Cascading, multiscale information and drought impacts in the humid tropics and sub-tropics: potential for prediction and risk mitigation

The El Nino Southern Oscillation (ENSO) is well identified as associated with drought or floods in much of the tropics and sub-tropics with very specific seasonality in the impact. ENSO is a multiscale spatio-temporal phenomena with a typical frequency of 2-7 years. However, the amplitude and phase of ENSO as well as its spatial manifestation in the tropical Pacific and ensuing season to sub-season global teleconnections, are modulated by decadal and century scale modes that can be discerned from long model runs as well as paleo-climatic records. The interaction with the Madden-Julian Oscillation and the resulting responses in the Inter-Tropical Convergence Zone are particularly important for the sub-seasonal character of the tropical and sub-tropical impacts. This multiscale climate variability due to ENSO and other climate phenomenon, has not been effectively translated into a corresponding spatio-temporal risk analysis for the global tropics and sub-tropics.

In this talk, we will present two types of analyses. First, we will present evidence of the ability to predict the seasonal and sub-seasonal precipitation in Florida, Colombia and Brazil using climate precursors using statistical and machine learning models, as input to drought analyses. Second, we will demonstrate the multiscale variations in the global co-occurrence of drought in the tropics and sub-tropics. This analysis is intended to inform strategies that global humanitarian relief agencies and insurance/ re-insurance organizations may need to consider for the design of their financial and resource management programs, since the collective demand for these services would reflect the variation in the aggregate drought/flood risk across this vulnerable region. Examples of sectoral impact - agriculture, mining and urban will be presented.

The two types of analyses referred to above will be used to demonstrate the need for and design of a multiscale stochastic modeling and resource allocation framework for portfolio risk management and institutional response. This approach will be sketched in the talk to provide a context for the two examples.