for the presence records of the species for which Crimmins et al. presented results. Bias remained considerable after the Crimmins et al. adjustment: 144 km for the plots and 104 km for the species. A linear regression between latitudinal bias and the reported change in elevation for each species (P = 0.04) had an intercept of –15 m and a slope of –79 m change in elevation per 100 km bias. This suggests that the latitudinal bias in the samples could explain much of the reported change in elevation.

I repeated Crimmins et al.’s procedure to compute the “optimum elevation” for their subset and the entire data set and obtained the same results as they did (–88 m elevation change) for their subset and –144 m for the entire data set. I attempted to remove geographic bias by pairing historic plots to their nearest current plot, using each plot only once, and only considering pairs within a maximum distance of 1 or 5 km of each other. For this much less geographically biased subset, I found an average change in elevation for the species studied of +3 m when using a maximum distance of 1 km between plots and of –15 m when using a maximum distance of 5 km (4). Neither was significantly different from zero (t test). I illustrate the effect on the two sampling methods in Fig. 1.

Fig. 1

Distribution of Juniperus occidentalis around 1935 (historic plots, circles) and around 2005 (modern plots, crosses) based on plot data sampled by Crimmins et al. (1) (A) and with my paired plot method (5-km maximum distance between plots) (B). Crimmins et al. estimated a decrease in optimum elevation of 745 m; I estimated a decrease of 207 m.

It is likely that some California plant species have shifted their range downward during the past 100 years, as has been observed for some mammal species (5). Identifying and analyzing such cases could help refine our understanding of species responses to climate change and to other factors such as fire suppression. However, the correction for geographic bias shows that the general downhill shift reported by Crimmins et al. (1) is not supported by their data. Moreover, they did not demonstrate that the species that moved downhill tracked water deficit better than those that did not move downhill, and the change in precipitation was only weakly supported (alpha = 0.10) and was strongly driven by drought in 1920 to 1935. My pairing method removed much of the geographic bias, but to draw general conclusions, more detailed analysis of bias in the data, the biology of the species, and the changes in the environments where they occur would be required. Finally, because of the effect of latitude on temperature, the concept of “optimal elevation” should generally be avoided.

References and Notes

  1. I thank S. Dobrowski for providing the data. My analysis of the data omits 16% of the modern plots because they did not have coordinates and could not be used to pair plots.
  2. Sampling changed the number of plots and species analyzed. I followed Crimmins et al.’s method to select species with a unimodal response curve. Their sample size was 8747 plots in each time period, in which they found 88 species with a unimodal response to elevation. There were 2843 plots with 71 species after paired plot sampling with a 5-km maximum distance, and 959 plots with 48 species after sampling with a 1-km maximum distance.
  3. Acknowledgments. I thank S. Harrison and B. Anacker for comments on the manuscript.
View Abstract

Navigate This Article