不同误差订正方法在中国区域气候模拟中的比较和应用

发布时间：2018-10-19 14:33

【摘要】：全球和区域尺度的气候模式,是进行气候变化模拟和预估研究的首要工具,但受气候系统复杂性和科学发展水平的制约,它们对当代气候的模拟与观测相比,也总是存在或多或少的偏差,包括在平均态和概率密度分布等多方面。若将气候模式结果直接应用于驱动如水文和农业等影响评估模式时,其偏差则会对模拟产生很大影响,因此需要进行误差订正工作。本研究主要通过分位数映射的方法,对再分析资料ERA-interm驱动下,RegCM4.4区域气候模式所模拟的当代日降水量、日平均气温、日最高气温和日最低气温进行订正。研究中,将全年划分为冬(12~2月)、春(3~5月)、夏(6~8月)和秋(9~11月)四个季节。首先以对逐日降水量进行订正试验,以模拟时段1991~2010年中的前半段(1991~2000年)作为参照时段,建立传递函数,对后一时段(2001~2010)进行订正并检验其效果。对使用参数和非参数所建立的6种不同传递函数方法进行对比,发现6种方法均可明显减少降水模拟的误差,其中RQUANT相对表现更为突出一些,即被选择做为主要订正方法,用于RegCM4所模拟降水和气温结果的订正。将RQUANT方法应用到对日降水量的订正上,发现经过误差订正后的结果可以大大减少模式的偏差,使中国大部分地区的降水量与观测的相对误差集中在-25%~25%之内;可以大幅度提升与观测的空间相关系数并减少误差标准差,使降水量无论从分布形态还是数值上都更接近观测数据;而对降水的年际变率也同样有一定的改善作用。由于模式本身对于气温的模拟效果要好于降水,因此误差订正对气温的订正结果与观测更接近,空间相关系数达到0.99以上。RQUANT方法对模式模拟的平均气温、日最高气温和日最低气温的订正效果都非常明显,中国大部分地区的订正结果与观测的偏差在±1°C之间;提高了空间相关系数,减少误差标准差;对气温年际变率的订正效果不太明显,但春季的空间相关系数提升明显,而且减少了年际变率的误差标准差。RQUANT方法对极端事件同样有着很好的订正效果,减少了模式对4个极端事件指数CDD(连续干旱日数)、SDII(强降水指数)、TXx(年极端最高气温)和TNn(年极端最低气温)的模拟偏差,提高了空间相关系数,减少了误差标准差,使订正结果更加符合观测数据。
[Abstract]:Climate models at the global and regional levels are the primary tools for modelling and predicting climate change. However, due to the complexity of the climate system and the level of scientific development, they are compared to contemporary climate modelling and observation. There are always more or less deviations, including average state and probability density distribution. If the climate model results are directly applied to driving the impact assessment models such as hydrology and agriculture, the deviation will have a great impact on the simulation, so it is necessary to carry out error correction work. In this study, the daily precipitation, daily mean temperature, daily maximum temperature and daily minimum temperature simulated by the RegCM4.4 regional climate model were revised by quantile mapping method driven by ERA-interm reanalysis data. In the study, the whole year is divided into four seasons: winter (December ~ February), spring (March ~ May), summer (June ~ August) and autumn (September ~ November). Firstly, the daily precipitation was revised, and the first half of the period from 1991 to 2010 (1991 ~ 2000) was used as the reference period. The transfer function was established, the later period (2001 ~ 2010) was revised and its effect was tested. By comparing the six different transfer function methods established by using parameter and non-parameter, it is found that the six methods can obviously reduce the error of precipitation simulation, and the relative performance of RQUANT is more prominent, that is, it is chosen as the main correction method. It is used to revise the simulated precipitation and temperature results of RegCM4. Applying the RQUANT method to the correction of daily precipitation, it is found that the deviation of the model can be greatly reduced after the error correction, and the relative error between precipitation and observation in most areas of China is concentrated within -25%. The spatial correlation coefficient can be greatly increased and the error standard deviation can be reduced, so that the precipitation is closer to the observed data in terms of distribution and numerical value, and the interannual variability of precipitation can also be improved to a certain extent. Because the effect of the model itself on temperature simulation is better than that of precipitation, the correction result of error correction is closer to that of observation, and the spatial correlation coefficient is more than 0.99. The average temperature simulated by RQUANT method is better than that of the model. The correction effect of daily maximum temperature and daily minimum temperature is very obvious, the deviation between the results of correction and observation in most areas of China is 卤1 掳C, the spatial correlation coefficient is increased, and the error standard deviation is reduced. The effect on the correction of temperature interannual variability is not obvious, but the spatial correlation coefficient in spring is obviously increased, and the error standard deviation of interannual variability is reduced. The RQUANT method also has a good correction effect for extreme events. The simulation deviation of four extreme event indices (CDD (continuous drought days), SDII (heavy precipitation index (), TXx () and TNn (annual extreme minimum temperature) is reduced. The spatial correlation coefficient is increased and the error standard deviation is reduced. The revised results are more in line with the observed data.
【学位授予单位】：中国气象科学研究院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P435

【相似文献】