Statistical Analyses to Support the National Land and Water Information Service System of Canada

J. O. Ramsay, McGill University

The National Land and Water Information Service (NLWIS) is an initiative of Canada's Agricultural Policy Framework, and is to provide land, soil, water, climatic and biodiversity resource information to land-use decision makers to support an environmentally sustainable agricultural sector. This major federally funded project is under Agriculture and Agri-Food Canada.

Agroclimatic Information Service is one of the seven components of the NLWIS, and has the goal of providing decision-making resources to Canadian food producers. A critical focus is the supplying of information that is spatially and temporally specific about the distribution of precipitation. This information base will feed into optimal risk management tools that are to be developed as a part of this project.

A project proposal was prepared in 2006 by Jim Ramsay, Sam Shen and Jim Zidek (McGill, San Diego State and UBC, respectively) to develop statistical analyses for this project using various sources of data, including daily historical data from thousands of weather stations distributed over the Canadian prairies. The proposal was made to the National Program for Complex Data Sets, and funding was provided for an initial workshop on the problem. This workshop will take place in June 2007 in Regina, and will bring together experts on the data to be used, spokespersons for the needs of the agricultural producer communities, and statisticians who have already worked in space/time data environments like these.

A statistical problem now being researched that would play an important role in this work is the estimated of distributed quantile functions. The quantile function is the inverse of the probability density function, and maps probabilities to data values, and the goal is to estimate this function as varying over space and time. A successful estimation strategy would provide users with localized information about the entire distribution in an easily consumable form.