New procedure for developing forecasting models
1996
Calvero, S.B. | Coakley, S.M. | Surin, A. | Kim, C.K. | Teng, P.S. (International Rice Research Inst., Los Banos, Laguna (Philippines))
A new procedure for generating empirical forecasting models is presented with rice-Pyricularia grisea (rice blast) as the pathosystem of interest. This procedure employs the use of weather factor searching computer software called WINDOW PANE, path analysis (PA), and multiple linear regression (MLR) analysis. Data on disease variables (such as leaf and panicle blast severity and incidence) on blast-susceptible rice cultivars were obtained from three blast hot spot sites in Asia: Cavinti in the Philippines during 1992-1993, Suphanburi in Thailand during 1993-94, and Icheon in South Korea during 1974-89 to serve as the response variables in the empirical models. Historical time-series weather data sets that include precipitation (rainfall, mm/d), temperature (minimum, maximum and mean), relative humidity (percent), wind speed (m/s) sunshine duration (h), and solar radiation (MJ/sq.m) were also obtained from each of the three sites to serve as the explanatory variables. The WINDOW PANE program showed that several weather factors were highly correlated with the disease variables. PA, however, revealed that not all of these factors have large direct effects on disease development. MLR models for large direct effects on blast, and these were consecutive days with mean temperature at optimum range (optimum temperature: 20 to 27 deg C) and total precipitation at Cavinti, mean maximum temperature and mean wind speed at Suphanburi, and number of days with precipitation greater than or equal to 84 mm/d and number of days RH greater than or equal to 80 percent at Icheon. Predictive abilities of the models were also tested by cross-validation procedures
显示更多 [+] 显示较少 [-]