Out of the total of 60 recorded species in the first period of fieldwork, 35 had the three or more datapoints required to construct a simple linear regression model. Of these, it was possible to create linear models with a p value < 0.05 for 22 species using untransformed data, as shown in table 3. Forcing the intercept through the origin seemed to have positive effects on the R2 values of the models, which were generally very high, with an average value of 0.92. The slope coefficients (b) of untransformed models ranged from roughly 0.02 to 1.03, however, the 1. quartile was 0.06 and the 3. quartile 0.14, meaning that half the species fall within this much narrower window.
 
In order to to extend model coverage to species with insufficient data, models for selected genera were constructed. In order for the genera-based models to be accurate, the constituent species must share largely similar morphology. If two or more morphological types in the same genera were sampled, each with more than one species, several models could potentially be constructed, although that was not necessary for this study. As seen in table 3, all models had p values <0.001, with high r2 values for Agrostis, Carex and Galium, but not for Festuca and Juncus.
 
A number of models based on a priori morphological types were also constructed, covering all the species in the dataframe, except for the two Juncus species, which were covered by a genus model. Species were assigned to a morphological type according to the following criteria.
 
 
·       Broad leaved graminoids
o   Graminoids with leaf blades averaging > 2 mm. at the widest point.
·       Narrow leaved graminoids
o   Graminoids with leaf blades averaging < 2 mm. at the widest point.
·       Forbs
o   All species of forbs.
·       Trees and shrubs
o   Plants with woody stems (Calluna vulgaris and Cytisus scoparius)
 
Forbs were originally divided into a model for ground-hugging forbs with basal rosettes or similar low growth patterns and those with more upright growth. The model estimates were so similar, however,  that they were combined into a single model with comparable R2 and lower standard error. All morphological type models had p values < 0.001 and, with the exception of “Narrow leaved graminoids” and “Effusus type rushes”, all had r2-values above 0.76, meaning that they would in theory be more accurate than the general model if used for modelling the species that had insufficient data for individual models.