Species distribution models and predictions tend to skew data trends and create bias predictions in biodiversity

Using species distribution models only may underestimate climate change impacts on future marine biodiversity

Summarized by Stephanie Sanders, a Geology major at the University of South Florida. She is currently a senior. Stephanie plans to attend Graduate school in paleobiology and once she graduates would like to work in the marine conservation field with a concentration in geographic information system (GIS) for mapping marine species. Outside of her studies, Stephanie enjoys the outdoors, drawing, and fostering kittens. 

Main point of paper: Anthropogenic (human-caused) climate change has affected a various number of species. We are seeing shifts in geological range, population counts, and environmental conditions. Species distribution has been affected due to these changes. Species distribution models (SDMs) are used to assess geological insight and predict distributions across different landscapes and time. SDMs use observations and estimates to make predictions on species occurrence and location. Many of these modeling software has ready-to-use generic software that is globally accessible. This ease of use has created databases that may not consider a wide range of important data sets that are crucial in interpreting the data. Some of these data sets include predation, species interaction, competition, and adaptation. This additional information that is excluded from many SDMs has created data bias. These biases include overestimating gains in data and underestimating losses in data. When only looking at the SDMs, we are losing vital data and creating data sets that may not accurately represent the species data. By only using SDMs data we may be overestimating the number of species present in certain times and locations which lead to distortion in data and predictions for future marine biodiversity. The consequences of this can be inadequate data for conservation efforts of species. The authors ask in this paper: Is generic SDM data enough to correctly predict future biodiversity or are additional data sets required to accurately represent species changes from anthropogenic climate change?

What data were used: Data from 100 species of various vertebrates and invertebrates from the Mediterranean Sea were used, downloaded from different online databases. Both a hybrid SDMs model of multispecies modeling and a hybrid OSMOSE-MED modeling (explained below), which allows for key life processes such as population growth, reproduction, and morality, were considered. 

Methods: The Mediterranean Sea served as a perfect spot for a comparison because it has a rare mix of biodiversity and is a global change hotspot. SDMs data is compared to other data modeling such as OSMOSE-MED. OSMOSE-MED allows for additional data sets to be considered (Figure 1). One hundred marine species such as fish, invertebrates and gastropods from the Mediterranean Sea, collected from the Global Biodiversity Information System (GBIF), the Ocean Biogeographical Information System (OBIS), the Food and Agriculture Organization’s Geonetwork Portal, and the FishMed database were compared with both SDMs modeling and OSMOSE-MED modeling. Two data sets were compared, a present day (2006-2013) and a future prediction (2071-2100). 

Results: Under the 100 species comparing SDMs to OSMOSE-MED, the following results were found: more species were found to have an increased geographical range in the SDMs model than the OSMOSE-MED model. Fewer species were found to have a decreased geographical range in the SDMs model than the OSMOSE-MED model, and three more species were projected to become extinct in the OSMOSE-MED model then the SDMs model. When using SDMs alone, a more optimistic projection of species distribution is observed. Without the inclusion of key life process data to predict future trends, the SDMs data alone could predict inaccurate data.  When we compare the two data sets, there is a clear discrepancy. SDM overestimates the positive data and underestimates the negative data. When looking to SDM alone, without the addition of extra relevant data sets, we can get a bias determination on future conservation data. This can include area prediction of where species are now inhabited or the number of species that are alive within a certain species. This can lead to improper protection areas or conservation efforts for a certain species. 

The figure above is a conceptual representation of a SDMs model showing the limited layers of species richness and dissimilarity index compared to the OSMOSE-MED model that allows for extra parameters to be added to fine tune the model to allow for more accurate representation of the model. OSMOSE-MED allows for more layers and is represented by a picture of those layers such as growth, morality, predation, and reproduction.
Conceptual representation of SDMs model vs. OSMOSE-MED model to accurately represent data and the projections that are compiled depending on the inclusion of extra data sets. OSMOSE-MED allows the addition of additional datasets that can more accurately represent the data of biodiversity that generic SDM data cannot.

Why is this study important?: Anthropogenic climate change is affecting a variety of species constantly. Many of these species will encounter a dramatic loss in population and even extinction. In order to put protective measures in place to help the longevity of threatened species, correct data is critical. If present and future studies are only utilizing generic models that don’t consider vital variables like predation, location predictions, and taxonomies. we may lose accurate calculations. The real-world implications of skewed data can be present in inaccurate locations of species, inaccurate population counts, and misinformation of vital data needed in the protection of threatened species. 

Broader Implications beyond this study: This specific study of SDMs vs. more in depth distribution models focuses on 100 species in the Mediterranean Sea. However, tracking anthropogenic climate change and the effects on species distribution is a global effort. In order to create accurate data sets, a global and local collaborative database is essential for comparative analysis of biodiversity. There are potentially major issues with inconsistent taxonomic standards applied when using SDMs data only. As the climate crisis begins to affect more species, additional data to create a global standard will be required. If we are to effectively create conservation efforts for threatened species, these standards need to be adapted and used regularly. 

Citation: Moullec, F., Barrier, N., Drira, S., Guilhaumon, F., Hattab, T., Peck, M. A., & Shin, Y.-J. (2022). Using species distribution models only may underestimate climate change impacts on future marine biodiversity. Ecological Modelling, 464, 109826. https://doi.org/10.1016/j.ecolmodel.2021.109826