甜玉米生育期推估及栽培管理之網路決策支援系統

Abstract

智慧農業結合資通訊技術與實際資料的使用，有望提升農民在氣候變遷下的調適能力。甜玉米的適採期的預測相當關鍵，太晚採收會使品質降低。氣溫是影響作物生長最重要的氣象因子，而生育度日 (GDD) 是用來評估溫度對於作物生長反應的重要工具。近期GDD模式則逐漸重視高溫對於作物的影響，但仍缺乏對於個別作物品種的參數校正，在預測甜玉米的生育期的誤差過大。為此，本研究提出了一套窮舉演算法來最佳化作物參數，參數最佳化模式的表現相較傳統GDD10,30算法，採收期預測日數與實際日數的R2自0.69 – 0.79提升到0.93 – 0.96，RMSE也從5.56至8.60天減少至2.39至3.90天。本研究並利用經最佳化的參數及GDD算法，發展出一套網路甜玉米決策支援系統，系統開發過程使用網路程式設計工具及開源氣象資料，設計出一套使用者友善的決策支援系統，可於手機及電腦上使用。其預測能力優於現行臺灣良好農業規範的生育日數法，及農試單位常用之GDD10,30生育度日法，解決傳統生育度日模式無法精確預測甜玉米生育期的問題。此外，使用者亦可自訂參數，未來若搭配實際物候資料與參數最佳化演算法校正，有潛力套用在不同作物的物候預測上。在氣候變遷下，提供農民、企業、學術及農政單位一個快速評估作物在不同氣候條件下的生長反應的工具。

Integration of information and communication technology (ICT) and real data in intelligence agriculture holds the potential to enhance farmers' adaptive capacity under climate change. Predicting the optimal harvesting period for sweet corn is crucial as late harvesting can result in reduced quality. Temperature is the most important weather factor influencing crop growth, and growing degree days (GDD) is an important tool for assessing the temperature response of crops. Recent GDD models have started considering the impact of high temperatures on crops but still lack parameter calibration for individual crop varieties, resulting in significant errors in predicting the growth stages of sweet corn. Therefore, this study proposes an exhaustive search method to optimize crop parameters, and the performance of the parameter optimization model is compared to the traditional GDD10,30 algorithm. The R2 value for predicting the harvesting period improved from 0.69-0.79 to 0.93-0.96, and the root mean square error (RMSE) decreased from 5.56-8.60 days to 2.39-3.90 days. Using the optimized parameters and GDD algorithm, a web-based decision support system for sweet corn was developed. The system utilizes web programming tools and open-source meteorological data to create a user-friendly decision support system that can be accessed through mobile phones and computers. The predictive capability of the system surpasses the current Taiwan Good Agricultural Practice's method for calculating growing days and the commonly used GDD10,30 method employed by agricultural research units, addressing the issue of inaccurate prediction of sweet corn growth stages by traditional growing degree day models. Additionally, users can customize the parameters, and future applications may include the calibration of actual phenological data and parameter optimization algorithms for different crops' phenological predictions. In the face of climate change, this system provides farmers, businesses, academia, and agricultural authorities with a tool to quickly assess crop growth responses under different climate conditions.