水稻生育階段預測模式之建立與驗證

Establishment and Validation of Prediction Model for Rice Growth Stages

林士閎(Shih-Hong Lin) ； 呂椿棠(Chun-Tang Lu) ； 卓緯玄(Woei-Shyuan Jwo) ； 呂秀英(Hsiu-Ying Lu)

《台灣農業研究》 63卷1期 (2014/03) Pp. 30-42

https://doi.org/10.6156/JTAR/2014.06301.03

摘要

本研究主要目的在建立國內現有3個重要中晚熟品種（「台稉9號」、「台農71號」及「台農秈22號」）之三個重要生育時期（50%始蘗期、幼穗分化期及50%抽穗期）的水稻生育階段預測模式，作為推估水稻重要特定生育時期的依據，以有效執行田間栽培管理。利用參試品種在2006－2010年各兩期作不同插秧日期試驗，以各生育階段期間內每日的發育速率與有效積溫（即生育度數，growing degree days; GDD）建立線性回歸模式，並驗證模式之預測能力。首先將所有試驗資料以內部驗證方式進行建模與驗證，結果發現預測與實際發育速率間之差異的95%信賴區間皆包含0，顯示預測誤差在可接受的5%顯著水準範圍內；再合併所有試驗資料建立各品種之3個生育階段的GDD預測模式，所得到截距與回歸係數之估值皆落在前述驗證程序所得模式之該兩個係數估值的95%信賴區間之內，顯示此等GDD模式是穩健的，由此可推知各品種達到各生育階段所需之GDD。本研究亦發現不同品種在各生育階段的發育速率不盡相同，相較於「台稉9號」，「台農71號」稍早熟而「台農秈22號」稍晚熟，乃由於「台農71號」在50%始蘗期至幼穗分化期間的發育速率較快，而「台農秈22號」在插秧至50%始蘗期間雖發育速率較快但之後到50%抽穗期間發育速率變慢所致。因此為提升稻穀產量，不同品種應配合其關鍵生育階段施以適當的田間栽培管理。

關鍵字

有效積溫 ； 生育度數 ； 生育速率 ； 線性回歸模式 ； 預測誤差

並列摘要

The aim of this study is to establish the models for predicting three major growth stages (i.e., 50% tillering, panicle initiation, and 50% heading) of three mid-late maturing Taiwan rice varieties (i.e., 'TK9', 'TNG71', and 'TNGS22'). Data sets were collected from the field experiments of rice varieties planted at various transplanting dates in a two-crop system in 2006-2010. For every growth stage of each variety, a linear regression model for predicting growth stage was established using daily development rate during the period as the dependent variable and its corresponding daily effective accumulated temperature (growing degree days; GDD) as the independent variable. The predictive capability of GDD model was tested by internal validation. It showed that all the 95% confidence intervals (CI) of the differences between predicted and actual values of development rate included zero, which reveals an acceptable predictive error at 5% significance level. Then, all data sets were pooled to build up the GDD models for predicting growth stage. The estimated values of intercept and regression coefficient in these GDD models were within the 95% CI of those estimated values obtained from above validation procedures. Results suggest that the established GDD models for predicting rice growth stages are robust. With these models, GDDs need to reach each growth stage in three rice varieties could be estimated. Results also indicated that these three rice varieties were not completely identical in development rates during the periods of different growth stages. Although they are midlate maturing varieties, the sequence of full maturity for these three mid-late maturing varieties is in the order of 'TNG71', 'TK9', and 'TNGS22'. It was due to the growth of 'TNG71' was faster during the period from 50% tillering to panicle initiation, while the growth of 'TNGS22' was slower during the period from 50% tillering to 50% heading in spite of rapid growth before 50% tillering. Therefore, to increase paddy yield, proper field management should be arranged in critical growth stages in respective rice varieties.

並列關鍵字

Effective accumulated temperature ； Growing degree days ； Linear regression model ； Development rate ； Predictive error