The Climate Change and Human Health Integrated Assessment Web

Background

There are several reasons why spatial analysis is key to our integrated assessment framework. First, different health outcomes require varying resolution of climate model projections, necessitating specialized statistical analysis to achieve realistic integration across the overall risk assessment. Second, many disease-carrying vectors disperse and spread disease spatially; the dispersal of infected individuals also spreads the disease spatially. This is apparent in the disease clusters observed for most infectious diseases. Finally, although one strives to incorporate all of the known covariates in the model, inevitably some will be missed. Spatially contiguous observations can serve as effective proxies for such missing covariates. The incorporation of the spatial component is known to have added not just to the realism of the models but also to more accurate prediction and prediction errors by borrowing from spatially contiguous observations (Clayton and Kaldor 1987; Yasui and Lele 1996; Lele and Taper 1996).

Proposed Analysis (September 1, 1997)

Spatial statistical analysis will help determine the spatial distribution of the regional populations at risk. Disease may shift in distribution, rather than simply expand or contract. For diseases such as dengue fever and Rocky Mountain Spotted Fever (RMSF), well validated mathematical models have proven useful (Focks 1993,1995; Mount 1989) but there are limitations to the purely deterministic models such as the Leslie Matrix model used for vector-borne diseases. They do not provide a measure for uncertainty of the predictions; such measures are central to supporting policy decisions. We will modify these mathematical models in two ways: First, we will make them stochastic, explicitly incorporating noise in the parameters. Such a modification, giving rise to Stochastic Differential Equations, has proven valuable in theoretical population ecology (May 1974; Dennis and Patil 1984). Second, we suggest explicitly incorporating spatial features to enhance ecologically-based disease models.

Applications

Methodology