Using fossil leaves as paleoprecipitation indicators : An Eocene example : Comment and Reply

Wilf et al. (1998) analyzed leaf sizes in modern vegetation and concluded that leaf sizes accurately predict mean annual precipitation (MAP). However, we question the methods employed by these authors to derive leaf size; the modern samples are not comparable to one another or to fossil leaf assemblages, and the influence of temperature on leaf size is ignored. Data for some samples were derived primarily from sizes cited in floral manuals, which cover the entire range of a species. Can this method produce even approximately valid size data? We compiled data for eight sites on the island of Yakushima in southern Japan. These sites were collected for CLAMP (Climate-Leaf Analysis Multivariate Program; Wolfe, 1993, 1995) samples; the full range of physiognomy, including leaf size, of each species of woody dicot was collected in limited areas analogous to areas represented by fossil leaf assemblages. Leaf sizes were calculated according to the method of Wilf et al.: The smallest and largest leaf sizes for species in the sample were based on two-thirds the length times the width for (1) the shortest and/or narrowest leaves and (2) the longest and/or broadest leaves; the resulting areas were converted to natural logs and then averaged to yield a mean natural log of the leaf area, which was then averaged for each sample. The same procedure was used for data taken from the pertinent floral manual (Ohwi, 1984). Obviously this procedure cannot validly delimit the smallest or the largest leaf sizes in every species, because in many instances, the shortest and narrowest measurements were not necessarily found on the same leaf, and likewise for the longest and widest measurements. In 25%–30% of the species analyzed, the low-end extreme measurements occurred on different leaves, as did the high-end extremes. Leaf areas derived from actual samples are markedly smaller (Table 1) than those derived from Ohwi (1984). Comparison was also made for sizes of a sample from Ketchikan, Alaska, to data from the pertinent manual (Vierick and Little, 1972). These comparisons indicate that (1) manual-derived data have a different mean leaf size than sample-derived data from a restricted climate zone and (2) size relates poorly to MAP, especially for temperate, high rainfall samples. Using Wilf et al.’s equations and mean leaf sizes, predicted MAP for the Yakushima samples is ~113–133 cm, and for Ketchikan, 64 cm. Neither the manual-derived plots nor the sample-derived plots for Yakushima fit Wilf et al.’s Figure 2 regression, and for Ketchikan, which has milder winters than the two Pennsylvania samples used by Wilf et al., both manualand sample-derived plots also deviate markedly from other plots. Regression of leaf size against MAP with the addition of the Yakushima and Ketchikan samples as in Wilf et al.’s Figure 2 reduces the r2 to 0.35. Leaf sizes from manuals do not yield valid site-specific data; more work may be involved in obtaining actual samples, as in CLAMP, but clearly actual collections are needed for valid calibration of physiognomy to climate. Leaf-size data employed by Wilf et al. from different samples are not comparable. Sarmiento (1972) measured only canopy leaves and Bongers and Popma (1990) measured only “sun-leaves” (presumably = canopy), and canopy and/or sun leaves are smaller than shade and/or subcanopy leaves (e.g., Richards, 1996). Only part of the flora was included in the size analyses, because subcanopy trees and shrubs were excluded. In contrast, data reported by Dolph and Dilcher (1980) represented all woody plants in a sample plot. The data compiled from Hall and Swaine (1981) were based on averages for each species of “leaves of mature plants not saplings” (p.105), although Hall and Swaine (p. 49) recognized that “the leaves of many species . . . are much larger . . . in the young sapling stage than in the canopy.” Dolph and Dilcher’s (1980) samples are comparable to what might be found in the fossil record, but the samples based on Hall and Swaine (1981) were derived from thousands of square kilometers. Leaf size, as in the instances of many other physiognomic character states, cannot be correlated in an isolated, univariate fashion to any one environmental parameter. Large leaves require both high moisture and high heat (Richards, 1996, p. 100). For example, lowland Yakushima samples have a lower score in the mesophyll categories (~16%–24%) than do samples from Fiji (32%–59%), although lowland Yakushima has considerably higher MAP (~430 vs. 200–300 cm) than does Fiji. However, Fiji is warmer, with a MAT of ~25 °C as opposed to ~19 °C, and thus the climate is more conducive to large leaves. Another example of the influence of temperature on leaf size is subalpine (including subarctic) mesic vegetation relative to non-subalpine mesic vegetation (Table 2). The reduction in leaf size in subalpine vegetation is surely not the result of reduction in MAP, because these subalpine samples have higher MAP, and this reduction cannot be attributed to winter cold, because these subalpine samples have higher CMMT (cold-month mean temperature). Low heat during, and brevity of, the growing season are major features of subalpine climates and reduce average leaf size; fossil assemblages that display physiognomy characteristic of subalpine vegetation are found in the Paleogene of Idaho and Colorado (Wolfe et al., 1998) and in the Miocene of Alaska and Kamchatka (Wolfe, 1994). The complexity of the interactions of various environmental parameters that produces various compromises in physiognomic adaptations demands a more sophisticated approach than presented by Wilf et al.