| 基于WheatGrow和CERES模型的区域小麦生育期预测与评价 |
| |
| 引用本文： | 吕尊富,刘小军,汤亮,刘蕾蕾,曹卫星,朱艳. 基于WheatGrow和CERES模型的区域小麦生育期预测与评价[J]. 中国农业科学, 2013, 46(6): 1136-1148. DOI: 10.3864/j.issn.0578-1752.2013.06.006 |
| |
| 作者姓名： | 吕尊富 刘小军 汤亮 刘蕾蕾 曹卫星 朱艳 |
| |
| 作者单位： | 南京农业大学农学院/国家信息农业工程技术中心/江苏省信息农业高技术研究重点实验室,南京,210095 |
| |
| 基金项目： | 国家自然科学基金（31271616）、国家科技支撑计划（2011BAD21B03）、江苏高校优势学科建设工程资助项目（PAPD） |
| |
| 摘 要： | 【目的】研究小麦生长模型在区域应用中的关键技术，并利用WheatGrow和CERES-Wheat两套模型模拟区域小麦生育期，以检验和评价模型区域应用方法的有效性。【方法】首先利用薄盘样条法（thin plate spline，TPS）对各站点逐日气象数据进行空间插值，得到研究区域气象要素表面数据；其次利用TPS方法对各站点历史多年小麦播种期进行空间插值，并将插值后的结果进行多年平均得到研究区域播种期表面数据；进一步将Markov Chain Monte Carlo（MCMC）方法与生长模型相结合，利用典型站点历史多年小麦生育期实测数据，估算出典型站点的品种参数，并将其作为各省份的代表性生态型品种参数；最后将生成的气象要素和播种期表面数据以及生态型品种参数等输入到WheatGrow和CERES-Wheat模型中，并以栅格为单元进行研究区域小麦生育期的模拟，进一步结合不同站点历史年份的生育期观测资料，检验和评价模型区域应用方法的有效性，并量化区域生育期模拟结果的不确定性。【结果】两个模型在区域尺度上的生育期预测效果均较好，区域尺度上拔节期、抽穗期、成熟期的预测值和观测值之间的R2分别为0.85、0.87和0.86（WheatGrow），0.87、0.85和0.82（CERES）；RMSE分别为9.6、7.2和6.3 d（WheatGrow），9.4、7.8和6.6 d（CERES）。另外，WheatGrow模型对抽穗期和成熟期区域预测的准确度略高于CERES-Wheat，但由品种参数导致的区域模拟结果的不确定性也相对较高。【结论】通过气象数据和小麦播期数据的TPS插值技术结合MCMC方法的品种参数估计技术，将基于机理的生育期模型拓展到区域尺度，较好地预测区域小麦生育期，量化了由品种参数导致模拟结果的不确定性，研究结果可为进一步量化中国小麦主产区的区域生产力提供技术支撑。
|
| 关 键 词： | 小麦 生育期 TPS插值方法 MCMC 区域应用 不确定性 |
| 收稿时间： | 2012-10-01 |
Regional Prediction and Evaluation of Wheat Phenology Based on the Wheat Grow and CERES Models |
| |
| Affiliation: | College of Agriculture, Nanjing Agricultural University/National Engineering and Technology Center for Information Agriculture/Jiangsu Key Laboratory for Information Agriculture, Nanjing 210095 |
| |
| Abstract: | 【Objective】The objectives of this study are to investigate the key techniques of wheat phenological model for regional application, and to evaluate the validity of regional application techniques through analyzing the simulated phenology results of WheatGrow and CERES-Wheat at regional scale. 【Method】First, the daily meteorological data at different weather stations were interpolated using the method of Thin Plate Spline (TPS) , and the gridded daily meteorological surface data was generated. Then, the historical sowing dates of wheat at different stations were interpolated using TPS method, and the multi-year average values of interpolated results were generated as regional sowing date surface data. In addition, by integrating the Markov Chain Monte Carlo (MCMC) method with the above two wheat models, the region-specific cultivar parameters were estimated with the measured phenology data from 1998 to 2003, and were taken as the typical cultivar parameters of each province. Finally, the regional daily meteorological data, sowing surface data and region-specific cultivar parameters were input into the above two models, which took grid as basic unit to simulate regional wheat phenology. Meanwhile, the simulated results were evaluated with observed data and the uncertainty coming from cultivar parameters was quantified at site and regional scales. 【Result】The estimated values from the above two models agreed well with the observed values at regional scale. R2 between estimated and observed values of jointing date, heading date and maturity date were 0.85, 0.87 and 0.86 (WheatGrow), 0.87, 0.85 and 0.82 (CERES), respectively. RMSE between the estimated and observed values of jointing date, heading date and maturity date were 9.6, 7.2 and 6.3 d (WheatGrow), 9.4, 7.8 and 6.6 d (CERES), respectively. By analyzing the simulated results of the two models at regional scale, the WheatGrow model had higher prediction accuracy than CERES-Wheat, while the uncertainty came from crop parameters were also higher than CERES-Wheat model.【Conclusion】Through combining the TPS interpolation method with MCMC-based parameter estimation techniques, the mechanism-based model could be effectively extended to regional scales to accurately predict the regional wheat phenology, and quantify the uncertainty coming from the parameter estimation. These results will provide a technology support for estimating crop productivity at regional scale in the future. |
| |
| Keywords: | wheat phenology TPS interpolating method MCMC regional application uncertainty |
| 本文献已被 万方数据 等数据库收录！ |
| 点击此处可从《中国农业科学》浏览原始摘要信息 |
|
点击此处可从《中国农业科学》下载全文 |
|
