Model predictions. Stephen E. Fick Rand R. Evett 10.1371/journal.pone.0194315.g004 https://plos.figshare.com/articles/figure/Model_predictions_/6087926 <p>Individual model predictions for soil phytolith content (percent dry weight), and the ensemble model. A) RandomForest, B) GAM, C) Lasso and D) Ensemble model predictions. Note that model predictions are extrapolated beyond the region of confidence (< 3 ‘environmental PC’ units) used in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0194315#pone.0194315.g003" target="_blank">Fig 3</a>. Interior blank areas are regions lacking soils data.</p> 2018-04-04 17:33:55 insight soil phytoliths Central Valley restoration Historical reconstructions pre-invasion plant species variability hypothesis extent species distribution models pre-Columbian Central Valley pre-Columbian California grasslands plant community distributions abundance soil phytolith data SDM