Ecological survey data are commonly collected for a particular task or purpose, usually with a heavily applied focus. These data often cover vast spatial and temporal scales yet, in their original form, are not always suitable for additional analysis. For example, plant abundance may be measured indirectly through proxies like crown-cover. Large scale, well-replicated and consistent vegetation survey data is lacking worldwide. We show that converting these existing sources of data to more versatile measures can be achieved with a fraction of the cost and effort required to resurvey.
Using a large dataset (n = 1,251 sites) originally collected to inform vegetation mapping in South East Queensland, Australia, we predicted the abundance of woody plant species from crown-cover estimates. To do this we sampled an additional 30 “calibration” sites replicating the original methodology but also measuring stem counts. Generalised additive mixed effects models were used to predict the stem counts for all woody species in the original dataset (n = 927 spp.). We used the predicted abundance data to investigate patterns of local and regional rarity between woody plant species of different heights.
Results/Conclusions
Non-linear models were effective at predicting stem counts with R2 values of 0.61 for the shrub model and 0.57 for the tree model. Stem densities increased rapidly with crown cover before plateauing at around 5% crown-cover for shrubs and 50% crown-cover for trees. Using the predicted abundance data, we found that shrubs were more abundant at local scales than trees, which were more regionally abundant. Additionally, we showed that there was a negative relationship within large trees (> 20 m maximum height) with local abundance declining with increasing regional abundance.
Many large datasets have potential for application beyond their initial intent. In this case, the original dataset included a proxy for abundance that was converted for a fraction of the investment (time and cost) that would be required to resurvey an equivalent number of sites. Analysis of the data following conversion revealed that trees are less abundant than shrubs locally but more abundant regionally, possibly due to properties of space filling and dispersal ability. These data can now be applied to a range of concepts such as functional ecology, metabolic theory, competition and climate change projections.