Abundance distributions for tree species in Great Britain: A two-stage approach to modeling abundance using species distribution modeling and random forest

Abstract

High-quality abundance data are expensive and time-consuming to collect and often highly limited in availability. Nonetheless, accurate, high-resolution abundance distributions are essential for many ecological applications ranging from species conservation to epidemiology. Producing models that can predict abundance well, with good resolution over large areas, has therefore been an important aim in ecology, but poses considerable challenges. We present a two-stage approach to modeling abundance, combining two established techniques. First, we produce ensemble species distribution models (SDMs) of trees in Great Britain at a fine resolution, using much more common presence-absence data and key environmental variables. We then use random forest regression to predict abundance by linking the results of the SDMs to a much smaller amount of abundance data. We show that this method performs well in predicting the abundance of 20 of 25 tested British tree species, a group that is generally considered challenging for modeling distributions due to the strong influence of human activities. Maps of predicted tree abundance for the whole of Great Britain are provided at 1 km² resolution. Abundance maps have a far wider variety of applications than presence-only maps, and these maps should allow improvements to aspects of woodland management and conservation including analysis of habitats and ecosystem functioning, epidemiology, and disease management, providing a useful contribution to the protection of British trees. We also provide complete R scripts to facilitate application of the approach to other scenarios.

Keywords: abundance distributions; abundance–occupancy relationships; biotic effects; mapping.

Figure 1

Schematic showing the outline of the two‐stage method for predicting abundance distributions. The first stage uses SDMs to produce maps of predicted probability of occupancy, while the second stage takes these maps as inputs and uses Random Forest regression to produce maps of predicted abundance. Distribution data inputs are shown in square boxes and model covariates in round boxes, and model outputs are shaded in solid gray and modeling processes in hashed gray

Figure 2

Observed abundance against abundance predicted by Random Forest regression, as used to assess model performance, shown for four tree species. The line on each graph is the 1:1 line showing perfect model fit

Figure 3

Maps of predicted abundance for four species, in hectares per km², or percent cover. Note the scale varies between species. Maps for all other successfully modeled species are available to download from Sylva Foundation website and Oxford University Research Archive (see Data Accessibility)

Figure 4

(a) Presence records of Acer campestre, downloaded from the BSBI database (some of the best available distribution data at the country‐wide level). The data are presence only on 2 × 2 km (tetrad) scale. Note that where presence is not recorded, it is impossible to say whether the species is truly absent. Compare with our modeled abundance distribution (b) showing modeled hectares covered by A. campestre per square kilometer for every 1 km square