Background Bridging biological and the socio-economic systems Social systems, such as cities and universities, are analogous to biological systems in their consumption of resources and production of waste. One obstacle to effectively measure the “sustainability” of university systems is the lack of meaningful metrics to compare universities. Rankings are typically based on per capita measurements The metabolism of complex systems tends to scale sublinear with size, so rankings based on per capita measurements may be flawed given these nonlinearities. Scaling laws as powerful tools to identify complex patterns Self-similar (fractal) patterns reflect underlying regularities of complex systems Power laws help to identify and describe these regularities Y(t) = Y0 X(t)β Y(t) is the dependent variable at time t, Y0 is a normalizing constant, X(t) is the independent variable at time t, and β, is the scaling exponent. β > 1 reveals increasing returns to scale, β = 1 demonstrates a linearity, and β < 1 reveals economies of scale. The scaling exponent is informative of the emergent non-linear properties that manifest across many orders of magnitude, implying an underlying ‘rule set’ that governs these patterns. Towards a more meaningful sustainability metrics The relationship of CO2 emissions to university population and physical size are general and quantitatively regular across universities. These scaling laws provide a null or the “generic university” across scales to which universities are compared. Deviations from the null serve as a new ranking system Results Log-transformed analyses: Total area (square footage of campus buildings) is the single best predictor of gross university emissions., with a R² value of Total population (students + faculty + staff) is also significant predictor of gross emissions., with R²=0.44 Combined linear regression of area and population increase the overall predictive value of gross emissions: to an R² value of 0.82 After area and population are accounted for, only temperature is a significant predictor of gross emissions; however, improves the model's predictive ability by approximately 3% Non-log-transformed analyses: similar to the results above, except R² values are lower, and temperature mean & standard deviation are not significant. Covariates that are not significant after college population and area are accounted for: Precipitation sum Precipitation days Mean humidity Humidity standard deviation Urban vs. suburban vs. rural Institution type (2-year vs. 4-year; private vs. public; doctoral- granting vs. bachelor's only) State Energetic scaling of U.S. academic institutions Meghan Balk¹, Mary Brandenburg¹, Robbie Burger¹, Michael Chang¹, Daniel Colman¹, Anna Davidson¹, Julian Davis¹, Tracy Diver¹, Trevor Fristoe¹, John Grady¹, Christian Gunning¹, Sarah Haft¹, Xiaoben Jiang¹, Jason Kimble¹, Jessica Martin¹ Kevin McCormick¹, Bryan McLean¹, Elizabeth Montano¹, Adeline Murthy¹, Melissa Pardi¹, Tatiana Paz¹ ², Shawn “Fred” Whiteman¹, and Natalie Wright¹ University of New Mexico, Department of Biology¹, University of New Mexico, Department of Computer Science² Figure 1. Location for all colleges and universities in our study. Relative differences in population sizes between academic institutions are represented by different size symbols. Population is based on the number of enrolled students, staff, and faculty for the year Acknowledgments Thanks to Jim Brown and Felisa Smith for their guidance and support. Figure 2. Gross emission of U.S colleges by building area (square feet) with an R² value of.77 and significant p-value of Figure 3. Gross emission of U.S colleges by building area (square feet) with an R² value of.44 and significant p-value of Table 1. Ranking of the ten most efficient and least efficient campuses based on total emissions, per capital emissions and emission residuals