Central America suffers from extremes of climate variability, with a limited societal capacity to mitigate these effects. Many rural people practice subsistence rain fed agriculture as a basic livelihood strategy, and as such are vulnerable to the impacts of drought or flood that can diminish or destroy a harvest. Furthermore, the drop in world coffee prices has led to the widespread loss of employment opportunities on the coffee plantations, which would otherwise mitigate dependence on subsistence agriculture. The heightened state of vulnerability in the region creates an opportunity for the effective application of climate information. With this food insecurity as a backdrop, ongoing collaboration between Central American3 and American1,2 researchers is applying new forecast interpretation techniques to probabilistic forecast. An example of this work is the interpretation of the seasonal climate forecast estimated by the Climate Outlook Forum, in Central America, coordinated by the Comit¨¦ Regional de Recursos Hidraulicos (CRRH). This adds to already existing information to help answer specific end-user questions. This presentation briefly describes the interpretation technique, the update of forecasts with observed precipitation as the season progresses and the evaluation of the use of forecast information in fields such as Agriculture, hydro electrical production, disaster mitigation, etc.

The UCSB/MFEWS forecast interpretation method, developed by Greg Husak and Joel Michaelsen, uses Monte Carlo resampling to translate probabilistic forecast terciles into a matching probability distribution function (pdf). The method begins with a 2 or 3 element vector of parameters, ¦Á, which typically define either a normal or conditional gamma distribution. The theoretical distribution is then resampled in accordance with a set of forecast probabilities. If a forecast calls for a 45%, 35%, 20% probability of rainfall within the wet, middle and dry terciles, the method might draw 45, 35, and 20 samples from the highest, middle and lowest terciles of the theoretical probability distribution. A new set of parameters is calculated and stored. The resampling process is repeated many times, and the median parameter values are used to define a new distribution (¦Á'). The new theoretical distribution can then be used to answer specific questions tailored to user end needs. Example question might be:

I manage a commercial bean agribusiness venture, and I know that beans generally require 700 mm of rainfall during August, September and October. What is the chance of getting at least this much rainfall?

This forecast interpretation method brings a new set of tools to decision makers and agroclimatic outlooks. This research effort will target a small set of end users, and track the utility of the interpreted seasonal forecasts through interviews. We describe several of these forecast applications, comment on their success and failures, and sketch the future scope of this collaborative activity.