基于遥感和AquaCrop作物模型的多同化算法比较

发布时间：2018-08-22 15:29

【摘要】：为了研究不同数据同化方法在AquaCrop(FAO Crop model to simulate yield response to water)模型模拟作物地上生物量(above ground biomass,AGB)、冠层覆盖度(canopy cover,CC)和产量过程的效率,以冬小麦为研究对象,利用2012-2013、2013-2014和2014-2015年冬小麦田间试验数据,将标定的Aqua Crop生长模型与遥感光谱信息相结合开展同化技术分析,应用粒子群优化(particle swarm optimization,PSO)、模拟退火(simulated annealing,SA)和复合型混合演化(shuffled complex evolution,SCE-UA)3种数据同化算法,以不同生育期、不同水分处理和不同氮肥水平的AGB和CC为双变量开展多同化算法的模拟分析,对3种数据同化算法的运算效率和同化结果进行对比分析。结果表明:1)3种数据同化算法达到的应度值0.26时,SCE-UA同化算法用时最少(833 s),SA数据同化算法用时最多(1433 s),表明SCE-UA同化算法效率最优,SA数据同化算法效率最低;2)不同生育期的同化结果,AGB的同化精度随着生育期的推进而降低,AGB的模拟值在拔节期和挑旗期高于实测值,被高估,在开花期和灌浆期被低估,总的AGB被低估;CC在拔节期和挑旗期被低估,在开花期和灌浆期被高估,总的CC被低估;3)不同水分处理的同化结果,AGB普遍被低估,CC在雨养(W0)条件下被高估,在正常灌溉(W1)和过量灌溉(W2)条件下被低估;产量均被低估;4)不同氮肥水平,AGB的模拟精度随着施N量的增加而降低,并且普遍被低估,CC普遍被高估,产量均被低估。以上结果表明,PSO、SA和SCE-UA 3种数据同化算法均能有效模拟冬小麦的AGB、CC和产量,其中SCE-UA数据同化算法无论在运算效率还是同化结果的精度上均优于PSO和SA数据同化算法。
[Abstract]:In order to study the efficiency of different data assimilation methods in simulating the aboveground biomass of (above ground biomass (above ground biomass AGB), canopy coverCC and yield process in AquaCrop (FAO Crop model to simulate yield response to water) model, winter wheat was used as the research object. Based on the data of winter wheat field trials in 2013-2014 and 2014-2015, the calibrated Aqua Crop growth model was combined with remote sensing spectral information to carry out assimilation analysis. Three data assimilation algorithms, particle swarm optimization (particle swarm), simulated annealing (SA) and compound hybrid evolutionary (shuffled complex evolution-SCE-UA (SCE-UA), were used to simulate and analyze the multivariate assimilation algorithms using AGB and CC with different growth stages, different water treatments and different nitrogen levels. The computational efficiency and assimilation results of three data assimilation algorithms are compared and analyzed. The results show that the three data assimilation algorithms have the lowest time (833s) and SA data assimilation algorithm (1433 s),) and the lowest SA data assimilation algorithm (1433 s), shows the best efficiency of SCE-UA assimilation algorithm and the lowest SA data assimilation algorithm efficiency 2) when the response value of the three data assimilation algorithms is 0.26 (833s) and (833s) SA data assimilation algorithm takes the most time (1433 s),). The assimilation accuracy of AGB decreased with the development of growth stage. The simulated value of AGB was higher than the measured value at jointing stage and flag picking stage. The total AGB was underestimated at jointing and flag stage, and was overestimated at flowering and filling stages. Total CC was underestimated. (3) the assimilation results of different water treatments were generally underestimated under rainfall (W0) condition and under normal irrigation (W1) and overirrigation (W2) condition. The simulation accuracy of AGB at different nitrogen levels decreased with the increase of N application rate, and the CC was generally underestimated and the yield was underestimated. The above results show that the three data assimilation algorithms PSOSA and SCE-UA can effectively simulate the AGBCC and yield of winter wheat, and the SCE-UA data assimilation algorithm is superior to the PSO and SA data assimilation algorithms in both operation efficiency and precision of assimilation results.
【作者单位】：北京农业信息技术研究中心;国家农业信息化工程技术研究中心;农业部农业信息技术重点实验室;北京市农业物联网工程技术研究中心;商丘师范学院测绘与规划学院;
【基金】：国家自然科学基金(41571416,41601346) 北京市农林科学院创新能力建设专项(KJCX20150409) 北京市自然科学基金(4152019)
【分类号】：S127

【参考文献】