[关键词]
[摘要]
玉米当季产量预测对农民制定栽培管理方案和政府决策者制定粮食战略都至关重要,作物过程模型与天气预报策略结合实现作物当季产量预测已经被广泛应用,然而该方法缺少在农户实际生产中的检验。本研究基于河北省曲周县2年(2017—2018年)农户跟踪数据和DSSAT模型,2017和2018年分别使用14位农户数据对当地主栽品种‘登海605’的遗传参数进行校准和验证;然后通过动态时间规整(DTW)算法检验气象数据时间序列的相似性,筛选出与预测年份气象数据相似度最高的历史年份,使用当季实时天气数据与历史年份数据结合的天气预报策略生成完整的玉米季天气数据,实现当季玉米产量预测。结果表明,校准后的DSSAT-CERES-Maize模型能够准确模拟玉米开花期日期(ARE: 2.19%, nRMSE: 2.53%),生物量(ARE: 7.55%, nRMSE: 9.50%)和产量(ARE: 5.70%, nRMSE: 6.60%);以DTW算法为基础的天气预报策略与DASST模型结合能够提前30—43天获得准确的预测产量(±8%)。该研究表明天气预报策略与作物过程模型结合的产量预测方法在农户管理的田块上表现较好,能够为农户当季产量预测和管理提供指导和借鉴。
[Key word]
[Abstract]
In-season maize yield prediction is important for farmers to make cultivation management plans and government decision-makers to formulate food strategies, while the combination of crop process model with meteorological forecast strategy has been widely used in crop yield prediction, however, this method lacks of testing in farmers' actual production. This study was conducted in Quzhou county, Hebei Province for 2 years (2017—2018) incorporated the DSSAT model based on the tracking data of householders. In 2017 and 2018, 14 farmers data were used to calibrate and verify the genetic parameters of the local main cultivar— ‘Denghai 605’. Then, dynamic time warping (DTW) algorithm was employed to test the similarity of time-series meteorological data between historical and forecasted years, and the most similar years were selected to generate the entire weather data in combination of real-time weather data during maize growing season. Finally, predicting the in-season maize yield with the entire weather data. The results showed that the calibrated DSSAT-CERES-Maize model could accurately simulate maize anthesis day (ARE: 2.19%, nRMSE: 2.53%); biomass (ARE: 7.55%, nRMSE: 9.50%) and yield (ARE:5.70%, nRMSE: 6.60%). Weather forecasting strategies based on the DTW algorithm combined with the DASST model can accurately predict yields (±8%) 30 -- 43 days ahead of harvest date. In conclusion, the method, combining weather forecasting strategy with process-based model, had better performances in farmers’ field, and supported in-season yield prediction and guided managements.
[中图分类号]
S145.6
[基金项目]
国家重点研发计划项目(2017YFD0200107)资助