中国农业气象

• 论文 • 上一篇    下一篇

水稻格点作物模型在中国区域的不确定性评估

孙擎,杨再强,车向红,杨世琼,王琳,郑晓辉   

  1. 1.南京信息工程大学气象灾害预报预警与评估协同创新中心,南京 210044;2.南京信息工程大学江苏省农业气象重点实验室,南京 210044;3.中国测绘科学研究院,北京 100830;4.北京师范大学全球变化与地球系统科学研究院,北京 100875
  • 收稿日期:2018-08-22 出版日期:2019-03-20 发布日期:2019-03-16
  • 基金资助:
    国家留学基金;江苏省研究生培养创新工程项目(KYLX16_0944)

Uncertainty Evaluation of Rice Gridded Crop Model in China

SUN Qing, YANG Zai-qiang, CHE Xiang-hong, YANG Shi-qiong, WANG Lin, ZHENG Xiao-hui   

  1. 1. Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, Nanjing University of Information Science & Technology, Nanjing 210044, China; 2. Jiangsu Key Laboratory of Agricultural Meteorology, Nanjing University of Information Science & Technology, Nanjing 210044; 3. Chinese Academy of Surveying & Mapping, Beijing 100830; 4. College of Global Change and Earth System Science, Beijing Normal University, Beijing 100875
  • Received:2018-08-22 Online:2019-03-20 Published:2019-03-16
  • About author:孙擎(1989-),博士生,主要从事作物模型、农业气象灾害及风险研究。E-mail: sunqingmeteo@gmail.com

摘要: 作物模型是评估气候变化对农业生产影响的主要手段之一,但中国对格点作物模型间的比较研究尚处于初始阶段。为全面评估不同作物模型在中国不同区域对水稻产量模拟的有效性,利用联合国粮农组织(FAO)和中国农业农村部种植业管理司(SYB)水稻年平均产量统计资料,对由2种气候资料(AgMERRA和WFDEI-GPCC)和 3种不同种植管理情景(Default、Fullharm和Harmnon情景)驱动的7种全球格点作物模型(CGMS-WOFOST、CLM-CROP、EPIC-BOKU、GEPIC、LPJML、PDSSAT和PEPIC模型)模拟的中国水稻产量进行了对比分析。结果表明:不同格点作物模型之间的模拟结果差异较大,在不同区域不同格点作物模型的模拟效果差异显著,不同格点作物模型对气候变化和种植管理情景的响应和敏感性不同,大部分模拟结果低估了水稻产量。使用不同水稻统计产量数据会对评估结果产生一定的影响。格点作物模型能够一定程度上模拟出水稻产量的年际变化和气候变化对产量的影响,但对于统计水稻产量上升的趋势较难模拟。通过综合分析产量在时间和空间上的波动情况,并利用2种评分方法对模拟表现打分,发现LPJML和PDSSAT在7种格点作物模型中模拟效果最好,同时也对不同气候数据和种植管理情景的变化最敏感,CLM-CROP的模拟效果最差。对不同种植管理情景,Default情景下的模拟效果显著高于Fullharm和Harmnon情景。多种格点作物模型集合平均(MME)可以降低单个格点作物模型模拟的误差,但需对MME中的集合模型进行挑选。

关键词: 全球格点作物模型(GGCM), 部门间影响模型比较计划(ISIMIP), 水稻产量, 多种作物模型集合平均(MME)

Abstract: Crop model is a widely-used strategy to evaluate the impact of climate change imposed on agriculture. The inter-comparison of gridded crop model is still at initial stage in China, thus making it crucial to comprehensively evaluate the performance of each global gridded crop model (GGCM). Using the statistics derived from Food and Agricultura Organization of the United Nations(FAO) and Ministry of Agriculture and Rural Affairs of the People’s Republic of China(SYB) statistical rice yearly mean yield, this paper compared the simulated rice yields of 7 GGCMs (i.e. CGMS-WOFOST, CLM-CROP, EPIC-BOKU, GEPIC, LPJML, PDSSAT and PEPIC), which are driven by 2 climate datasets (AgMERRA and WFDEI-GPCC) and 3 management scenarios (Default, Fullharm and Harmnon) from 1980 to 2009 in China. The comparisons show that the simulated rice yields from different GGCMs have significant discrepancy in different regions of China. Each GGCM has different response and sensitiveness to climate datasets and management scenarios. The majority of simulations underestimate rice yield in China even though different statistical rice yield data will affect evaluation results. To some degree, GGCMs are able to simulate the inter-annual variation of rice yield and climate change effects, but hardly simulate the pattern of rice yield increase of statistics. The analyses of rice yield fluctuation on temporal and spatial aspect demonstrate LPJML and PDSSAT perform better among 7 GGCMs using 2 skill score approaches, and are most sensitive to different climate datasets and management scenarios, while CLM-CROP have lowest stimulation accuracy. In terms of management scenarios, the simulation on Default scenario performs dramatically better than that of Fullharm and Harmnon scenarios. In addition, Multi-gridded crop model ensemble (MME) could reduce simulation error compared to single GGCM, but requires suitable members to precisely perform MME.

Key words: Global gridded crop model (GGCM), The Inter-Sectoral Impact Model Intercomparison Project (ISIMIP), Rice yield, Multi-crop model ensemble (MME)