以線性模型與時間序列分析建立雲林縣落花生產量預測模型

A Study on Applications of General Linear Models and Statistical Time Series to Prediction of the Peanut Production in Yun-Lin County

陳欣宜(Sin-Yi Chen)

指導教授 ： 劉仁沛

國立臺灣大學/生物資源暨農學院/農藝學研究所/碩士(2017年)

https://doi.org/10.6342/NTU201701336

摘要

落花生為臺灣重要栽培作物之一，為期作性作物，一年可種植二期作。落花生主產地為雲林、彰化、嘉義及臺中等地區，因常與其他作物輪作，栽培面積常隨輪作作物及前期市場價格而變動。生長環境部分，落花生性喜高溫，生長較不受日照長短及雨量分佈的限制，以種植於富含有機質的砂質壤土或砂土最為適宜。由於此項作物總產量易受種植面積及氣候影響，使價格因產量增減而波動，進而影響農民收益，未來期能藉由產量預測模型，於收穫前更精準掌握當年度落花生產量，以規劃適宜的產銷調節措施，穩定農產品價格，亦維護農民收益。
農作物單位面積產量易受氣候影響，為影響總產量關鍵因素。目前中央氣象局於全國各地設立觀測站，透過觀測站內相關儀器記錄氣溫、雨量及輻射量等資訊，然而在以儀器測量同時，易產生測量誤差，影響蒐集資料準確性，亦影響迴歸模型之建立。因此，本研究蒐集過去15年落花生產業資料及氣象資料，除以迴歸分析建立種植面積預測模型外，另以測量誤差分析及時間序列分析等2種方法建立單位面積產量預測模型，建立最終的總產量預測模式。

關鍵字

農情調查 ； 農情預測 ； 落花生 ； 測量誤差 ； 時間序列 ； 對數常態分布

並列摘要

Peanut is one of the most important crops in Taiwan, and it can be cultivated twice in a year. Yunlin, Changhua, Chiayi and Taichung are the main areas for producing the crop. Due to the habit of crop rotation, its planting areas are influenced by its previous price and the rotating item. In the case of appropriate environment, peanut prefers to higher temperature, and it is less sensitive to photoperiod and percipetation. In addition, sandy or sandy loam soil with abundant organic matter would be more suitable for peanut. With the influence in its planting area and climate change, the price would be susceptable to the total yield, thus having an impact on farmers’ income. In the future, it is expected to develope yield-prediction models which can precisely estimate the yield in the particular time. With these tools, the government would have been able to make a proper decision in agricultural market regulation, stabilize product prices and protect farmers.
Crop yield per hacter is easily affected by local climate, which become one of the key factor for total yield. By now, Central Weather Bureau has set up several weather stations around Taiwan. By meterological instruments, local temperature, percipetation and radiation can be recorded. While using measurement instruments, it leads to measurement error and affects the model-building of regression analysis. In this study, we applied the data from agricultural survey of peanut and local climate in Yunlin for the last 15 years. We used regression analysis to build the model of planting area. On the other hand, measurement error model and time series model is used to build the yield per hacter, thus having the prediction model of total yield.

並列關鍵字

Archis hypogaea ； Agricultural survey ； Agricultural prediction ； Measurement error model ； Time series model ； Lognormal distribution

參考文獻

2. 行政院農業委員會(2002)。90年農業統計年報。臺北市：行政院農業委員會。

11. 行政院農業委員會(2010)。98年農業統計年報。臺北市：行政院農業委員會。

12. 行政院農業委員會(2011)。99年農業統計年報。臺北市：行政院農業委員會。

13. 行政院農業委員會(2012)。100年農業統計年報。臺北市：行政院農業委員會。

14. 行政院農業委員會(2013)。101年農業統計年報。臺北市：行政院農業委員會。