The Random Forests classification algorithm was used to predict the occurrence of the realized climate niche for two sub-specific varieties of Pinus ponderosa and three varieties of Pseudotsuga menziesii from presence-absence data in forest inventory ground plots. Analyses were based on ca. 271,000 observations for P. ponderosa and ca. 426,000 observations for P. menziesii, with ca. 6% of the observations in each dataset recording the presence of one of the varieties. Classification errors to the respective databases attributable to fitting the models were ca. 5%, most of which were from falsely predicting varietal occurrence. Confusion in classifying varieties was nil. The primary drivers of the niche model were summer precipitation, winter precipitation and summer degree-days >5 C for the varieties of P. ponderosa and the summer-winter temperature differential, summer maximum temperatures and summer precipitation for the varieties of P. menziesii. Projected impacts of global warming using output from an ensemble of 17 general circulation models were greater for P. ponderosa than for P. menziesii and for varieties of both species from inland climates than from coastal. Projected impacts imply dire consequences for the varieties of P. menziesii occurring in Mexico.
All Science Journal Classification (ASJC) codes
- Nature and Landscape Conservation
- Management, Monitoring, Policy and Law