不同空间分辨率驱动数据对作物模型区域模拟影响研究

Influence of Different Resolutions Data on Regional Simulation of Crop Model

为了探究不同空间分辨率驱动数据对作物模型区域模拟结果的影响,以中国北方冬麦区为研究区域,采用薄盘样条插值方法生成4种空间分辨率(5、10、25、50 km)的气象驱动数据,采用空间聚集方法构建相应空间分辨率的土壤参数集,以农业气象观测站数据为基础,通过泰森多边形方法扩展农田管理和作物模型品种参数,在以上基础上建立不同空间分辨率的WOFOST模拟平台,结合区域统计产量数据,诊断不同空间分辨率驱动数据对作物模型区域模拟的影响。研究表明:对于模拟的开花期、成熟期、潜在产量水平的地上生物量和穗质量、雨养产量水平的地上生物量和穗质量,4种空间分辨率的区域平均值模拟结果之间无显著差异;高分辨率驱动数据下,模拟结果分布上有更多的极值。不同空间分辨率的模拟结果均能反映冬小麦生长的空间分布规律;与同区域统计产量相比,不同空间分辨率下WOFOST雨养产量可以解释观测产量年际变异的75.4%~85.4%。不同空间分辨率的潜在产量和雨养产量与气候因子相关分析表明,生育期辐射可以解释16.6%~29.6%的潜在产量变化,生育期降水可以解释13.3%~17.8%的雨养产量变化。高空间分辨率的数据存贮量和计算时间分别是低空间分辨率的80和100倍以上。研究结果可以为作物模型区域应用,尤其是空间分辨率的选择提供理论依据。

Crop model is constructed on the field scale. There are simulated errors due to the up-scaling when it is applied from filed scale to larger scale. Regional applications of crop model are becoming universal methodologies to explore the interactions between crop,climate and management in agriculture systems to provide policy and decision making for food production. A spatial resolution should be determined before using crop model to simulate yield in the regional scale. There is a dilemma in considering the spatial resolution. A high spatial resolution simulation needs more hardware resource and expensive computing cost while a coarse resolution simulation would result in loss of spatial detail of variability. Therefore,exploring the uncertainties of regional simulation of crop model due to different spatial resolutions is essential to application of crop model in large scale. A WOFOST regional simulation platform was constructed at different spatial resolutions to quantify the simulation errors by spatial resolution of model force data. Five spatial resolution meteorological data（ 5 km,10 km,25 km and50 km） were interpolated by the thin plate smoothing spline method which was provided by the software ANUSPLIN. The corresponding resolution soil parameters were extracted from a fine resolution soil database by spatial aggregation. Spatial management and crop cultivar parameters were calculated from observed agro-meteorological sites and then extended to different resolutions by Thiessen polygon method.The simulated yields were compared at different resolutions and statistical yields,and it was found thatthe average value of anthesis,maturity date,total above ground production and yield at potential yield level and water limited level did not have significant difference. However,there were more extreme values in the high resolution. Each simulation of four resolutions can perform the spatial distribution of crop development. Compared with the regional statistical yields,simulated water limited yield at different resolutions contributed to the variation of statistical yields by 75. 4% ~ 85. 4%. The correlation analysis between potential yield and water-limited yield and climate factor indicated that the irradiation in the growing stage contributed to 16. 6% ~ 29. 6% of variability of potential yield,the precipitation in the growing stage contributed to 13. 3% ~ 17. 8% of variability of water-limited yield. The data storage capacity and computing cost at high resolution,i. e. 5 km was a hundred times more than coarse resolution,i. e. 50 km. The results provided theoretical and scientific basis for regional application,especially for selecting suitable spatial resolution for regional simulation.