3.2. Environmental variables
Simple variate – covariate plots showed that C. reticulata seed oil content decreased with elevation, slope and rainfall, and increased with temperature (Fig. 2a–d). It should however be noted that temperature was strongly correlated with elevation (Pearson’s R = −0.93), and there was also a correlation between slope and rainfall (Pearson’s R = 0.71). Significant predictors in the final models were elevation and soil type with both forward and backward selection converging on the same model (Fig. 3). The regression coefficients show that, on average, the seed oil content decreased by ∼1% with every ∼45 m increase in elevation (Fig. 4). At a given altitude seed oil content was higher on haplic
Acrisols (ACh) and, to a lesser degree, on humic Acrisols (ACu) and haplic Alisols (ALh); it was on average 8–11% less on haplic Luvisols (LVh) and chromic Cambisols (CMx) (Fig. 4). Overall the variation in the seed oil content explained by soil type and elevation was 24% (F = 23.47; d.f. = 473; p < 0.001; R2 = 0.244), whereby soil type accounted for 15% of the variation on its own, and elevation for 7% on its own. This final model was extrapolated to highlight areas of similar macro-environments to areas where Camellia seed oil production is demonstrably high, i.e. potentially suitable habitat for the production of Camellia seed oil (Fig. 1). Areas in red and yellow have potential for high oil production, and areas in green and blue are likely to be less productive. The prediction is conservative in that it is clamped to cells with the same soil types and from the same environmental range as the sample data points, i.e. there may be other suitable areas for the production of Camellia seed oil that are not captured in our input data.