BRT results and supporting simple regression biplots
Despite the observed patterns of increase, decline, or unimodal response to latitude for P, D, and D&P respectively (Fig. 1b) this variable was not found to be a significant predictor in any of the BRT analyses (Fig. 3–5; Appendix 3). Instead, for both haploid/diploid and polyploid species, temperature (either annual average or maximum of the warmest month) had strong predictive power as did evapotranspiration. The simplified models, using only the significant predictors for haploidy/diploidy and polyploidy, explained 92% and 81% of the variance respectively. The standard regression trees also explained a high proportion of variance (Fig. 3–5) and they indicated that macrophyte species richness had a positive influence in the case of haploidy/diploidy and negative for polyploidy. Species number explained some variation in percentage of haploidy/diploidy at average temperatures above 17 °C. Conversely the percentage of polyploidy was influenced by species number below 15.7 °C with the highest incidence of polyploidy at low temperature and low species number (Fig. 3 and 4).
The models for mixed ploidy were not as strong, though 75% of the variance was explained, by two variables – endemic species richness of macrophytes and average temperature (Fig. 5). The standard regression tree included other variables (evapotranspiration and temperature range) but the association with endemic species richness was marked, with mixed ploidy only occurring once endemic richness in a gridcell fell below a value of 150, with a negative association thereafter, and the percentage of mixed ploidy rising as endemic richness declined.
Climatic and biotic variables were stronger drivers of ploidy state than location in either human-impacted landscapes (CROP) or highly-stressed areas of the planet (CCV and ALT). Three current climate variables (potential evapotranspiration: ET0; temperature range: TRG; annual precipitation: PCP) were strong predictors of ploidy state, though varying in importance between the three models.