Statistical Regression

Regression techniques estimate crop yields using yield-affecting variables. A comprehensive list of possible variables that affect yield is provided in chapter 1. Usually, the weather variables routinely available for a historical period that significantly affect the yield are included in a regression analysis. Regression techniques using weather data during a growing season produce short-term estimates (e.g., Sakamoto, 1978; Idso et al., 1979; Slabbers and Dunin, 1981; Diaz et al., 1983; Cordery and Graham, 1989; Walker, 1989; Toure et al., 1995; Kumar, 1998). Various researchers in different parts of the world (see other chapters) have developed drought indices that can also be included along with the weather variables to estimate crop yield. For example, Boken and Shaykewich (2002) modifed the Western Canada Wheat Yield Model (Walker, 1989) drought index using daily temperature and precipitation data and advanced very high resolution radiometer (AVHRR) satellite data. The modified model improved the predictive power of the wheat yield model significantly. Some satellite data-based variables that can be used to predict crop yield are described in chapters 5, 6, 9, 13, 19, and 28.

The short-term estimates are available just before or around harvest time. But many times long-term estimates are required to predict drought for next year, so that long-term planning for dealing with the effects of drought can be initiated in time. Long-term predictions are especially useful for food-grain exporting countries, such the United States and Canada, because grain-exporting agencies are interested in knowing the amount of food grains available for export next year to help set the export targets. For such prediction, time series analysis can be performed.

0 0

Post a comment