Spatial sampling bias in the Neotoma paleoecological archives affects species paleo-distribution models

QUATERNARY SCIENCE REVIEWS

卷:198

Article

Inman, Richard^1,2  Franklin, Janet³  Esque, Todd¹  Nussear, Kenneth⁴ 

[1] US Geol Survey, Western Ecol Res Ctr, Las Vegas Field Stn, 160 N Stephanie St, Henderson, NV 89074 USA

[2] Arizona State Univ, Sch Geog Sci & Urban Planning, POB 875302, Tempe, AZ 85287 USA

[3] Univ Calif Riverside, Dept Bot & Plant Sci, 900 Univ Ave, Riverside, CA 92521 USA

[4] Univ Nevada, Dept Geog, 1664 N Virginia St, Reno, NV 89557 USA

关键词: Holocene;    Paleogeography;    North America;    Data treatment;    Data analysis;    Neotoma;    Spatial sampling bias;    Species distribution modeling;    Niche breadth;    Macrofossils;   

DOI  :  10.1016/j.quascirev.2018.08.015

来源: Elsevier

【 摘 要 】

The ability to infer paleo-distributions with limited knowledge of absence makes species distribution modeling (SDM) a useful tool for exploring paleobiogeographic questions. Spatial sampling bias is a known issue when modeling extant species. Here we quantify the spatial sampling bias in a North American packrat midden archive and explore its impact on estimating paleo-distributions. We test whether (1) spatial sampling bias inherent in this macrofossil record can influence estimates of paleodistributions, (2) this bias can alter the ability to measure shifts in distributions and climatic niche breadth from the Northgrippian subdivision of the Holocene (8.3 ka - 4.2 ka) to present day (1950 - 2000 yr), and (3) bias correction methods can improve estimates of paleo-distributions and analyses of range shifts and niche breadth. We estimate spatial sampling bias for the mid-Holocene period with a three-stage statistical model, each representing a hypothesized source of bias: fossil site availability, preservation and accessibility. This approach enables the use of SDM to evaluate three separate paleodistributions calibrated on the packrat midden archive: those without bias correction (sigma-naive), those created with a standard method (sigma-standard), and those created with a novel alternative (sigma-modeled) incorporating the three-stage model of bias. We find that paleo-distributions modeled for the mid Holocene without bias correction (sigma-naive) provided poor estimates of hindcast paleo-distributions, and that the sigma-modeled correction method improved paleo-distributions for our six species with, on average, 50% higher overlap to hindcast distributions than sigma-naive paleo-distributions (sigma-standard results fell between sigma-naive and sigma-modeled). Published by Elsevier Ltd.

【 授权许可】

Free