COS 7-2
Modelling net primary production using plant traits: Development and testing of a finely-resolved predictive model for temperate ecosystems

Monday, August 10, 2015: 1:50 PM
321, Baltimore Convention Center
Simon Smart, Land Use Group, Centre For Ecology and Hydrology, Lancaster, United Kingdom

Large-scale dynamic models can reproduce gradients of above-ground Net Primary Production (aNPP) at coarse resolution but are less sensitive to the patch-scale effects of land-use interacting with plant trait variation. To facilitate testing and coupling of large-scale dynamic and patch-scale empirical models across scales we developed a model predicting aNPP in terms of abundance-weighted plants traits associated with the global Leaf Economics Spectrum. Our aims were three-fold:

1) To derive a Minimum Adequate Model that could predict aNPP across temperate vegetation types for the purposes of down-scaling and enhancing dynamic land-atmosphere models for policy scenario testing.

2) To test whether prediction of aNPP was significantly improved when subordinate species were included or whether adequate prediction could be achieved based on co-dominants only.

3) To test whether prediction was significantly improved by using trait values measured in the same plots as aNPP was determined or whether adequate prediction could be achieved using trait values from existing databases for British species. 

We measured aNPP in fixed vegetation plots in ecosystem types distributed across the full range of successional and productivity gradients in Britain. Model building was based on combinations of Specific Leaf Area (SLA), Leaf Dry Matter Content (LDMC), bare ground, bryophyte cover, woody species cover, canopy height and climate variables.


The most parsimonious and best-performing model was based on omission of subordinate plant species but incorporating field rather than database values of traits. This model predicted the natural log of aNPP in terms of abundance-weighted LDMC as the only covariate.

Analysis of residuals showed that observed aNPP was increasingly under-predicted as agricultural land-use intensity increased. This could reflect the higher productivity of modern land-use regimes compared to the environments in which present-day agriculturally favoured species evolved. An alternative hypothesis is that, if measured and introduced into the model, intraspecific trait variation could account for this residual variation.

We show that patch-scale aNPP can be satisfactorily estimated by abundance-weighted plant traits. Our statistical model can be used to usefully estimate fine resolution aNPP in temperate ecosystems and therefore the amount of sub-grid variation not accounted for by large-scale land-atmosphere models.