Towards maize yield prediction in the Katumani area of Kenya
In this study, attempts were made to relate yield of maize grown at Katumani Agrometeorological Station to rainfall and air temperature variables by who approaches. In the first approach, yield was regressed on interphase rainfall totals. The results indicate that the flowering to wax ripeness interphase period is the most Sensitive to unit changes in rainfall. They also indicate that the approach cannot be used to predict yield till end of the season when the best combination of variable's, SER, ELR and FWR, can be obtained. A relatively stage error was found presumably due to the use of data for a short run. In the second approach, the yield was regressed on rainfall alone, then on rainfall and temperature combinations employing Fisher's (1924) regression technique as modified by Hendricks and Scholl (1943). The seasons were divided into 3, 5, 7 and 1J-day periods~ First and second degree equations were tried. The results were much improved. Out of all the combinations of variables and periods, rainfall and temperature range in 3-day periods in second degree equation was best, giving a high coefficient of determination and minimum error. Unlike the best equation in the first approach this equation can at any tjme of the season be used to assess the effect of weather v6riables on maize yield. Much of the yield variability seems to be vii due to little and/or bad distribution of rainfall. In using this equation, it was noted that the maize is affected differently during its different stages of growth and development. Above average 3-day totals of rainfall had a favourable effect up to about tasseling/ flowering time, thereafter the effect was reduced, and negative up to maturity. Above average 3-day temperature range had a similar effect, being favorable up to about tasseling time and then a progressively reduced and negative effect up to maturity.