基于序贯高斯模拟的日最低气温空间不确定性评估

Assessment for spatial uncertainty of daily minimum temperature by using sequential Gaussian simulation

明确日最低气温对于评估低温对作物的危害、指导人们及时采取补救措施、保障粮食安全具有重要意义.克里格是近地面温度场插值的主流方法,但其平滑效应会导致低值区域被过高估计而高值区域被过低估计.对2011年12月12日冷空气影响下的海南岛日最低气温,采用交叉验证法评估了普通克里格和带漂移的克里格两种插值法的预测精度;并对克里格插值法和序贯高斯模拟法产生的当日海南岛最低气温的空间分布进行对比分析.结果表明:带漂移克里格法的预测精度(r=0.86)并不显著优于普通克里格法(r=0.86);序贯高斯模拟能产生多个等概率的符合数据整体分布和方差函数的模拟结果,模拟结果克服了克里格插值的平滑效应,能够比克里格插值更加真实地反映当日最低气温的空间分布;在低温区域,气温变化小,序贯高斯模拟结果的条件方差小于普通克里格方差;潜在寒害区的空间不确定性能够通过多个序贯高斯模拟实现并加以量化.序贯高斯模拟在低温导致农业气象灾害的评估中具有较高的应用价值.

Understanding daily minimum temperature is of great importance for assessing low tem- perature damages to crops and guiding people to take timely remedial measures to ensure food secur- ity. Kriging is a widely used technology for mapping the spatial distribution of the near-surface tem- perature. However, the smoothing effect, commonly found in the Kriging maps, leads to low values to be overestimated and high values to be underestimated. For daily minimum temperature on Hai- nan Island which was affected by cold air on December 12, 2011, cross-validation was adopted to evaluate the prediction accuracy of ordinary Kriging （OK） and Kriging with external drift （KED）. The spatial distribution maps of daily minimum temperature on Hainan Island on December 12, 2011 produced by OK and sequential Gaussian simulation （SGS） were compared. Results showed that the prediction accuracy of KED （ r = 0.86 ） was not superior to OK （ r = 0.86 ） significantly. SGS could generate multiple equiprobable simulation realizations, and the distribution and variance function of the original data could be reproduced in the realizations. The simulation realizations gen- erated by SGS overcame the smoothing effect of Kriging and could more truly reflect the spatial dis- tribution of minimum temperature on the day on Hainan Island. In the region where daily minimum temperature was low, and the temperature change was small, the conditional variance of the SGS re- sults was less than the ordinary Kriging variance. Spatial uncertainty of a potential chilling damage area could be quantified by multiple simulation realizations generated by SGS. SGS was a valuabletool for assessing agro-meteorological disasters caused by low temperature.