16 4
All temperature and precipitation data used in the analyses were based on cli-
mate data collected from 1961 to 1990 that were associated with locations listed in
species collection records. The online GrassPortal database (Osborne et al. 2011 )
was used to access these data. GrassPortal is a collaborative project that allows spe-
cies data from Kew Gardens GrassBase to be combined with phylogenetic data
developed at the University of Lausanne, Lausanne, France, collection records
available in the Global Biodiversity Information Facility (GBIF 2013 ), and average
climate data compiled by the University of East Anglia Climatic Research Unit
(Osborne et al. 2011 ). Appendix 6.1 contains a list of data sets and data providers
for collection records accessed through GrassPortal and GBIF. We defi ned the fol-
lowing variables based on data available in GrassPortal : minimum temperature
(mean annual minimum temperature), maximum temperature (mean annual maxi-
mum temperature), temperature range (maximum temperature minus minimum
temperature), minimum precipitation (minimum annual mean precipitation), maxi-
mum precipitation (maximum annual mean precipitation), and precipitation range
(maximum precipitation minus minimum precipitation).
We ran a series of univariate linear logistic regressions with SAS 9.2 using
Procedure Logistic to analyze relationships between the dependent variables and
the independent variables. Each of these regressions used one dependent variable
and one independent variable. The number of species used for each univariate
regression depended on available data.
We ran four logistic multiple regressions (each with one of the four dependent
variables and multiple independent variables) for the 54 species for which multifac-
tor data were available. For these analyses, we used Procedure Logistic with a back-
ward stepwise elimination and the following independent variables : year, life span,
mean seed mass, maximum awn length, ploidy level, human use, cultivar availabil-
ity, minimum temperature, maximum temperature, minimum precipitation, and
maximum precipitation. We then repeated the four multiple regressions substituting
temperature range and precipitation ranges for the minimum and maximum tem-
perature and precipitation variables. When we evaluated the results of the logistic
multiple regressions, we considered factors to be signifi cantly correlated if p -values
were less than 0.05.
6.2.3 Histograms
We created histograms showing species counts for section, year, maximum awn
length, mean seed mass, ploidy level, human use, cultivar availability, and climate
variables. All species were included in the histograms, not just those included in the
statistical analyses. For some of these histograms, we separated species into three
groups: sections Bromus , Genea , and Nevskiella (Eastern Hemisphere species with
mostly annual life spans), sections Ceratochloa and Neobromus (polyploid Western
Hemisphere species with annual to perennial life spans), and section Bromopsis
(with perennial life spans).
S.Y. Atkinson and C.S. Brown