Improving Forecasts Made by Weather and Climate Models
Weather and climate model forecast errors grow with time as a result of atmospheric instabilities that amplify uncertainties in the initial conditions, and deficiencies in the numerical model which introduce errors during integration. Weather predictions have benefited dramatically from advances in the science of data assimilation (DA), essentially the attempt to synchronize a computational solution with physical observations of the modeled system by a clever combination of the uncertainties associated with each. However, climate models suffer from errors related to the difficulties associated with resolving multiple temporal scales, a problem DA methods have yet to appropriately address.
The present research project aims to develop and test a new advanced mathematical method for improving predictions made by climate models, focusing on DA and the coupling of slow- and fast-scale phenomena. The techniques are being developed using simple (N ~ 10 variable ODE) and sophisticated (N ~ 10^6 variable CFD) models of a thermal convection loop, a toy climate analogous to the convection system studied by Lorenz, to predict regime changes (flow reversals) in a laboratory fluid dynamics experiment.
We are also testing methods to improve forecasts on the National Center for Environmental Prediction's state-of-the-art Global Forecast System (GFS, N ~ 10^9) to improve short-range forecasts and suggest improvements to the model's physical parameterizations.For information about his research and more, please see Chris Danforth's homepage.
Peter Dodds and Chris Danforth
Measuring Population-Level Happiness Using Social Networks
The importance of quantifying the nature and intensity of emotional states at the level of populations is evident: we would like to know how, when, and why individuals feel as they do if we wish, for example, to better construct public policy, build more successful organizations, and, from a scientific perspective, more fully understand economic and social phenomena.
Normally measured through self-report, happiness has often been indirectly characterized and overshadowed by more readily quantifiable economic indicators, such as gross domestic product. By incorporating direct human assessment of words, we quantify happiness levels on a continuous scale for a diverse set of large-scale texts including song titles and lyrics, weblogs, and status updates expressed by over 20 million users in the online, global social network Twitter.
Our method is transparent, improvable, capable of rapidly processing Web-scale texts, and moves beyond approaches based on coarse categorization. We use a real-time, remote-sensing, non-invasive approach a kind of hedonometer to uncover collective dynamical patterns of happiness levels. With a growing Twitter data set comprising more than 10 billion words, we are exploring temporal happiness variations at time scales of hours, days and months, as well as the relationship between happiness and perceived popularity (number of followers), geography, diversity of word usage, and social network connections.
For more information, see "Happiness Study Featured in 'Times,' 'Science,' and 'Chronicle.'"
Biological Control Systems
Biological systems use feedback to respond to changes in their environment. Even the simplest, single-celled microorganisms can quickly adapt to new conditions. We are interested in studying the feedback networks and gene regulation that allow microbes to respond to changing conditions. The time scale of regulation can span milliseconds to hours, indicating complexity within even the simplest organisms.
To study feedback and regulation, we express fluorescent proteins (similar to those in a firefly) in microbes and use them to measure gene expression within single cells. By taking time-lapse images under the microscope, we can track how the dynamics of gene expression change with time. We are interested in understanding what causes regulation to be robust or variable.
By studying simple biological systems we can gain insight into how more complicated organisms use feedback to quickly respond to changing environments. In addition, minimal feedback networks have important engineering applications for devices that are limited in size or computational ability.For information about her research and more, please see Mary Dunlop's homepage.