Linking earth observations, citizen science data, long-term monitoring, and functional traits to improve species abundance distribution models

Describing the distribution and density of species across landscapes is a fundamental, but challenging endeavor. The approaches to this problem can be broadly broken down into top-down or bottom-up methods. Top-down methods use regressions to relate species presence or absence to underlying environmental conditions before extrapolating to unmeasured points in space or into novel environmental conditions under scenarios of change. This allows for the immense spatial and taxonomic scope, but typically biases towards a climate-centric view of species distributions. In contrast, local, bottom-up, studies often show hugely important, and highly localized, impacts of habitat availability on organism presence and abundance.

In this talk, we lay out what we hope to be an effective framework to potentially bridge these two approaches. We combine the data products traditionally used in top-down approaches with a high-quality local dataset of bird counts conducted along 48 transects continuously for 20 years. The local sites are located within four climactically distinct regions and within each region along a forest-agriculture gradient, which represents the principle driver of land-use change worldwide. We make use of remotely sensed data layers, continentally distributed presence records, our long-term monitoring program, and functional traits to improve both the fit and resolution of species abundance distribution models.

We were able to obtain ranges that overlapped Costa Rica and valid trait values to estimate forest response curves for 740 species. We mapped the sum of both the number of species predicted to occur in a pixel (A) and the total abundance of birds in that pixel (B) under current forest cover. Shown for comparison is a recently published estimate of bird species richness at a global level (C; data from Jenkins et al. 2013). Our findings on forest affinity in various bird species also align well with known biological patterns. For instance, we show that fruit and nectar specialists are likely to benefit from the conversion of forest to agriculture, while insectivores will be negatively impacted. We believe this pipeline of data processing and modeling can be extrapolated to other taxa and land-use change drivers in different regions and perhaps, with enough data, globally.