Comparacao de tres modelos de previsao da safra de café no Estado de Minas Gerais
In order to predict coffee yield for the State of Minas Gerais, three models were developed and evaluated. They are: simple model, biennial model, which separates odd and even years, and composed model, which weights three regional models. The models were constructed through multiple regression. The climatic factors including such as precipitation; maximun temperature, minimun temperature, relative humidity, potential evapotranspiration, and excess and deficit water were considered in the process of model construction. The monthly climatic data from the years of 1964 to 1986 at the stations of Patos de Minas, Lavras and Machado were used in this study. The data within the period of 1964 to 1983 were used for model construction; and the rest of the years were used for model testing. The simple model (Ycs3) had a R2 of 0.880 with a mean error of 11.4 per cent ranging from 0.3 per cent to 30.9 per cent. The biennial model with even year (Ycs5) had a R2 of 0.972 and a mean error of 2.9 per cent ranging from 0.4 to 9.5 per cent; with odd year (Ycs6) had a R2 of 0.988 and mean error of 2.2 per cent ranging from 0.3 to 7.9 per cent. The composed model (YmI) had a mean error of 15.8 per cent ranging from 1.5 percent to 68.2 per cent. The prediction error of years 1984 and 1985 were 3.8 per cent and 8.1 per cent by the simple model 13.5 per cent and 13 per cent by the biennial model and 14.3 per cent by the composed model. It was concluded that the simple model had better prediction precision; the biennial model had better statistical and the composed model worked well for the recent years. The composed model could be improved by including a model of macroregion of Mata and Rio Doce
| Main Authors: | , |
|---|---|
| Format: | biblioteca |
| Published: |
Ago
|
| Subjects: | COFFEA, PRONOSTICO DE COSECHAS, MODELOS, PRONOSTICO DEL RENDIMIENTO, EFECTOS DEL MEDIO AMBIENTE, BRASIL, |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|