作物生長模式及統計模式於台灣氣候變化對水稻產量預測之比較

A Comparison of the Crop Process Models and Statistical Models to Predict Rice Yield to Climate Changes

洪偉勝(Wei-Sheng Hung)

指導教授 ： 劉仁沛

共同指導教授 ： 蔡政安(Chen-An Tsai)

國立臺灣大學/生物資源暨農學院/農藝學研究所/碩士(2012年)

https://doi.org/10.6342/NTU.2012.10865

摘要

近年全球氣候暖化，台灣位處亞洲季風氣候區，亦受氣候變遷的影響，氣候改變勢必影響水稻產量。本研究使用1985年至2009年台灣實際水稻產量及天氣資料建構迴歸模式和時間序列模式等兩個統計模型，並應用統計模型及作物生長模式ORYZA2000預測台灣2000至2009年過去10年水稻平均產量，並與實際平均產量作比較。
應用ORYZA2000與統計模式預測2000-2009年溫度上升1-2℃或下降1-2℃對產量的影響。因為無台灣本土水稻生理資料、田間試驗資料及土壤資料等執行作物生長模式必需之相關參數資料，所以應用ORYZA2000預測時是使用內建之水稻生理、田間試驗資料及土壤資料，唯獨天氣資料是使用台灣本土實際天氣數據。ORYZA2000預測結果為當溫度上升，水稻產量並無一致變化趨勢，但一般而言溫度下降時水稻產量出現增加現象，由於目前參數資料來源不足，預測結果可能不一致。
統計模型著重於利用平均溫度及平均降雨量預測溫度變化對水稻產量的影響，迴歸分析結果發現當溫度上升1-2℃時，一、二期作之每公頃平均總產量至多會減少約4.2-8.4%，約0.17-0.34噸；溫度下降1-2℃時，產量會增加4.2-8.4 %，約0.17-0.34噸。時間序列分析結果發現當溫度上升1-2℃時，每公頃平均總產量至多會減少4.19-8.4%，約0.17-0.34噸；溫度下降1-2℃時，產量會增加約4.19-8.4%，約0.17-0.34噸。
本研究發現迴歸模式表現較好且預測平均溫度變化對平均總產量改變之影響於統計上為顯著，時間序列預測結果之平均溫度變化對平均總產量改變雖有影響但在統計上為不顯著，ORYZA2000因無台灣實際資料，故表現不理想。因ORYZA2000執行所需要大量參數資料，但台灣本土水稻生理參數及田間試驗等資料來源不足或缺乏，故目前可考量以統計模式暫時替代作物生長模式預測溫度變化下的水稻產量。

關鍵字

水稻 ； 產量預測 ； 作物生長模式 ； ORYZA2000 ； 氣候變遷 ； 迴歸分析 ； 時間序列分析

並列摘要

Recently, since the global climate changes and Taiwan is located in the Asian monsoon climate region, the weather in Taiwan would be changed by the global warming. Climate changes can affect the rice yields. In the study, we collected the actual average yield in rice and different weather variables for 14 regions in Taiwan from 1985 to 2009. Then we applied one process model- ORYZA2000, and two statistical models- regression model and time series model to predict the impact of weather changes on the average rice yield in Taiwan.
All three models were applied to predict the effects of the rises or decline from 1 to 2 degrees centigrade in average temperature on the average rice yields. Because there are no local crop data, experiment data and soil data were collected in Taiwan, we used the default crop data file, experiment data file, soil data files provided in ORYZA2000 program and the actual weather data in Taiwan to predict the effect of changes in average temperature on the average yield. The results show no constant trend of yield changes will respect to the changes in average temperature.
The results from regression analysis show the average rice yields per year would reduce by 4.2 to 8.4%, approximately from 0.17 to 0.34 tons per hectare per year as the average temperature rises from 1℃ to 2℃. The average yield per year will increase from 4.2 to 8.4%, approximately from 0.17 to 0.34 tons per hectare as the average temperature declines from 1℃ to 2℃. The results from time-series analysis show the average rice yields per year would reduce by 4.19 to 8.4%, approximately from 0.17 to 0.34 tons per hectare as the average temperature rises from 1℃ to 2℃. The average rice yield per year will increase from 4.19 to 8.4%, approximately from 0.17 to 0.34 tons per hectare as the average temperature declines from 1℃ to 2℃.
Among the three models, the performance of the regression model in predicting yields is the best and the impact of average temperature on the average rice yields is statistical significant, followed by the time series model but the results of the impact of average temperature on the average rice yields of time series model is not statistical significant. Because of no actual data file in Taiwan, the performance of ORYZA2000 is not consistent. Because the local data for parameter required by ORYZA2000 have not been collected in Taiwan, the statistical models may currently substitute the crop process model for predicting the impact of changes in weather on the rice yield.

並列關鍵字

Rice ； Predicted Yield ； Crop Process models ； ORYZA2000 ； Climate Changes ； Regression analysis ； Time-series analysis

參考文獻

行政院農業委員會農糧署統計室(2010)。台灣糧食統計要覽。行政院農業委員會農糧署。

行政院農業委員會統計室(2004)。台灣糧食統計要覽。行政院農業委員會。

行政院農業委員會統計室(2003)。台灣糧食統計要覽。行政院農業委員會。

行政院農業委員會統計室(2002)。台灣糧食統計要覽。行政院農業委員會。

行政院農業委員會統計室(2001)。台灣糧食統計要覽。行政院農業委員會。