Structured Predictive Models for Climate Informatics

Abstract:

Predictive modeling of complex climatological phenomenon has
emerged as an active area of research in climate informatics. From
the perspective of accuracy as well as interpretability, a key
requirement for success in such modeling is the ability to find
good predictors for climatological phenomenon of interest. The
challenges in finding such good predictors come from the fact that
one has to explore a wide range of spatial and temporal features,
different temporal lags, spatial teleconnections, etc., using a
small number of samples of the phenomenon of interest.
Further, there could be multiple mechanisms affecting the
phenomenon, possibly under different phases or even as a
superposition. In this talk, we briefly discuss recent advances in
structured predictive modeling, along with promising applications
for modeling regional temperature and precipitation.