2021 ESA Annual Meeting (August 2 - 6)

census2sci4MGWR 0.0.900- an R package for improving researcher’s access to socioeconomic data for inclusion in Multiscale Geographically Weighted Regression (MGWR)

On Demand
Trey Hull, Integrated and Applied Sciences, Saint Louis University;
Background/Question/Methods

Ecological processes are inherently spatial and multiscale and can be dramatically influenced or altered by anthropogenic factors. Evidence of this can be found in the ecological impacts of urbanization. In many human-dominated systems, urbanization underlines the vital connections between “natural ecology” and “human ecology.” Despite this, it is not uncommon for US ecological research to neglect anthropogenic and spatial relationships. This is due in large part to limitations in the availability of socioeconomic data at different spatial scales, as well as the complexity of analyzing data collected at different spatial scales. To overcome these problems, we are developing an R package that will enable researchers to more efficiently access socioeconomic data via the US Census API (application programming interface), and to then mitigate potential error due to scale discrepancies through multiscale geographically weighted regression (MGWR) models. Our package was tested using mosquito data collected in St. Louis, MO, USA, from residential gardens participating in a conservation program, along an urbanization gradient.

Results/Conclusions

In developing this package, our primary goal is to provide researchers with the ability to apply MGWR to identify spatially dependent relationships that exist between ecological and socioeconomic variables. Results from testing package functions using mosquito data are consistent with known relationships. Additionally, a preliminary needs assessment was conducted to determine the viability of such a package. Initial results of the needs assessment identified four R packages for accessing and managing US census data, and two R packages with MGWR functions. However, the functionality of the equations within these packages may limit the usability of these features due to complex and lengthy input requirements, creating a potential obstacle to their use. So as to avoid creating a similar obstacle, our package features aggregate functions in order to reduce the complexity of user input. Packages such as ours will encourage researchers to include socioeconomic factors in their research, leading to an improved understanding of anthropogenic and spatial relationships on ecological processes.